Split (3)

train · 4.35k rows

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
Autophagy defects are implicated in multiple late-onset neurodegenerative diseases including Amyotrophic Lateral Sclerosis ( ALS ) and Alzheimer’s , Huntington’s , and Parkinson’s diseases . Since aging is the most common shared risk factor in neurodegeneration , we assessed rates of autophagy in mammalian neurons during aging . We identified a significant decrease in the rate of constitutive autophagosome biogenesis during aging and observed pronounced morphological defects in autophagosomes in neurons from aged mice . While early stages of autophagosome formation were unaffected , we detected the frequent production of stalled LC3B-negative isolation membranes in neurons from aged mice . These stalled structures recruited the majority of the autophagy machinery , but failed to develop into LC3B-positive autophagosomes . Importantly , ectopically expressing WIPI2B effectively restored autophagosome biogenesis in aged neurons . This rescue is dependent on the phosphorylation state of WIPI2B at the isolation membrane , suggesting a novel therapeutic target in age-associated neurodegeneration . Aging is a complex process that often impairs physiological and tissue function . Further , age is the most relevant risk factor for many prominent diseases and disorders , including cancers and neurodegenerative diseases ( Niccoli and Partridge , 2012 ) . Macroautophagy ( hereafter referred to as autophagy ) is an evolutionarily conserved , cytoprotective degradative process in which a double membrane engulfs intracellular cargo for breakdown and recycling ( Cuervo et al . , 2005; Rubinsztein et al . , 2011 ) . The autophagy pathway has been directly implicated in aging in model organisms ( Cuervo , 2008; Rubinsztein et al . , 2011 ) . Neurons are post-mitotic , terminally differentiated cells that must maintain function in distal compartments throughout the lifetime of a human . These maintenance mechanisms may wane as a person ages , potentially contributing to neuronal dysfunction and death . Accordingly , misregulation of autophagy has been associated with multiple age-related neurodegenerative diseases , including Alzheimer’s disease ( AD ) , Parkinson’s disease , Huntington’s disease , and amyotrophic lateral sclerosis ( ALS ) ( Menzies et al . , 2017; Nixon , 2013; Yamamoto and Yue , 2014 ) . Furthermore , specifically disrupting autophagy in neurons results in neurodegeneration in animal models ( Hara et al . , 2006; Komatsu et al . , 2006; Zhao et al . , 2013 ) . Despite the implication of this pathway in neurodegenerative disease , autophagy is best understood for its roles in maintaining cellular homeostasis in yeast and mammalian cells in response to acute stressors such as starvation ( Abada and Elazar , 2014; Hale et al . , 2013; Mariño et al . , 2011; Reggiori and Klionsky , 2013; Son et al . , 2012; Wu et al . , 2013 ) . Much less is known about how autophagy is regulated in neurons . Robust , constitutive autophagy functions at a constant , basal level in neurons both in vitro and in vivo . Autophagosomes are generated distally at the axon terminal or synapse and are then actively transported back to the soma during maturation to a fully-acidified degradative compartment ( Fu et al . , 2014; Hara et al . , 2006; Hollenbeck , 1993; Komatsu et al . , 2007; Maday et al . , 2012; Neisch et al . , 2017; Soukup et al . , 2016; Stavoe et al . , 2016; Yang et al . , 2013; Yue et al . , 2009 ) . In contrast to the pronounced induction of autophagy in other systems by cellular stressors , there is little evidence that neuronal autophagy is substantially upregulated by either proteomic stress ( Maday et al . , 2012; Wong and Holzbaur , 2014 ) or nutrient deprivation ( Maday and Holzbaur , 2016 ) . While recent progress has firmly linked autophagy and aging ( Chang et al . , 2017; Hansen et al . , 2018 ) , little is known about how this essential homeostatic mechanism in neurons is affected by aging . Autophagosome biogenesis , conserved from yeast to humans , involves over 30 proteins that act in distinct protein complexes to engulf either bulk cytoplasm or specific cargo within a double-membrane . A signature autophagy protein , LC3B , is used to label autophagosomes , as it is processed to a lipidated form that becomes tightly associated with the limiting membrane of the developing autophagosome . We have previously used live imaging to examine autophagosome biogenesis in dorsal root ganglion ( DRG ) neurons from transgenic mice expressing GFP-LC3B ( Maday et al . , 2012; Maday and Holzbaur , 2014 ) . Importantly , the spatially-specific pathway for constitutive axonal autophagy that has been extensively characterized in DRG neurons ( Fu et al . , 2014; Maday et al . , 2012; Maday and Holzbaur , 2014; Wong and Holzbaur , 2014 ) has been confirmed across multiple models , including hippocampal and cortical neurons in vitro ( Lee et al . , 2011; Maday and Holzbaur , 2014 ) and motor , touch , and interneurons in vivo in Drosophila and C . elegans ( Chang et al . , 2017; Neisch et al . , 2017; Soukup et al . , 2016; Stavoe et al . , 2016 ) . However , unlike hippocampal or cortical neurons , which are typically isolated from embryonic or early postnatal rodents , DRG neurons can be isolated from mice of any age and grow robustly in culture following dissection . As rates of autophagosome biogenesis in DRG neurons model those seen in vivo ( Soukup et al . , 2016; Stavoe et al . , 2016 ) and longitudinal studies indicate biogenesis rates remain constant over time in neurons in vitro ( Maday and Holzbaur , 2014 ) , DRGs represent a powerful model system to investigate autophagosome formation in mammalian neurons from aged mice with high temporal and spatial resolution . Young neurons appear to clear dysfunctional organelles and protein aggregates very efficiently ( Boland et al . , 2008 ) , but few studies have examined autophagy in aged neurons . Since age is the most relevant shared risk factor in neurodegenerative disease ( Niccoli and Partridge , 2012 ) , elucidating how autophagy changes in neurons with age is crucial to understanding neurodegenerative diseases . Here , we examine how autophagy is altered with age in primary neurons from mice . We find that the rate of autophagosome biogenesis decreases in neurons with age . This decrease is not due to a change in the kinetics of either initiation or nucleation during autophagosome formation . Instead , we find that the majority of autophagosome biogenesis events in neurons from aged mice exhibit pronounced stalling , remaining ATG13-positive and failing to recruit lipidated LC3B with normal kinetics . We observe pronounced morphological differences in autophagic vesicles in neurons from aged mice , including an increased frequency of multilamellar membranes , similar to observations of neurons from the brains of aging Alzheimer’s patients ( Nixon et al . , 2005 ) . Importantly , depletion of WIPI2 in neurons from young adult mice was sufficient to decrease the rate of autophagosome biogenesis to that of aged mice , while overexpression of WIPI2B in neurons from aged mice was sufficient to return the rate of autophagosome biogenesis to that found in neurons from young adult mice . Further , we find that the rescue of autophagosome biogenesis depends on the phosphorylation state of WIPI2B at the isolation membrane , suggesting that dynamic phosphorylation of WIPI2B regulates autophagosome biogenesis . Thus , while the rate of autophagosome biogenesis decreases in aged neurons , this decrease can be rescued by the restoration of a single autophagy component , suggesting a novel therapeutic target for future studies . Since impaired autophagy has been implicated in the pathogenesis of neurodegeneration and age is the most relevant risk factor for neurodegenerative disease , we used the GFP-LC3B probe to assess how biogenesis rates change with age in primary DRG neurons dissected from mice of four different ages: 1-month-old young mice , 3-month-old young adult mice , 16–17 month-old aged mice , and 24-month-old advanced aged mice . We produced robust cultures of DRG neurons harvested from mice aged from 1 to 24 months; neurons harvested from all ages extended long neurites , and we did not detect significant loss of viability with age ( data not shown ) . We used live-cell spinning disk fluorescence microscopy to examine autophagosome biogenesis at the axon tips of DRG neurons in culture with high spatial and temporal specificity . We identified autophagosome biogenesis events as the formation of discrete GFP-LC3B puncta visible above the background cytoplasmic GFP-LC3B signal ( Figure 1A ) . These puncta enlarged over approximately three minutes to form a 1 μm autophagosome . Strikingly , we found that the rate of autophagosome biogenesis significantly decreased with age , corresponding to a 53% decrease in autophagosome biogenesis in aged neurons compared to neurons from young adult mice . Furthermore , the decrease was even more pronounced in neurons from advanced aged mice ( Figure 1B ) . These data indicate that the rate of autophagosome biogenesis , as detected by the generation of GFP-LC3B-positive puncta , decreases in axon terminals with increasing age . In subsequent experiments , we focused on 16–17 month-old aged mice , given the significant decrease in the rate of autophagosome biogenesis observed at this time point relative to young adult mice and the relevance of this time point to the age of onset for age-associated neurodegenerative diseases such as ALS and AD . To further characterize changes in autophagic vesicle ( AV ) biogenesis in neurons during aging , we used transmission electron microscopy to compare the ultrastructure of AVs at axon terminals of neurons from young adult and aged mice . We observed stereotypical double-membrane structures with heterogeneous contents in the axonal tips of neurons from young adult mice ( Figure 1C–E , Figure 1—figure supplement 1A and C–E ) . However , in neurons from aged mice we more frequently observed aberrant AVs with a multilamellar ( onion skin-like ) structure ( Figure 1F–H , Figure 1—figure supplement 1B and F–M ) . Quantitative analysis indicated that only 34 . 0% of AVs ( n = 153 AVs ) observed in the distal tips of neurons from 16 to 17 month-old mice were morphologically normal , significantly different than the 80 . 4% of AVs judged to be morphologically normal in the axon tips of neurons from young adult mice ( n = 56 AVs; p<0 . 0001 by unpaired two-tailed Fisher’s exact test ) . The aberrant morphology of AVs observed in aged mice suggested misregulated membrane extension during AV biogenesis , consistent with previous observations that failure to lipidate LC3 at isolation membranes prevented closure and inhibited the degradation of the inner autophagosome membrane ( Tsuboyama et al . , 2016 ) . Furthermore , these aberrant AVs were reminiscent of AVs previously observed in aged rodents ( Majeed , 1993; Majeed , 1992 ) and in cortical biopsy specimens from patients with Alzheimer’s disease ( Nixon et al . , 2005 ) . We next queried whether we could detect these age-related morphological differences in intact neuronal tissues , focusing on the prominent synapses that form between motor neurons and muscle at the neuromuscular junction ( NMJ ) . We used NMJs from young adult and aged mice to assess any age-related changes in autophagosomes in vivo . Again , we observed stereotypical double-membrane structures with heterogeneous contents in neurons from young mice ( Figure 1I , Figure 1—figure supplement 1O ) . As we observed in DRG neurons in culture , we identified multilamellar structures in NMJs from aged mice in vivo ( Figure 1J , Figure 1—figure supplement 1N and P ) . Autophagosome biogenesis can be divided into stages: initiation/induction , nucleation , elongation , and membrane closure ( Figure 2A ) . The initiation complex , including ATG13 and ULK1/ATG1 , induces autophagosome biogenesis by phosphorylating other autophagy components ( Feng et al . , 2014; Kamada et al . , 2000; Reggiori et al . , 2004 ) . The nucleation complex , including VPS34 and ATG14 , generates phosphatidylinositol 3-phosphate ( PI3P ) at the site of autophagosome biogenesis ( Kihara et al . , 2001; Obara et al . , 2006 ) . Subsequently , the elongation complex , composed of two conjugation complexes , including ATG5 , ATG12 , and ATG16L1 , is required to conjugate phosphatidylethanolamine ( PE ) to LC3 to yield LC3-II ( Tanida et al . , 2004 ) . LC3-II is recruited to autophagosomes as the isolation membrane elongates during biogenesis and remains associated with autophagosomes until degradation of the internalized components . ATG9 , a six-pass transmembrane protein , is thought to shuttle to the growing membrane with donor membrane ( Koyama-Honda et al . , 2013; Orsi et al . , 2012; Sekito et al . , 2009; Suzuki et al . , 2015; Yamamoto et al . , 2012; Young et al . , 2006 ) . The ATG2 and WIPI4 complex is thought to work in concert with ATG9 to tether and provide lipids to the growing membrane ( Chowdhury et al . , 2018; Gómez-Sánchez et al . , 2018; Osawa et al . , 2019; Valverde et al . , 2019; Wang et al . , 2001 ) . Finally , the limiting membrane closes and fuses with itself to generate the unique double-membrane organelle . The autophagosome then undergoes retrograde transport along microtubules and subsequent fusion with lysosomes to degrade engulfed contents ( Figure 2A ) ( Xie and Klionsky , 2007 ) . The observed decrease in the rate of autophagosome biogenesis that we measured by monitoring GFP-LC3B-positive puncta could result from alterations to the initiation , nucleation , or elongation complexes . To determine which stage of autophagosome biogenesis is affected by age , we used live-cell imaging to compare the kinetics of each step of the pathway ( Figure 2A ) . We monitored the recruitment of the initiation complex by quantifying the appearance of fluorescent mCherry ( mCh ) -ATG13 puncta and observed similar kinetics of mCh-ATG13 recruitment in neurons from young , young adult , aged , and advanced aged mice ( Figure 2B–F ) . To examine the kinetics of nucleation , we examined the recruitment of Double FYVE-containing protein 1 ( DFCP1 ) , which binds to PI3P , the product of the autophagy nucleation complex ( Figure 2A ) . We did not detect a change in DFCP1 puncta formation with age ( Figure 2G–K ) . We examined elongation with the marker mCh-ATG5 , and again observed no change between neurons from young , young adult , and aged mice; we did observe a decrease in the rate of mCh-ATG5 puncta in neurons from advanced aged mice compared to young adult mice ( Figure 2L–P ) . Together , these data demonstrate that the decrease in the rate of autophagosome formation with age is not due to an alteration in the kinetics of the early stages of biogenesis . Next we used dual-color live cell imaging to compare assembly dynamics in neurons from young adult and aged mice co-expressing GFP-LC3B and initiation component mCh-ATG13 . We observed ‘productive’ biogenesis events in neurons from aged mice ( Figure 3A , Video 1 ) , very similar to those previously described in neurons from young adult ( 4–6 month-old ) mice ( Maday and Holzbaur , 2014 ) . Quantitative analysis of the change in fluorescence intensity over time ( Figure 3C ) indicated that mCh-ATG13 transiently localizes to these puncta for 100 to 150 s; subsequent recruitment of GFP-LC3B to a mCh-ATG13-positive punctum coincided with a loss in mCh-ATG13 signal intensity . Frequently , however , dual labeling of autophagosome biogenesis in neurons from aged mice revealed ‘stalled’ events , in which mCh-ATG13 puncta formed and were stably maintained for at least 5 min of a 10 min video; we observed that these stalled events also failed to recruit GFP-LC3B within the imaging window ( Figure 3B and D and Video 2 ) . In neurons from young adult mice , greater than 75% of observed events were productive AVs ( Figure 3E ) . In striking contrast , we found that stalled events dominated in neurons from aged mice , representing greater than 75% of total events in aged neurons ( Figure 3E ) . We observed a similar distinction between productive and stalled events when we compared the recruitment kinetics of GFP-LC3B with elongation complex component mCh-ATG5 ( Figure 3F–G; Videos 3 , 4 ) . In neurons from aged mice , we observed stereotypical AV kinetics , in which the transient recruitment of mCh-ATG5 over approximately 100 s is followed by a steady increase in GFP-LC3B intensity ( Figure 3H ) , similar to our observations with mCh-ATG13 . Again , productive events predominated ( >70% of total events ) in neurons from young adult mice , while stalled events predominated ( ~80% of total events ) in neurons from aged mice ( Figure 3I–J ) . While these stalled events did not go on to produce GFP-LC3B-positive autophagosomes , stalled AVs remained dynamic within the confines of the axon tip rather than remaining tethered in place ( Figure 3B and G , Video 2 , 4 ) . Furthermore , both stalled and productive AVs could be found within the same axonal tip in neurons from both young adult and aged mice ( Figure 3—figure supplement 1 , Videos 5 and 6 ) . These data suggest that aging does not impair the initial steps of autophagosome biogenesis . However , there is a striking block in LC3B recruitment downstream from the recruitment of both ATG13 and ATG5 that occurs infrequently in neurons from young adult mice , but predominates in neurons from aged mice . To further characterize stalled events in neurons from aged mice , we asked if other autophagy components colocalize with stalled AVs . ATG9 is the only multi-pass transmembrane protein in the core autophagy machinery ( Lang et al . , 2000; Noda et al . , 2000; Young et al . , 2006 ) and is thought to transit to the growing isolation membrane with donor membranes ( Sekito et al . , 2009; Suzuki et al . , 2015; Yamamoto et al . , 2012; Young et al . , 2006 ) . Normally , ATG9 is only transiently associated with the developing autophagosome ( Koyama-Honda et al . , 2013; Orsi et al . , 2012 ) . We used multi-color live-cell imaging to assess colocalization between autophagy components in neurons from young adult or aged mice co-expressing fluorescently labeled LC3B , ATG9 , and ATG13 or ATG5 ( Figure 4A ) . As expected , we did not observe significant colocalization of SNAP-ATG9 with productive autophagosome biogenesis events in neurons from either young or aged mice ( Figure 4B and D ) . However , we did observe the robust and persistent colocalization of SNAP-ATG9 with mCh-ATG13 or mCh-ATG5 in the majority of stalled events in neurons from aged mice ( Figure 4A , C and D , Video 4 , 7 ) . Further , we noted persistent SNAP-ATG9 colocalization with the very rare stalled events seen in neurons from young adult mice ( Figure 4D ) . These data suggest that ATG9 may only transiently associate with productive biogenesis events , whereas the majority of stalled AVs aberrantly accumulate or retain ATG9 . We next sought evidence for a similar stalling of autophagosome biogenesis in vivo . As above , we used NMJs to examine AVs in intact tissues . We identified LC3-positive AVs at these synapses in muscle tissue dissected from both young adult and aged mice ( Figure 4E ) . We used ATG9 colocalization with either ATG13 or ATG5 as a marker for stalled events in fixed tissue . In NMJs from young adult mice , we observed ATG13 puncta within the presynaptic compartment , but those puncta did not colocalize with ATG9 ( Figure 4F , left ) . In contrast , in NMJs from aged mice , we observed the colocalization of ATG13 with ATG9 , indicating the persistence of stalled AV formation within the presynaptic compartment in vivo ( Figure 4F , right ) . We then quantified the number of stalled AVs in the NMJ motor axon terminal . While we observed stalled AVs ( using ATG9 colocalization with ATG13 as a stalled AV marker ) only rarely in motor axon terminals from young adult mice , NMJ axon terminals from aged mice consistently contained several stalled AVs ( Figure 4G ) . In contrast , we did not detect a change with age in the number of ATG13-positive puncta that did not co-recruit ATG9 ( Figure 4H ) . Thus , the fraction of stalled AVs to total ATG13-positive puncta significantly increased with age ( Figure 4I ) . These data suggest that our observations of stalled events in cultured primary DRG neurons from aged mice can also be seen in other neuronal types in vivo . Since LC3B is not recruited to stalled AVs , we asked whether this defect was due to a failure to recruit the elongation stage constituents required for LC3B lipidation . Using multi-color immunocytochemistry , we examined the localization of endogenous elongation stage components ATG12 , ATG7 , ATG16L1 , and ATG3 . We used colocalization of endogenous ATG9 with ATG5 to identify stalled AVs in fixed neurons from aged mice . We observed that the lipidation machinery was successfully recruited to stalled AVs ( Figure 5A–E ) . Given the lack of LC3B recruitment to stalled AVs harboring intact lipidation machinery , we asked whether other LC3B homologs could be recruited to stalled AVs in aged mice . There are multiple orthologs of yeast Atg8 expressed in mammals ( mAtg8s ) , including LC3A , LC3B , LC3C , γ-aminobutyric acid receptor-associated protein ( GABARAP ) , GABARAP-Like 1 ( GABARAPL1/GEC1 ) , and GABARAPL2/GATE16 ( Schaaf et al . , 2016 ) . Mice do not appear to have a gene encoding LC3C , but may express a LC3 isoform related to human LC3C ( Liu et al . , 2017 ) . Both immunocytochemistry ( Figure 5—figure supplement 1A–D ) and live cell imaging ( Figure 5F–I ) revealed that LC3A , GABARAP , GABARAPL1/GEC1 , and GABARAPL2/GATE16 can each associate with stalled AVs in neurons from aged mice , with each mCherry-mAtg8 colocalizing with persistent Halo-ATG5 puncta ( Figure 5J ) . These data indicate that the deficit in LC3B recruitment to stalled AVs in neurons from aged mice is specific and that the recruitment of other mAtg8s is not sufficient to convert stalled AVs into productive AVs in neurons from aged mice . These observations are consistent with a growing literature indicating that mAtg8s are not fully functionally redundant ( Nguyen et al . , 2016 ) . Also using live-cell imaging , we asked if ectopic expression of the mAtg8s altered the assembly kinetics of AVs in neurons from aged mice . Surprisingly , we found that overexpression of mScarlet-LC3A , but not other mAtg8s , caused GFP-LC3B recruitment to 84 . 2% of stalled AVs ( persistent Halo-ATG5 puncta ) , significantly different from control neurons ( Figure 5—figure supplement 1E; p=0 . 0002; 21 . 1% of stalled AVs in control ) . However , this induced recruitment of GFP-LC3B to stalled AVs did not resolve the stalled event ( data not shown ) , further implying that while the failure to recruit LC3B is a hallmark of stalled AV events , it is not the principal defect involved . PROPPINs ( β-propellers that bind phosphoinositides ) are essential PI3P effectors in autophagy and are conserved from yeast to humans ( Michell et al . , 2006; Polson et al . , 2010; Proikas-Cezanne et al . , 2004 ) . In mammals , there are four PROPPINs , termed WD-repeat protein interacting with phosphoinositides ( WIPI1 through WIPI4 ) ( Polson et al . , 2010; Proikas-Cezanne et al . , 2004 ) . WIPI1 and WIPI2 are closely related and orthologs of yeast Atg18 , while WIPI3 and WIPI4 form a separate paralogous group ( Behrends et al . , 2010; Polson et al . , 2010; Proikas-Cezanne et al . , 2004 ) . WIPI1 , the first family member to be identified to have a role in autophagy , is recruited to autophagosomal membranes upon autophagy induction ( Gaugel et al . , 2012; Itakura and Mizushima , 2010; Proikas-Cezanne et al . , 2007; Proikas-Cezanne et al . , 2004; Vergne et al . , 2009 ) . WIPI2 links PI3P production by the autophagy nucleation complex to LC3 interaction with the isolation membrane as WIPI2 binds to both PI3P and ATG16L1 ( Figure 6A ) ( Dooley et al . , 2014; Lamb et al . , 2013; Polson et al . , 2010 ) . Thus , we hypothesized that alterations in WIPI2 function may result in lower levels of LC3B recruitment and deleteriously affect productive biogenesis . First we confirmed the importance of WIPI2 in autophagosome biogenesis in primary neurons . Depletion of WIPI2 by RNAi ( Figure 6B ) did not alter rates of AV initiation ( determined by mCh-ATG5 puncta generation ) ( Figure 6C ) , but led to a significant deficit in autophagosome biogenesis , which was fully restored by expression of an RNAi-resistant human Halo-WIPI2B construct ( Figure 6D ) . WIPI2 binds PI3P via a conserved FRRG motif ( Baskaran et al . , 2012; Dove et al . , 2004; Gaugel et al . , 2012; Jeffries et al . , 2004; Krick et al . , 2006; Proikas-Cezanne et al . , 2007; Proikas-Cezanne et al . , 2004; Watanabe et al . , 2012 ) . This interaction can be abolished by mutating the positively charged arginine residues in the motif to uncharged threonine residues ( FTTG ) ( Figure 6A ) ( Dooley et al . , 2014 ) . Overexpression of Halo-WIPI2B ( FTTG ) was unable to rescue the deficit , consistent with a key role for phosphoinositide signaling in autophagosome biogenesis ( Figure 6D ) . WIPI2B also interacts with ATG16L1 , an essential component of the LC3 conjugation complex . The interaction between WIPI2B and ATG16L1 can be abrogated by switching a positively charged arginine to a negatively charged glutamate ( R108E ) in WIPI2B ( Dooley et al . , 2014 ) . Ectopic expression of Halo-WIPI2B ( R108E ) in WIPI2-depleted neurons from young adult mice did not affect rates of AV initiation but did not rescue the deficit in the rates of formation of GFP-LC3B-positive autophagosomes ( Figure 6C–D ) . Furthermore , overexpression of the WIPI2 paralog SNAP-WIPI1A was also unable to compensate for the loss of WIPI2 in young adult neurons ( Figure 6D ) . These data confirm that WIPI2 , including its known functional domains , is required in autophagosome biogenesis in DRG neurons . Next , we ectopically expressed Halo-tagged WIPI2B in neurons from aged mice . Halo-WIPI2B colocalized with the early autophagosome marker mCh-ATG5 in neurons ( Figure 6E , Video 8 ) and did not affect rates of AV initiation ( Figure 6F ) . Strikingly , ectopic WIPI2B expression increased rates of autophagosome biogenesis in neurons from aged mice from 0 . 21 AVs per minute to 0 . 47 AVs per minute ( Figure 6G ) , a rate similar to that observed in neurons from young adult mice ( Figure 1B ) . Furthermore , overexpression of WIPI2B did not alter the kinetics of productive AV biogenesis ( Figure 6H ) . In contrast to wild type WIPI2B , expression of WIPI2B constructs with targeted mutations in either the PI3P or ATG6L1 binding motifs did not restore autophagosome biogenesis in neurons from aged mice ( Figure 6F–G ) . These data suggest that overexpression of WIPI2B in neurons from aged mice restores autophagosome biogenesis and that this rescue requires both the PI3P-binding and ATG16L1-binding functions of WIPI2B . Given that ectopically expressing WIPI2B in neurons from aged mice rescues rates of autophagosome biogenesis , we initially hypothesized that WIPI2 levels decrease in neuronal tissues with age . However , we observed no significant deficits in WIPI2 expression levels at the level of RNA or protein , or those of any of the WIPI family members with age in either whole brain lysates or DRG lysates ( Figure 6—figure supplement 1 ) . We also observed no significant changes in expression levels of several other autophagy components ( ULK1 , P-ULK1 , ATG14 , P-ATG14 , Beclin1 , ATG3 , ATG5 , ATG7 , ATG10 , ATG16L1 , LC3B , LC3A , GABARAP , GABARAPL1 , GABARAPL2 , p62 , WIPI3 , and WIPI4 ) with age in whole brain or DRG lysates ( Figure 6—figure supplement 1A and B , Figure 6—figure supplement 2 ) . Additionally , using immunocytochemistry on fixed DRG neurons from aged mice , we observed endogenous WIPI1 and WIPI2 localized to stalled AVs ( Figure 5—figure supplement 1F–G ) . These results indicate that the decrease in autophagosome biogenesis with age is not due to an age-related loss of WIPI2 or its paralogs and that endogenous WIPIs can be recruited to stalled AVs in neurons from aged mice . Next we looked to post-translational modification of WIPI2 . WIPI2B is known to be phosphorylated at serine 395 ( S413 in WIPI2A ) ( Hsu et al . , 2011; Wan et al . , 2018 ) , although the mechanistic effects of this phosphorylation have not been fully explored . We used two independent phosphorylation-sensitive antibodies ( Figure 7—figure supplement 1A , C and D ) to confirm that phosphorylated WIPI2 is found in neuronal tissues ( Figure 7A ) . Additionally , we confirmed that we could detect phospho-WIPI2 on AVs in DRG distal neurites by immunocytochemistry ( Figure 7B , Figure 5—figure supplement 1H ) . Our data ( Figure 7—figure supplement 1B ) agreed with a previous report ( Wan et al . , 2018 ) that phosphorylation of WIPI2B at serine 395 does not affect its ability to bind to PI3P or Atg16L1 . Next , we asked how WIPI2 phosphorylation affects autophagosome biogenesis . We ectopically expressed a RNAi-resistant nonphosphorylatable construct , Halo-WIPI2B ( S395A ) , or a RNAi-resistant phospho-mimetic construct , Halo-WIPI2B ( S395E ) , in WIPI2-depleted neurons from young adult mice . Similar to our previous results , overexpression of Halo-WIPI2B constructs did not affect rates of AV initiation ( Figure 7C ) . We found that the phospho-dead construct , Halo-WIPI2B ( S395A ) , rescued rates of autophagosome biogenesis similar to the wild type Halo-WIPI2B construct . In contrast , overexpression of the phospho-mimetic construct , Halo-WIPI2B ( S395E ) , did not restore rates of autophagosome biogenesis in WIPI2-depleted neurons from young adult mice ( Figure 7D ) . When we expressed these constructs in neurons from aged mice , we obtained similar results; the phospho-dead construct , Halo-WIPI2B ( S395A ) , restored the rate of autophagosome biogenesis in neurons from aged mice , while the phospho-mimetic construct , Halo-WIPI2B ( S395E ) , did not ( Figure 7E ) . We did not detect differences in the levels of overexpression for the different Halo-WIPI2B constructs or changes in overexpression of ectopic Halo-WIPI2B constructs with age ( Figure 7—figure supplement 1E ) . These data suggest that WIPI2 must be dephosphorylated to enable productive AV biogenesis . These results led us to hypothesize that levels of phosphorylated WIPI2 increase with age in neuronal tissues . However , just as we saw with total WIPI2 levels , we did not see an overall change in phosphorylated WIPI2 protein with age in either whole brain or DRG lysates ( Figure 6—figure supplement 1C–E ) . These results suggest that WIPI2 may be found in its phosphorylated form throughout the cytosol and only transiently dephosphorylated at the isolation membrane , masking any functional age-related change in WIPI2 phosphorylation in bulk assays . This hypothesis is consistent with our data indicating that stalled and productive AVs occur in the same axonal tip ( Figure 3—figure supplement 1 , Video 6 ) , suggesting that AV stalling results from a highly localized defect . To further characterize the role of phosphorylated WIPI2B in autophagosome biogenesis , we examined AV events in neurons from aged mice overexpressing the phospho-dead construct , Halo-WIPI2b ( S395A ) or the phospho-mimetic construct , Halo-WIPI2B ( S395E ) in conjunction with GFP-LC3B and mCh-ATG13 , by live-cell microscopy . The phospho-dead construct Halo-WIPI2B ( S395A ) was cytoplasmic , but also colocalized with mCh-ATG13 and GFP-LC3B on productive AVs ( Figure 7H ) . Next , we examined neurons from aged mice ectopically expressing the phospho-mimetic Halo-WIPI2B ( S395E ) construct . We identified GFP-LC3B-positive AVs and determined whether WIPI2B ( S395E ) was ever associated with the AV during the video . GFP-LC3B-positive AV events that failed to recruit Halo-WIPI2B ( S395E ) did not increase in size ( Figure 7F , G and I , Figure 7—figure supplement 2 ) , suggesting that the isolation membrane could not successfully extend . In contrast , GFP-LC3B-positive AV events that did recruit Halo-WIPI2B ( S395E ) increased in size to form a GFP-LC3B ring structure , consistent with a mature autophagosome ( Figure 7F , G and J ) . Taken together , these results suggest that WIPI2B is dephosphorylated at the isolation membrane to allow autophagosome biogenesis to initiate . Further , these results suggest that WIPI2B is then dynamically rephosphorylated at the AV to enable the autophagosome to grow and complete biogenesis . If this hypothesis is correct , phosphorylation of WIPI2B at serine 395 might affect its affinity for membranes . To determine whether the phosphorylation state of WIPI2B affects its ability to interact with membranes , we performed crude fractionation experiments . We collected whole brain lysates from young adult and aged nontransgenic mice and separated the cytosolic and membrane fractions by centrifugation . We then compared the endogenous levels of phospho-WIPI2 and total WIPI2 associated with each fraction by immunoblot ( Figure 8A ) . The ratio between phospho-WIPI2 and total WIPI2 in the membrane fraction was reduced by approximately 50% compared to the cytosolic fraction for both ages ( Figure 8B ) . These results suggest that phosphorylation of WIPI2B at serine 395 decreases its affinity for membranes . We also tested our hypothesis that phosphorylation of WIPI2B at serine 395 causes WIPI2B to disassociate from the AV membrane in live-cell imaging of both the phospho-dead WIPI2B ( S395A ) and phospho-mimetic WIPI2B ( S395E ) constructs in the same neurons . We examined the dynamics of ectopically expressed RNAi-resistant Halo-WIPI2B ( S395A ) and SNAP-WIPI2B ( S395E ) in WIPI2-depleted DRG neurons from young adult mice . We measured the length of time each WIPI2B construct resided at a given punctum . Since our time-lapse videos were captured over 10 min , the maximum residence time we could measure was 600 s . We found that Halo-WIPI2B ( S395A ) remained associated for nearly 600 s on average . In contrast , in the same neurons , SNAP-WIPI2B ( S395E ) only remained associated with puncta for approximately 300 s , or half as long as the phospho-dead construct ( Figure 8C ) . These data are consistent with our lysate fractionation data , indicating that phosphorylation of WIPI2B at serine 395 correlates with a decreased affinity for membranes . Similar to our observations of other aspects of AV biogenesis , within a given DRG axonal tip , we could observe both an expanding AV that co-recruited Halo-WIPI2B ( S395E ) and an AV that was Halo-WIPI2B ( S395E ) -negative that failed to enlarge ( Figure 8D ) . Thus , the rephosphorylation of WIPI2B ( S395 ) may also be a highly localized process . The dynamic phosphorylation of WIPI2 during AV biogenesis at the isolation membrane could allow for tight , spatially and temporally localized regulation of autophagy ( Figure 9 ) . Here , we investigated the dynamics of autophagy during aging in primary neurons and demonstrated that the rate of autophagosome biogenesis significantly decreases in neurons with age . Surprisingly , the deficit was specific , as the initial stages of autophagosome formation , initiation and nucleation , were not altered with age in mammalian neurons . Instead , we found that the majority of AVs in aged neurons successfully initiated but then stalled . EM analysis suggested that this deficit was correlated with a morphological defect in autophagosome formation , characterized by excess membrane accumulation within the autophagic vacuole , detectable in aged neurons both in vitro and in vivo . WIPI2B overexpression in neurons from aged mice increased the rate of autophagosome biogenesis , restoring this rate to that found in neurons from young adult mice . Further , we propose that the dynamic regulation of WIPI2B phosphorylation at the isolation membrane may be integral to autophagosome biogenesis . Our results indicated that the nonphosphorylatable S395A form of WIPI2B was sufficient to rescue AV biogenesis upon depletion of endogenous WIPI2 , while recruitment of the phosphomimetic WIPI2B ( S395E ) mutant correlated with expansion of the nascent autophagosome , suggesting that both dephosphorylation and rephosphorylation of WIPI2B are key regulatory steps . Ultimately , we showed that the rate of autophagosome biogenesis decreased in neurons during aging , but we mitigated this decrease by overexpressing a single autophagy component , WIPI2B ( Figure 6G ) . In neurons from aged mice , the majority of stalled AVs aberrantly accumulated ATG9 ( Figure 4 ) . Our results showing that WIPI2B overexpression in aged neurons restored autophagosome biogenesis are consistent with previous studies indicating that WIPI2 downregulation induced the localized accumulation of ATG9 at AVs ( Orsi et al . , 2012 ) . We hypothesize that the multilamellar structures we detected by TEM ( Figure 1 ) correlate with the stalled AVs we observed by fluorescence microscopy . Work from other groups suggest that this hypothesis could be correct . ATG9 interacts with ATG2 , a conserved core autophagy protein ( Barth and Thumm , 2001; Gómez-Sánchez et al . , 2018; Shintani et al . , 2001; Wang et al . , 2001 ) . ATG2 also interacts with WIPI4 ( Behrends et al . , 2010; Chowdhury et al . , 2018; Lu et al . , 2011; Velikkakath et al . , 2012 ) . The speculation that the ATG2-ATG9 complex transfers lipids to the membrane-hungry growing autophagosome ( Gómez-Sánchez et al . , 2018; Kumar et al . , 2018 ) was recently confirmed in yeast ( Osawa et al . , 2019 ) and mammalian cells ( Valverde et al . , 2019 ) . Thus , the prolonged association of ATG9 with stalled AVs detected in neurons from aged mice ( Figure 4 ) may indicate a prolonged association of ATG2 with stalled AVs . This extended residency at the AV could enable unregulated lipid transfer to the stalled AV , resulting in the multilamellar structures we observed in neurons from aged mice ( Figure 1 ) . We also found that stalled AVs failed to recruit LC3B ( Figure 3 ) , while the recruitment of other mAtg8s was not affected ( Figure 5 ) . Of note , lipidation of LC3B is less efficient on less curved and less PE-rich membranes than lipidation of GABARAPL1 ( Nath et al . , 2014 ) , suggesting that stalled AVs contain sufficient curvature and have a sufficient PE concentration to allow the lipidation and incorporation of all mAtg8s except LC3B . Furthermore , inducing the recruitment of LC3B to AVs by LC3A overexpression was not sufficient to rescue autophagosome formation . These results , in conjunction with recent observations ( Nguyen et al . , 2016; Tsuboyama et al . , 2016 ) , suggest that LC3B recruitment is neither strictly required nor sufficient for AV generation and elongation . Instead , LC3B recruitment may regulate membrane expansion or membrane fusion to form a double-membrane structure . We propose that upon perturbation of dynamic WIPI2 phosphorylation , membrane extension may proceed in an unrestricted manner , generating the multilamellar structures observed by TEM in neurons from aged mice ( Figure 1 ) ( Majeed , 1993 ) and in aged human AD brain ( Nixon et al . , 2005 ) . Here , we chose to focus on WIPI2 as a master regulator of autophagosome biogenesis . However , it will be interesting to determine in the future how LC3B incorporation into the isolation membrane relates to lipid incorporation and autophagosome membrane extension and expansion . The autophagy pathway has been extensively studied in non-neuronal cells , where autophagy can be induced by starvation or other cellular stressors ( Abada and Elazar , 2014; Hale et al . , 2013; Mariño et al . , 2011; Reggiori and Klionsky , 2013; Son et al . , 2012; Wu et al . , 2013; Zhang and Baehrecke , 2015 ) . Conversely , in vivo and in vitro studies in neurons indicate that neuronal autophagy is not significantly induced by starvation ( Fox et al . , 2010; Maday and Holzbaur , 2016; Mizushima et al . , 2004; Tsvetkov et al . , 2010 ) or by proteotoxic stress ( Maday et al . , 2012; Wong and Holzbaur , 2014 ) . These studies suggest that autophagosome biogenesis is regulated differentially in neuronal and non-neuronal cells . Our results indicate that induction of autophagosome biogenesis is constitutive and remains robust during aging in neurons ( Figure 2 ) . Further , our data identify a novel age-related regulation of neuronal autophagosome biogenesis , suggesting that autophagy can be regulated at distinct steps apart from the autophagy initiation complex . Consistent with our finding that WIPI2 regulates basal autophagy , WIPI1 and WIPI2 recruitment to AVs is independent of glucose starvation ( McAlpine et al . , 2013; Pfisterer et al . , 2011 ) . Our data ultimately suggest that neuronal autophagy may be more easily modulated ectopically via WIPI2B than the better-studied , starvation-sensitive ULK1-ATG13 initiation complex . Stalled AVs in neurons from aged mice provide a unique opportunity to tease apart the role of WIPI2B in autophagosome biogenesis . WIPI2 interacts with PI3P at the isolation membrane and is required for subsequent WIPI1 localization ( Bakula et al . , 2017 ) . Both WIPI1 and WIPI2 are predicted to form an amphipathic α-helix upon lipid binding , similar to the yeast homolog , Atg18 ( Gopaldass et al . , 2017 ) , which can promote membrane deformation . Further , Atg18 forms oligomers upon membrane binding ( Gopaldass et al . , 2017; Scacioc et al . , 2017 ) . Considering our data in conjugation with these data from cell lines , we now propose that WIPI2 dephosphorylation at S395 may allow robust recruitment of WIPI2 and WIPI1 to the isolation membrane . The recruitment of WIPI1 and formation of WIPI1-WIPI2 hetero-oligomers promote membrane deformation . Of note , dephosphorylation of yeast Atg18 is required for Atg18 association with the membrane ( Tamura et al . , 2013 ) , suggesting that the dynamic phosphorylation of PROPPINs could be a conserved regulatory mechanism for autophagy . Further support for a sequential phosphorylation model would come from the identification of the kinase ( s ) responsible for WIPI2 ( S395 ) phosphorylation . One WIPI2 kinase , mTORC1 , has recently been identified ( Hsu et al . , 2011; Wan et al . , 2018 ) . While this study found that phosphorylated WIPI2B led to its degradation by the proteasome in HEK293T cells , our data indicate a more nuanced role for phosphorylated WIPI2 in autophagosome biogenesis in neurons . Our results indicated that phosphorylation at WIPI2B serine 395 lowered the affinity of WIPI2B for membrane and shortened the residence time at the nascent autophagic membrane ( Figure 8A–C ) . Recruitment of WIPI2B ( S395E ) to the isolation membrane correlated with expansion of the autophagosome membrane ( Figure 7F–J ) , so we speculate that dissociation of WIPI2B might be a critical and regulated step during autophagosome biogenesis ( Figure 9 ) . Thus , it will be interesting to determine how a specific post-translational modification could have such opposite effects on the same pathway . We also propose that a localized phosphatase may regulate WIPI2 activity at the developing autophagosome . One compelling candidate is PP2A , which has been implicated in autophagy regulation ( Magnaudeix et al . , 2013; Neisch et al . , 2017; Yeasmin et al . , 2016 ) and shown to interact with WIPI2 via a PP2A regulatory subunit , PPP2R1A ( Bakula et al . , 2017 ) . An age-related decrease in phosphatase levels in neural tissues has been implicated in Alzheimer’s disease ( Sontag and Sontag , 2014 ) , and PP2A has been shown to decrease in neural tissues with age ( Veeranna et al . , 2011 ) , suggesting that declining PP2A activity could contribute to the defect we observed . Our data suggest that successful autophagosome biogenesis is dependent on the localized environment surrounding the isolation membrane . We observed that a given DRG distal tip could contain both productive and stalled events ( Figure 3—figure supplement 1 ) . We also detected WIPI2B ( S395E ) -positive and –negative AVs within a given DRG axonal tip ( Figure 8D ) . Since the rate of AV biogenesis decreased with age in neurons ( Figure 1 ) , at least one component critical to the process is likely altered during aging . However , we now propose that rather than a change in protein level , the critical component might become mislocalized during aging . This age-related mislocalization would prevent an isolation membrane from developing into a productive autophagosome , explaining the observed decreased rate of AV biogenesis with age . In this study , we focused on the effects of aging on neuronal autophagosome biogenesis . However , there are critical steps in the autophagy pathway downstream from initial autophagosome biogenesis , including autophagosome closure , fusion with lysosomes , retrograde transport of autophagosomes and autolysosomes to the soma , and degradation of cargo . Aging also likely affects these later stages of autophagy . For example , lysosomal integrity has been shown to decrease with age in neurons ( Nixon , 2017 ) . In non-neuronal cells , retrograde transport of autophagosomes appears to decrease with age in primary mouse fibroblasts ( Bejarano et al . , 2018 ) . It will be interesting to investigate how the later stages of autophagy are altered with age in neurons . Misregulation of autophagy has been implicated in many neurodegenerative diseases and disorders ( Haack et al . , 2012; Komatsu et al . , 2007; Komatsu et al . , 2006; Nixon et al . , 2005; Saitsu et al . , 2013 ) . While ectopic induction of autophagy has met with some success in attenuating aggregated mutant huntingtin and Tau in neurodegeneration models ( Ravikumar et al . , 2002; Wang et al . , 2009 ) , our data suggest that targeting the autophagy initiation complex may not be generally effective for treatment of age-related neurodegenerative disease . Rather , modulating other stages of autophagosome biogenesis , such as dynamic WIPI2 phosphorylation at the isolation membrane , may produce more successful therapies . GFP-LC3B transgenic mice ( strain: B6 . Cg-Tg ( CAG-EGFP/LC3 ) 53Nmi/NmiRbrc ) were generated by N . Mizushima ( Tokyo Medical and Dental University , Tokyo , Japan; Mizushima et al . , 2004 ) and obtained from RIKEN BioResource Center in Japan . These mice were bred with C57BL/6J mice obtained from The Jackson Laboratory . Hemizygous and wild type littermates were used in experiments . Constructs used include: mCherry-ATG13 ( subcloned from Addgene 22875 ) , mCherry-ATG5 ( Addgene 13095 ) , Halo-ATG5 ( subcloned from Addgene 13095 ) , SNAP-ATG9 and Halo-ATG9 ( subcloned from Addgene 60609 ) , SNAP-WIPI1A ( subcloned from Addgene 38272 ) , Halo-DFCP1 ( subcloned from Addgene 38269 ) , mSarlet-LC3A ( subcloned from Addgene 73946 ) , mCherry-GABARAP ( subcloned from Addgene 73948 ) , mCherry-GEC1 ( GABARAPL1 , subcloned from Addgene 73945 ) , and mCherry-GATE16 ( GABARAPL2 , subcloned from Addgene 73518 ) . SNAP-WIPI2B and Halo-WIPI2B were subcloned from GFP-WIPI2B ( Dooley et al . , 2014 ) . SNAP- and Halo- WIPI2B ( FTTG ) , WIPI2B ( R108E ) , WIPI2B ( S395A ) , and WIPI2B ( S395E ) were generated via quick change and subcloned into original plasmids . The SNAP backbone was originally obtained from New England Biolabs ( NEB ) , and the Halo backbone was originally obtained from Promega . DRG neurons were isolated as previously described ( Perlson et al . , 2009 ) and cultured in F-12 Ham’s media ( Invitrogen ) with 10% heat-inactivated fetal bovine serum , 100 U/mL penicillin , and 100 μg/mL streptomycin . For live-cell microscopy , DRGs were isolated from P21-28 ( young ) , P90-120 ( young adult ) , P480-540 ( aged ) , or P730-760 ( advanced aged ) mice and plated on glass-bottomed dishes ( MatTek Corporation ) and maintained for 2 days at 37°C in a 5% CO2 incubator . Prior to plating , neurons were transfected with a maximum of 0 . 6 μg total plasmid DNA using a Nucleofector ( Lonza ) using the manufacturer’s instructions . Relevant siRNA was co-transfected with plasmid DNA ( 25 pmol ON-TARGET plus SMARTpool Wipi2 siRNA , L-057690–01 from Dharmacon ) . For imaging experiments with siRNA , control neurons were transfected with 30 pmol Cy5-labeled non-targeting siRNA ( Dharmacon ) per dish and experimental neurons were co-transfected with 5 pmol Cy5-labeled non-targeting siRNA ( Dharmacon ) to identify which neurons received siRNA . For biochemistry siRNA experiments , control neurons were transfected with 25 pmol ON-TARGET plus non-targeting siRNA ( D-001810–01 from Dharmacon ) . Microscopy was performed in low fluorescence nutrient media ( Hibernate A , BrainBits ) supplemented with 2% B27 and 2 mM GlutaMAX . For nucleofected constructs that yielded Halo- or SNAP-tagged proteins , DRG neurons were incubated with 100 nM of the appropriate Halo or SNAP ligand ( SNAP-Cell 647-SiR , SNAP-Cell TMR-Star , or SNAP-Cell 430 from NEB; HaloTag TMR Ligand from Promega , silicon-rhodamine-Halo ligand from K . Johnsson , École Polytechnique Federale de Lausanne , Lausanne , Switzerland , or JF646-Halo ligand from Luke Levis , Janelia Farms , HHMI ) for at least 30 min at 37°C in a 5% CO2 incubator . After incubation , DRGs were washed three times with complete equilibrated F-12 media , with the final wash remaining on the DRGs for at least 15 min at 37°C in a 5% CO2 incubator . Mice of either sex within the indicated postnatal range ( 1 month , 3 months , 16–17 months or 24 months ) were euthanized prior to dissection . All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania . Microscopy was performed on a spinning-disk confocal ( UltraVIEW VoX; PerkinElmer ) microscope ( Eclipse Ti; Nikon ) with an Apochromat 100x , 1 . 49 NA oil immersion objective ( Nikon ) at 37°C in an environmental chamber . The Perfect Focus System was used to maintain Z position during time-lapse acquisition . Digital micrographs were acquired with an EM charge-coupled device camera ( C9100; Hammamatsu Photonics ) using Volocity software ( PerkinElmer ) . Time-lapse videos were acquired for 10 min with a frame every 3 s to capture autophagosome biogenesis . Multiple channels were acquired consecutively , with the green ( 488 nm ) channel captured first , followed by red ( 561 nm ) , far-red ( 640 nm ) , and blue ( 405 nm ) . DRGs were selected for imaging based on morphological criteria and low expression of transfected constructs . To minimize artifacts from overexpression , neurons within a narrow range of low fluorescence intensity were chosen for imaging , ensuring the analyzed neurons expressed low levels of the ectopic tagged proteins . Time-lapse micrographs were analyzed with FIJI ( Schindelin et al . , 2012 ) . ‘Stalled’ biogenesis events were defined as mCherry-ATG13 , mCherry-ATG5 , or Halo-ATG5 puncta that remained visible for at least 5 min . ‘Productive’ biogenesis events were defined as mCherry-ATG13 , mCherry-ATG5 , or Halo-ATG5 puncta that persisted for less than 5 min and recruited GFP-LC3B . Brains or DRGs of non-transgenic mice were dissected and subsequently homogenized and lysed . Brains were homogenized individually in RIPA buffer [50 mM NaPO4 , 150 mM NaCl , 1% Triton X-100 , 0 . 5% deoxycholate , 0 . 1% SDS , 1x complete protease inhibitor mixture ( Roche ) , and 1x Halt protease and phosphatase inhibitor cocktail ( Thermo ) ] . DRGs were homogenized in RIPA buffer with a 1 . 5 mL pestle . Homogenized samples were lysed for 30 min on ice . For the siRNA and overexpression controls , isolated DRGs were plated at 120 , 000 neurons per 35 mm dish as described above for 2 DIV . For the Halo-WIPI2B overexpression controls , where indicated , neurons were treated with 100 nM BaflomycinA1 ( BafA ) for 4 hr prior to lysis . Neurons were washed with PBS ( 50 mM NaPO4 , 150 mM NaCl , pH 7 . 4 ) and then lysed as above . Samples were centrifuged at 17 , 000 x g at 4°C for 15 min . Total protein in each lysate was determined by BCA assay ( ThermoFisher Scientific ) so that equal amounts of protein were loaded into each lane . All supernatants were analyzed by SDS-PAGE , transferred onto FL PVDF membrane , and visualized with fluorescent secondary antibodies ( Li-Cor ) on an Odyssey CLx imaging system ( Li-Cor ) . The specificity of relevant antibodies was confirmed by immunocytochemistry as described below or by immunoblot . See Table for antibodies used . All western blots were analyzed with Image Studio ( Li-Cor ) . Total protein was used as a loading control to control for differences in sample loading . The normalization factor is listed below each blot as a percent . For brain lysate cytosolic and membrane fractions , brains were dissected from non-transgenic mice and subsequently homogenized and lysed . Brains were homogenized individually in Motility Assay Buffer ( MAB ) [10 mM PIPES , 50 mM K-Acetate , 4 mM MgCl2 , 1 mM EGTA , 2 mM PMSF , 210 μM leupeptin , 1 . 5 μM pepstatin-A , 52 . 8 μM N-p-Tosyl-L-arginine methyl ester , 20 mM DTT , and 1x Halt protease and phosphatase inhibitor cocktail ( Thermo ) ] . The homogenate was spun at 17 , 000 x g for 30 min at 4°C . The resultant supernatant was then spun 95 , 000 x g for 20 min at 4°C . The supernatant was the cytosolic fraction , and the pellet was resuspended in an equal volume of MAB . HeLa-M ( A . Peden , Cambridge Institute for Medical Research ) and HEK293 ( ThermoFisher , R70507 ) cells were cultured in complete medium ( DMEM supplemented with 10 fetal bovine serum and 2 mM GlutaMAX ) . Lipofectamine 2000 ( Invitrogen ) and FuGENE ( Promega ) were used to transiently transfect HEK293 and HeLa-M cells , respectively . Immunoprecipitation was performed 24 hr after transfection . Cells were permeabilized with TNTE buffer ( 20 mM Tris HCl , pH 7 . 5 , 150 mM NaCl , 1% triton TX-100 , 5 mM EDTA , and 1X Halt Phosphatase and Protease Inhibitor , ThermoFisher ) . The HeLa-M lysates were immunoprecipitated with 3 μg mouse monoclonal Anti-Halo ( Promega , G9211 ) and Dynabeads Protein G ( Invitrogen ) . The immunoprecipitated sample was cleaved from the Dynabeads by boiling in Orange Protein Loading Buffer ( Li-Cor ) . GFP-labeled proteins ectopically expressed in HEK293 cells were immunoprecipitated with GFP-Trap beads . The HEK293 cells were routinely tested for mycoplasma and authenticated by STR ( short tandem repeat ) profiling by The Francis Crick Cell Services . HEK293 cells were used because they are of human origin , fast growing , easy to transfect , and express ectopic proteins without high toxicity . The HeLa cells were routinely tested for mycoplasma using the MycoAlert detection kit ( Lonza , LT07 ) and authenticated by STR profiling using the GenePrint 10 system ( Promega , B9510 ) at the University of Pennsylvania Perelman School of Medicine DNA Sequencing Facility . Whole brains were collected from euthanized non-transgenic 3-month-old or 16–17 month-old mice and immediately frozen at −80°C . Brains were homogenized in 2 mL TRIzol reagent ( ThermoFisher Scientific , 15596018 ) . 2 mL TRIzol reagent and 800 μL chloroform were added after homogenization . Solution was vortexed for 15 s , incubated at room temperature for 5 min , and centrifuged at 12 , 000 x g for 15 min at 4°C . Clear aqueous phase was mixed with one volume of 200 proof ethanol . Mixture was transferred to Zymo Quick-RNA miniprep kit ( Zymo Research , R1057 ) . Total RNA was immediately transferred to Polytract mRNA Isolation System III ( Promega , Z5310 ) to isolate mRNA . Isolated mRNA was immediately transformed into cDNA using M-MuLV Reverse Transcriptase ( New England Biolabs ( NEB ) , M0253L; other NEB reagents: S1330S , M0314S ) . Nucleic acid concentration and purity was monitored throughout isolation . 10 ng total cDNA was added to a 50 μL qPCR reaction with Luna Universal qPCR Master Mix ( NEB , M3003G ) . Each biological sample was loaded in triplicate into qPCR plate . All biological samples for each gene tested were loaded into a single qPCR plate ( Phenix Research Products , MPC-3425 and LMT-RT2 ) , with a reference gene loaded into the same qPCR plate . All primers were initially identified through Primer Bank ( https://pga . mgh . harvard . edu/primerbank/index . html ) ( Spandidos et al . , 2010; Spandidos et al . , 2008; Wang , 2003; Wang et al . , 2012 ) . Primers were optimized to have melting temperatures at 62°C and tested to ensure appropriate dynamic range . Final qPCR primers used were: Pgk1 Fwd ( 5’-ATGTCGCTTTCCAACAAGCTGACTTTGGAC ) , Pgk1 Rev ( 5’-GGACTTGGCTCCATTGTCCAAGCAGAATTTG ) , Rplp0 Fwd ( 5’-GGGCATCACCACGAAAATCTCCAGAGG ) , Rplp0 Rev ( 5’-CTGCCGTTGTCAAACACCTGCTGG ) , Ulk1 Fwd ( 5’-GCAAGTTCGAGTTCTCTCGCAAGGACC ) , Ulk1 Rev ( 5’-CCACGATGTTTTCGTGCTTTAGTTCCTTCAGG ) , Wipi1 Fwd ( 5’-GCTGCTTCTCTTTCAACCAAGACTGCACATC ) , Wipi1 Rev ( 5’-CACGTCAGGGATTTCATTGCTTCCATGGAC ) , Wipi2 Fwd ( 5’-CCAGGATAACACGTCCCTAGCTGTTGG ) , Wipi2 Rev ( 5’-CTCTCCACAATGCAGACATCTTCAGTGTCAG ) , Wdr45b ( Wipi3 ) Fwd ( 5’-CGGGTGTTTTGCATGTGGAATGGAAAATGG ) , Wdr45b ( Wipi3 ) Rev ( 5’-CAGATCATCACTTTGTTGGGAGGGTATTTCGG ) , Wdr45 ( Wipi4 ) Fwd ( 5’-GCGCCATTCACTATCAATGCACATCAGAGTG ) , Wdr45 ( Wipi4 ) Rev ( 5’-GGAGGAGTCGTGGCTGAAGTTAATGCAG ) , Map1lc3b ( Lc3b ) Fwd ( 5’-CCCAGTGATTATAGAGCGATACAAGGGGGAG ) , Map1lc3b ( Lc3b ) Rev ( 5’-CTGCAAGCGCCGTCTGATTATCTTGATGAG ) , Atg5 Fwd ( 5’-GGCACACCCCTGAAATGGCATTATCC ) , Atg5 Rev ( 5’-CCTCAACCGCATCCTTGGATGGAC ) , Atg2a Fwd ( 5’-CTATCTGTTCCCAGGTGAACGGAGTGG ) , and Atg2a Rev ( 5’-CTGGATGCAGCTGTGTCACGATGG ) . qPCR was performed on a QuantStudio 3 Real-Time PCR System ( ThermoFisher Scientific ) controlled by QuantStudio Design and Analysis Software ( ThermoFisher Scientific ) . Normalized target gene expression level is 2^ΔΔCt for each gene relative to the indicated reference gene . DRGs from non-transgenic mice were isolated as above and plated as spot cultures on glass-bottomed dishes ( MatTek Corporation ) and maintained for 2 days at 37°C in a 5% CO2 incubator . DRGs were fixed with 2 . 5% glutaraldehyde , 2 . 0% paraformaldehyde in 0 . 1M sodium cacodylate buffer , pH 7 . 4 , overnight at 4°C . For NMJs , non-transgenic mice were euthanized and subsequently perfused with 2 . 5% glutaraldehyde , 2 . 0% paraformaldehyde in 1X PBS . EDL muscles were dissected and post-fixed in 2 . 5% glutaraldehyde , 2 . 0% paraformaldehyde in 1X PBS overnight at 4°C . EDL muscles were further post-fixed in 2 . 5% glutaraldehyde , 2 . 0% paraformaldehyde in 0 . 1M sodium cacodylate buffer , pH 7 . 4 . Fixed DRGs and NMJs were then transferred to the Electron Microscopy Resource Laboratory at the University of Pennsylvania , where all subsequent steps were performed . After subsequent buffer washes , the samples were post-fixed in 2 . 0% osmium tetroxide for 1 hr at room temperature and then washed again in buffer , followed by dH2O . After dehydration through a graded ethanol series , the tissue was infiltrated and embedded in EMbed-812 ( Electron Microscopy Sciences , Fort Washington , PA ) . Thin sections were stained with lead citrate and examined with a JEOL 1010 electron microscope fitted with a Hamamatsu digital camera and AMT Advantage image capture software . Regions between DRG cell body densities with maximum neurite invasion were chosen for imaging . All image analysis was performed on raw data . Images were prepared in FIJI ( Schindelin et al . , 2012 ) ; contrast and brightness were adjusted equally to all images within a series . Figures were assembled in Adobe Illustrator . Prism 6 ( GraphPad ) was used to plot graphs and perform statistical tests . Prism 8 ( GraphPad ) was used to plot graphs and perform statistical tests for Figure 4G–I . Statistical tests are indicated in the text and figure legends . To quantify AV biogenesis , GFP-LC3B puncta were tracked manually using FIJI . An AV biogenesis event was defined as the de novo appearance of a GFP-LC3B punctum based on changes in fluorescence intensity over time . For GFP-LC3B puncta that were present at the start of the time-lapse series , only those puncta that increased in fluorescence intensity and/or area with time were counted as AV biogenesis events .	Unlike most of the cells in our body , our neurons are as old as we are: while other cell types are replaced as they wear out , our neurons must last our entire lifetime . The symptoms of disorders such as Alzheimer's disease and ALS result from neurons in the brain or spinal cord degenerating or dying . But why do neurons sometimes die ? One reason may be that elderly neurons struggle to remove waste products . Cells get rid of worn out or damaged components through a process called autophagy . A membranous structure known as the autophagosome engulfs waste materials , before it fuses with another structure , the lysosome , which contains enzymes that break down and recycle the waste . If any part of this process fails , waste products instead build up inside cells . This prevents the cells from working properly and eventually kills them . Aging is the major shared risk factor for many diseases in which brain cells slowly die . Could this be because autophagy becomes less effective with age ? Stavoe et al . isolated neurons from young adult , aging and aged mice , and used live cell microscopy to follow autophagy in real time . The results determined that autophagy does indeed work less efficiently in elderly neurons . The reason is that the formation of the autophagosome stalls halfway through . However , increasing the amount of one specific protein , WIPI2B , rescues this defect and enables the cells to produce normal autophagosomes again . As long-lived cells , neurons depend on autophagy to stay healthy . Without this trash disposal system , neurons accumulate clumps of damaged proteins and eventually start to break down . The results of Stavoe et al . identify one way of overcoming this aging-related problem . As well as providing insights into neuronal biology , the results suggest a new therapeutic approach to be developed and tested in the future .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "neuroscience" ]	2019	Expression of WIPI2B counteracts age-related decline in autophagosome biogenesis in neurons
T cell receptor ( TCR ) engagement opens Ca2+ release-activated Ca2+ ( CRAC ) channels and triggers formation of an immune synapse between T cells and antigen-presenting cells . At the synapse , actin reorganizes into a concentric lamellipod and lamella with retrograde actin flow that helps regulate the intensity and duration of TCR signaling . We find that Ca2+ influx is required to drive actin organization and dynamics at the synapse . Calcium acts by promoting actin depolymerization and localizing actin polymerization and the actin nucleation promotion factor WAVE2 to the periphery of the lamellipod while suppressing polymerization elsewhere . Ca2+-dependent retrograde actin flow corrals ER tubule extensions and STIM1/Orai1 complexes to the synapse center , creating a self-organizing process for CRAC channel localization . Our results demonstrate a new role for Ca2+ as a critical regulator of actin organization and dynamics at the synapse , and reveal potential feedback loops through which Ca2+ influx may modulate TCR signaling . Soon after a T cell encounters cognate antigen on an antigen-presenting cell ( APC ) , it spreads out over the cell’s surface , forming a tightly apposed structure known as the immune synapse ( Bromley et al . , 2001; Yokosuka and Saito , 2010; Dustin , 2008 ) . The synapse regulates T cell activation by maximizing the contact area and organizing the T cell receptors ( TCR ) and associated signaling proteins into zones . Strong antigenic stimuli create three concentric regions ( Monks et al . , 1998; Grakoui et al . , 1999 ) : a central supramolecular activation cluster ( cSMAC ) , an intermediate zone ( the peripheral SMAC , or pSMAC ) , and a zone at the synapse edge ( the distal SMAC , or dSMAC ) ( Freiberg et al . , 2002 ) . TCRs assemble with scaffolding and signaling proteins to form microclusters in the dSMAC which migrate centripetally towards the cSMAC ( Grakoui et al . , 1999; Krummel et al . , 2000; Campi et al . , 2005; Varma et al . , 2006; Yokosuka et al . , 2005 ) . As they move , TCR microclusters activate a MAP kinase cascade and Ca2+ influx through Ca2+ release-activated Ca2+ ( CRAC ) channels , both of which are essential to initiate gene expression programs that drive T cell proliferation and differentiation ( Feske et al . , 2001 ) . Signaling by TCR microclusters is terminated as they enter the cSMAC by the dissociation of signaling proteins ( Yokosuka et al . , 2005; Campi et al . , 2005; Varma et al . , 2006 ) and endocytosis of TCRs ( Lee et al . , 2003; Liu et al . , 2000; Das et al . , 2004 ) . Thus , the strength of signaling at the synapse is thought to reflect a dynamic balance between formation of new microclusters in the dSMAC/pSMAC and their disassembly and internalization in the cSMAC . Actin reorganization at the synapse is crucial for TCR microcluster assembly , movement and signaling ( Babich et al . , 2012; Campi et al . , 2005; Delon et al . , 1998; Kaizuka et al . , 2007; Liu et al . , 1995; Valitutti et al . , 1995; Varma et al . , 2006; Yi et al . , 2012; Kumari et al . , 2015 ) . In the dSMAC , actin is dense and highly branched ( Parsey and Lewis , 1993; Bunnell et al . , 2001 ) and exhibits rapid retrograde flow similar to actin in the lamellipod of migrating cells . In the neighboring pSMAC region , actin is less dense and resembles a lamella with actin organized into concentric arcs by myosin IIA ( Babich et al . , 2012; Yi et al . , 2012; Yu et al . , 2012 ) . Actin is sparse in the actin-depleted zone ( ADZ ) corresponding to the cSMAC . Centripetal actin flow regulates TCR function in at least two ways . First , it transports TCR microclusters towards the cSMAC where they are disassembled , limiting the signaling lifetime of each microcluster to a few minutes ( Yu et al . , 2010; Varma et al . , 2006; Yokosuka et al . , 2005 ) . Second , actin polymerization and depolymerization are critical for microcluster formation and function , based on the ability of cytochalasin D ( an actin polymerization inhibitor ) and jasplakinolide ( an actin depolymerization inhibitor ) to rapidly quell microcluster formation , MAP kinase signaling , and Ca2+ influx at the synapse ( Valitutti et al . , 1995; Varma et al . , 2006; Rivas et al . , 2004; Babich et al . , 2012; Yi et al . , 2012 ) . Thus , the mechanisms that control actin organization and flow at the synapse are key to understanding synapse formation as well as T-cell signaling . TCR stimulation is known to drive actin reorganization by activating the Rho-family GTPases Rac1 and Cdc42 , which function via Wiscott-Aldrich syndrome protein ( WASp ) and WASp-family verprolin homologous protein ( WAVE2 ) to initiate actin nucleation through the Arp2/3 complex ( Billadeau et al . , 2007 ) . Recent studies have shown that actin polymerization collaborates with myosin IIA contractility to drive retrograde actin flow from the lamellipod to the ADZ , although there is some disagreement as to their relative contributions ( Babich et al . , 2012; Yi et al . , 2012 ) . The mechanisms that control retrograde flow at the synapse are still not fully understood , and the possibility remains that a master regulator of some kind may act on a global scale to organize this process . Indirect evidence suggests that intracellular Ca2+ may regulate actin organization and dynamics at the synapse . Elevated intracellular Ca2+ ( [Ca2+]i ) in T cells has been associated with such cytoskeleton-dependent processes as motility arrest ( Negulescu et al . , 1996; Bhakta et al . , 2005 ) , cell rounding ( Donnadieu et al . , 1994 ) , cell spreading ( Bunnell et al . , 2001 ) and synapse stabilization ( Negulescu et al . , 1996; Krummel et al . , 2000; Delon et al . , 1998 ) . In addition , T cells express a range of Ca2+-sensitive proteins known to regulate actin depolymerization , severing , bundling , and capping ( Babich and Burkhardt , 2013; Joseph et al . , 2014; Janmey , 1994 ) . TCR engagement is known to elicit Ca2+ influx through CRAC channels via a cascade in which PLCγ generates inositol 1 , 4 , 5-trisphosphate ( IP3 ) , releasing Ca2+ from the ER and causing the ER Ca2+ sensor STIM1 to redistribute to ER-plasma membrane ( PM ) junctions ( Wu et al . , 2006; Luik et al . , 2006 ) where it traps and activates Orai1 , the pore-forming subunit of the CRAC channel ( Luik et al . , 2006; Wu et al . , 2014 ) . STIM1 and Orai1 colocalize at early times at the immune synapse ( Lioudyno et al . , 2008; Barr et al . , 2008 ) and later at the distal pole of the cell ( Barr et al . , 2008 ) , but functional CRAC channel complexes as indicated by Ca2+ influx have only been shown at the synaptic contact zone ( Lioudyno et al . , 2008 ) . The precise localization of CRAC channels at the synapse , the mechanisms that control their localization , and their possible effects on actin organization and dynamics are all unknown . In this study , we applied an in vitro model system to investigate the localization of CRAC channels and the role these channels may play in regulating the actin cytoskeleton at the immune synapse . We found that Ca2+ influx through CRAC channels acts at multiple levels to organize actin and promote retrograde flow , which in turn drives ER remodeling and the localization of STIM1 and Orai1 to the center of the synapse . In this way , Ca2+ self-organizes CRAC channels at the synapse while creating feedback loops that may help regulate T cell sensitivity to antigen . To study the location and redistribution of the population of STIM1/Orai1 complexes positioned at the synapse , Jurkat T cells expressing STIM1 labeled with mCherry ( mCh-STIM1 ) and Orai1 labeled with EGFP ( Orai1-EGFP ) were stimulated on coverslips coated with anti-CD3 mAb ( Bunnell et al . , 2001 ) . Under these conditions , the cells spread over the coverslip to form a structure resembling an immune synapse and time-lapse TIRF microscopy can be used to obtain high resolution 2-dimensional images of the cell region within 200 nm of the coverslip . Previous studies have shown that cells stimulated in this way reorganize their cytoskeleton similarly to T cells forming conjugates with APCs or binding to peptide-MHC complexes in supported planar bilayers ( Parsey and Lewis , 1993; Bunnell et al . , 2001; Yi et al . , 2012 ) . After settling on stimulatory coverslips , cells spread over several minutes until they reached a constant size and roughly circular shape . Puncta containing STIM1 and Orai1 appeared at the contact zone beginning within seconds of initial contact and continuing through the spreading phase . After cells had spread fully ( 3–7 min after contact with the coverslip ) , colocalized puncta of STIM1 and Orai1 continued to increase in number and intensity over the next several minutes and appeared to be confined to the center of the synapse ( 68 of 82 cells; Figure 1A and Video 1 ) . While the great majority of cells had centralized puncta , the abundance varied from only 5 to an array too densely packed to accurately count , possibly reflecting cell-to-cell variations in STIM1 and Orai1 expression and the degree of ER [Ca2+] depletion . In a minority of cells ( 27 of 68 cells ) , puncta containing STIM1 and Orai1 were detected near the periphery and moved toward the center of the synapse ( Figure 1B , C , Video 1 ) with an average velocity of 47 ± 3 nm/s ( n = 24 puncta; mean ± SEM ) . These motile puncta were more frequently detected in 0 . 5–0 . 8 mM extracellular Ca2+ ( Ca2+o; 48% of 33 cells ) than in 2 mM Ca2+o ( 31% of 35 cells ) , probably because a greater degree of ER Ca2+ depletion is expected under the reduced [Ca2+]o conditions . Thus , we suspect that our experiments actually underestimate the number of motile STIM1-Orai complexes at the synapse because they are dim and difficult to detect when ER [Ca2+] is only partially depleted . Puncta of colocalized STIM1 and Orai1 correspond to ER-PM junctions where STIM1-bound Orai1 conducts Ca2+ into the cell ( Luik et al . , 2006; Wu et al . , 2006 ) . Thus , our results suggest that as the synapse matures ER-PM junctions become concentrated in the center of the contact zone , and individual ER-PM junctions loaded with STIM1 and Orai1 translocate from the periphery to further increase the density of Ca2+ influx sites in the center . 10 . 7554/eLife . 14850 . 003Figure 1 . STIM1 and Orai1 accumulate in puncta in the actin-depleted zone of the immune synapse . ( A ) TIRF images of Jurkat cells stimulated on anti-CD3 coated coverslips in 0 . 8 mM Ca2+o . mCh-STIM1 ( green ) and Orai1-EGFP ( red ) puncta accumulate in the center of the synapse over time . Images taken from Video 1 . Scale bar , 5 µm . ( B ) Magnification of the boxed region in A shows a STIM1/Orai1 punctum ( arrows ) moving toward the center of the synapse . Gamma was adjusted to highlight puncta ( mCh-STIM1 gamma = 1 . 3 and Orai1-EGFP gamma = 1 . 5 ) . ( C ) Centripetal trajectories of STIM1 and Orai1 puncta overlaid on a single image of Orai1-EGFP . The frame-to-frame punctum velocity was 47 ± 3 nm/s ( n = 24 particles , mean ± SEM ) . Dashed line indicates the cell edge . ( D ) ER tubules containing mCh-STIM1 ( green ) move centripetally with contraction of the EGFP-actin ( red ) ring . The dashed line indicates the boundary of the ADZ . ( E ) Kymograph analysis along the indicated line ( left ) from the cell in D ( see Video 2 ) . STIM1 moves at the same velocity as the edge of the actin ring . In all panels , time after initial image acquisition is indicated in min:sec; scale bar , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 00310 . 7554/eLife . 14850 . 004Video 1 . STIM1 and Orai1 accumulate in puncta in the center of the synapse . Time-lapse TIRF movie of a Jurkat cell expressing mCh-STIM1 ( left ) and Orai1-EGFP ( center ) stimulated on an anti-CD3-coated coverslip . A merge of the STIM1 ( green ) and Orai1 ( red ) channels is shown at right . Images acquired every 5 s and time compressed 35x . Scale bar , 5 µm . This video supplements Figure 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 004 To address the possible role of actin in the localization of STIM1 and Orai1 puncta , we examined STIM1 and actin dynamics simultaneously in cells expressing mCh-STIM1 and actin labeled with GFP ( GFP-actin ) . In agreement with previous reports ( Bunnell et al . , 2001; Kaizuka et al . , 2007; Yu et al . , 2010; Babich et al . , 2012; Yi et al . , 2012 ) , cells formed a peripheral lamellipod characterized by a bright band of actin that appeared striated and ruffled in and out of the TIRF plane . At the inner edge of the lamellipod , actin density dropped off sharply , marking the transition into the lamella region where actin formed arc-like structures encircling a central ADZ ( Video 2 ) . Actin moved continually in a radial retrograde direction at velocities that declined from ~100 nm/s at the cell edge to near 0 nm/s at the border of the ADZ ( data not shown ) . The highest density of STIM1 puncta occurred within the ADZ while dimmer , more dynamic STIM1-containing tubules extended into the lamella ( 21 of 21 cells; Figure 1D and Video 2 ) . Kymograph analysis shows that STIM1 puncta in the lamella move centripetally with and at the same velocity as F-actin ( Figure 1E ) . These observations suggest that the advancing actin cytoskeletal network moves STIM1/Orai1 puncta and the associated ER-PM junctions towards the ADZ . 10 . 7554/eLife . 14850 . 005Video 2 . STIM1 puncta accumulate in the ADZ of the synapse . Time-lapse TIRF movie of a Jurkat cell expressing mCh-STIM1 ( left ) and GFP-actin ( center ) stimulated on an anti-CD3-coated coverslip . A merge of the STIM1 ( green ) and actin ( red ) channels is shown at right . Images acquired every 5 s and time compressed , 35x . Scale bar , 5 µm . This video supplements Figure 1D and E . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 005 ER organization and behavior at the immune synapse has not been well studied . To better understand the mechanisms underlying CRAC channel positioning we examined ER localization and dynamics and their potential links to the actin cytoskeleton . We labeled actin and the ER membrane by expressing GFP-actin and mCherry tail-anchored to the ER membrane ( ER-mCh ) ( Bulbarelli et al . , 2002 ) . The ER appeared in the TIRF evanescent field within minutes of cell contact with the stimulatory coverslips and expanded peripherally as cells spread ( Figure 2A and Video 3 ) . The ER near the PM was highly enriched in the ADZ ( 20 of 20 cells ) in both tubular and sheet-like structures that became more centrally concentrated and immobile as the ring of actin surrounding the ADZ contracted . Dynamic ER tubules extended from the ADZ toward the lamellipod ( 16 of 20 cells; Figure 2B , pink arrows ) and occasionally traversed the lamella/lamellipod border , then either rapidly retracted along a similar trajectory ( 10 of 20 cells; Figure 2Bii , green arrows ) , or appeared to bend before moving centripetally ( 20 of 20 cells; Figure 2Biii , cyan arrows ) . A subset of tubules that penetrated the lamellipod remained relatively immobile in actin-sparse regions ( 8 of 20 cells; Figure 2Biii , yellow arrows ) . 10 . 7554/eLife . 14850 . 006Figure 2 . Synaptic ER tubules extend from the ADZ and are moved centripetally by actin . ( A ) TIRF images of a Jurkat cell coexpressing GFP-actin ( red ) and ER-mCh ( green ) , after spreading on an anti-CD3-coated coverslip . ( B ) Magnification of the boxed regions in A depicting an extending ER tubule ( i , pink arrows ) , a tubule extending and retracting along the same trajectory ( ii , green arrows ) , a tubule bending and moving centripetally between actin filaments ( iii , cyan arrows ) and an immobile tubule in an actin-poor region ( iii , yellow arrows ) . ( C ) Kymograph analysis of the cell from A along the line shown ( left ) demonstrating coordinated centripetal movement of the ER and actin ( see Video 3 ) . Time after initial image acquisition is indicated in min:sec; scale bar , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 00610 . 7554/eLife . 14850 . 007Figure 2—figure supplement 1 . The ER extends at the tips of dynamic microtubules that move radially toward the lamella/lamellipod border . ( A ) TIRF images of a Jurkat cell stimulated on anti-CD3 coated coverslips expressing EB1-EGFP ( red ) and ER-mCh ( green ) . Magnified view of the boxed region shows an ER tubule moving peripherally ( green arrow ) with EB1 at the tip ( red arrow ) . Images are from Video 4 . ( B ) A projection of the standard deviation of 40 images of EB1 acquired at 1-s intervals overlaid on a single image of F-tractin-P-tdTom , indicating radial EB1 movement in the ADZ and movements perpendicular to cell edge at lamella/lamellipod border . Images are from Video 5 . Time after initial image acquisition is indicated in min:sec; scale bar , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 00710 . 7554/eLife . 14850 . 008Video 3 . ER tubules extend from the ADZ into the lamella and are moved centripetally by actin . Time-lapse TIRF movie of a Jurkat cell expressing GFP-actin ( left ) and ER-mCh ( middle ) stimulated on an anti-CD3-coated coverslip . A merge of the actin ( red ) and ER ( green ) channels is shown at right . Images acquired every 5 s and time compressed 35x . Scale bar , 5 µm . This video supplements Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 008 What mechanisms determine ER dynamics at the synapse ? In general , ER tubules can extend by sliding along the sides of microtubules or by attaching to the tips of growing microtubules ( Waterman-Storer and Salmon , 1998 ) through an interaction between STIM1 and the microtubule tip attachment proteins EB1 and EB3 ( Grigoriev et al . , 2008 ) . In Jurkat cells expressing EB1 labeled with EGFP ( EB1-EGFP ) and ER-mCh , EB1 was seen at the tips of many extending ER tubules ( Figure 2—figure supplement 1A and Video 4 ) , confirming that ER tubules can extend toward the synapse periphery by attaching to the tips of growing microtubules . ER tubules rarely extended into the lamellipod , likely reflecting infrequent microtubule forays into the lamellipod ( Figure 2—figure supplement 1B and Video 5 ) , as has been reported for migrating epithelial cells ( Waterman-Storer and Salmon , 1997 ) . Like ER tubules , EB1 moved roughly radially though the ADZ and lamella , but at the lamellipod/lamella border the majority reoriented and moved parallel to the synapse edge or disappeared as they moved above the TIRF evanescent field ( Figure 2—figure supplement 1B and Video 5 ) . These findings suggest that ER tubules infrequently enter the lamellipod because microtubules cannot easily penetrate this thin , actin-dense compartment . 10 . 7554/eLife . 14850 . 009Video 4 . ER tubules extend toward the synapse edge on the tips of microtubules . Time-lapse TIRF movie of a Jurkat cell expressing ER-mCh ( left ) and EB1-EGFP ( middle ) stimulated on an anti-CD3-coated coverslip . A merge of the ER ( green ) and EB1 ( red ) channels is shown at right . Images acquired every 5 s and time compressed 35x . Scale bar , 5 µm . This video corresponds to the boxed region in Figure 2—figure supplement 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 00910 . 7554/eLife . 14850 . 010Video 5 . EB1 moves radially in the ADZ but parallel to the cell edge at the lamella/lamellipod border . Time-lapse TIRF movie of a Jurkat cell expressing F-tractin-P-tdTom ( left ) and EB1-EGFP ( middle ) stimulated on an anti-CD3-coated coverslip . A merge of the actin ( red ) and EB1 ( green ) channels is shown at right . Images acquired every 1 s and time compressed 7x . Scale bar , 5 µm . This video corresponds to Figure 2—figure supplement 1B . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 010 Whereas ER tubule elongation was closely associated with microtubule extension , retrograde ER movement was linked to centripetal actin flow . ER tubules undergoing retrograde movement in the lamella were commonly sandwiched between actin arcs ( 17 of 20 cells; Figure 2Biii , cyan arrows , and Video 3 ) , and both moved at the same velocity ( Figure 2C ) . Moreover , the similar retrograde velocities of isolated STIM1/Orai1 puncta ( 47 ± 3 nm/s , mean ± SEM , n = 24 particles; Figure 1C ) and lamellar actin ( 36 ± 6 nm/s , mean ± SEM , n = 18 cells; Figure 6F ) suggest that retrograde actin flow also sweeps ER-PM junctions into the ADZ . Based on these results , we conclude that the ER at the synapse is dynamic , with peripheral extension on microtubules continually balanced by retraction imposed by retrograde actin flow . Given the high density of Ca2+ influx sites in the ADZ and the known ability of Ca2+ to regulate actin dynamics in many cells ( Janmey , 1994 ) , we asked whether Ca2+ influx might acutely regulate retrograde actin flow at the synapse . To this end , we expressed in Jurkat cells the low affinity F-actin binding domain of inositol trisphosphate 3-kinase A labeled with a tandem dimer of fluorescent Tomato ( F-tractin-P-tdTom ) , which allows visualization of filamentous actin without alteration of actin dynamics or function ( Johnson and Schell , 2009; Yi et al . , 2012 ) . In the presence of Ca2+o , cells formed a clearly defined lamellipod and lamella ( Figure 3A , left ) with extensive ruffling of the lamellipod and continuous bulk retrograde actin flow as described above in cells expressing GFP-actin ( Video 2 ) . 10 . 7554/eLife . 14850 . 011Figure 3 . Calcium influx organizes synaptic actin and promotes retrograde flow . ( A ) TIRF images of a Jurkat cell expressing F-tractin-P-tdTom after spreading on anti-CD3 in 2 mM Ca2+o ( left ) , 3 . 25 min after Ca2+o removal ( center ) , and 1 min after readdition of 2 mM Ca2+o ( right ) . Ca2+ alters F-actin organization and density . Images taken from Video 6 . ( B ) Spatiotemporal image correlation spectroscopy ( STICS ) analysis ( Hebert et al . , 2005 ) of the cell in A , depicting the direction and relative velocity of actin movement before ( left ) and after Ca2+o removal ( center ) and after readdition of 2 mM Ca2+o ( right ) . Color scale represents relative velocities; numerical values were not assigned because small immobile features cause underestimation of velocity by STICS . ( C , D ) Blocking Ca2+ influx with 2-APB has the same effect on actin as removal of Ca2+o . A representative cell is shown before and 2 . 5 min after treatment with 100 µM 2-APB , and STICS analysis is shown in D . ( E ) Spinning disk confocal images of a primary human T lymphoblast expressing Lifeact-GFP after spreading on anti-CD3 and ICAM-1 in 0 . 5 mM Ca2+o ( left ) , 3 min after Ca2+o removal ( center ) , and 1 . 5 min after readdition of 2 mM Ca2+o ( right ) . The width of the lamellipod ( indicated by the green carets ) was reduced in 0 Ca2+o . Images are maximum intensity projections of 3 successive 0 . 25 µm sections of the cell footprint taken from Video 7 . ( F ) Kymograph analysis of the cell in E along the indicated yellow line ( left ) demonstrates centripetal actin flow rate of 426 nm/s in 0 . 5 mM Ca2+o ( left , velocity calculated from the slope of the red dashed lines ) that slows to 94 nm/s upon Ca2+o removal ( center ) and accelerates to 130 nm/s following readdition of 2 mM Ca2+o . Time is indicated in min:sec; scale bars , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 011 Perfusion with Ca2+-free medium to terminate Ca2+ influx through CRAC channels caused several rapid and profound changes in the organization and dynamics of F-tractin-P at the synapse . In the majority of cells ( 38 of 45 ) , the distinguishing features of the lamellipod and lamella were lost: ruffling at the periphery was greatly reduced and lamella actin arcs became less apparent , the lamella/lamellipod boundary disappeared as F-actin became more uniformly distributed across the synapse , and in some cells ( 15 of 45 ) F-actin extended into the ADZ ( Figure 3A and Video 6 ) . Most strikingly , the centripetal movement of actin was severely reduced and any remaining movement was less radial and more randomly oriented ( Figure 3B and Video 6 ) . These effects all reversed within seconds of restoring Ca2+o , and were also observed in cells expressing GFP-actin ( data not shown ) . Pharmacological inhibition of CRAC channel function with 2-aminoethyldiphenyl borate ( 2-APB ) in the presence of Ca2+o produced a similar response ( 11 of 14 cells; Figure 3C , D ) , indicating that changes in actin organization and dynamics result from the inhibition of Ca2+ influx through CRAC channels rather than from the removal of Ca2+o itself . 10 . 7554/eLife . 14850 . 012Video 6 . Calcium influx organizes synaptic actin and promotes retrograde flow in Jurkat cells . Time-lapse TIRF movie of a Jurkat cell expressing F-tractin-P-tdTom after spreading on an anti-CD3 coverslip in 2 mM Ca2+o , followed by perfusion with 0 Ca2+o and 2 mM Ca2+o . Images acquired every 5 s and time compressed 35x; scale bar , 5 µm . This video supplements Figure 3A and B . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 012 A critical question is whether these effects of Ca2+ on actin in Jurkat leukemic T cells extend to primary human T cells making synapses in a more physiological setting . A recent report has described WASp-associated actin foci in primary T cell synapses that were not detectable in Jurkat cells , suggesting that actin organization in primary T cells may be more complex than previously recognized ( Kumari et al . , 2015 ) . Jurkat cells also lack the lipid phosphatase PTEN , which may affect actin dynamics by enhancing PIP3 accumulation in the plasma membrane ( Shan et al . , 2000 ) . Finally , primary T cells are normally activated by APCs displaying the integrin ICAM-1 , which modestly alters actin organization and slows retrograde actin flow at the synapse ( Comrie et al . , 2015 ) . To investigate effects of Ca2+ on actin dynamics in a more physiological model , we transduced primary human CD4+ T lymphoblasts with Lifeact-GFP , a short F-actin binding peptide from Abp140 tethered to GFP ( Riedl et al . , 2008 ) . T lymphoblasts plated on anti-CD3- and ICAM-1-coated coverslips formed a distinct lamellipod with rapid retrograde actin flow as described previously ( Comrie et al . , 2015 ) and Ca2+ removal altered the distribution of synaptic actin and reduced ruffling and retrograde flow ( 20 of 21 cells; Figure 3E , Video 7 ) . Kymograph analysis showed that on average , Ca2+ removal slowed actin flow by 44% and Ca2+o reperfusion restored it to 97% of the initial velocity ( Figure 3F , Table 1 ) . Ca2+o removal narrowed the lamellipod by 61% , while Ca2+o restoration returned the lamellipod to 82% of its initial width ( Figure 3E , Table 1 ) . T lymphoblasts on coverslips coated with anti-CD3 alone produced similar responses ( Video 8 , 40 of 45 cells , Table 1 ) . Thus , primary T lymphoblasts and Jurkat T cells responded similarly to changes in Ca2+o , supporting the use of Jurkat T cells as a physiologically relevant model system for studying the effects of Ca2+ on actin dynamics at the synapse . 10 . 7554/eLife . 14850 . 013Video 7 . Calcium influx organizes synaptic actin and promotes retrograde flow in primary human T lymphoblasts plated on anti-CD3 and ICAM-1 . Time-lapse spinning disk confocal movie of a human T lymphoblast expressing Lifeact-GFP after spreading on anti-CD3 and ICAM-1 Fc in 0 . 5 mM Ca2+o ( left ) , 1 . 3 min following perfusion with 0 Ca2+o ( center ) and 2 . 2 min after perfusion with 2 mM Ca2+o ( right ) . Images are displayed as maximum intensity projections of 3 image planes separated by 0 . 25 µm that were acquired at 2 s intervals . Time compressed 40x; scale bar , 5 µm . This video supplements Figure 3E and F . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 01310 . 7554/eLife . 14850 . 014Table 1 . Effects of calcium on actin dynamics at the primary T cell immune synapseDOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 014Anti-CD3Anti-CD3 + ICAM-10 . 5 Ca2+0 Ca2+2 Ca2+0 . 5 Ca2+0 Ca2+2 Ca2+Velocity ( nm/s ) 243 ± 8 ( 9 ) 145 ± 6 ( 9 ) 208 ± 8 ( 9 ) 167 ± 4 ( 11 ) 94 ± 3 ( 11 ) 162 ± 5 ( 11 ) Lamellipod width ( µm ) 3 . 0 ± 0 . 2 ( 9 ) 1 . 5 ± 0 . 1 ( 9 ) 2 . 3 ± 0 . 2 ( 9 ) 2 . 8 ± 0 . 1 ( 10 ) 1 . 1 ± 0 . 1 ( 10 ) 2 . 3 ± 0 . 2 ( 8 ) [Ca2+]o indicated in mM in the order in which the solutions were applied ( see text ) . Means ± SEM; number of cells indicated in parentheses . Velocities are from a total of 104-133 measurements from kymographs made at 3 different locations per cell10 . 7554/eLife . 14850 . 015Video 8 . Calcium influx organizes synaptic actin and promotes retrograde actin flow in primary human T lymphoblasts plated on anti-CD3 without ICAM-1 . Time-lapse spinning disk confocal movie of a primary human T lymphoblast expressing Lifeact-GFP after spreading on anti-CD3 in 0 . 5 mM Ca2+o ( left ) , 4 . 3 min following perfusion with 0 Ca2+o ( center ) and 2 min following perfusion with 2 mM Ca2+o ( right ) . Images are displayed as maximum intensity projections of sets of 3 image planes separated by 0 . 25 µm that were acquired at 2 s intervals . Time compressed 40x; scale bar , 5 µm . This video supplements Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 015 Because ER tubule movement is influenced by centripetal actin flow ( Figure 2C ) , we examined the effect of Ca2+o removal on ER tubule distribution and dynamics at the Jurkat cell synapse . In the absence of Ca2+o , very few ER tubules were visible in the lamella/lamellipod region by TIRF although small segments of ER were seen near the cell edge , suggesting that under these conditions , peripheral ER tubules extend in the Z dimension out of the TIRF evanescent field ( Figure 4A , B ) . Readdition of Ca2+o initiated a rapid increase in the density of ER tubules in the lamella consistent with their reentry into the evanescent field , and tubules moved centripetally as retrograde actin flow resumed ( 6 of 6 cells; Figure 4B , C , and Video 9 ) . These results demonstrate that Ca2+ influx through CRAC channels helps to corral extended peripheral ER tubules back to the center of the synapse by promoting retrograde actin flow . 10 . 7554/eLife . 14850 . 016Figure 4 . Calcium influx promotes ER corralling . ( A ) A cell expressing F-tractin-P-tdTom ( red ) and ER-GFP ( green ) on an anti-CD3 coverslip is shown in 0 Ca2+o and after readdition of 2 mM Ca2+o . Peripheral ER tubules in the TIRF images are sparse in 0 Ca2+o , but Ca2+o readdition causes peripheral tubules to appear as they move into the evanescent field . Pink dotted lines outlining the edge of the ADZ in 0 Ca2+o serve as a landmark to highlight centripetal ER movement following Ca2+o readdition . Images taken from Video 9 . ( B ) Peripheral ER tubules were traced in 5 images acquired at 10 s intervals , then color-coded for time and overlaid to indicate movement between frames in 0 Ca2+o ( top ) and immediately following re-addition of 2 mM Ca2+o ( bottom ) . In 0 Ca2+o , peripheral tubules are sparse , extended and move in radial and non-radial directions . Peripheral tubules appearing upon readdition of Ca2+o move centripetally . ( C ) Kymograph analysis of the cell in A along the indicated line ( left ) demonstrates centripetal movement of ER tubules between actin structures upon Ca2+o readdition . Black horizontal lines indicate bath exchange . Gamma adjusted to 0 . 7 to highlight ER tubules . Time after initial image acquisition is indicated in min:sec; scale bar , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 01610 . 7554/eLife . 14850 . 017Video 9 . Calcium-dependent retrograde actin flow corrals the ER in the ADZ . Time-lapse TIRF movie of a Jurkat cell expressing ER-GFP ( left ) and F-tractin-P-tdTom ( center ) that had spread in 2 mM Ca2+o before perfusion with 0 Ca2+o . Video begins with the cell in 0 Ca2+o and shows the effect of restoring 2 mM Ca2+o . A merge of the ER ( green ) and F-tractin-P ( red ) channels is shown at right . Images acquired every 5 s and time compressed 35x . Scale bar , 5 µm . This video supplements Figure 4A–C . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 017 One clue to the mechanism of calcium’s effects on actin dynamics at the synapse was that blocking Ca2+ influx increased F-tractin-P fluorescence intensity ( and thus F-actin density ) by 20–30% ( Figure 5A , Figure 5—figure supplement 1A ) . Elevated F-tractin-P fluorescence did not appear to be a consequence of bulk movement of cellular structures into the evanescent field ( such as might result from changes in cell shape ) because Ca2+o removal also increased the F-tractin-P fluorescence at the synapse when viewed by spinning disk confocal microscopy , which samples a much thicker optical section ( Figure 5—figure supplement 1B , C ) . 10 . 7554/eLife . 14850 . 018Figure 5 . Intracellular calcium reduces the density of F-actin at the synapse . Jurkat T cells expressing F-tractin-P-tdTom and loaded with fura-2 were stimulated on anti-CD3-coated coverslips . ( A ) Pseudocolor image of F-tractin-P-tdTom intensity in a cell exposed sequentially to 0 . 5 mM Ca2+o ( left ) , 0 Ca2+o ( center ) and 5 mM Ca2+o ( right ) , indicating a Ca2+-dependent decrease in F-actin density . Linear color scale indicates fluorescence intensity ( 0–1 a . u . ) ; scale bar , 5 µm . ( B ) Change in F-tractin-P-tdTom fluorescence ( green; relative to fluorescence in 0 Ca2+o ) and fura-2 ratio ( blue ) from the cell in A . The data are replotted on the right with an inverted F-tractin-P axis to highlight the delay between changes in [Ca2+]i and F-tractin-P intensity upon Ca2+o removal ( top ) and readdition ( bottom ) . ( C ) Change in F-tractin-P-tdTom fluorescence ( relative to fluorescence in 0 Ca2+o or 100 µM 2-APB in 0 . 5 mM Ca2+o ) as a function of fura-2 ratio . Each point is an average single-cell value measured at constant fura-2 ratio and F-tractin-P fluorescence in the presence of 0 . 5–10 mM Ca2+o ( green ) . The red dot indicates the average baseline fura-2 ratio ( ± s . d . ) for all cells in 0 Ca2+o or 2-APB . A linear fit to the data is shown ( r2 = 0 . 83 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 01810 . 7554/eLife . 14850 . 019Figure 5—figure supplement 1 . Ca2+ influx through CRAC channels reduces F-actin density at the synapse . ( A ) Jurkat cells expressing F-tractin-P-tdTom were allowed to spread on anti-CD3 coated coverslips in 2 mM Ca2+o , followed by perfusion with 100 µM 2-APB to block Ca2+ influx through Orai1 . 2-APB causes a 20% increase in F-actin at the synapse as indicated by the F-tractin-P-tdTom intensity vs . time normalized to the average fluorescence of the last five images ( means ± SEM; n = 8 cells ) . ( B ) Spinning disk confocal images of a Jurkat cell expressing F-tractin-P-tdTom stimulated on anti-CD3 coated coverslips in 0 . 5 mM Ca2+o ( left ) and 100 s after perfusion of 0 Ca2+o ( right ) . Scale bar , 5 µm; color scale indicates fluorescence intensity ( 0–1 a . u . ) . ( C ) The average F-tractin-P-tdTom intensity vs . time normalized to the average fluorescence of the last five images for cells stimulated and imaged as in B ( means ± SEM; n = 10 cells ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 019 To better understand the relationship between F-actin density and [Ca2+]i , we made measurements from single cells expressing F-tractin-P-tdTom and loaded with the Ca2+-sensitive dye fura-2 . Cells spreading on anti-CD3 in the presence of Ca2+o had variable [Ca2+]i consistent with known cell-to-cell variation in proximal TCR signaling in Jurkat cells ( Lewis and Cahalan , 1989 ) . When both [Ca2+]i and F-tractin-P fluorescence reached steady-state , Ca2+o was removed . In some cells stimulated in 2 mM Ca2+o , [Ca2+]i declined slowly following Ca2+o removal and did not reach a plateau , probably due to the slow release of mitochondrial Ca2+ into the cytosol ( Hoth et al . , 1997 ) . To avoid this complication , we studied synapses formed in 0 . 5 mM Ca2+o , for which Ca2+o removal or 2-APB application evoked a rapid and monotonic [Ca2+]i decline to a similar plateau level in all cells ( fura-2 ratio of 0 . 39 ± 0 . 04 , mean ± SEM , n = 26 cells ) . The decline in [Ca2+]i was closely followed by an increase in F-tractin-P fluorescence that plateaued ~30 s after [Ca2+]i ( Figure 5B , right ) . Similarly , readdition of Ca2+o caused [Ca2+]i to rise and F-actin to decline , demonstrating a rapidly reversible effect on F-actin density . Quantifying the relationship between [Ca2+]i and F-actin concentration is complicated by variation in the expression of F-tractin-P among cells . We therefore quantified the change in F-tractin-P fluorescence in each cell relative to the value in 0 Ca2+o or following 2-APB application , which produced a similar minimum [Ca2+]i in all cells . F-tractin-P fluorescence and [Ca2+]i were measured during the initial response to the TCR stimulus ( in 0 . 5 mM Ca2+o ) and following readdition of 2 , 5 , or 10 mM Ca2+o ( Figure 5B , C ) . In this group of 26 cells the level of F-actin declined as [Ca2+]i increased , and this relationship was similar regardless of whether measurements were made before or after Ca2+ removal . The level of F-actin was highly correlated with [Ca2+]i ( R2 = 0 . 83 ) but not with [Ca2+]o ( R2 = 0 . 22 ) , and 2-APB application and Ca2+o removal had similar effects on F-tractin-P density , demonstrating that intracellular Ca2+ reversibly regulates the density of F-actin at the synapse . Given that the steady-state level of F-actin in cells reflects a balance between the overall rates of actin polymerization and depolymerization , Ca2+ could reduce the density of F-actin by increasing the rate of depolymerization and/or by reducing the rate of polymerization . We first examined the effect of Ca2+ on synaptic actin depolymerization in cells coexpressing F-tractin-P-tdTom and photoactivatable GFP-labeled actin ( PAGFP-actin ) . After brief photoactivation of a small region , the GFP fluorescence indicates only F-actin , because monomeric PAGFP-actin rapidly escapes the region by diffusion; thus , the subsequent decay of fluorescence provides a measure of the actin depolymerization rate ( McGrath et al . , 1998 ) . 3–7 min after settling onto stimulatory coverslips , images of F-tractin-P-tdTom were used to identify cells with steady-state treadmilling actin and small ( ~1 by 3 µm ) regions in the lamella and lamellipod were defined for photoactivation ( Figure 6A , left , red ovals ) . PAGFP actin was first photoactivated in the presence Ca2+o and again ~2 min following Ca2+o removal when F-actin had reached a new steady-state ( Figure 6A , Figure 6—figure supplement 1A–D and Video 10 ) . After each photoactivation , the resulting GFP fluorescence was measured over time by widefield microscopy rather than TIRF in order to prevent any change in signal due to possible movement of actin in the Z direction , and translocation of the photoactivated region was used to measure centripetal velocity . 10 . 7554/eLife . 14850 . 020Figure 6 . Calcium accelerates actin depolymerization and centripetal velocity at the synapse . Jurkat T cells expressing F-tractin-P-tdTom and PAGFP-actin were stimulated on anti-CD3 coated coverslips and TIRF images of F-tractin-P-tdTom ( A , left ) were used to identify regions in the lamellipod and lamella ( red ovals ) to photoactivate . The lamella/lamellipod border in 2 mM Ca2+o and cell edge are indicated by pink dashed lines . ( A ) Widefield epifluorescence images of PAGFP-actin after photoactivation in 2 mM Ca2+o ( top ) and a subsequent photoactivation in 0 Ca2+o ( bottom ) . Images are from Video 10; color scale indicates fluorescence intensity ( 0–1 a . u . ) . Time after photoactivation indicated in min:sec . Scale bar , 5 µm . ( B , C ) Position of peak PAGFP-actin fluorescence as a function of time after photoactivation in the lamellipod ( B ) and the lamella ( C ) ( see figure supplement 1A-D ) . Data are plotted in the presence ( blue ) and absence ( red ) of Ca2+o for the cell pictured in A . Linear fits to the data indicate lamellipod velocities of 72 nm/s ( 2 Ca ) and 22 nm/s ( 0 Ca ) and lamella velocities of 41 nm/s ( 2 Ca ) and 1 nm/s ( 0 Ca ) . ( D , E ) The fluorescence decay of photoactivated PAGFP-actin in the lamellipod ( D ) and lamella ( E ) for the cell in A was fitted by a single exponential . In the lamellipod , τ = 8 . 3 s ( 2 Ca ) and 12 . 0 s ( 0 Ca ) ; in the lamella , τ = 9 . 2 s ( 2 Ca ) and 12 . 4 s ( 0 Ca ) . F/Fmax is the fluorescence intensity after photoactivation relative to the peak . ( F ) The centripetal velocity of photoactivated actin in the lamellipod ( n = 10 cells ) and lamella ( n = 18 cells ) in the presence of absence of Ca2+o , calculated as described in B , C . Error bars indicate SEM; p-values from Student’s two-tailed t-test . ( G ) Actin filament half-life calculated from the exponential rate of fluorescence decay in photoactivated regions in the lamellipod ( n = 10 ) and lamella ( n = 19 ) with and without Ca2+o . P-values are from paired Student’s two-tailed t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 02010 . 7554/eLife . 14850 . 021Figure 6—figure supplement 1 . Actin filament velocity and half-life at the synapse . Jurkat cells were stimulated on anti-CD3 coated coverslips and PAGFP-actin was photoactivated in small regions within the lamellipod and lamella . ( A-D ) Fluorescence intensity profiles of photoactivated PAGFP-actin in the lamellipod ( A , B ) and the lamella ( C , D ) in the presence of 2 mM or 0 Ca2+o as indicated . Each trace is the intensity along a line perpendicular to the direction of actin movement and averaged across its 19-pixel width , displayed every 2 s . Data are from the cell in Figure 6 and Video 10 . The 0 position indicates the cell edge , and the dotted line marks the lamellipod/lamella border . Velocity was calculated as in Figure 6 from the peak position versus time . ( E ) Velocity and ( F ) actin filament half-life of photoactivated regions were measured in cells before and after perfusion of 2 mM Ca2+o . P-values from paired Student’s two-tailed t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 02110 . 7554/eLife . 14850 . 022Figure 6—figure supplement 2 . Calcium influx alters actin organization and density independently of myosin activity . ( A ) TIRF images of Jurkat cells expressing F-tractin-P-tdTom pretreated for 30 min with 50 µM blebbistatin , then stimulated on anti-CD3 coated coverslips in 2 mM Ca2+o and blebbistatin ( left ) followed by perfusion with 0 Ca2+o and blebbistatin ( right ) . Scale bar , 5 µm . ( B ) F-tractin-P-tdTom intensity at the synapse versus time for cells as in A , normalized to the average fluorescence of the last five images ( means ± SEM; n = 12 cells ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 02210 . 7554/eLife . 14850 . 023Video 10 . Calcium increases actin depolymerization and centripetal velocity at the synapse . Time-lapse TIRF movie of a Jurkat cell expressing F-tractin-P-tdTom and PAGFP-actin . Bars of PAGFP-actin are photoactivated in the lamella and lamellipod of the same cell in 2 mM Ca2+o ( left ) and 0 Ca2+o ( right ) . Intensity is rendered in pseudocolor using the scale in Figure 6A . Images acquired every 2 s and time compressed 6x . Scale bar , 5 µm . The video supplements Figure 6A . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 023 With Ca2+o present , photoactivated actin moved toward the center of the synapse at velocities comparable to those previously reported from kymograph measurements ( Yi et al . , 2012; Babich et al . , 2012 ) , and moved faster in the lamellipod than in the lamella as expected ( Figure 6B , C , F and Table 2 ) . In the absence of Ca2+o , mean actin velocity decreased by 41% in the region of the lamellipod and by 64% in the lamella ( Figure 6B , C , F and Table 2 ) . 10 . 7554/eLife . 14850 . 024Table 2 . Effects of calcium on actin dynamics at the Jurkat cell immune synapseDOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 024LamellipodLamella2 Ca2+0 Ca2+2 Ca2+0 Ca2+Half-life ( s ) 6 . 5 ± 0 . 4 ( 10 ) 8 . 0 ± 0 . 4 ( 10 ) 6 . 4 ± 0 . 4 ( 19 ) 7 . 5 ± 0 . 5 ( 19 ) Velocity ( nm/s ) 58 ± 6 ( 10 ) 34 ± 5 ( 10 ) 36 ± 6 ( 18 ) 13 ± 3 ( 18 ) [Ca2+]o indicated in mM . Means ± SEM; number of cells indicated in parentheses . We measured the rate of actin filament depolymerization from the single exponential decay of GFP fluorescence with time ( Figure 6D , E and Figure 6—figure supplement 1A–D ) . The decay kinetics in 2 mM Ca2+o were similar in the lamella and lamellipod , and Ca2+ removal extended the mean actin filament half-life by 17% in the lamella and 23% in the lamellipod region ( Figure 6G and Table 2 ) . To control for possible effects of shear force during perfusion or Ca2+-independent changes in actin dynamics over time , we photoactivated regions before and after perfusing the cell with a solution containing the same [Ca2+] and found no significant change in either velocity or half-life ( Figure 6—figure supplement 1E , F ) . Thus , our findings indicate that Ca2+ influx accelerates actin depolymerization at the synapse . Ca2+ can enhance actin depolymerization through many effectors . Myosin IIA seemed a likely candidate because it is present at the synapse ( Ilani et al . , 2007; Jacobelli et al . , 2004; Babich et al . , 2012; Yi et al . , 2012 ) , is known to disassemble actin filaments in the lamella of epithelial cells ( Wilson et al . , 2010 ) , and can be activated by Ca2+ ( Kamm and Stull , 1985 ) . However , after inhibition of myosin ATPase activity with blebbistatin , Ca2+o removal induced a 30% increase in F-actin density ( Figure 6—figure supplement 2A , B ) , similar to its effect in the absence of the drug ( Figure 5C ) . This was not due to a failure to inhibit myosin ATPase activity because blebbistatin treatment caused actin arcs to accumulate in the ADZ as previously reported ( Yi et al . , 2012 ) . These results suggest that Ca2+ regulates F-actin density and depolymerization at the synapse independently of myosin . A second potential mechanism by which Ca2+ could alter F-actin density at the synapse is by influencing actin polymerization . To test for such an effect , we photoactivated PAGFP-actin to release a pool of fluorescent actin monomers and monitored their incorporation into actin filaments . Cells expressing F-tractin-P-tdTom and PAGFP-actin were stimulated on anti-CD3 coverslips and F-tractin-P-tdTom images were used to identify cells with steady-state treadmilling actin . PAGFP-actin was photoactivated within the ADZ , where the majority of actin is expected to be monomeric and freely diffusible , in the presence of Ca2+o or ~1 . 5 min after its removal , a time when [Ca2+]i would be expected to reach a constant minimum ( see Figure 5B; Video 11 ) . After each photoactivation , the incorporation of fluorescent PAGFP-actin monomers into filaments throughout the cell was visualized over time by TIRF . 10 . 7554/eLife . 14850 . 025Video 11 . Calcium restricts actin polymerization to the distal edge of the synapse . Time-lapse TIRF movie of a Jurkat cell expressing F-tractin-P-tdTom and PAGFP-actin . Monomeric PAGFP-actin is photoactivated in the ADZ of two different cells in 2 mM Ca2+o ( left ) and 0 Ca2+o ( right ) . Intensity is rendered in pseudocolor using the scale in Figure 7A . Images acquired every 500 ms and time compressed 3 . 5x . Scale bar , 5 µm . This video supplements Figure 7A and B . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 025 In the absence of Ca2+o , GFP fluorescence increased immediately in the ADZ upon photoactivation , followed by a slower rise throughout the lamella as fluorescent actin monomers diffused through the cytosol and incorporated into F-actin . The fluorescence rise reached the cell’s edge within 3 s of photoactivation , consistent with the rapid diffusion of monomeric actin in cells , and covered most of the actin-rich area of the synapse , reflecting widespread polymerization ( Figure 7A , C ) . In the presence of Ca2+o , actin polymerization was strikingly different . Within 2 s of photoactivation in the ADZ , fluorescence increased selectively in a narrow band around the periphery of the lamellipod ( Figure 7B ) . Fluorescence at the periphery peaked within ~6 s , then declined slightly as the combination of peripheral incorporation and centripetal flow labeled the entire lamellipod , generating a wide band of fluorescent actin that dropped off sharply at the lamella/lamellipod border ( Figure 7B , D ) . In contrast , fluorescence in the lamella increased only minimally during the 30 s following photoactivation . The reduced polymerization in the lamella was not due to the inability of PAGFP-actin to incorporate into lamellar structures ( Yi et al . , 2012 ) because PAGFP-actin incorporated efficiently into the lamella in 0 Ca2+o . These results demonstrate that Ca2+ influx effectively promotes actin polymerization at the distal edge of the lamellipod while suppressing polymerization elsewhere throughout the synapse . 10 . 7554/eLife . 14850 . 026Figure 7 . Calcium restricts actin polymerization to the distal edge of the synapse . ( A , B ) Two Jurkat T cells expressing F-tractin-P-tdTom and PAGFP-actin were stimulated on anti-CD3 coated coverslips in 0 . 5 mM Ca2+ , and PAGFP-actin was photoactivated in the ADZ regions indicated in the F-tractin-P-tdTom TIRF images ( left , yellow circles ) 2 min after perfusion of 0 Ca2+o ( A ) or 2 mM Ca2+o ( B ) . Incorporation of fluorescent PAGFP-actin is shown as a function of time after photoactivation . The lamella/lamellipod border in 2/0 . 5 mM Ca2+o and cell edge are indicated by pink dashed lines . Images are from Video 11 . Time after photoactivation is in min:sec; scale bar , 5 µm; color scale indicates fluorescence intensity ( 0–1 a . u . ) . ( C , D ) Normalized PAGFP-actin fluorescence intensity ( see Materials and methods ) along the line indicated ( top right ) as a function of radial position . The fluorescence profile before photoactivation is shown in black; the color scale applies to subsequent profiles acquired every 1 . 5 s after photoactivation . The cell edge ( red arrowhead ) , the lamellipod/lamella border ( blue arrowhead ) and the edge of the ADZ ( green arrowhead ) are indicated . Data are representative of 12–13 cells . ( E ) Representative TIRF images of Jurkat cells stimulated on anti-CD3 in 0 . 5 mM Ca2+then transferred to 0 . 5 Ca2+ ( left ) or 0 Ca2+ ( right ) for 2 . 5 min , labeled with Alexa-594 phalloidin ( top ) and anti-WAVE2 ( bottom ) ( see Materials and methods ) . Ca2+ promotes localization of WAVE2 to the edge of the lamellipod . ( F ) Average anti-WAVE2 fluorescence in a 1-µm band around the perimeter of the synapse ( Fperimeter ) relative to the average fluorescence across the whole synapse ( Fcell ) in 0 and 0 . 5 mM Ca2+ ( n = 63 cells each ) . Error bars indicate SEM; p-values from Student’s two-tailed t-test . ( G ) TIRF images of a Jurkat cell expressing F-tractin-P-tdTom ( top ) and EGFP-Abi1 ( bottom ) stimulated on anti-CD3 in 0 . 5 mM Ca2+o ( left ) , 1 . 5 min after Ca2+o removal ( center ) , and 1 . 5 min after readdition of 2 mM Ca2+o ( right ) . Ca2+ promotes Abi1 localization to the edge of the lamellipod . Images are taken from Video 12 . Scale bar , 5 µm; color scale indicates fluorescence intensity ( 0–1 a . u . ) . ( H ) The average fluorescence of EGFP-Abi1 in a 1-µm band around the perimeter of the synapse ( Fperimeter ) relative to the average fluorescence across the entire synapse ( Fcell ) versus time ( n = 8 cells ) . ( I ) The fluorescence intensity ( a . u . ) of EGFP-Abi1 along the line indicated ( top right , pink ) in 0 . 5 mM Ca2+o ( blue ) and 1 . 5 min after Ca2+o removal ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 026 One potential mechanism for directing actin polymerization to the lamellipod edge is through selective localization of F-actin nucleation complexes . The nucleation promotion factor WAVE2 is essential for normal F-actin accumulation at the synapse , and a protein complex including WAVE2 and Abi1 has been detected at the synapse periphery ( Zipfel et al . , 2006; Nolz et al . , 2007 ) , but whether Ca2+ influences the localization of this complex is not known . Using immunocytochemistry we found that Ca2+o enriched the level of endogenous WAVE2 at the synapse periphery ( Figure 7E , F ) . To examine the timing of the WAVE2 complex response to Ca2+ , we visualized EGFP-Abi1 in live cells . Abi1 was highly enriched at the synapse periphery in the presence of Ca2+o as expected ( Video 12 , Figure 7G , left , right panels ) , and Ca2+o removal caused Abi1 to become diffusely distributed throughout the synapse in a reversible manner ( Figure 7G , middle , Figure 7H , I ) . The finding that Ca2+ localizes the WAVE2 complex to the lamellipod edge suggests that Ca2+ restricts actin polymerization to the synapse periphery at least in part through controlling the location of nucleation . 10 . 7554/eLife . 14850 . 027Video 12 . Calcium promotes localization of Abi1 to the distal edge of the synapse . Time-lapse TIRF movie of a Jurkat cell expressing F-tractin-P-tdTom and EGFP-Abi1 stimulated on anti-CD3 coated coverslip in 0 . 5 mM Ca2+o , followed by perfusion with 0 Ca2+o and 2 mM Ca2+o . Images acquired every 15 s and time compressed 45x; scale bar , 5 µm . This video supplements Figure 7G , H and I . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 027 Extensive reorganization of the actin cytoskeleton underlies the formation of the immune synapse , and retrograde actin flow from the lamellipod towards the ADZ is critical for maintaining TCR signaling and [Ca2+]i elevation ( Valitutti et al . , 1995; Varma et al . , 2006; Rivas et al . , 2004; Babich et al . , 2012; Yi et al . , 2012 ) . While the critical role of [Ca2+]i elevation in regulating gene expression during T cell activation is well established ( Feske et al . , 2001 ) , the current study reveals several essential new functions for Ca2+ in determining synapse form and function . Ca2+ influx through CRAC channels organizes actin into distinct lamella and lamellipod zones , stimulates retrograde actin flow and concentrates active CRAC channels in the center of the synapse , in part through its action to return extending ER tubules to the ADZ . Our findings extend upon previous work implicating Ca2+ in actin remodeling at the synapse . In a pioneering study , Bunnell et al demonstrated that intracellular Ca2+ is required for actin accumulation and cell spreading during the early phase of synapse formation; however , the failure of the cells to spread in the absence of Ca2+o precluded study of calcium’s effects on the mature synapse ( Bunnell et al . , 2001 ) . Likewise , Ca2+-sensitive proteins including L-plastin ( Wabnitz et al . , 2010 ) , gelsolin ( Morley et al . , 2007 ) , calpain ( Watanabe et al . , 2013 ) and myosin IIA ( Ilani et al . , 2009; Yi et al . , 2012 ) have been implicated in actin remodeling at the synapse , but in these studies protein expression levels or activity were perturbed prior to synapse initiation , and thus effects on cell adhesion and spreading could not be distinguished from possible effects on actin dynamics in the mature synapse . We were able to address the role of Ca2+ in the mature synapse by acutely blocking Ca2+ influx after the synapse was fully formed as indicated by its stable contact area , well-defined lamellipod and lamella actin zones , and retrograde actin flow . In both Jurkat cells and primary T lymphoblasts , Ca2+o removal triggered similar changes in actin organization and retrograde flow . Ca2+o removal slowed retrograde flow in the lamellipod to nearly the same degree in both cells ( 41% decrease in Jurkat versus 44% in primary T cells; Tables 1 and 2 ) . Ca2+ removal diminished the lamellipod in primary T cells , reducing its width by 61% , while in Jurkat cells the effect was somewhat more pronounced with the lamellipod becoming indistinguishable from the rest of the actin network . Overall , these results demonstrate that Ca2+ effects on actin organization and dynamics are not specific to Jurkat cells but apply also to primary T cells . Furthermore , T lymphoblasts responded similarly to Ca2+o removal when stimulated on surfaces including ICAM-1 in order to more closely resemble an APC , indicating that our findings extend to more physiological surface interactions . New optical technologies such as light sheet microscopy may enable further studies of Ca2+ effects on actin dynamics in the most physiological setting , at the synapse between a primary T cell and an APC ( Ritter et al . , 2015 ) . We found that Ca2+ acts at multiple levels to organize actin into a lamellipod and lamella with sustained retrograde flow . First , Ca2+ directs actin polymerization largely to the distal edge of the lamellipod while suppressing it elsewhere ( Figure 7 ) . By effectively restricting polymerization to the edge , Ca2+ suppresses filament growth at random angles throughout the synapse and promotes retrograde vectorial movement of actin filaments ( Figure 8 ) . Second , Ca2+ accelerates actin depolymerization ( Figure 6 ) , which is expected to enhance the rate of actin flow further by increasing the level of monomeric actin and therefore the rate of actin addition to free barbed ends at the lamellipod edge ( Figure 8 ) . Overall , these findings demonstrate that Ca2+ adds a second level of regulation to actin remodeling downstream of TCR triggering . The net effect of Ca2+ influx is to suppress the density of F-actin at the synapse ( Figure 5 ) , which was somewhat surprising given that elevating intracellular Ca2+ has been reported to increase the level of F-actin in unstimulated T cells ( Dushek et al . , 2008 ) . This discrepancy may have resulted from the different imaging methods that were used; flow cytometry may indicate an increased level of F-actin globally ( Dushek et al . , 2008 ) , while TIRF or confocal imaging at the cell footprint may instead detect a local decrease in F-actin at the synapse ( this study ) . Alternatively , TCR triggering may initiate Ca2+-sensitive actin regulatory pathways that are quiescent in resting cells , and thus Ca2+ engages a different set of actin remodeling proteins after TCR stimulation . 10 . 7554/eLife . 14850 . 028Figure 8 . Effects of calcium on actin dynamics and retrograde flow at the synapse . ( A ) Retrograde actin flow at the immune synapse ( yellow arrows ) continually removes extended ER tubules ( purple ) from the periphery , thereby concentrating the ER in the ADZ . An expanded view of the red-boxed region ( top ) depicts Ca2+ effects on actin regulation ( bottom ) . Ca2+ drives centripetal actin flow in two ways: ( 1 ) by restricting polymerization to the lamellipod edge ( green chevrons ) , it enforces vectorial movement of the actin network; and ( 2 ) by increasing the rate of depolymerization , it increases the pool of free actin monomers ( grey chevrons ) , thus enhancing polymerization on free barbed ends at the lamellipod edge ( green chevrons ) . Ca2+ restricts polymerization to the lamellipod edge by localizing WAVE2 and Abi1 to this site where they promote ARP2/3-mediated actin nucleation ( blue triangles ) and possibly by capping free barbed ends elsewhere ( pink circles ) . ( B ) Experimentally terminating Ca2+ influx reduces retrograde actin flow such that extended ER tubules are no longer effectively pushed into the ADZ . In the absence of Ca2+o , actin depolymerization is reduced ( bottom ) , nucleation occurs more uniformly throughout the lamellipod/lamella and capping of free barbed ends may be reduced . The overall result is a slowed , non-directional polymerization throughout the lamellipod and lamella resulting in reduced retrograde flow . DOI: http://dx . doi . org/10 . 7554/eLife . 14850 . 028 Calcium acts in two ways to spatially restrict actin polymerization at the mature synapse: by promoting polymerization around the cell perimeter and by suppressing polymerization throughout the rest of the contact area . The extensive polymerization at the lamellipod edge is closely paralleled by the recruitment of the WAVE2/Abi1 complex to the periphery ( Figure 7 ) , where it presumably activates Arp2/3 to initiate actin polymerization and branching ( Takenawa and Suetsugu , 2007 ) . A central question arising from these findings is how Ca2+ directs WAVE2/Abi1 to the periphery . One possible mechanism is suggested by the ability of Ca2+ to stimulate PI3K localization to the lamellipod of migrating cells , where generation of phosphatidylinositol ( 3 , 4 , 5 ) -trisphosphate ( PIP3 ) may recruit the WAVE2 complex to the membrane ( Oikawa et al . , 2004 ) . Ca2+ may exert additional effects through its ability to promote GTP loading of Rac , a GTPase essential for WAVE2-mediated actin nucleation ( Fleming et al . , 1999 ) . The mechanism by which Ca2+ suppresses polymerization elsewhere in the lamella and lamellipod is also unknown , although Ca2+-sensitive capping proteins expressed in T cells such as gelsolin ( Yin , 1987 ) or CapG ( Yu et al . , 1990 ) are attractive candidates . Our results indicate that Ca2+-dependent acceleration of depolymerization is unlikely to involve myosin ( Figure 6—figure supplement 2A , B ) , but the actin-severing proteins cofilin ( Maus et al . , 2013; Meberg et al . , 1998; Wang et al . , 2005 ) and gelsolin ( Yin , 1987 ) remain viable candidates as both are expressed in T cells and respond to physiological levels of [Ca2+]i ( Lin et al . , 2000 ) . Because the ER-PM junction forms the physical site for STIM1-Orai1 assembly into active CRAC channels , the location and dynamics of the ER are critical factors that determine when and where Ca2+ influx sites arise when T cells contact their targets . Our results provide the first view of ER dynamics at the synapse and how the actin cytoskeleton restricts both the ER and CRAC channel distribution to the cSMAC/ADZ . As the synapse forms , STIM1 and Orai1 appear in the cSMAC/ADZ by two mechanisms . The first is related to the movement of the centrosome and associated MTOC to the synapse , as EM tomography has shown enrichment of the ER around the centrosome at synaptic contact sites ( Ueda et al . , 2011 ) . This mechanism appears to account for the bulk of ER localization and STIM1-Orai1 complexes as cells settled onto coverslips . Once a stable synapse formed , microtubule extension carried ER tubules toward the periphery , and these were repeatedly returned to the ADZ by an advancing front of actin . A similar action of actin to oppose the extension of microtubules and associated ER tubules has also been described at the leading edge of migrating epithelial cells ( Terasaki and Reese , 1994; Waterman-Storer and Salmon , 1997 ) . Our observations that STIM1/Orai1 puncta and actin move towards the ADZ at similar speeds suggest that nascent ER tubules may form ER-PM junctions in peripheral regions , which then enable the assembly of active STIM1-Orai1 complexes that traverse the lamella before they are collected in the ADZ . Interestingly , intracellular Ca2+ binding to membrane phospholipids is thought to enhance TCR signaling by exposing membrane-associated CD3 ITAM motifs ( Shi et al . , 2013 ) . Thus , an intriguing possibility is that mobile CRAC channel complexes create local sites of high [Ca2+]i needed to fully activate mobile TCR microclusters . This study illustrates two new levels of signal regulation at the immune synapse . Because actin dynamics are required to sustain TCR activity at the synapse ( Kaizuka et al . , 2007; Valitutti et al . , 1995; Varma et al . , 2006; Rivas et al . , 2004; Babich et al . , 2012; Yi et al . , 2012; Kumari et al . , 2015 ) , the action of Ca2+ influx to promote actin turnover and flow creates a positive feedback loop that would be expected to maintain or enhance the activation of CRAC channels . This loop creates the potential for nonlinear effects , such that graded increases or decreases in [Ca2+]i may act through actin to modulate TCR activity and bias the cell towards all-or-none , threshold-like behavior in response to antigen . At the same time , Ca2+ influx effectively limits the lifetime of active TCR microclusters by increasing the rate at which they are transported to the synapse center , where signaling is terminated ( Yu et al . , 2010; Varma et al . , 2006; Yokosuka et al . , 2005 ) . In addition , the accumulation of STIM1 and Orai1 in the ADZ reveals a new type of CRAC channel self-organization . At the level of single ER-PM junctions , STIM1 and Orai1 complexes self-organize through a diffusion trap mechanism based on STIM1 binding to the PM and Orai1 ( Wu et al . , 2014 ) . At the synapse CRAC channels self-organize in a second way , by promoting the retrograde flow of actin that concentrates ER-PM junctions and CRAC channels in the ADZ . Given evidence that Ca2+ locally regulates exocytosis in T cells and mast cells ( Pores-Fernando and Zweifach , 2009; Holowka et al . , 2012 ) CRAC channel self-organization may ensure that Ca2+ is optimally positioned to serve critical Ca2+-dependent functions including the directional secretion of cytokines like interleukin 2 and interferon-γ ( Huse et al . , 2006 ) that drive subsequent phases of the immune response . Cells were cultured at 37°C in a humidified incubator with 5% CO2 . Jurkat E6 . 1 cells ( ATCC ) were maintained in RPMI 1640 supplemented with 1% L-alanyl-glutamine and 10% fetal bovine serum ( all from Gemini Bioproducts , West Sacramento , CA ) . Primary human peripheral blood CD4+ T cells were obtained without donor identifiers from the University of Pennsylvania’s Human Immunology Core under an Institutional Review Board approved protocol . Lymphoblasts were generated by stimulating primary T cells for 24 hr with human T-Activator CD3/CD28 magnetic beads ( Dynabeads , Life Technologies ) and cultured in RPMI 1640 supplemented with 1% GlutaMAX , 1% penicillin-streptomycin ( all from Invitrogen , Carlsbad , CA ) and 10% fetal bovine serum ( Atlanta Biologicals , Norcross , GA ) prior to lentiviral transduction . Sulfinpyrazone , ( - ) -blebbistatin , and 2-APB were from Sigma-Aldrich ( St . Louis , MO ) , and fura-2/AM was from Invitrogen . IL-2 was obtained through the AIDS Research and Reference Reagent Program , Division of AIDS , National Institute of Allergy and Infectious Diseases , National Institutes of Health; human rIL-2 was from M . Gately , Hoffmann-LaRoche , Nutley , NJ . Cloning of mCh-STIM1 was as described ( Luik et al . , 2006 ) . Orai1-EGFP was a gift from T . Xu ( Xu et al . , 2006 ) , F-tractin-P-tdTomato was a gift from J . A . Hammer III ( Yi et al . , 2012 ) , PAGFP-actin was a gift from C . G . Galbraith ( Galbraith et al . , 2007 ) , GFP-actin was from Clontech ( Mountain View , CA ) , and ER-GFP ( GFP-17 ) was a gift from N . Borgese ( Bulbarelli et al . , 2002 ) . ER-mCh was made using site-directed mutagenesis to introduce a Not1 restriction site after GFP in GFP-17 ( primers: 5’GAT GAA CTA TAC AAA GCG GCC GCT GAG CAG AAG CTG ATC T 3’ and reverse complement ) , then cloning mCherry into Kpn1/Not1 sites of the resulting plasmid . EB1-EGFP was a gift from L . Cassimeris ( Addgene plasmid #17234; Piehl and Cassimeris , 2003 ) and EGFP-MyH9 was a gift from R . S . Adelstein ( Addgene plasmid #11347; Wei and Adelstein , 2000 ) . EGFP-Abi1 was cloned by digesting Abi1 from p-EYFP-Abi1 ( gift from A . M . Pendergast; Courtney et al . , 2000 ) and cloning into BglII site of p-EGFP-C1 ( Clontech , Mountain View , CA ) . cDNA encoding Lifeact-GFP was a gift from R . Wedlich-Soldner ( Riedl et al . , 2008 ) and was subcloned into pDONR221 and subsequently into the lentiviral expression vector pLX301 using Gateway Technology . Jurkat cells at a density of 4–6 x 107/ml in Ingenio electroporation solution ( Mirus Bio LLC , Madison , WI ) were electroporated in 0 . 4 cm cuvettes with 6–20 μg of plasmid DNA 40–48 hr prior to imaging . Primary human T lymphoblasts were transduced with Lifeact-GFP lentivirus 24 hr after stimulation . Lentivirus and 8 μg/ml polybrene ( Sigma-Aldrich ) were mixed with 5–10×106 T cells in 5 ml round bottom polystyrene tubes and centrifuged at 1 , 200 g for 2 hr at 37°C . Lentivirus-containing medium was then replaced with primary human T cell culture medium , and the cells were returned to the incubator . Two days after transduction , the medium was supplemented with 2 µg/ml puromycin , and cells were cultured for an additional four days before magnetic removal of the activator beads . Cells were cultured for an additional 1–2 days in medium with 2 µg/ml puromycin and 10 U/ml IL-2 before use ( day 8–9 after activation ) . Stimulatory coverslips were washed with 100% ethanol , then coated overnight at 4°C with 10 µg/ml monoclonal anti-CD3 ( OKT3 from eBiosciences , San Diego , CA for Jurkat cells and from BioXCell , Lebanon , NH for primary T lymphoblasts ) in PBS and washed thoroughly with PBS . Where indicated , coverslips were subsequently coated with 2 μg/ml human ICAM-1 Fc chimera ( R&D Systems , Minneapolis , MN ) for 2 hr at 20–22°C then washed thoroughly with PBS . Unless otherwise noted , cells were stimulated on the microscope at 37°C in Ringer’s solution containing ( in mM ) : 155 NaCl , 4 . 5 KCl , 2 CaCl2 , 1 MgCl2 , 10 D-glucose and 5 Na-HEPES ( pH 7 . 4 ) . In solutions with >2 mM CaCl2 , [NaCl] was reduced to maintain normal osmolarity , and in solutions with <2 mM Ca2+ , MgCl2 was substituted for CaCl2 . In Ca2+-free Ringer’s solution , 1 mM EGTA and 2 mM MgCl2 were substituted for CaCl2 . In Ca2+ imaging experiments , all solutions contained 250 μM sulfinpyrazone to inhibit fura-2 extrusion . Cell imaging commenced within 3–7 min after loading cells onto coverslips , and all images were collected from cells having a constant , maximal diameter , a ruffling edge and in cells expressing fluorescently labeled actin or F-tractin-P , a clearly defined actin ring with retrograde flow in Ca2+-containing Ringer’s solution . [Ca2+]i sometimes failed to decline to a minimum baseline level following Ca2+o removal ( described in Figure 5 ) , and the persistent [Ca2+]i elevation was associated with actin treadmilling and a ruffling lamellipod . Therefore , in the experiments of Figures 4A , 6 and 7 we limited our analysis to cells that lost the ruffling lamellipod and retrograde flow upon Ca2+o removal . In Figure 1 and Figure 2—figure supplement 1 , TIRF images were acquired at 32–37°C on a custom-built through-the-objective TIRF microscope using an Axiovert S100TV base and a Fluar 100X , 1 . 45 NA oil-immersion objective ( Carl Zeiss , Oberkochen , Germany ) . For simultaneous acquisition of GFP and mCherry/tdTomato , a Di01-R488/561 dual-band dichroic mirror ( Semrock , Rochester , NY ) directed excitation light from Sapphire 488-nm and Compass 561-nm lasers ( Coherent , Santa Clara , CA ) to the cells . Two bands of fluorescence emission were collected onto an Andor iXon DU897E EMCCD camera using an Optosplit-II ( Cairn Research , Kent , UK ) image splitter containing a dichroic mirror ( FF580-FDi01 , Semrock ) and emission filters for GFP ( FF02-525/50 , Semrock ) and mCherry/tdTomato ( E600LP , Chroma , Bellows Falls , VT ) . Laser shutters and image acquisition were controlled by Micro-Manager ( Edelstein et al . , 2010 ) . All other TIRF and photoactivation experiments ( Figure 2–7 and associated supplements ) were performed in the Stanford Cell Sciences Imaging Facility on a Nikon Eclipse-TI inverted microscope platform with a PLAN APO-TIRF 100X 1 . 49 N . A . oil-immersion objective , an environmental chamber for acquisition at 37°C , and a Perfect Focus System ( Nikon , Tokyo , Japan ) . Images were collected with a Neo sCMOS camera ( Andor , Belfast , UK ) , with 2x2 binning for photoactivation experiments only . A Lambda XL lamp and Lambda 10–3 filter wheel ( Sutter , Novato , CA ) were used for widefield epifluorescence illumination , and 488- and 561-nm lasers were used for through-the-objective TIRF imaging of GFP and tdTomato , respectively . GFP was imaged using a TRF49904-ET-488-nm laser bandpass filter set , while F-tractin-P-tdTom was imaged using a TRF49909-ET-561-nm laser bandpass filter set ( Chroma ) . For all photoactivation experiments , a constant exposure time and illumination intensity was used , and photobleaching was less than 10% over the duration of each experiment . For photoactivation of PAGFP , a Mosaic digital illumination system ( Andor ) was used to steer a 405-nm laser to a user-defined region on the coverslip and photoactivate for 100 ms . All equipment was controlled using NIS-Elements software ( Nikon ) . Transfected cells were loaded with 2 . 5 µM fura-2/AM at 22–25oC for 30 min in RPMI 1640 without phenol red or sodium bicarbonate . After washing , cells remained for 30 min in RPMI before they were resuspended in Ringer’s solution immediately prior to loading onto coverslips . Cells were stimulated on anti-CD3 in 0 . 5 mM Ca2+o , followed by either Ca2+o removal or application of 100 µM 2-APB in 0 . 5 mM Ca2+o , then perfusion with 2 , 5 , or 10 mM Ca2+ . F-tractin-P-tdTom and fura-2 were imaged on a Zeiss Observer Z1 inverted microscope at 37°C using an αPlan-Apochromat 100X , 1 . 46 N . A . oil immersion DIC objective ( Carl Zeiss ) . Fura-2 imaging was performed using a Lambda XL lamp ( Sutter ) , 380/15 and 357/10 excitation filters , 400-nm dichroic and 480-nm long pass emission filter ( Omega Optical , Brattleboro , VT ) . Data are displayed as the ratio of emissions in response to excitation at 357 and 380 nm ( 357/380 ratio ) . F-tractin-P-tdTom images were acquired using through-the-objective TIRF with 561-nm laser excitation and a Zeiss 74HE filter set and an ImagEM-1K EMCCD camera ( Hamamatsu , Hamamatsu City , Japan ) , and all equipment was controlled using Zeiss Axiovision software . Confocal imaging of Jurkat cells was performed in the Stanford Cell Sciences Imaging Facility on a Nikon Eclipse-TI inverted microscope platform with a CFI Plan Apochromat λ 60X , 1 . 4 N . A . oil-immersion objective , a CSU-X1 spinning disk ( Yokogawa , Tokyo , Japan ) , an environmental chamber for acquisition at 37°C , and a Perfect Focus System ( Nikon ) . Cells were illuminated with a 561-nm laser ( Spectral Applied Research , Ontario , Canada ) and images were projected via a 405/488/568/647 dichroic mirror and a 600/37 emission filter ( Semrock ) to an iXon Ultra 897 EMCCD camera ( Andor ) . All equipment was controlled using NIS-Elements software . Primary human CD4+ lymphocytes expressing Lifeact-GFP were imaged on an Axiovert 200M microscope ( Carl Zeiss ) equipped with a spinning disk confocal system ( Yokogawa ) , a 63x Plan Apo , 1 . 4 N . A . oil immersion objective and an environmental chamber for acquisition at 37°C . Cells were illuminated with a 488 nm laser ( Melles Griot , Carlsbad , CA ) and images were projected via a 405/488/561/640 dichroic mirror and a 527/55 emission filter ( Chroma ) to an Orca ER CCD camera ( Hamamatsu ) . Sets of 3 image planes collected at 0 . 25 µm increments were collected every 2 s and displayed as maximum intensity projections . All equipment was controlled using Volocity v . 6 . 3 imaging software ( Perkin Elmer , Waltham , MA ) . Jurkat T cells were allowed to settle onto stimulatory coverslips in 0 . 5 mM Ca2+ Ringer’s solution at 37°C for 4 min , then transferred to either 0 or 0 . 5 mM Ca2+ for 2 . 5 min before fixation with 4% paraformaldehyde in 10 mM PBS for 20 min . Cells were washed with 10 mM PBS containing 50 mM glycine , then permeabilized for 5 min with 0 . 1% Triton X-100 and blocked for 1 hr in 10 mM PBS , 50 mM glycine , and 10% fetal bovine serum . Cells were incubated at 20-22oC with anti-WAVE2 ( H-110 , Santa Cruz Biotechnology , Dallas , TX ) diluted to 4 μg/ml in blocking buffer for 1 hr followed by 1-hr incubation with 2 μg/ml Alexa Fluor-488 goat anti-rabbit secondary antibody and 0 . 2 units/ml Alexa Fluor-594 phalloidin ( Thermo Fisher , Grand Island , NY ) . Cells were washed extensively with blocking buffer between incubations and imaged immediately by TIRF . All images were background-corrected and analyzed using ImageJ ( Schneider et al . , 2012 ) . In kymographs , each pixel represents the average intensity across the 5-pixel width of the scan line . Spatiotemporal image correlation spectroscopy ( STICS ) analysis was used to determine the direction and velocity of actin movement by applying an ImageJ plugin ( STICS map jru v2 , developed by Jay Unruh at Stowers Institute for Medical Research in Kansas City , MO ) based on methods developed by Hebert et al . ( Hebert et al . , 2005 ) . Sample sizes were determined based on previous in situ measurements of actin depolymerization rates ( Theriot and Mitchison , 1991 ) and live-cell imaging studies of the immune synapse ( Babich et al . , 2012 ) . To measure the [Ca2+]i dependence of F-tractin-P fluorescence , F-tractin-P intensity was normalized to an average of the final five images collected in 0 Ca2+ , at which time the fura-2 ratio had reached a minimum value of 0 . 39 ± 0 . 04 ( mean ± SEM , n = 26 cells ) . Data points represent the mean F-tractin-P fluorescence and fura-2 ratio when both signals were at steady-state and the sample size was selected to ensure that the cell population represented a broad range of [Ca2+]i . PAGFP fluorescence decay was measured within an ROI encompassing the photoactivated region that was moved centripetally to remain centered on the fluorescent bar of actin . The time course was fitted by a single exponential function using IgorPro ( Wavemetrics , Portland , OR ) . At each time point , the mean intensity across the width of the photoactivated bar was measured along a line perpendicular to the direction of movement , and displacement of the peak fluorescence value versus time was used to calculate velocity . To create fluorescence profile plots of PAGFP-actin incorporation , the fluorescence intensity as a function of radial position was calculated by averaging across 19 pixel-wide scan lines , subtracting the average of 10 scans acquired before photoactivation , and plotting the result relative to this pre-photoactivation signal .	An effective immune response requires the immune system to rapidly recognize and respond to foreign invaders . Immune cells known as T cells recognize infection through a protein on their surface known as the T cell receptor . The T cell receptor binds to foreign proteins displayed on the surface of other cells . This interaction initiates a chain of events , including the opening of calcium channels embedded in the T cell membrane known as CRAC channels , which allows calcium ions to flow into the cell . These events ultimately lead to the activation of the T cell , enabling it to mount an immune response against the foreign invader . As part of the activation process , the T cell spreads over the surface of the cell that is displaying foreign proteins to form an extensive interface known as an immune synapse . The movement of the T cell's internal skeleton ( the cytoskeleton ) is crucial for the formation and function of the synapse . Actin filaments , a key component of the cytoskeleton , flow from the edge of the synapse toward the center; these rearrangements of the actin cytoskeleton help to transport clusters of T cell receptors to the center of the synapse and enable the T cell receptors to transmit signals that lead to the T cell being activated . It is not entirely clear how the binding of T cell receptors to foreign proteins drives the actin rearrangements , but there is indirect evidence suggesting that calcium ions may be involved . Hartzell et al . have now investigated the interactions between calcium and the actin cytoskeleton at the immune synapse in human T cells . T cells were placed on glass so that they formed immune synapse-like connections with the surface , and actin movements at the synapse were visualized using a specialized type of fluorescence microscopy . When calcium ions were prevented from entering the T cell , the movement of actin stopped almost entirely . Thus , the flow of calcium ions into the T cell through CRAC channels is essential for driving the actin movements that underlie immune synapse development and T cell activation . In further experiments , Hartzell et al . tracked the movements of CRAC channels and actin at the synapse and found that actin filaments create a constricting “corral” that concentrates CRAC channels in the center of the synapse . Thus , by driving cytoskeleton movement , calcium ions also help to organize calcium channels at the immune synapse . Future work will focus on identifying the actin remodeling proteins that enable calcium ions to control this process .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "immunology", "and", "inflammation" ]	2016	Calcium influx through CRAC channels controls actin organization and dynamics at the immune synapse
Previous reports have shown that individual neurons of the brain can display somatic genomic mosaicism of unknown function . In this study , we report altered genomic mosaicism in single , sporadic Alzheimer's disease ( AD ) neurons characterized by increases in DNA content and amyloid precursor protein ( APP ) gene copy number . AD cortical nuclei displayed large variability with average DNA content increases of ∼8% over non-diseased controls that were unrelated to trisomy 21 . Two independent single-cell copy number analyses identified amplifications at the APP locus . The use of single-cell qPCR identified up to 12 copies of APP in sampled neurons . Peptide nucleic acid ( PNA ) probes targeting APP , combined with super-resolution microscopy detected primarily single fluorescent signals of variable intensity that paralleled single-cell qPCR analyses . These data identify somatic genomic changes in single neurons , affecting known and unknown loci , which are increased in sporadic AD , and further indicate functionality for genomic mosaicism in the CNS . The genome has been classically viewed as being constant from cell to cell in the same individual , with genomic differences passed on through the germline . However , within neurons of the brain , numerous studies have reported somatic variability producing complex genomic mosaicism but having unknown function . Identified forms of somatically arising genomic mosaicism include aneuploidy ( reviewed in Bushman and Chun , 2013 ) , LINE elements ( Muotri et al . , 2005; Baillie et al . , 2011; Evrony et al . , 2012 ) , copy number variations ( CNVs ) ( Gole et al . , 2013; McConnell et al . , 2013; Cai et al . , 2014 ) , and DNA content variation ( DCV ) ( Westra et al . , 2010; Fischer et al . , 2012 ) . AD is the most common form of dementia and is characterized by the presence of amyloid plaques , synaptic loss , and cell death ( Alzheimer's Association , 2013 ) , notably affecting the prefrontal cortex . The major component of these plaques is β-amyloid ( Aβ ) , a protein encoded by APP ( Goldgaber et al . , 1987; St George-Hyslop et al . , 1987; Tanzi et al . , 1987 ) . Familial AD accounts for less than 5% of all cases and has been genetically linked to mutations in APP and two presenilins ( PSEN ) , PSEN1 and PSEN2 , the catalytic components of γ-secretase and the units responsible for cleavage of APP ( Price and Sisodia , 1998; Bertram et al . , 2010 ) . In addition , APP gene dosage is strongly associated with AD pathogenesis based on multiple lines of evidence . First , Down syndrome ( DS ) , with three copies of APP , produces neuropathology virtually identical to AD ( Glenner and Wong , 1984; Delabar et al . , 1987 ) and APP locus duplications are sufficient to cause familial AD ( Rovelet-Lecrux et al . , 2006; Sleegers et al . , 2006; McNaughton et al . , 2012 ) . Moreover , AD-protective effects have been reported in DS with APP deletion via partial trisomy ( Prasher et al . , 1998 ) , as well as in familial AD with an APP partial loss-of-function mutation ( Jonsson et al . , 2012 ) . However , seminal studies in the 1980s failed to detect evidence of APP amplification in sporadic Alzheimer's disease peripheral blood and whole brain ( Podlisny et al . , 1987; St George-Hyslop et al . , 1987; Tanzi et al . , 1987 ) despite strong linkage in familial AD , thus linkage between sporadic AD and APP remains unclear . The existence of region specific genomic mosaicism in the normal brain ( Westra et al . , 2010 ) raised the possibility that DCV , defined as variations in the total DNA amount present in a single cell or population , might play a functional role in sporadic brain diseases by altering pathogenic loci in individual cells . The validated pathogenicity of APP in familial AD suggested that mosaic alterations in APP copy number within single neurons may play a role in producing sporadic AD . Through the use of five independent experimental approaches , we report increased somatic genomic variation within individual sporadic AD neurons involving mosaic increases in both DNA content and APP copy number . Previous work identified DNA content changes with regional variability in the non-diseased brain , where prefrontal cortical nuclei—particularly neuronal nuclei—displayed increased DCV compared to nuclei of the cerebellum and non-brain controls ( Westra et al . , 2010 ) . The current study utilizes the same techniques for DNA content analysis by flow cytometry using propidium iodide ( PI ) . PI staining is the predominant methodology for quantitatively differentiating nuclei or cells with variable DNA content and is routinely utilized as a gold standard in multiple fields including genomic comparisons across species in botany ( Dolezel et al . , 2007 ) , studies of the cell cycle ( Krishan , 1975 ) , and DNA degradation produced by apoptosis ( Riccardi and Nicoletti , 2006 ) . Prior analyses ruled out the effects of DNA dyes , nuclear size , mitochondrial contamination , and autofluorescent lipofuscin on DNA content and validated genomic increases in DNA content using quantitation of CENP-B PNA probes against human centromere repeats ( Westra et al . , 2010 ) which have been further substantiated by identification of copy number gains in single human neurons ( Gole et al . , 2013; McConnell et al . , 2013 ) . To further validate DCV in human neurons , whole genome amplification ( WGA ) was used on cortical neuronal nuclei sorted into populations of high or low DNA content based upon PI intensity . Nuclei with high or low DNA content were sorted into 12 replicates of 1000 , 500 , or 100 and were then subject to DNA content assessment by WGA to assess starting amounts of DNA template in each sample ( Figure 2A ) . Nuclei were denatured and amplified by multiple displacement amplification ( MDA ) during which DNA synthesis was continually measured by SYBR Green fluorescence ( Figure 2B ) . In every case , nuclei with high PI intensity also showed increased DNA synthesis over those with low PI intensity . These results independently confirm , as expected , that PI staining intensity faithfully reports DNA content . 10 . 7554/eLife . 05116 . 004Figure 2 . AD cortical nuclei show increased DNA content variation ( DCV ) by flow cytometry . ( A ) Histogram displaying gating parameters used in sorting ‘high’ and ‘low’ DNA content populations for validation of DNA content . ( B ) Validation of DNA content analyses using semi-quantitative MDA whole-genome amplification ( WGA ) on ‘high’ and ‘low’ DNA content populations of 1000 , 500 , and 100 nuclei . ( C and D ) Representative DNA content histograms for lymphocytes ( LYM ) , AD cerebellum ( CBL ) , and AD prefrontal cortex ( CTX ) . Each colored histogram represents a separate sample in each set; CTX and CBL samples are from paired brains . Chicken erythrocyte nuclei ( CEN ) were used as internal calibration controls . ( E ) Representative orthogonal view of DNA content vs forward scatter width ( FSC-W ) . For each brain sample , the area to the right of the vertical line indicates a DNA content increase of the population of nuclei . AD-6 CTX is a representative right-hand peak shift and AD-7 is a representative right-hand shoulder ( see Figure 3A for more examples ) . ( F ) DNA content changes for all human LYM , ND , and AD brain samples examined ( AD CTX N = 32 , AD CBL N = 16 , LYM N = 15 [20 meta analysis] , ND CTX = 21 [36 meta analysis] , ND CBL = 11 [12 meta analysis] ) . Red bars denote average for each group relative to lymphocytes . Averages are as follows ( including metadata from Westra et al . ( 2010 ) ) : AD CTX 8 . 219%; AD CBL −0 . 1104%; LYM −0 . 2915%; ND CTX 2 . 239%; ND CBL −3 . 358% . ( G ) DNA content changes of the current study ( AVOVA p < 0 . 0001 ) . ( H ) DNA content changes of the current study combined with metadata from Westra et al . ( 2010 ) ( ANOVA p < 0 . 0001 ) . ( I ) Comparison of mean coefficient of variation ( CV statistic from FlowJo of the population , included metadata from Westra et al . , 2010 ) demonstrates that there is an average increase in the variation of AD samples ( ANOVA p < 0 . 0001 ) . p = 0 . 05 , p = 0 . 01 , p = 0 . 001 , p < 0 . 0001 , See Figure 2—source data 1 for exact p values . See Figure 2—figure supplement 1 for age , PMI and Braak score correlations . See Figure 2—figure supplement 2 for control of nuclear size analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 00410 . 7554/eLife . 05116 . 005Figure 2—source data 1 . DNA Index ( DI ) and percent change values and statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 00510 . 7554/eLife . 05116 . 006Figure 2—figure supplement 1 . DNA content shows no correlation with age or post-mortem index ( PMI ) . ( A ) Comparison of mean skew values for each sample group , skew determined as: ( Mean − Mode/Standard Deviation of the diploid DNA content peak ) . ( B ) No correlation was observed between DNA content and Braak score . ( C–E ) No correlation was observed between DNA content and age across all brains analyzed . ( F–H ) No correlation was observed between DNA content and post-mortem index across all brains analyzed . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 00610 . 7554/eLife . 05116 . 007Figure 2—figure supplement 2 . Analysis of nuclear size and DNA content . ( A–C ) Representative flow cytometry scatter plots of nuclei . ( A ) Lymphocytes ( LYM ) , ( B ) CTX nuclei , ( C ) CBL nuclei . ( D ) Overlay of red boxes shown in ( A–C ) , demonstrating that cortical nuclei similar in size to LYM and CBL consistently display a DNA content shift . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 007 AD neuropathology strongly affects the prefrontal cortex . We therefore first interrogated the DNA content of pathologically confirmed prefrontal cortices ( N = 32 ) , using previously described methodologies for DCV analyses in neuronal nuclei ( Figure 1B ) ( Westra et al . , 2010 ) . In control experiments , AD cerebellar nuclei showed DNA content profiles similar to lymphocytes , characterized by histograms with sharp peaks and narrow bases ( Figure 2C , E ) . By comparison , AD cortices displayed high variability characterized by right hand shoulders ( Figure 2D , E [AD-7] ) , large right hand peak shifts ( Figure 2D , E [AD-6] ) , and wide bases . Of the AD cortex samples examined , greater than 90% displayed a net DNA content increase , averaging approximately 8% gain over human lymphocyte controls ( Figure 2F–H ) . Notably , the AD cortex displayed increases beyond those observed in non-diseased cortices , with an average gain of approximately 6% over age and sex-matched samples ( Figure 2F–H ) . AD cortices also displayed an increased coefficient of variation over non-diseased cortical nuclei and all cerebellar nuclei ( Figure 2I ) as well as a consistent skewed distribution compared to AD cerebellum ( Figure 2—figure supplement 1A ) . In addition , we examined 14 paired sets of AD cortex and cerebellum from the same individual ( Figure 3A ) and 12 paired sets from non-diseased individuals ( Figure 3B ) . Each AD cortex showed unique cortical histograms and increases in total cortical DNA compared to the cerebellum . In AD and non-diseased brain samples , DNA content changes did not correlate with age , Braak score , or postmortem index ( Figure 2—figure supplement 1B–H ) , and DNA content was independent of nuclear size ( Figure 2—figure supplement 2 ) ( Westra et al . , 2010 ) . Importantly , the increase observed was significantly less than a 4N tetraploid genome ( ∼12 , 800 Mb ) , yet significantly more than what would be expected from a hypersomy of even the largest chromosome ( chr1: ∼250 Mb or 3 . 9% ) . These results indicate that neuroanatomically selective increases in DNA content represent a distinct , reproducible , and prevalent characteristic of the AD brain . 10 . 7554/eLife . 05116 . 008Figure 3 . Pairwise DNA content analyses in AD cortical nuclei vs AD cerebellum . ( A ) Pairwise analysis of overlaid DNA content histograms ( CTX = solid red , CBL = black dashed lines ) in the same AD individual ( each graph represents a unique AD individual ) . ( B ) Pairwise analysis of overlaid DNA content histograms ( CTX = solid blue , CBL = black dashed lines ) in the same ND individual . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 008 In the non-diseased cortex , neurons were identified as the predominant cell type contributing to increased DCV ( Westra et al . , 2010 ) . To determine whether neurons were also responsible for increased DCV in sporadic AD , nuclei were immunolabeled for the neuronal nuclear antigen , NeuN , and flow cytometry was used to analyze DCV in NeuN-positive and NeuN-negative nuclei ( Figure 4A–C ) . NeuN-positive AD cortical neurons showed right hand DNA content shifts ( Figure 4D ) and increased DCV , displaying ∼9% gain over that observed in NeuN-negative nuclei ( Figure 4E ) . A comparison of NeuN-positive nuclei in paired samples from the same brain also showed significantly increased DNA content in cortical neurons over cerebellar neurons ( Figure 4F ) . These distinctions do not rule out AD-specific effects on DCV from non-neuronal cells , but do implicate neurons as the major cellular locus for increased DCV . 10 . 7554/eLife . 05116 . 009Figure 4 . DNA content increases observed in AD cortical nuclei are attributable to neurons . ( A–C ) The gating procedure used for NeuN-positive flow cytometry analysis . ( A ) DNA content peak for identified nuclei . ( B ) A sample of unlabeled neuronal nuclei that display no NeuN-positive signal . ( C ) Selection for NeuN-positive nuclei for downstream analysis . ( D ) DNA content histograms of four AD samples displaying NeuN-positive nuclei ( solid purple ) vs NeuN-negative nuclei ( black dashed line ) . NeuN-positive populations display distinct cortical histograms with prominent right-shifted peaks ( arrows ) . ( E ) Comparison of DNA index ( DI ) increases from NeuN-positive nuclei ( solid purple ) vs NeuN-negative nuclei ( black dashes ) from AD CTX samples . NeuN-positive nuclei ( DI = 1 . 10 ) showed an average gain of 9% over NeuN-negative nuclei ( DI = 1 . 01 ) , p = 0 . 0011 . ( F ) Comparison of DNA content in NeuN-positive nuclei from AD CTX ( DI = 1 . 10 ) ( red ) vs AD CBL ( DI = 0 . 94 ) ( pink ) from the same individual; CTX nuclei displayed a 15 . 6% gain over CBL nuclei , p = 0 . 0335 . Statistics are paired two-tailed t-test . Bars indicate ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 009 The DNA content increases in sporadic AD neurons raised the question of what genomic loci may be specifically increased in AD . The most highly validated gene associated with AD is APP , wherein constitutive gains have previously been linked to AD through familial AD and DS ( Prasher et al . , 1998; Rovelet-Lecrux et al . , 2006; Jonsson et al . , 2012; McNaughton et al . , 2012 ) . However , prior large-scale analyses in peripheral , non-neural tissues from sporadic AD cases have not reported APP copy number changes ( Bertram and Tanzi , 2009; Harold et al . , 2009 ) . The normal existence of genomic mosaicism might produce ‘hot-spots’ of cells with increased APP copy number , and we therefore interrogated small cohorts of brain nuclei from paired cortex and cerebellum . Three of six AD samples displayed significant increases in APP copy number ( 3 . 9–6 . 3 copies compared to reference genes ) over paired cerebellar samples ( Figure 5A ( AD-1 , AD-3 , AD-6 ) ) . Similar APP amplification was not detected in non-diseased brains ( Figure 5B ) , and DS controls along with reference gene assessments demonstrated expected results ( Figure 5—figure supplement 1A , B ) . Notably , AD samples displayed substantial variability in APP copies ( Figure 5C ) , revealing inherent limitations in precisely quantifying copy numbers in small , genomically mosaic cell populations and highlighting the necessity for single-cell analyses . 10 . 7554/eLife . 05116 . 010Figure 5 . Mosaic amplification of the APP locus in small cohorts of AD cortical neurons unrelated to trisomy 21 . ( A ) Comparison of relative copy number of APP in CBL and CTX fractions from six AD brains . APP locus-specific amplification was determined relative to reference gene SEMA4A; paired CBL nuclei were used as a calibrator sample for each brain , normalized to 2 . 00 for a diploid cell . Differences in ΔΔCt ± SEM of APP in the cortex vs cerebellum were assessed in each individual using an unpaired , two-tailed t-test ( **p = 0 . 0001 , p = 0 . 0165 , p = 0 . 0489 ) ( B ) Comparison of relative copy number of APP in CBL and CTX fractions from 4 non-diseased brains . ( C ) Average relative copy number in non-diseased vs AD brains . Control genes and DS individuals were also examined ( Figure 5—figure supplement 1 ) . ( D–J ) FISH strategy of chromosome 21 counting through simultaneous labeling using chr 21 q arm ‘whole’ chromosome paint ( WCP , green ) and chr 21 regional FISH probe for 21q22 . 13-q22 . 2 ( red ) ( see Figure 5—source data 1 for raw counts ) . ( D and E ) The ability to detect aneuploidy was validated using interphase nuclei from a human trisomy 21 brain , where three regional spots ( red , encompassing the APP gene ) were seen , despite WCP spatial variation ( see also Rehen et al . ( 2005 ) ) . ( F–I ) Chromosome 21 aneusomy was examined in prefrontal cortical nuclei . Examples of chr 21 ( F ) monosomy , ( G ) disomy , ( H ) trisomy , and ( I ) tetrasomy ( please note tetrasomy is not an example of aneuploidy ) . ( J ) Quantification of individual FISH signals showed no significant differences in monosomy , disomy , trisomy , or tetrasomy . 5 control brains and 9 AD brains were used . At least 450 nuclei were quantified per brain sample . Scale bar = 10 um . 4974 total nuclei examined . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01010 . 7554/eLife . 05116 . 011Figure 5—source data 1 . Raw dual point-paint probe FISH counts . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01110 . 7554/eLife . 05116 . 012Figure 5—figure supplement 1 . Controls for small population qPCR . ( A ) Reference genes validated in small population qPCR via examination of APP exon 14 in Down Syndrome ( DS ) nuclei as a positive control . ( B ) Representative males ( AD-1 and AD-6 ) displayed reduced copy number of PCDH11X , a gene located on the X chromosome , while a representative female demonstrates two copies of PCDH11X and CCL18 , a second single copy control gene . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 012 APP gene dosage in familial AD and DS has driven hypotheses connecting AD pathogenesis with increased incidence of trisomy 21 ( Heston and Mastri , 1977; Potter , 1991; Geller and Potter , 1999 ) . We therefore examined AD cortical nuclei using a highly liberal protocol for calling aneusomies whereby borderline FISH profiles suggestive of aneusomy were always included in quantitative assessments , in an effort to detect possible differences between AD and control brains . Importantly , all analyses were conducted blind to the identity of samples , an approach made possible by interrogating purified nuclei rather than cells or tissues that themselves show identifying increases of plaques and tangles in AD . Nuclei were double-labeled with a commercially available chromosome 21 q-arm ‘whole chromosome paint’ ( WCP ) and a regional ‘point’ probe for 21q22 . 13-q22 . 2 ( 220 Kb ) ( Figure 5D–I ) . The ability of this technique to detect aneuploidy was validated using interphase nuclei from a human trisomy 21 brain ( Svendsen et al . , 1998 ) revealing three nuclear signals ( Figure 5D , E ) ( see also ( Rehen et al . , 2005 ) ) . Three separate , blinded observers counted each sample . Despite using liberal counting criteria , no statistically significant changes in chromosome 21 aneuploidy rates , including trisomies , between AD ( N = 4974 nuclei , N = 9 brains ) and non-diseased brains ( N = 2576 nuclei , N = 5 brains ) were observed ( Figure 5J ) . Comparably high levels of aneuploidy have been reported for chromosome 21 displaying no difference between AD and non-diseased hippocampal cells ( Thomas and Fenech , 2008 ) . Thus , results from dual probe FISH analyses do not support increased trisomy 21 in AD but are consistent with alternative CNV mechanisms that could produce APP copy number values exceeding 3 , as was observed in 75-genome qPCR analyses ( Figure 5 ) . Mosaic single-cell increases in APP copy number could explain variations observed in small cohort qPCR data . To assess APP copy number in single neurons , an optimized microfluidic protocol to assess genomic copy number via qPCR using a Biomark HD 48 . 48 Dynamic Array integrated fluidic circuit ( IFC ) ( Fluidigm , South San Francisco ) was adapted from gene expression protocols . Nuclei were isolated and interrogated at two reference genes ( see Materials and methods ) and two APP exons that flanked the majority of the coding sequence at the 5′ and 3′ ends of APP ( exons 3 and 14 ) . Samples were run in triplicate and assays in sextuplicate , which generated high numbers of replicate data points allowing for improved copy number resolution to an extent not possible by conventional qPCR ( Weaver et al . , 2010; Whale et al . , 2012 ) . A total of 154 neuronal nuclei were individually examined from three AD and three non-diseased brains , within which nuclei from the cerebellum and prefrontal cortex were separately analyzed . In single AD cortical nuclei , significant increases in APP copy number , ranging up to 12 copies in a single nucleus ( Figure 6A , B ) , were observed with a high concordance rate between exons ( Figure 6—figure supplement 1 ) . AD cortical nuclei displayed increased average copy numbers ( ∼3 . 8–4 copies ) over control samples ( ∼1 . 7–2 . 2 copies ) ( Figure 6A , B ) , with increased frequencies of high copy number nuclei ( six or more copies ) ( Figures 6C and 5E , red lines ) primarily occurring in prefrontal cortex samples ( Figure 6C–F ) . The AD cortex showed an approximately fourfold increase over the non-diseased cortex of nuclei with greater than 2 APP copies ( 55–66% vs 12–15% ) ( Figure 6G , I ) . In addition , nuclei with fewer than 2 APP copies were also identified . In neuronal nuclei with greater than two copies ( Figure 6G , I , gold bars ) , AD cortical nuclei showed statistically significant increases ( ∼5 copies for both exons ) over the other samples ( ∼3 copies for both exons ) ( Figure 6H , J ) . These data identify mosaic , neuroanatomically enriched and disease-associated increases in APP copy number in single , sporadic AD neurons . 10 . 7554/eLife . 05116 . 013Figure 6 . Mosaic APP locus amplification in single neurons from AD brains . ( A ) Single nuclei relative copy numbers for exon 3 of APP from non-diseased ( ND ) CBL , ND CTX , AD CBL , and AD CTX; each black diamond represents one neuron . For each group , the mean is displayed in red and bars represent 95% confidence intervals . AD CTX showed a mean APP copy number of 3 . 80; this is significantly higher than AD CBL ( 2 . 23 ) , ND CTX ( 1 . 60 ) , and ND CBL ( 2 . 28 ) . p = 0 . 0147 , p = 0 . 0015 , p < 0 . 0012 , ANOVA p < 0 . 0001 ( see Figure 6—source data 1 for raw numbers and statistics ) . ( B ) Single nuclei relative copy numbers for exon 14 of APP , similar to ( A ) . The two exons showed a high concordance ( Figure 6—figure supplement 1 ) where the AD CTX showed a mean APP copy number of 3 . 40 while the AD CBL ( 2 . 34 ) , ND CTX ( 1 . 44 ) , and ND CBL ( 1 . 92 ) remained closer to 2 copies . p = 0 . 0163 , p = 0 . 0016 , **p < 0 . 0001 , ANOVA p < 0 . 0001 . ( C–F ) Distribution of copy number calls for exon 3 ( C and D ) and exon 14 ( E and F ) binned by relative copy number . The AD CTX for both exons displayed unique distributions , with more nuclei falling into the high copy number bins . ( G and I ) Distribution of nuclei with copy numbers less than , equal to , and greater than two copies . ( H and J ) Average copy number increases in nuclei binned with greater than two copies ( gold columns in G ) ( AD CTX: Exon 3 = 5 . 01 , Exon 14 = 4 . 96 , p = 0 . 0361 ) . All statistics represent an ANOVA with a Tukey's multiple comparison test . Bars indicate ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01310 . 7554/eLife . 05116 . 014Figure 6—source data 1 . Single Cell qPCR Data and Statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01410 . 7554/eLife . 05116 . 015Figure 6—figure supplement 1 . Concordance of APP exon 3 and 14 from single cell qPCR . ( A ) Relative copy numbers ( RCN ) for APP exon 3 and APP exon 14 displaying concordance between exons . 100 of 115 nuclei examined for both exons display copy numbers within one copy number call . 10 of the 15 remaining nuclei , while more than one copy number apart , were both called as gains . Bars represent RCN Min and RCN Max . ( B ) Scatter plot of average relative copy numbers . The data remain consistent with those displayed for individual exons ( Figure 6A , B ) . Statistics represent an ANOVA with a Tukey's multiple comparison test . p < 0 . 01 , p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 015 The identification of high APP copy number nuclei by single-cell qPCR suggested the possibility that amplified loci could produce detectable FISH signals . In chromosome 21 non-quantitative regional point probe FISH experiments , we occasionally observed variable puncta sizes ( Figure 5H , I ) ( Rehen et al . , 2005 ) that were initially dismissed as technical hybridization variability but which might potentially indicate sub-chromosomal variation of loci containing target sequences . However , conventional point-probe FISH analysis , while useful for assessments of aneusomy or rearrangements , cannot quantitatively evaluate copy number changes occurring as contiguous copies in close proximity to one another . Moreover , other conventional techniques for identifying copy number in single cells , such as single cell sequencing , rely heavily on genome amplification , which may introduce bias or mask CNVs . We therefore developed a detection assay based upon PNA chemistry . PNA probes hybridize with single base discrimination and have been used to quantify short repeat sequences such as those found on telomeres ( Buchardt et al . , 1993; Lansdorp et al . , 1996 ) . This raised the possibility that amplified copies of APP could be identified using multiple PNA probes simultaneously hybridized to unique gene sequences , provided that amplification occurred in a physically constrained locus rather than being dispersed throughout the genome . Nine PNA probes of 12–18 residues , each conjugated to a single Alexa-488 fluor , were designed against multiple sites of the same APP exons examined by single-cell qPCR , and validated both in silico and by blotting for specificity and linear quantitative behaviors ( Figure 7A , Figure 7—figure supplement 1A–C ) . Relative probe sequence locations were also designed to avoid fluorochrome proximity quenching . Single PNA probes did not produce a detectable signal in any samples using standard fluorescence microscopy used for aneusomy FISH ( Rehen et al . , 2001 , 2005 ) , contrasting with clearly detectable telomeric signals , which led to evaluation of more sensitive microscopic techniques including confocal , deconvolution ( Westra et al . , 2010 ) , and ultimately super resolution structured illumination microscopy ( N-SIM , Nikon ) ( Gustafsson , 2000 ) , which all failed to detect possible signals in AD nuclei . However , hybridization of increasing numbers of distinct PNA probes combined with N-SIM visualization resulted in the empirical determination of a threshold that identified one , and rarely two , punctate signals that were much more frequent in AD nuclei ( Figure 7B ) , and with a frequency consistent with single-cell qPCR data ( Figure 6 ) . AD neuronal nuclei showed single purely green puncta ( Figure 7B , #1 arrow , Video 1–6 ) and rarer two puncta signals ( Figure 7B , #3 arrow , Video 6 ) , all of which could be differentiated from lipofuscin puncta that fluoresced in all channels ( Figure 7B , #2 arrow , Video 5 ) . Internal positive control signals identified telomeres ( Figure 7 , red fluorescent puncta ) . Use of 3D-SIM enabled acquisition and analysis of super resolution projections of individual green puncta ( Figure 7B ) revealing a range of morphologies and intensities ( Figure 7B , D , Figure 7—figure supplement 1D , E ) consistent with varied APP copy number and possibly distinct , intrachromosomal genomic organization . This threshold detection using multiple PNA probes targeting APP visualized green puncta in 56% of AD neuronal nuclei compared to 22% in non-diseased and 14% in DS ( Figure 7C ) . There was no evidence of single-copy detection in any sample , contrasting with standard FISH data ( Figure 5D–I ) , despite always showing telomeric signals . Compared to single-cell qPCR data , which reports both endogenous alleles in addition to amplification , copy numbers at a single locus below ∼2 were not visualized with PNA-FISH , allowing focused analyses on positive profiles resulting in fluorescence intensity plots ( Figure 7D and Figure 7—figure supplement 1D ) that were highly reminiscent of APP copy number plots produced by single-cell qPCR ( Figure 6 ) . 10 . 7554/eLife . 05116 . 016Figure 7 . Visualization of APP copy number increases in neuronal nuclei from AD brain samples . ( A ) Peptide nucleic acid probes ( PNA ) were developed against nine separate sites on APP ( 4 sites within exon 3 and 5 sites within exon 14 ) . Each PNA probe consists of a peptide backbone conjugated to a single fluorophore , with separately conjugated nucleotides , substantially increasing specificity ( Lansdorp et al . , 1996 ) . Single copies of APP are not detectable because of fluorophore detection limits . Detection of increased copy number by PNA probes can be visualized as copies of APP increase ( Figure 7—figure supplement 1B , C ) . Positive internal controls using PNA probes directed against telomere sequences were simultaneously hybridized . ( B ) Visualization of copy number increases in neuronal nuclei . Green puncta ( arrow 1 , insets ) indicate visualized APP increases . Telomere labeling ( red puncta ) was present in all nuclei , demonstrating probe accessibility and template fidelity . Lipofuscin ( arrow 2 , orange puncta ) was detected in nuclei , visualized by extensive fluorescence signal in all channels , but was eliminated from quantifications . Limited nuclei displayed two green puncta ( arrow 3 ) . V1-6 Refers to the supplemental videos where 3-D projections can be visualized . ( C ) Graphic representation of non-diseased ( blue ) and DS ( grey ) brains displayed limited numbers of threshold-detected increases in APP ( Figure 7—source data 1 ) . AD ( red ) brains displayed significant and consistent threshold-detected increases in APP . ( D ) Individual threshold-detected APP increases were quantified and plotted on a relative intensity scale ( blue diamonds: non-diseased , red diamonds: AD ) . Dotted line represents the threshold below which APP copy number was undetectable , only limited puncta were identified in non-diseased nuclei . Bars indicate ± SEM , p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01610 . 7554/eLife . 05116 . 017Figure 7—source data 1 . Data and statistics for PNA-FISH counts . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01710 . 7554/eLife . 05116 . 018Figure 7—figure supplement 1 . PNA FISH controls . ( A ) PNA probe specificity was verified via dot blots of APP sequence followed by PNA probe hybridization and immunoblotting against the Alexa-488 fluorophor . Probes designed against exon 3 exhibited specific binding to the 5′ region of APP , and probes designed against exon 14 exhibited specific binding to the 3′ region of APP , while probes did not display significant binding to non-specific sequences . ( B ) Plasmids containing all 9 APP PNA binding sites were blotted at 1× ( 1 . 8 µg ) , 2× ( 3 . 6 µg ) , and 3× ( 5 . 4 µg ) DNA concentration , and PNA probes were hybridized and an empty plasmid at 1 . 8 μg was used for a negative control in lane 1 . Fluorescent output demonstrated a linear increase with increasing DNA concentration . ( C ) 10 μg of plasmids containing 0 , 3 , 6 , and 9 copies of the APP PNA binding sites was blotted onto membrane and probes were hybridized . Fluorescent output showed an expected linear increase with the number of PNA probe binding sites . ( D ) Quantification of the variable APP signal increases observed across four brains . ( E ) Representative APP signals visualized and verified using super-resolution 3D projections displayed a range of variable intensities . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01810 . 7554/eLife . 05116 . 019Video 1 . PNA-Fish analysis of APP in nuclei from non-diseased cortical neurons . Video of 3-D projection from Figure 7B , V1 . Green puncta indicating APP increases were infrequently visualized in non-diseased brains . Red puncta indicate telomere labeling with separate telomere-specific PNA probes and were visualized as a positive control for PNA hybridization . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 01910 . 7554/eLife . 05116 . 020Video 2 . PNA-Fish analysis of APP in nuclei from non-diseased cortical neurons with lipofuscin . Video of 3-D projection from Figure 7B , V2 . Green puncta indicating APP increases were infrequently visualized in non-diseased brains . Lipofuscin ( orange puncta ) , visualized by extensive fluorescence signal in all channels , was detected in some nuclei , but was excluded from analysis . Red puncta indicate telomere labeling with separate telomere-specific PNA probes and were visualized as a positive control for PNA hybridization . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 02010 . 7554/eLife . 05116 . 021Video 3 . PNA-Fish analysis of APP in nuclei from non-diseased cortical neurons . Video of 3-D projection from Figure 7B , V3 . Green puncta indicating APP increases were infrequently visualized in non-diseased brains . Red puncta indicate telomere labeling with separate telomere-specific PNA probes and were visualized as a positive control for PNA hybridization . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 02110 . 7554/eLife . 05116 . 022Video 4 . PNA-Fish analysis of APP in nuclei from AD cortical neurons . Video of 3-D projection from Figure 7B , V4 . Green puncta indicate visualized APP increases . Red puncta indicate telomere labeling with separate telomere-specific PNA probes and were visualized as a positive control for PNA hybridization . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 02210 . 7554/eLife . 05116 . 023Video 5 . PNA-Fish analysis of APP in nuclei from AD cortical neurons with lipofuscin . Video of 3-D projection from Figure 7B , V5 . Green puncta indicate visualized APP increases . Lipofuscin ( orange puncta ) , visualized by extensive fluorescence signal in all channels , was detected in some nuclei , but was excluded from analysis . Red puncta indicate telomere labeling with separate telomere-specific PNA probes and were visualized as a positive control for PNA hybridization . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 02310 . 7554/eLife . 05116 . 024Video 6 . PNA-Fish analysis of APP in nuclei from AD cortical neurons . Video of 3-D projection from Figure 7B , V6 . Green puncta indicate visualized APP increases . Limited nuclei in AD displayed two green puncta . Red puncta indicate telomere labeling with separate telomere-specific PNA probes and were visualized as a positive control for PNA hybridization . DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 024 Our results support a mechanism for the development of sporadic AD whereby the validated familial pathogenic APP gene is mosaically amplified in neurons . Mosaicism was documented based on multiple independent criteria: increased DCV in 90% of examined AD brains , varied and increased APP copy number in small cohorts , increased APP CNVs in single neurons identified by single-cell qPCR , and APP copy number increases in single nuclei visualized by PNA-FISH . These data are consistent with our inability to substantiate increases in trisomy 21 in cells of the sporadic AD brain . The concomitant increase in DCV suggests effects on other loci that are also altered in AD , the most obvious of which would be copy gains , but which could also involve losses based upon the presence of increased DCV range observed in AD; such loci might contribute to the progeric presentation of sporadic AD and possibly familial forms of AD that still require decades for disease to manifest . Our working hypothesis ties the currently accepted pathogenicity of APP gene dosage , established by familial AD and DS , to sporadic AD through a mechanism of somatic , mosaically increased APP copy number in some neurons . Our data do not exclude the possibility that the changes observed are a downstream effect of causative factors in the disease . However , if mosaic genomic changes are downstream of disease onset , APP copy number changes would likely play a significant role in disease progression . Advances in single-cell genomic sequencing had initially suggested its use for identifying APP CNVs in single AD neurons . However , existent technology is limited to published sequence resolution between ∼2-5 Mb and ∼0 . 025X genome coverage ( Evrony et al . , 2012; Gole et al . , 2013; McConnell et al . , 2013; Cai et al . , 2014 ) . Considering the <0 . 3 Mb size of the APP locus , single cell sequencing would be incapable of identifying all but the rarest APP CNVs observed here . In addition , whole genome single cell sequencing is limited by throughput , demonstrated by the low number of neurons reported in this field , and is complicated further by variable results ( Cai et al . N = 82 ( QC neurons ) , Evrony et al . N = 6 , Gole et al . N = 6 , McConnell et al . N = 110 ) . Notably , DNA losses predominated in both Cai et al . ( 2014 ) and McConnell et al . ( 2013 ) however , Gole et al . ( 2013 ) observed that two thirds of somatic CNVs were gains , consistent with increases in DNA content reported previously , and all three studies support a range of DCV amongst neurons , although none of these prior studies assessed AD neurons . While future advances will improve sequence resolution , throughput , and amplification fidelity in single cells , the distinct single cell strategy employed here allowed copy number assessment of a single targeted gene , APP , an approach which may be generalizable to other loci and diseases . Compared to the prior single-neuron reports , 320 single , neuronal nuclei were assessed here for APP CNVs wherein single cell qPCR data markedly resembled PNA-FISH data . Notably , these single cell techniques also possess limitations . For example , single cell qPCR requires normalization to single copy reference genes and control populations , which presents unique difficulties in assessing mosaic neuronal populations . For example , changes in copy number could reflect changes in both the reference and target gene , which might result in artifactual increases ( or decreases ) in a target gene . Similarly , assessment of copy number from an amplified template could also produce artifactual changes , and thus results require independent techniques for verification , especially since the original template is consumed in the single cell reaction and cannot be further assessed . The use of PNA-FISH does not require any amplification or normalization , and yet PNA-FISH and single-cell qPCR produced a highly similar distribution of APP copy number amplifications . The employed techniques were not capable of assessing intact neurons or their histological organization , which will require technological modifications . Similarly , assessments of genomic and expression data in a single neuron also await technological advancements . Finally , these techniques are currently capable of interrogating only parts of the entire APP locus; therefore , at this time the precise structure of APP CNVs and the mechanisms leading to their existence are unknown . Our data have bearing on at least 3 AD hypotheses . First , the prevailing amyloid hypothesis in AD posits that Aβ deposition drives AD ( Hardy and Selkoe , 2002 ) . Increased incidences of amyloid senile plaques are observed in all forms of AD and appear to be directly linked to APP dosage in DS and familial AD . The mosaic increases in neuronal APP copy number reported here provide an explanation for the universal presence of Aβ senile plaques in sporadic forms—indeed , all forms—of AD despite an absence of constitutive copy number gain . The presence of rare neurons with APP amplifications in non-diseased brains also provides an explanation for the observed presence of senile plaques in otherwise normal , aged brains ( Gibson , 1983; Cras et al . , 1991; Mackenzie et al . , 1996 ) , consistent with the ages of non-diseased brains examined here , which exceeded 75 years . Along with the occurrence and augmentation of DCV , APP amplification in AD is consistent with dysregulation of normally occurring processes in the etiology of AD pathogenesis . A second hypothesis proposes increased APP dosage through trisomy 21 based largely upon the neuropathology of DS ( Heston and Mastri , 1977; Potter , 1991; Geller and Potter , 1999 ) . However , dual chromosome point-paint FISH for chromosome 21 using liberal calling criteria on ∼5000 AD brain cells ( N = 14 brains ) showed no relationship between total mosaic aneusomies and AD . Moreover , the lack of observed increases of trisomy 21 in sporadic AD are consistent with prior studies in both peripheral non-brain and intact brain tissues , which reported APP levels approximating 2N ( Podlisny et al . , 1987; St George-Hyslop et al . , 1987; Tanzi et al . , 1987; Bertram et al . , 2010 ) , and is further consistent with mosaic CNVs observed here . The purposeful use of liberal counting criteria resulted in generally higher percentages of aneusomies than reported previously ( Rehen et al . , 2005; Iourov et al . , 2009 ) , however the increases were observed in both AD and non-diseased samples without linkage to AD . Our results do not eliminate roles for mosaic aneuploidy in AD , but do not support trisomy 21 as a specific mechanism for increasing APP copy number in sporadic AD . A third hypothesis is that abnormal cell cycle reentry in post-mitotic neurons contributes to AD ( Yang et al . , 2001; Herrup and Arendt , 2002; Kruman II et al . , 2004; Copani et al . , 2006; Mosch et al . , 2007; Herrup , 2012 ) . Our DCV analyses detected maximum gains of ∼21% and average gains of ∼8 . 2% over lymphocytes , representing a fraction of the 100% increase expected in a 4N cell . The subgenomic increases observed here are also consistent with the reported absence of adult neurogenesis in the normal cerebral cortex ( Rakic , 2002; Bhardwaj et al . , 2006 ) , and could also be relevant to reports of nucleotide incorporation in adult cortical neurons ( Gould et al . , 1999 ) . Concepts involving changes in DNA synthesis in AD without cell-cycle progression or neurogenesis could be rectified through subgenomic DNA synthesis in post-mitotic neurons that might involve DNA synthetic proteins reported in AD ( Copani et al . , 2002 , 2006 ) . The observed genomic alterations in AD could occur through both developmental as well as aging processes as reported for at least one form of mosaicism , aneuploidies ( Bushman and Chun , 2013 ) that have been linked to caspase-mediated cell death ( Peterson et al . , 2012 ) and could have relevance to developmental APP functions ( Nikolaev et al . , 2009 ) . Perhaps tellingly , these developmental cell death processes also involve DNA fragmentation and double strand breaks ( Blaschke et al . , 1996; Staley et al . , 1997; Blaschke et al . , 1998 ) that have recently been reported in neurons exposed to physiological and AD stimuli ( Suberbielle et al . , 2013 ) . It is possible that somatic mosaicism may arise through dysfunction of normally operative DNA repair mechanisms that produce DCV in the normal brain , with augmentation in the diseased brain . We speculate that increased DCV and APP CNVs observed in AD may in part reflect neurons that have delayed or averted cell death , wherein somatic genomic changes could provide a survival advantage at a cost of altered neurophysiological functions . The DNA content changes observed here showed not only disease relationships but also varied with brain region and cell type . It is important to note that the interrogated neurons are likely terminal in their fate: they will not divide further , contrasting with lymphocytes that , perhaps surprisingly , showed the least amount of DCV despite being a proliferating population . Unlike neurons , proliferating lymphocytes may be similar to stem and progenitor populations that can maintain germline genomes to promote population expansion . By contrast , neurons are by definition post-mitotic , and therefore not subject to this possible requirement . We think it likely that many dividing cells that deviate significantly from the germline genome—and thus would be incapable of further division—are eliminated by cell death and would be removed from the population interrogated for DNA content . Support for this view includes reports on aneuploidies in neuroprogenitor cells , whereby the most extremely aneuploid forms appear to be eliminated by caspase-mediated cell death ( Blaschke et al . , 1996 , 1998; Peterson et al . , 2011; Peterson et al . , 2012 ) . DNA content changes may therefore be contextual , enabling diverse functions in different cell types . The large differences observed even between brain regions and amongst neuronal nuclei would suggest the existence of distinct neuronal functions produced by DCV , and possibly specific processes or stimuli that can genomically alter neurons . Diverse stimuli previously implicated as AD risk factors including age and trauma ( Sparks et al . , 1990; Plassman et al . , 2000 ) , might share common endpoints via influences on genomic mosaicism . The consequences of somatic APP amplification in AD may support functionality of other somatically altered genomic loci observed in single neurons that could contribute to the progeric presentation of AD as well as aspects of the disease itself . We further hypothesize that other sporadic or idiopathic brain diseases could arise through altered genomic mosaicism that includes somatic variations at both known and unknown pathogenic loci . All human tissue protocols were approved by the Scripps Office for the Protection of Research Subjects ( SOPRS ) at The Scripps Research Institute ( TSRI ) and conform to National Institutes of Health guidelines . Fresh-frozen brain tissue was provided by the NICHD Brain and Tissue Bank for Developmental Disorders at the University of Maryland , the University of California Alzheimer's Disease Research Center ( UCI-ADRC ) , and the Institute for Memory Impairments and Neurological Disorders , the Johns Hopkins School of Medicine Alzheimer's Disease Research Center , and Dr Edward Koo at the University of California , San Diego . Lymphocytes from human peripheral blood were obtained from healthy donors at TSRI's Normal Blood Donor Services . Detailed information about the samples used can be found in the table below ( Table 1 ) . 10 . 7554/eLife . 05116 . 025Table 1 . Human samples used in each experimentDOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 025SexAgePaper codeSampleExperimentsPost mortem intervalBraak scoreAlzheimer's disease prefrontal cortex , N = 32F811521D23VIF831562D6UF74AD-131866DTUUF79AD-51868DSTUUF82AD-101875DSTUUF831893D9UF871899D5VIF62AD-91912DSTUUF80AD-81913DTUUF54AD-41916DTBUUF72AD-21921DTBUUF772400D3 . 7VF80AD-112500DP2 . 3VIF10150341D19VF9161788D11VF9862405D11VF8962439D11VF7762509D18VIM88AD-1102DSBP3IVM90268D91VM83736D27UM821211D18UM79AD-151252D9UM921748D5 . 5VM85AD-121861DTUUM85AD-141870DTUUM82AD-32401DSBP3VIM84AD-62499DS3 . 4VIM63AD-74199DP3VIM8013173D22IVM9430022D20VM9160987D22 . 5VMean82 . 1Alzheimer's disease cerebellum , N = 15F74AD-131866DUUF79AD-51868DSUUF82AD-101875DSUUF80AD-81913DUUF62AD-91912DSUUF54AD-41916DBUUF72AD-21921DBUUM88AD-1102DSB3IVM79AD-151252D9UM701625D1UM85AD-121861DUUM85AD-141870DUUM82AD-32401DSB3VIM84AD-62499DS3 . 4VIM63AD-74199D3VIMean75 . 9Non-diseased prefrontal cortex , N = 40F74ND-21901DSBP2 . 3IIF74ND-8299D2 . 8IIF84ND-3703S5 . 8IIIF53ND-111379D15IIIF73713TUUF9560831D9IIF511568P22UF171230P16UF87ND-101502D5IIF8060728D13IIM79ND-9827DTUUM96ND-11102DSB3 . 4IIM83ND-52501DB1 . 7IIM95ND-41301DS3 . 5IM871471TUUF711571TUUM531344DTUUF93318D2 . 3VIF92955D20 . 5IIIF564238D12M704534D28F9111488D16IIF7913188D12 . 5F9013204D9 . 5IIF10360329D5IIIF8560428D8 . 5IIIF9960524D15IIF9562043D20 . 5M71389M15F83719M17IIIM69946M12M872039M6 . 3IIIF864546M22M9160772M16IIF8061218M5 . 5M8761334M8IIM88PDC2MUM80PDC5MUM75PDC8MUMean79 . 54Non-diseased cebellum , N = 15F74ND-21901DSB2 . 3IIF74ND-8299D2 . 8IIF84ND-3703S5 . 8IIIF771569D8IIIF83719DUUF531379D15IIIF711571DTUUF87ND-101502D5IIM531344DTUUM96ND-11102DSB3 . 4IIM83ND-52501DB1 . 7IIM95ND-41301DS3 . 5IM79ND-9827DUUM871471DUUF864546M22UMean78 . 8DS/AD , N = 3F51DS-1M1864B19UF47DS-2M3233S24UF44DS-31258S13UMean47 . 3LYM , N = 21F405162DN/AN/AF403963DN/AN/AF634984DN/AN/AF604519DN/AN/AM55Lym 1DN/AN/AM354651DN/AN/A29DN/AN/A187DN/AN/A4781DN/AN/A4801DN/AN/A4903DN/AN/AM285259DN/AN/AM5683MF521344MF564603MM544609MM58Lym 2MF51Lym 3MM52Lym 4MMean50 . 0MSamples in bold are paired CBL and CTX . Denotes mid frontal gyrus ( MFG ) , D = DNA content analyses , S= Small population qPCR , T = FISH Analysis , B = single cell qPCR on Biomark HD , P=PNA FISH , M = Westra et al . DNA content metadata . Human brain nuclei were isolated and prepared for FCM and FACS as previously described ( Westra et al . , 2008 , 2010 ) . Isolated nuclei were fixed with 2% paraformaldehyde ( or 70% ethanol for microfluidic qPCR ) , labeled with mouse anti-NeuN antibody ( 1:100 ) ( Millipore , Germany ) and Alexa Fluor 488 goat anti-mouse IgG secondary ( 1:250 ) ( Life Technologies , Carlsbad , CA ) , and counterstained with propidium iodide ( 50 μg/ml ) ( Sigma , St . Louis , MO ) in solution containing 50 μg/ml RNase A ( Qiagen , Valencia , CA ) and chicken erythrocyte nuclei ( CEN ) ( Biosure , Grass Valley , CA ) . Electronically gated diploid neuronal nuclei , determined by PI fluorescence and immunolabeling , were analyzed and sorted either in bulk for standard qPCR and PNA-FISH or singly in 96-well plates for microfluidics-based qPCR . FCM and FACS were performed at the TSRI Flow Cytometry Core using a Becton Dickinson ( BD Biosciences , San Jose , CA ) LSRII and FACS-Aria II , respectively . Post hoc DNA content analyses were performed using FlowJo software ( TreeStar Inc . , Ashland , OR ) . For validation of DNA content analyses , nuclei were sorted by FACS in populations of 1000 , 500 , and 100 into 96-well plates according to relative DNA content ( Figure 2A ) . Nuclei were denatured by multiple freeze–thaw cycles and potassium hydroxide , the solution was neutralized , and phi29 DNA polymerase ( Illumina , San Diego , CA ) , dNTPs , SYBR green , and random hexamers ( IDT , Coralville , IA ) were added ( Gole et al . , 2013 ) . The MDA reaction was performed at 30°C and SYBR Green intensity was recorded every 2 min on a QuantStudio RT-PCR 12K Flex ( Life Technologies , Carlsbad , CA ) . 12 replicates were run and averaged for each group . All qPCR was performed according to MIQE guidelines ( Bustin et al . , 2009 ) . For standard qPCR , primers against the APP gene exon 14 and reference genes SEMA4A , CCL18 , and PCDH11X were designed in house and optimized to an annealing temperature of 59°C using Primer3 software ( University of Massachusetts Medical School ) and synthesized by Valuegene ( San Diego , CA ) ( Table 2 ) . For microfluidics-based qPCR , TaqMan assays against APP exons 3 and 14 ( 178 kb apart ) ( designed from NCBI Reference Sequence: NG_007376 . 1 ) , SEMA4A and PCDH11X were synthesized by Applied Biosystems ( Life Technologies , Carlsbad , CA ) or Integrated DNA Technologies ( APP exon 3 ) ( San Diego , CA ) and optimized to an annealing temperature of 60°C . Primers were assessed in silico to determine specificity of the primer set for a single genomic region and for the presence of SNPs in the targeted genomic region which could decrease primer binding and amplification efficiency . The specificity of all qPCR assays was assessed by gel electrophoresis to confirm a single PCR product of the expected length and sequenced to confirm amplification of the expected product . 10 . 7554/eLife . 05116 . 026Table 2 . Primers used for small population and single nuclei qPCRDOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 026GeneProteinLocusAssay typePrimer sequenceProbeProduct lengthEfficiencyAPP , Exon 14Amyloid precursor protein21q21 . 3SYBR GreenF-TGCACGTGAAAGCAGTTGAAG , R-AAAGATGGCATGAGAGCATCGN/A2140 . 973SEMA4ASemaphorin 4A1q22SYBR GreenF- ATGCCCAGGGTCAGATACTAT , R-TTCTCCGAGATCCTCTGTTTCN/A1770 . 997CCL18Chemokine ( C–C motif ) ligand 1817q11 . 2SYBR GreenF-TTCCTGACTCTCAAGGAAAGG , R-CTGGCACTTACATGACACCTGN/A2091 . 006PCDH11XProtocadherin 11 X-linkedXq21 . 3SYBR GreenF-TCTTTTGGTCAGTGTTGTGCG , R-CAACAAGTCGCCTATCAGGACN/A1880 . 993APP , Exon 14Amyloid precursor protein21q21 . 3TaqManCGGTCAAAGATGGCATGAGAGCATC* , Assay Hs01255859_cnFAM-MGB911 . 040APP , Exon 3Amyloid precursor protein21q21 . 3TaqManF-GCACTTCTGGTCCCAAGCAT , R-CCAGTTCTGGATGGTCACTGROX-IB1400 . 992SEMA4ASemaphorin 4A1q22TaqManGTTCAAGGGTATGTGAGGTGAGATG* , Assay Hs00329046_cn_VICVIC-MGB901 . 016*Denotes probe sequence provided by Life Technologies . Standard curves were used to determine the amplification efficiency of all primer sets used for qPCR . For standard qPCR primers , curves were created by serially diluting purified pGEM-T Easy plasmid DNA ( Promega , Madison , WI ) containing a single copy of the gene of interest ( Pfaffl , 2001 ) ; serial dilutions of genomic DNA were used for TaqMan primers ( D'Haene et al . , 2010 ) . DNA concentrations were converted to gene copy number by calculating the weight ( in g/mol ) of the DNA used for generating the standard curve . A linear regression of the curve comparing the log of the gene copy number vs the crossing threshold ( Ct ) of the primer set was determined from primer efficiency ( E ) = 10−1/slope − 1 ( Pfaffl , 2001 ) . Only standard curves with R2 values of greater than 0 . 99 were used . Genomic DNA from bulk-sorted nuclei ( described above ) was isolated using the DNeasy Blood and Tissue Kit ( Qiagen , Valencia , CA ) and quantified using Quant-iT PicoGreen ( Life Technologies , Carlsbad , CA ) ; genomic DNA was stored at −20°C before use . Standard qPCR reactions using SYBR Green ( Promega , Madison , WI ) fluorescence detection ( ex: 494 nm; em: 529 nm ) were performed in triplicate using 0 . 5 ng of sample gDNA per reaction . Reactions were run on a Rotor-Gene RG-3000 72-well thermocycler ( Qiagen , Valencia , CA ) using GoTaq qPCR master mix ( Promega , Madison , WI ) and the following parameters: denaturation ( 95°C for 5 min ) , amplification ( 95°C for 25 s , 59°C for 30 s and 72°C for 30 s ) , and quantification through 40 cycles; and a melting curve determination ( 55–99°C , 30 s on the first step , 5 s for each subsequent step ) . The crossing threshold ( Ct ) was determined for each primer set within the linear region of the amplification curve . Down Syndrome nuclei were used as a control . Single neuronal nuclei from FACS were sorted directly into a 96-well plate containing QuickExtract DNA Extraction Solution ( Epicentre , Illumina , San Diego , CA ) according to the manufacturer’s instructions . Multiple independent sorts were completed for each group and each individual . Prior to analysis on the Fluidigm Biomark HD ( South San Francisco , CA ) , single neuron genomic DNA was pre-amplified as per Fluidigm protocols ( Fan and Quake , 2007; Dube et al . , 2008; Qin et al . , 2008; Jones et al . , 2011; White et al . , 2011; Whale et al . , 2012 ) on a Veriti thermocycler ( Life Technologies , Carlsbad , CA ) using locus-specific Taq-man primer sets ( primer Table above ) ( 20× initial concentration , 18 µM primer , 5 µM probe; combined and diluted to 0 . 2× ) ( 95°C denaturation for 5 min; 18 amplification cycles of a 95°C denaturation for 15 s , followed by a 60°C annealing and extension step for 4 min; and a final extension step at 72°C for 7 min ) . Locus-specific pre-amplification was confirmed on a Roche LightCycler ( Roche Applied Science , Indianapolis , IN ) using one targeted primer set prior to analysis on the Biomark 48 . 48 Dynamic Array integrated fluidic circuit ( IFC ) ( Fluidigm , South San Francisco , CA ) . Samples were diluted 1:5 and loaded into the 48 . 48 Dynamic Array IFC according to the manufacturer's protocol . DNA was loaded in triplicate and assays in sextuplicate for a total of 18 replicates per assay per nucleus . Samples were run across multiple arrays for quality control between runs ( Table 3 ) and multiple individuals were run on each array . The thermocycling program was performed on the Biomark: 95°C for 10 min , then 55 cycles of 95°C denaturation , and 60°C annealing and extension . Fluorescent probes used for these assays were 5′-FAM or 5′-VIC with a 3′-minor groove binding ( MGB ) non-fluorescent quencher , or 5′-ROX with a 3′ Iowa Black quencher . Ct values were determined using Fluidigm's Real-Time PCR Analysis Software . Only nuclei with 10 or more replicates per assay were used for analysis . 10 . 7554/eLife . 05116 . 027Table 3 . Quality control between 48 . 48 Dynamic Array runsDOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 027Cell 1Cell 2Cell 3Cell 4Cell 5Cell 6Cell 7Cell 8APP 14Run 121 . 3718 . 7718 . 0218 . 3417 . 1119 . 3320 . 1218 . 02Run 220 . 6018 . 7718 . 1418 . 4117 . 2519 . 5020 . 6018 . 26APP3Run 122 . 7920 . 3518 . 5319 . 11Run 221 . 9719 . 8418 . 6519 . 28SEMA4ARun 125 . 5424 . 2922 . 4424 . 2322 . 8125 . 7424 . 4624 . 67Run 225 . 2624 . 5323 . 9224 . 3923 . 0325 . 4224 . 1124 . 30 For both qPCR strategies , relative copy number ( RCN ) for a diploid sample was calculated as RCN = 2 × ( 1 + E ) −ΔΔCt where E is primer efficiency ( E = 10−1/slope − 1 ) ( Weaver et al . , 2010; Pfaffl , 2001; D'Haene et al . , 2010; Livak and Schmittgen , 2001 ) . Paired cerebellar samples were used as a calibrator to determine ΔΔCt values . For single cell RCN determinations , cerebellar ΔCt values for the paired cerebellum were averaged . RCNs were modeled for each copy number 1–6 , assuming a system standard deviation of 0 . 25 and a 95% CI equal to the standard error of the mean multiplied by the critical t value for a two-tailed t-distribution ( degrees of freedom = 68 ) with p = 0 . 05 ( Weaver et al . , 2010 ) . 95% confidence interval ( CI ) was also determined for the RCN of each assay for each single nucleus giving RCN = 2 × ( 1 + E ) −ΔΔCt ± CI , and the upper and lower bounds were used to call copy numbers for APP exons 3 and 14 such that a CI that overlapped with the modeled RCN range was considered as belonging in that copy number bin ( See Table 4 for modeled upper and lower bounds ) . Copy number bins were 1–6 and >6 , as beyond 6 copies the CIs for the given degrees of freedom begin to overlap , allowing for assessments of significant increases in copy number but limiting the ability to distinguish between nuclei with , for example , 8 vs 9 copies . There was high concordance between the APP exons 3 and 14 demonstrating minimal amplification bias between the primer sets . 10 . 7554/eLife . 05116 . 028Table 4 . Confidence Intervals for ( CI ) calling Copy Number ( CN ) DOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 028CNCIRCN , value1Lower0 . 92156Upper1 . 085122Lower1 . 84312Upper2 . 170233Lower2 . 76468Upper3 . 255354Lower3 . 68624Upper4 . 340475Lower4 . 60780Upper5 . 425596Lower5 . 52936Upper6 . 51070 For DNA content FCM , DNA indices ( DIs ) were determined by taking the ratio of the mean from the diploid ( 2N ) peak from the brain sample to the mean of the lymphocyte control peak , both normalized to the mean of the CEN standard ( Darzynkiewicz and Huang , 2004; Darzynkiewicz et al . , 2004; Westra et al . , 2010 ) . The percent change was calculated from DI values assuming a 2N diploid would have a DI of 1 . p values for comparison of mean percent change , skew and coefficients of variation comparisons were determined by one-way ANOVA and Tukey's multiple comparison tests . Linear regression analysis was used to determine age-percent change correlation . For comparison of average percent change from NeuN-positive vs negative cortical nuclei , or NeuN-positive cortical and cerebellar nuclei , p values were determined by unpaired , two-tailed t-test . For standard qPCR analyses , differences in ΔΔCt ± SEM of APP in the cortex vs cerebellum were assessed in each individual using an unpaired , two-tailed t-test . For single cell qPCR , p values were determined by one-way ANOVA and Tukey's multiple comparison tests . Isolated nuclei were stained with DAPI and hybridized using dual color FISH as described previously ( Rehen et al . , 2001; Kaushal et al . , 2003 ) . FISH paints against the whole q arm of chromosome 21 and a point probe against a region on the q arm of 21 ( 21q22 . 13-q22 . 2 ) ( Vysis . Downer's Grove , IL ) were used . The mounted slides were examined on a Zeiss Axioskop microscope and Axiocam CCD camera ( Carl Zeiss , Thornwood , NY ) . Approximately , 500 nuclei were blindly counted for each brain , by two independent observers , on 14 samples ( 5 non-diseased , 9 diseased ) . Total AD cortical nuclei were examined using a highly liberal protocol for calling aneusomies whereby borderline FISH profiles suggestive of aneusomy were always included in quantitative assessments . All analyses were conducted blind to the identity of samples by interrogating purified nuclei . The ability of this technique to detect aneuploidy was validated using interphase nuclei from a human trisomy 21 brain ( Svendsen et al . , 1998 ) revealing three nuclear signals ( Figure 5D , E ) . Peptide nucleic acid probes were custom designed in coordination with PNA Bio , Inc . Nucleic acid sequences were identified and analyzed in silico to ensure binding to only one specific genomic region on APP . Nine unique probes ( 4 on exon 3 and 5 on exon 14 ) were designed and conjugated to a single fluorophore of Alexa-488 ( Table 5 ) . Specificity of probes was confirmed using dot blot DNA detection paired with immunoblot via antibodies against the Alexa-488 fluorophore ( AF-488 ) ( Figure 5—figure supplement 1A ) . Increasing DNA concentration ( 1 . 8 , 3 . 6 and 5 . 4 μg ) of plasmids containing one copy of each exon ( all nine PNA binding sites ) was used to verify linear increases in AF-488 signal ( Figure 7—figure supplement 1B ) . For the APP copy number curve , DNA concentration remained consistent but plasmids containing 0 , 3 , 6 , and 9 copies of all APP PNA binding sites were used . DNA was denatured in 0 . 1 M NaCl at 50°C and dot blotted onto a positively charged nylon membrane in distilled deionized water ( DDW ) and washed in 6× SSC . DNA was cross linked to the membrane , and PNA probes were hybridized to the membrane using conditions consistent with probe hybridization to nuclei on slides . Briefly , probes were prepared in 20 mM Tris , 60% formamide , and 0 . 1 μg/ml salmon sperm . Probes and membrane were heated 85°C for 5 min , and probes were added to the membrane and incubated at 85°C for 10 min , followed by room temperature overnight . Membrane was then washed twice at 60°C in 2× SSC + 0 . 1% Tween-20 , and then in successive washes of 2× SSC , 1× SSC , and DDW . Probes were visualized on the membrane using a Typhoon fluorescence scanner from General Electric ( Fairfield , CT ) . For quantification , color data were removed , image was inverted , brightness/contrast was adjusted evenly across the entire image , and average pixel intensity was acquired for a region of interest with a standardized area across sample comparisons . 10 . 7554/eLife . 05116 . 029Table 5 . Peptide nucleic acid ( PNA ) probe sequencesDOI: http://dx . doi . org/10 . 7554/eLife . 05116 . 029GeneProteinLocusAssaySequenceAPP Exon 3Amyloid precursor protein21q21 . 3PNA FISHA488-GATGGGTCTTGCACTG , A488-CCCCGCTTGCACCAGTT , A488-GGTTGGCTTCTACCACA , A488-CAGTTCAGGGTAGACAPP Exon 14Amyloid precursor protein21q21 . 3PNA FISHA488-CTCCATTCACGG , A488-GTGGTTTTCGTTTCGGT , A488-ACTGATCCTTGGTTCAC , A488-ACTGATCCTTGGTTCAC , A488-ACGTCATCTGAATAGTTTelomereN/ATelomeresPNA FISHTelC-Cy3 ( F1002 , PNA BIO ) Probes were hybridized according to the manufacturer's instructions ( Lansdorp et al . , 1996 ) . Neuronal nuclei from non-diseased and AD brains were sorted for NeuN positivity and dried onto slides . Slides were washed in PBS , fixed with 4% PFA , and treated with RNase ( Qiagen , Valencia , CA ) for 20 min at 37°C . Slides were then digested with 200 µg/ml proteinase K ( Roche Applied Science , Indianapolis , IN ) for 5 min at 37°C . Slides were dehydrated in ethanol series and denatured at 85°C for 5 min . PNA probes in hybridization buffer ( 20 mM Tris pH 7 . 4 , 60% formamide , 0 . 1 µg/ml salmon sperm DNA ) were then added for 10 min at 85°C . Slides were then removed and placed at room temperature for 2 hr . Slides were then washed twice in 2× SSC + 0 . 1% tween at 60°C for 30 min each and mounted using progold mounting media . Z-stacks were acquired using a Nikon N-SIM ( structured illumination microscopy ) super resolution microscope with an Andor iXon3 back-illuminated high sensitivity EMCCD camera with single photon detection capability . Projections were rendered using 3D-SIM Elements software from Nikon . Counts were averaged and analyzed by unpaired , two-tailed t-test	The instructions for living cells are contained in certain stretches of DNA , called genes , and these instructions have been largely considered to be invariant , such that every cell in the body has the same DNA . However , research has revealed that many neurons in the human brain can contain different amounts of DNA compared to other cells . When cells with varied DNA are present in the same person , it is referred to as mosaicism . The effects of this mosaicism are unknown , although by altering the instructions in brain cells , it is suspected to influence both the normal and diseased brain . The brains of patients with Alzheimer's disease often contain deposits of proteins called amyloids . The precursor of the protein that makes up most of these deposits is produced from a gene called the amyloid precursor protein gene , or APP . Having an extra copy of the APP gene can cause rare ‘familial’ Alzheimer's disease , wherein the APP duplication can be passed on genetically and is present in all the cells of a patient's body . By contrast , ‘sporadic’ Alzheimer's disease , which constitutes around 95% of cases , does not show any difference in the number of APP genes found in tissue samples , including whole brain . The early studies that discovered this were conducted before an appreciation of brain mosaicism , and thus single neurons were not investigated . This raises the possibility that the number of APP genes may be mosaically increased , which would not be detected by examining non-brain or bulk brain tissue . Bushman , Kaeser et al . used five different types of experiments to examine the DNA content of single neurons and investigate whether mosaicism could explain the discrepancy in the results of the previous studies . The neurons from people with Alzheimer's disease contained more DNA—on average , hundreds of millions of DNA base pairs more—and more copies of the APP gene , with some neurons containing up to 12 copies . Bushman , Kaeser et al . 's findings present evidence of a way that mosaicism can affect how the brain works by altering the number of gene copies , and how this impacts the most common form of Alzheimer's disease . Many questions arise from the work , including when does mosaicism arise , and what promotes its formation ? How does this relate to age ? What parts of the genome are changed , what genes are affected , and how do these changes alter neuronal function ? Furthermore , Bushman , Kaeser et al . 's work suggests that mosaicism may also play a role in other brain diseases , and could also provide new insights into the normal , complex functions of the brain . In the future , this knowledge could help to identify new treatments for brain diseases; for example , by identifying new molecular targets for therapy hidden in the extra DNA or by understanding how to alter mosaicism .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2015	Genomic mosaicism with increased amyloid precursor protein (APP) gene copy number in single neurons from sporadic Alzheimer's disease brains
Reduced susceptibility to infectious disease can increase the frequency of otherwise deleterious alleles . In populations of African ancestry , two apolipoprotein-L1 ( APOL1 ) variants with a recessive kidney disease risk , named G1 and G2 , occur at high frequency . APOL1 is a trypanolytic protein that confers innate resistance to most African trypanosomes , but not Trypanosoma brucei rhodesiense or T . b . gambiense , which cause human African trypanosomiasis . In this case-control study , we test the prevailing hypothesis that these APOL1 variants reduce trypanosomiasis susceptibility , resulting in their positive selection in sub-Saharan Africa . We demonstrate a five-fold dominant protective association for G2 against T . b . rhodesiense infection . Furthermore , we report unpredicted strong opposing associations with T . b . gambiense disease outcome . G2 associates with faster progression of T . b . gambiense trypanosomiasis , while G1 associates with asymptomatic carriage and undetectable parasitemia . These results implicate both forms of human African trypanosomiasis in the selection and persistence of otherwise detrimental APOL1 kidney disease variants . Infectious disease is a major driving force of natural selection on human populations . Such evolutionary pressures can select for genetic variants that confer increased resistance to infectious agents , but may also predispose to specific genetic disorders , as exemplified by Plasmodium selection for the sickle-cell trait ( Allison , 1954 ) . Like sickle-cell disease , chronic kidney disease also affects millions worldwide ( Global Burden of Disease Study 2013 Collaborators , 2015 ) , with a disproportionate risk in populations of recent sub-Saharan African ancestry ( National Institutes of Health and National Institute of Diabetes and Digestive and Kidney Diseases , 2010; Norris and Agodoa , 2002; McClellan et al . , 1988 ) . In African-Americans a large component of this disparity has been attributed to two common genetic variants of APOL1 ( MIM 603743 ) , known as G1 and G2 ( Genovese et al . , 2010; Tzur et al . , 2010 ) . These variants are closely spaced in the C-terminal domain of APOL1 but are located on separate haplotypes ( Genovese et al . , 2010 ) ( Figure 1 ) . Individuals possessing a high-risk G1/G1 , G2/G2 or G1/G2 genotype composed of two risk alleles ( approximately 13% of African-Americans [Friedman et al . , 2011] ) , are strongly predisposed to a wide spectrum of chronic kidney disorders that includes focal segmental glomerulosclerosis ( Genovese et al . , 2010; Kopp et al . , 2011 ) , HIV-associated nephropathy ( Kopp et al . , 2011; Kasembeli et al . , 2015 ) and end-stage renal disease ( Genovese et al . , 2010; Tzur et al . , 2010; Freedman et al . , 2014 ) . APOL1 G1 and G2 are prevalent only in populations of recent African heritage ( Genovese et al . , 2010; Kopp et al . , 2011 ) , with evidence for a selective sweep within the last 10 , 000 years ( Genovese et al . , 2010 ) , indicative of strong positive selection . Human African trypanosomiasis ( HAT ) , a deadly parasitic disease endemic to sub-Saharan Africa , has been proposed as the source of this positive selective pressure ( Genovese et al . , 2010 ) . HAT is caused by two tsetse fly-transmitted African trypanosomes , Trypanosoma brucei rhodesiense and T . b . gambiense , which are responsible for the acute East African form and more chronic West Africa form of the disease , respectively ( Kennedy , 2013 ) . Both parasites have been responsible for widespread fatal epidemics in sub-Saharan Africa throughout recorded human history ( Steverding , 2008 ) suggesting the potential to exert potent selection pressure on the human genome . A heterozygous advantage model has been proposed for APOL1 G1 and G2 ( Genovese et al . , 2010 ) in which recessive susceptibility to chronic kidney disease is balanced by dominant resistance to one or both forms of human African trypanosomiasis . 10 . 7554/eLife . 25461 . 003Figure 1 . Schematic of G1 and G2 polymorphisms in human apolipoprotein L1 . Human apolipoprotein-L1 ( APOL1 ) is a 398-amino acid protein consisting of a cleavable N-terminal signal peptide , a pore-forming domain , a membrane-addressing domain , and a serum resistance-associated ( SRA ) -interacting domain . The polymorphisms that characterize the G1 and G2 renal risk variants are located in the SRA-interacting domain , the target site for binding of the SRA protein expressed by the human-infective T . b . rhodesiense parasite , which results in loss of APOL1 lytic function . The location of the critical binding region ( residues 370–392 ) for this interaction is indicated by a helical graphic . G1 consists of two missense SNPs rs73885319 ( p . Ser342Gly ) and rs60910145 ( p . Ile384Met ) while the G2 polymorphism , rs71785313 ( p . Asn388_Tyr389del ) , is found on an alternative APOL1 haplotype , and represents an in-frame two amino acid deletion . DOI: http://dx . doi . org/10 . 7554/eLife . 25461 . 003 Prior to the discovery of its association with kidney disease , APOL1 had already been recognised for encoding the pore-forming serum protein Apolipoprotein L1 , which inserts into trypanosome membranes and effectively lyses the Trypanosoma species that cause disease in animals ( Vanhamme et al . , 2003; Pérez-Morga et al . , 2005; Molina-Portela et al . , 2005; Thomson and Finkelstein , 2015; Vanwalleghem et al . , 2015 ) . However , the two human-infective subspecies have evolved independent mechanisms to resist APOL1-mediated lysis . In T . b . rhodesiense , this is the result of an APOL1-binding protein ( Xong et al . , 1998; De Greef et al . , 1989 ) whereas for T . b . gambiense the mechanism of APOL1 resistance appears more complex and multifactorial ( Capewell et al . , 2013; Uzureau et al . , 2013; DeJesus et al . , 2013; Kieft et al . , 2010 ) . It has been hypothesised that APOL1 G1 and G2 variants could overcome one or more of these resistance mechanisms to protect against HAT . Indeed , previous studies have shown that APOL1 G2 ( and to a lesser extent G1 ) plasma is lytic to East African T . b . rhodesiense parasites in vitro ( Genovese et al . , 2010 ) , but not West African T . b . gambiense ( Genovese et al . , 2010 ) . Consequently , T . b . rhodesiense is considered the most likely candidate for positive selection of both APOL1 variants in African populations ( Genovese et al . , 2010 ) . Notably , however , the G1 variant appears significantly less effective at killing T . b . rhodesiense , and is found at very high frequency in West Africa ( Genovese et al . , 2010; Kopp et al . , 2011; Ko et al . , 2013; Thomson et al . , 2014 ) , where only T . b . gambiense is endemic ( Simarro et al . , 2010 ) . Furthermore , a class of asymptomatic individuals has been recently identified in T . b . gambiense disease foci , who exhibit a long-term T . b . gambiense-specific serological response but low or undetectable parasitemia indicative of a latent asymptomatic infection ( Koffi et al . , 2006; Jamonneau et al . , 2012; Bucheton et al . , 2011; Ilboudo et al . , 2011; Jamonneau et al . , 2010; Garcia et al . , 2000 ) . Parasites from such individuals appear genetically indistinguishable from those of T . b . gambiense clinical cases ( Kaboré et al . , 2011 ) , suggesting disease outcome may be mediated by , as yet unidentified , host genetic factors . Field studies are therefore warranted to fully evaluate the contribution of variants of the host protein APOL1 to HAT susceptibility . Here , we present a retrospective association study to test the relationship between APOL1 G1 and G2 variants and susceptibility to the two different forms of human African trypanosomiasis , T . b . rhodesiense in East Africa and T . b . gambiense in West Africa . In Uganda , an association analysis was performed between T . b . rhodesiense-infected individuals and controls in a major disease focus . In the principal T . b . gambiense focus in Guinea , the presence of both clinical patients and asymptomatic individuals permitted a two-stage analysis . Firstly , the association between APOL1 variants and susceptibility to T . b . gambiense infection ( infected versus controls ) , and secondly the association with disease outcome following infection ( clinical cases versus asymptomatic carriage ) . We report that the association of APOL1 chronic kidney disease variants with HAT susceptibility are markedly different for the two subspecies . As hypothesised , a dominant protective association was detected for the G2 variant against T . b . rhodesiense infection . Conversely , we found that the APOL1 G1 variant was not associated with resistance to T . b . rhodesiense infection , but with protective asymptomatic carriage of T . b . gambiense . We consider the implications of these strikingly different susceptibilities in the context of human co-evolution with African trypanosomes and the distribution , selection and persistence of these kidney disease risk variants in sub-Saharan Africa . To test the heterozygous advantage hypothesis proposed for these APOL1 variants against T . b . rhodesiense infection ( Genovese et al . , 2010 ) , 180 controls and 184 clinically confirmed T . b . rhodesiense patients from a principle disease focus in central-eastern Uganda were genotyped for G1 and G2 polymorphisms . The G1 haplotype comprises of two non-synonymous substitutions , rs73885319 and rs60910145 situated just 128 bp apart and in near-perfect linkage disequilibrium ( Genovese et al . , 2010; Kopp et al . , 2011 ) . In this study , as reported by others ( Kopp et al . , 2011; Behar et al . , 2011 ) , a small number of individuals were identified with only a partial G1 haplotype ( the kidney disease risk genotype at one of the G1 polymorphism positions but the non-risk genotype at the other ) and were excluded from the G1 haplotype association analysis . The second chronic kidney disease risk variant , G2 ( rs71785313 ) , is found on an alternative haplotype and represents a six base pair in-frame deletion . Comparing genotype frequencies in confirmed T . b . rhodesiense-infected individuals with uninfected controls found no association between the G1 haplotype and T . b . rhodesiense infection ( p=0 . 50; Table 1 ) . In contrast , we observed a significant dominant protective association for the G2 variant , with an odds ratio of 0 . 20 ( 95% CI: 0 . 07 to 0 . 48 , p=0 . 0001; Table 1 ) . This indicates a five-fold reduced susceptibility to T . b . rhodesiense infection for individuals that possess a single copy of the G2 variant , compatible with a model of heterozygous protection . 10 . 7554/eLife . 25461 . 004Table 1 . Association between APOL1 kidney disease risk variants and T . b . rhodesiense infectionDOI: http://dx . doi . org/10 . 7554/eLife . 25461 . 00410 . 7554/eLife . 25461 . 005Table 1—source data 1 . APOL1 genotype data for T . b . rhodesiense-infected individuals and controlsIndividuals excluded from the APOL1 G1 association analysis . T . b . r: T . b . rhodesiense , G0: genotype compatible with the non-risk G0 allele for both rs73885319 and rs60910145 , G1: genotype compatible with the G1 CKD risk allele for both rs73885319 and rs60910145 , G1M: genotype compatible with the G1 CKD risk allele for rs60910145 and the non-risk G0 allele for rs73885319 , G1G: genotype compatible with the G1 CKD risk allele for rs73885319 and the non-risk G0 allele for rs60910145 , G2: genotype compatible with the G2 CKD risk allele for rs71785313 . DOI: http://dx . doi . org/10 . 7554/eLife . 25461 . 00510 . 7554/eLife . 25461 . 006Table 1—source data 2 . Association between individual APOL1 G1 kidney disease risk variants and T . b . rhodesiense infection Two-tailed Fisher's exact test with mid-P method using a dominant genetic model ( carriage of 1 or 2 copies of the designated APOL1 SNP ) . CKD: chronic kidney disease , T . b . r: T . b . rhodesiense , OR: odds ratio , CI: confidence interval . Raw data for Table 1—source data 2 can be found in Table 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25461 . 006Dominant model - InfectionAPOL1 haplotypeT . b . r infectedControlAssociation analysis T . b . r infected/ControlNumber%Number%OR [95% CI]PG0 Ancestral Haplotype rs73885319 ( A ) + rs60910145 ( T ) + rs71785313 ( TTATAA ) G0184100 . 017999 . 4N . C0 . 49Non-G000 . 010 . 6Total184100 . 0180100 . 0G1 Haplotype† rs73885319 ( A>G ) + rs60910145 ( T>G ) G194 . 9126 . 70 . 73 [0 . 29 to 1 . 79]0 . 50Non-G117395 . 116893 . 3Total182100 . 0180100 . 0G2 Haplotype rs71785313 ( TTATAA>del6 ) G263 . 32614 . 40 . 20 [0 . 07 to 0 . 48]0 . 0001Non-G217896 . 715485 . 6Total184100 . 0180100 . 0*Two-tailed Fisher's exact test with mid-P method using a dominant genetic model ( carriage of 1 or 2 copies of the designated APOL1 haplotype ) , †Individuals with only a partial G1 haplotype were excluded from the analysis . T . b . r: T . b . rhodesiense , OR: odds ratio , CI: confidence interval , N . C: not calculable . All raw data for Table 1 can be found in Table 1—source data 1 . The association analysis of the two individual component SNPs of the G1 haplotype can be found in Table 1—source data 2 . To visualize the geographic distribution of APOL1 variants in relation to HAT endemicity ( Figure 2B ) , data generated by this study were merged with previously reported allele frequencies for 38 other sub-Saharan African populations , to produce a cohort of 5287 genotyped samples . Frequency distributions for G1 and G2 were transformed into geographical contour maps using the Kriging algorithm for data interpolation ( Figure 2C and D ) . The allele frequencies from the Ugandan population and the mangrove foci in Guinea appear consistent with the general geographical distribution pattern for these variants in sub-Saharan Africa . Both variants are reported at higher prevalence in T . b . gambiense endemic West Africa , particularly G1 , which reaches frequencies as high as 49% in the Ibo ( Thomson et al . , 2014 ) and Esan ( Abecasis et al . , 2012 ) tribes of Nigeria , decreasing to complete absence in Northeast Africa ( Tzur et al . , 2010; Behar et al . , 2011 ) . Allele frequency is moderately inversely correlated with longitude for both G1 ( Pearson correlation: r = −0 . 526 , p=2 . 0 × 10−4 , N = 40 ) and G2 ( r = −0 . 593 , p=5 . 7 × 10−5 , N = 37 ) but not latitude ( G1 , p=0 . 33 , G2 , p=0 . 30 ) , indicating a significant decreasing relative frequency for both APOL1 variants from West to East across the continent . Here , we report the first association study between APOL1 G1 and G2 kidney disease risk variants and T . b . rhodesiense and T . b . gambiense , revealing a more complex relationship with human African trypanosomiasis susceptibility than was originally predicted . The implications of these findings in relation to each of the human-infective parasite subspecies are considered in turn . The zoonotic T . b . rhodesiense parasite has been responsible for a number of severe HAT outbreaks in recent human history in East Africa that have claimed hundreds of thousands of lives ( Hide , 1999; Fèvre et al . , 2004 ) . Our data indicate that the APOL1 G2 variant was strongly associated with protection against T . b . rhodesiense infection in a Ugandan disease focus . The observed five-fold reduced susceptibility for individuals possessing a single copy of the APOL1 G2 variant is consistent with laboratory studies reporting in vitro lysis of T . b . rhodesiense for APOL1 G2 plasma and recombinant protein ( Genovese et al . , 2010 ) , and increased survival of APOL1 G2 transgenic mice in a T . b . rhodesiense infection model ( Thomson et al . , 2014 ) . T . b . rhodesiense parasites are defined by the potential to express the serum-resistance-associated ( SRA ) protein ( Xong et al . , 1998; De Greef and Hamers , 1994 ) which binds to ancestral APOL1 ( G0 ) , inhibiting its formation of lethal pores in trypanosome membranes ( Vanhamme et al . , 2003; Pérez-Morga et al . , 2005; Molina-Portela et al . , 2005; Thomson and Finkelstein , 2015; Vanwalleghem et al . , 2015 ) . The two-amino acid deletion that characterises the G2 haplotype ( rs71785313 , [p . N388_Y389del] ) , is situated within a C-terminal region of APOL1 demonstrated to be essential for SRA binding ( Lecordier et al . , 2009 ) ( residues 370–392; Figure 1 ) . Studies indicate that G2 shifts the position of a critical lysine residue within the binding region that virtually abolishes the interaction with SRA ( Genovese et al . , 2010; Thomson et al . , 2014 ) . This implicates evasion of the SRA virulence protein as the probable mechanism by which G2 restores APOL1 lytic function and protects the host against T . b . rhodesiense infection . The results of this case-control study add substantial support to the proposed heterozygous advantage model of dominant protection against T . b . rhodesiense infection for this recessive kidney disease risk variant . For the G1 variant , no protective association against T . b . rhodesiense infection was detected . This finding is somewhat at odds with the reported moderate in vitro T . b . rhodesiense lytic activity for G1 donor plasma and recombinant protein ( Genovese et al . , 2010 ) , and delayed parasitemia in an APOL1 G1 mouse model ( Thomson et al . , 2014 ) . Notably however , the trypanolytic effect for G1 in both studies was significantly inferior to the G2 variant by up to several orders of magnitude . The G1 haplotype is composed of two closely positioned missense mutations ( rs73885319; [p . S342G] and rs60910145; [p . I384M] , Figure 1 ) , of which the latter is also located in the crucial SRA-binding region . However , the rs60910145 point mutation results in an isoleucine to methionine substitution that only slightly weakens SRA-APOL1 interaction ( Genovese et al . , 2010 ) and this substitution alone did not extend survival in a mouse model ( Thomson et al . , 2014 ) . Together these data suggest that the G1 variant is not able to confer substantial protection from T . b . rhodesiense infection . However , our data do not preclude an effect of G1 on the time course or severity of T . b . rhodesiense disease . Additionally , for both the G1 variant and the small number of G2-possessing individuals that were infected with T . b . rhodesiense it is possible that dosage effects or inactivating mutations may be present which have abrogated the trypanolytic ability of these APOL1 variants . No stop codons or mutual non-synonymous variants were observed within the exomes of these individuals during sequence verification of the APOL1 G1 and G2 genotypes ( Supplementary file 1 ) , but this possibility cannot be definitively excluded . In contrast to T . b . rhodesiense , T . b . gambiense primarily infects humans , and is the pathogen responsible for the majority of human disease ( Simarro et al . , 2010 ) and significant and widespread epidemics of a slower progressing form of sleeping sickness in Central and West Africa . No lytic ability for either APOL1 variant was reported against T . b . gambiense tested by in vitro assays with donor plasma or recombinant protein ( Genovese et al . , 2010 ) . Consistent with this observation , in the Guinea focus neither variant demonstrated a resistance association with T . b . gambiense infection . Instead , as summarized in Figure 2A , contrasting associations were observed with infection outcome . APOL1 G1 was associated with a predisposition to latent asymptomatic carriage , while individuals possessing the G2 variant were more likely to progress to clinical disease . The association between APOL1 variants and infection outcome for T . b . gambiense implicates this molecule as a critical modulating factor in disease control . APOL1 is a high-density lipoprotein-associated serum protein , the expression of which is up-regulated by pro-inflammatory stimuli including IFN-γ ( Sana et al . , 2005 ) and TNF-α ( Monajemi et al . , 2002 ) . In accordance with this , APOL1 expression is demonstrably increased during T . b . gambiense infection ( Ilboudo et al . , 2012 ) . However , no association has been observed between APOL1 expression levels and blood parasite density or clinical outcome of T . b . gambiense infection ( Ilboudo et al . , 2012 ) . Instead , the results of our study indicate that particular APOL1 variants , rather than modulation of global APOL1 protein level , contribute to differential susceptibility to disease . How these variants influence T . b . gambiense is less perceptible than for T . b . rhodesiense . The mechanism of T . b . gambiense APOL1 resistance does not involve SRA , but three independent contributing components have been implicated: a sub-species-specific protein , TgsGP ( Capewell et al . , 2013; Uzureau et al . , 2013 ) , which alters trypanosome membrane rigidity ( Uzureau et al . , 2013 ) ; reduced uptake of APOL1 ( DeJesus et al . , 2013 ; Kieft et al . , 2010 ) ; and proposed faster degradation of APOL1 within the endocytic system of the parasite ( Uzureau et al . , 2013; Alsford et al . , 2014 ) . It is possible that alterations to the APOL1 molecule conferred by G1 and G2 polymorphisms affects one or more of these processes with opposing downstream consequences . Furthermore , the strengthened risk association of G2 with clinical disease when individuals who possess both haplotypes ( G1/G2 compound heterozygotes ) were excluded indicates a potential dominance for the protective G1 haplotype that might be able to mitigate the disease progressive effects of the G2 variant . Despite its critical function in human innate resistance to most trypanosomes , the role of APOL1 in T . b . gambiense disease progression appears complex . Contrasting inflammatory cytokine profiles reported between individuals with clinical stage disease and latent carriers ( Ilboudo et al . , 2014 ) suggests that an intricate multi-gene interplay between host immune factors , APOL1 , and the parasite ultimately determines disease outcome for this subspecies . For the G1 variant , the relationship with T . b . gambiense appears more akin to the well- established association between Plasmodium and the sickle haemoglobin S ( HbS ) polymorphism ( Allison , 1954 ) . In this classic example of heterozygous advantage , the heterozygous HbS genotype does not protect from Plasmodium infection per se but reduces the risk of severe malaria once infected ( Allison , 1954; Taylor et al . , 2012 ) . This advantage has selected and maintained prevalence of the HbS polymorphism in malaria-endemic sub-Saharan Africa , despite the high penetrance of life-threatening sickle cell disease in homozygotes ( Allison , 1954; Piel et al . , 2010 ) . In T . b . gambiense , possession of a G1 allele is associated with the capacity to sustain the asymptomatic latent period of what is normally a fatal disease . This moderation of disease severity could plausibly confer greater survival and reproductive opportunities for individuals possessing the G1 variant than for their G0- or G2-carrying counterparts , who typically progress more rapidly to severe disease ( G2 > G0 > G1 ) . Such a selection advantage may explain the high allele frequency of G1 recorded in T . b . gambiense-endemic West Africa ( up to 49% ( Thomson et al . , 2014; Abecasis et al . , 2012 ) ;Figure 2C ) , which in some populations exceeds even the maximum global HbS alleles frequencies ( Piel et al . , 2010 ) . This is consistent with a strong positive selective force on G1 ( Genovese et al . , 2010 ) , and conceivably , a less powerful opposing deleterious pressure from kidney disease in homozygotes , which is typically of late onset , and incompletely penetrant ( Kruzel-Davila et al . , 2016 ) . Population genetics studies of T . brucei indicates that both human-infective sub-species likely arose independently and relatively recently from the animal pathogen T . b . brucei ( Tait et al . , 1985; Gibson et al . , 2002; Balmer et al . , 2011; Weir et al . , 2016 ) . Molecular clock analysis dates the emergence of T . b . gambiense as a human pathogen in West Africa from a single progenitor approximately 1 , 000–10 , 000 years ago ( Weir et al . , 2016 ) . During this time a pivotal lifestyle transition was occurring with the development of agriculture and larger , more densely populated permanent settlements that provided favorable conditions for the emergence of many animal-derived human pathogens ( Harper and Armelagos , 2010; Wolfe et al . , 2007 ) . This also coincides with the timeline for a robust selective sweep on G1 detected in the Nigerian Yoruba population ( Genovese et al . , 2010 ) , at the geographical hotspot for this allele ( Figure 2C ) in West Africa . A plausible scenario is that within the last 10 , 000 years an animal-infective T . b . brucei predecessor of T . b . gambiense evolved the essential human serum resistance gene TgsGP ( Capewell et al . , 2013 ) , facilitating its transmission to humans in the ancestral Bantu population of the Nigeria-Cameroon region . Over time , as T . b . gambiense has undergone progressive adaptation into a predominantly human pathogen ( Wolfe et al . , 2007 ) , selection for the human APOL1 G1 variant may have occurred in turn , which was able to mitigate the lethal progression of disease and promote long-term asymptomatic carriage . The T . b . gambiense-protective APOL1 G1 haplotype could then have spread with human migration and introgression into other sub-Saharan populations during the Bantu expansions ( Tishkoff et al . , 2009 ) , or along commercial routes within the last 4000 years , to reach its current distribution across sub-Saharan Africa ( Figure 2C ) . For APOL1 G2 , the increased risk of clinical T . b . gambiense disease contrasts with the strong protective association observed for this variant against T . b . rhodesiense . Puzzlingly , as for G1 , some of the highest frequencies of G2 are also found in T . b . gambiense-endemic West Africa ( Figure 2D ) , raising speculation about the evolutionary history of these two variants . Studies of genetic diversity at the APOL1 locus are consistent with a older ( 2 , 000–7 , 000 years ) , less intensively selected allele ( Genovese et al . , 2010; Pinto et al . , 2016 ) for G2 , and a more recent , rapid sweep for the G1 allele in West Africa ( Genovese et al . , 2010; Pinto et al . , 2016; Limou et al . , 2015 ) . One possible interpretation of the available data is that T . b . rhodesiense preceded T . b . gambiense in West Africa and was responsible for driving positive selection of the G2 variant in the Nigeria-Cameroon region . Rising frequencies of this protective variant ( or other unrelated epidemiological factors ) could have then forced an eastward shift in T . b . rhodesiense endemicity to an approximation of its current distribution in East and Southern Africa . Subsequently , when T . b . gambiense emerged in West Africa , the relative fitness of the APOL1 G2 allele in the exposed population would have been diminished , providing an opportunity for the robust selective sweep of an alternate APOL1 variant , G1 , which was able to reduce the disease severity of this new pathogen . While this is an attractive theory , there is little epidemiological support for a shift in the endemicity of T . b . rhodesiense , which has only been detected in East Africa and has no recorded history in West Africa ( Gibson et al . , 2002; Radwanska et al . , 2002; Balyeidhusa et al . , 2012; Picozzi et al . , 2005 ) . Moreover , isolates of T . b . rhodesiense from across East Africa show a strong genetic relationship with sympatric T . b . brucei strains , compatible with a predominantly East African origin ( Balmer et al . , 2011; Godfrey et al . , 1990; MacLeod et al . , 2001 ) . An alternative model is that selection in favour of the G2 variant may have originated from a different source in West Africa , and it is only more recently , as the G2 variant spread eastwards with the Bantu expansion ( Tishkoff et al . , 2009 ) , that it has fortuitously proved advantageous against T . b . rhodesiense . Indeed , beyond its proven capacity for trypanolysis , APOL1 was shown to limit Leishmania major infections in mice ( Samanovic et al . , 2009 ) and suppress HIV-replication in macrophages ( Taylor et al . , 2014 ) , hinting at a much broader role for APOL1 in innate immunity to infectious disease . The association between APOL1 chronic kidney disease risk variants and human African trypanosomiasis reveals a more complex picture of selection and human evolution than was originally hypothesized . Despite their close genetic proximity APOL1 G1 and G2 polymorphisms confer very different , and even opposing , dominant associations with human African trypanosomiasis susceptibility , yet appear convergent in their deleterious recessive contribution to kidney pathology . While the origins of the G2 allele remain speculative , a model of dominant protection against T . b . rhodesiense infection is supported . For G1 , the strong West African allele distribution bias and evidence for recent , rapid , positive selection , suggest an alternative evolutionary ancestry for this allele , which we propose involves protection from the lethal consequences of the T . b . gambiense parasite . Participants were identified through healthcare providers , community engagement and active field surveillance in association with the national control programmes . Written informed consent for sample collection , analysis and publication of anonymised data was obtained from all participants by trained local healthcare workers . Subjects or their legal guardian gave consent as a signature or a thumbprint after receiving standardised information in English , French or their local language , as preferred , and were free to withdraw from the study at any time . Efforts were made to ensure the engagement of all local stake holders and approval was obtained from local leaders in each study area where appropriate . Ethical approvals for the study were obtained from within the TrypanoGEN Project following H3Africa Consortium guidelines for informed consent ( H3Africa Consortium , 2013 ) , from Comité Consultatif de Déontologie et d’Ethique ( CCDE ) at the Institut de recherche pour le développement ( IRD; 10/06/2013 ) for the Guinea study , and from the Uganda National Council for Science and Technology ( UNCST; 21/03/2013 ) for the Uganda study . Research procedures were also approved by the University of Glasgow MVLS Ethics Committee for Non-Clinical Research Involving Human Subjects ( Reference no . 200120043 ) . Statistical analyses of association between APOL1 genotypes and human African trypanosomiasis in this case-control study were performed by contingency table analyses using Fisher’s exact test with mid-P method . Statistical tests were computed using Open-epi . Calculation of the minimum detectable odds ratios was performed for the study sample size in Uganda ( <0 . 144 , >2 . 662 [G1] , < 0 . 350 , >2 . 116 [G2] ) and Guinea ( <0 . 448 , >2 . 008 [G1] , <0 . 471 , >1 . 971[G2] ) using Sampsize software with the parameters of 80% power , 5% alpha risk and a two-sided test . To visualize the geographical distribution of APOL1 G1 and G2 polymorphisms in sub-Saharan Africa , a contour map was generated by collating data from this study with previously published datasets to produce a cohort of 5287 individuals from across 40 African populations . Published datasets with a low sample size ( n ≤ 19 ) were excluded . G1S342G ( rs73885319 ) was used as a proxy for G1 , where G1 frequency data were unavailable ( rs73885319 and rs60910145 are in almost complete positive linkage disequilibrium ) ( Genovese et al . , 2010; Kopp et al . , 2011 ) . The contour map was drawn using Surfer 8 . 0 ( Golden Software Inc . , Golden , Colorado ) applying the Kriging algorithm for data interpolation . Interpolation may be inaccurate where there are few data points . G1 and G2 allele frequencies were analysed for an association with the geographical coordinates ( absolute latitude and longitude ) using Pearson's correlation test ( GraphPad Prism version 6 . 0 , RRID:SCR_002798 ) . The map of T . b . rhodesiense and T . b . gambiense endemicity was drawn from the Human African trypanosomiasis endemicity classification of the Global Health Observatory data repository ( World Health Organization , 2015 )	African-Americans have a greater risk of developing chronic kidney disease than Americans with European ancestry . Much of this increased risk is explained by two versions of a gene called APOL1 that are common in people with African ancestry . These two versions of the gene , known as G1 and G2 , suddenly became much more common in people in sub-Saharan Africa in the last 10 , 000 years . One theory for their rapid spread is that they might protect against a deadly parasitic disease known as African sleeping sickness . This disease is caused by two related parasites of a species known as Trypanosoma brucei , one of which is found in East Africa , while the other affects West Africa . Laboratory studies have shown that blood from individuals who carry the G1 and G2 variants is better at killing the East African parasites . However , it is not clear if these gene versions help people living in the rural communities , where African sleeping sickness is common , to fight off the disease . Now , Cooper , Ilboudo et al . show that G1 and G2 do indeed influence how susceptible individuals in these communities are to African sleeping sickness . Individuals with the G2 version were five-times less likely to get the disease from the East African parasite . Neither version could protect individuals from infection with the West African parasite , but infected individuals with the G1 version had fewer parasites in their blood and were less likely to become severely ill . The ability of the G1 version to control the disease and prolong life could explain why this gene version has become so common amongst people in West Africa . Unexpectedly , the experiments also revealed that people with the G2 version were more likely to become severely unwell when they were infected by the West African parasite . This indicates that whether this gene variant is helpful or harmful depends on where an individual lives . The next step following on from this work will be to investigate exactly how the G1 version reduces the severity of the West African disease . This may aid the development of new drugs for African sleeping sickness and kidney disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "genetics", "and", "genomics" ]	2017	APOL1 renal risk variants have contrasting resistance and susceptibility associations with African trypanosomiasis
Aging is a major risk factor in many forms of late-onset neurodegenerative disorders . The ability to recapitulate age-related characteristics of human neurons in culture will offer unprecedented opportunities to study the biological processes underlying neuronal aging . Here , we show that using a recently demonstrated microRNA-based cellular reprogramming approach , human fibroblasts from postnatal to near centenarian donors can be efficiently converted into neurons that maintain multiple age-associated signatures . Application of an epigenetic biomarker of aging ( referred to as epigenetic clock ) to DNA methylation data revealed that the epigenetic ages of fibroblasts were highly correlated with corresponding age estimates of reprogrammed neurons . Transcriptome and microRNA profiles reveal genes differentially expressed between young and old neurons . Further analyses of oxidative stress , DNA damage and telomere length exhibit the retention of age-associated cellular properties in converted neurons from corresponding fibroblasts . Our results collectively demonstrate the maintenance of age after neuronal conversion . Increasing evidence suggests that in addition to genetic susceptibility , age-related neurodegeneration may be caused in part by cellular aging processes that result in accumulation of damaged DNA and proteins in neurons ( Mattson and Magnus , 2006 ) . However , because of the inaccessibility to neurons from elderly individuals , studying these age-related cellular processes in human neurons remains a difficult task . Cellular reprogramming approaches have explored generating populations of human neurons by inducing pluripotent stem cells ( iPSC ) from human fibroblasts and subsequent differentiation into neurons ( Takahashi and Yamanaka , 2006; Takahashi et al . , 2007; Hanna et al . , 2008; Hu et al . , 2010 ) . Importantly , this induction of pluripotency in adult fibroblasts reverts cellular age to an embryonic stage ( Lapasset et al . , 2011; Patterson et al . , 2012 ) which remains even after differentiation into neurons ( Patterson et al . , 2012; Miller et al . , 2013 ) . While this is useful for modeling early developmental phenotypes ( Lafaille et al . , 2012; Lee et al . , 2009 ) , iPSC-derived cells have been reported to be unsuitable in recapitulating phenotypes selectively observed in aged cells ( Mattson and Magnus , 2006; Vera and Studer , 2015 ) . Recently , experimental manipulations to accelerate aging in iPSC-derived cells have been explored , for instance , by overexpressing progerin , a mutant form of lamin A observed in progeria syndrome , to force the detection of age-related pathophysiology of neurodegenerative disease ( Arbab et al . , 2014; Cornacchia and Studer , 2015 ) . Alternatively , we previously described a reprogramming paradigm using neuronal microRNAs ( miRNAs ) , miR-9/9* and miR-124 ( miR-9/9-124 ) , that exert reprogramming activities to directly convert human fibroblasts to specific mature neuronal subtypes ( Richner et al . , 2015; Victor et al . , 2014; Yoo et al . , 2011 ) . Because this neuronal conversion is direct and bypasses pluripotent/multipotent stem cell stages , we reasoned that miR-mediated directly reprogrammed neurons would retain the age signature of the original donor . To assess the cellular age , we used the epigenetic clock method , which is a highly accurate biomarker of age based on DNA methylation ( Horvath , 2013 ) . Further , we evaluated age-associated signatures based on gene expression levels , miRNAs , and cellular readouts considered to be hallmarks of aging ( López-Otín et al . , 2013 ) . Our thorough investigation into multiple age-associated signatures collectively demonstrate the maintenance of cellular age of the original donor during neuronal conversion and strongly suggest that directly converted human neurons can be advantageous for studying age-related neuronal disorders . Establishing robust reprogramming efficiency is essential prior to assessing age-related phenotypes in reprogrammed neurons . We therefore elected to test our recently developed conversion approach that utilizes miR-9/9-124 and transcription factors to robustly generate a highly enriched population of striatal medium spiny neurons ( MSNs ) from the fibroblasts of donors of varying ages ( Richner et al . , 2015; Victor et al . , 2014 ) . Fibroblast samples from donors ranging from 3 days to 96 years of age were collected and expanded to match population doubling level ( PDL ) to eliminate any confounding variability introduced by sequential passaging ( Campisi and d'Adda di Fagagna , 2007; Pazolli and Stewart , 2008 ) , then subsequently transduced with miR-9/9-124 with CTIP2 , DLX1/2 , and MYT1L ( CDM ) following our established protocol ( Richner et al . , 2015; Victor et al . , 2014 ) ( Figure 1A ) . Reprogrammed cells were then stained for neuronal markers , MAP2 , TUBB3 , NeuN and MSN markers , DARPP32 and GABA ( Figure 1B–C ) . MAP2 and TUBB3-positive reprogrammed neurons exhibiting extensive neurite outgrowth represented approximately 70–80% of the cell population ( Figure 1D ) , demonstrating the consistency of reprogramming efficiency in all fibroblast samples . Neurons reprogrammed from young and old fibroblasts exhibited fast inward currents and action potentials in monocultures without necessitating coculturing with glial cells or primary neurons ( n = 11 ) ( Figure 1E ) . In addition , the consistent upregulation of neuronal genes , including MAP2 , NCAM , and voltage-gated sodium channels , and downregulation of fibroblast-associated genes were observed in reprogrammed neurons from both young and old cells ( Figure 1—figure supplement 1 ) . These results suggest the consistent applicability of miRNA-based neuronal reprogramming in fibroblasts of all ages . 10 . 7554/eLife . 18648 . 003Figure 1 . MicroRNA-mediated direct neuronal conversion applied to fibroblasts across the age spectrum . ( A ) Schematic diagram of neuronal conversion of human fibroblast samples from individuals ranging from three days to 96 years of age . Primary fibroblasts were transduced with microRNA-9/9-124 and the transcription factor cocktail CTIP2 , DLX1/2 , MYT1L ( CDM ) and analyzed for neuronal properties after 30 days . ( B ) Expression of pan-neuronal markers MAP2 ( top ) and TUBB3 ( bottom ) after neuronal conversion of human fibroblasts ranging in age . Scale bar = 50 µm . ( C ) Expression of pan-neuronal marker NeuN and medium spiny neuron-specific markers GABA and DARPP32 in reprogrammed neurons from fibroblasts aged 91- , 72- , 92-years respectively . ( D ) Immunostaining analysis of percentage of reprogrammed neurons positive for neuronal markers TUBB3 and MAP2 over DAPI signals ( n = 200–300 per cell line ) . ( E ) Representative whole-cell current clamp recording of converted neurons from young ( 29 years old , left ) and old ( 94 year old donor , right ) donors . Converted human neurons in monoculture displayed multiple action potentials in response to step current injections at four weeks post-transduction . All reprogrammed neurons from old fibroblasts recorded ( n = 11 ) fired APs in response to current injections ( top ) . Representative traces of fast-inactivating inward currents recorded in voltage-clamp mode . Voltage steps ranged from +10 to +70 mV ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 00310 . 7554/eLife . 18648 . 004Figure 1—figure supplement 1 . Transcriptome analyses between converted neurons and fibroblasts of young and old age groups . ( A ) Volcano plots show global transcriptomic differences between reprogrammed neurons compared to starting human fibroblasts from young ( three day , five month and one year-old , top ) and old ( 90 , 92 and 92 year-old , bottom ) donors . Upregulated DEGs ( in red ) and downregulated DEGs ( in blue ) after direct neuronal reprogramming are shown . ( B ) Heatmap of representative DEGs: upregulation of neuronal genes ( top ) and downregulation of fibroblast associated genes ( bottom ) in reprogrammed neurons from both young and old fibroblasts ( right ) when compared to corresponding fibroblasts ( left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 004 Aging largely influences the epigenetic landscape of cells ( Oberdoerffer and Sinclair , 2007 ) with a number of genomic loci becoming differentially methylated with age ( Horvath et al . , 2012; Christensen et al . , 2009 ) . The epigenetic clock , which analyzes the methylation status of 353 specific CpG loci , has been shown to be a highly accurate age estimator that applies to all human organs , tissues , and cell types ( Horvath , 2013 ) . The epigenetic clock leads to an age estimate ( in units of years ) which is referred to as epigenetic age or DNA methylation ( DNAm ) age . We analyzed DNA methylation levels of neurons converted from 16 fibroblast samples aged three days to 96 years alongside 37 fibroblast samples aged three days to 94 years ( Figure 2A ) . The actual chronological donor age was highly correlated with the estimated DNAm age of fibroblasts ( correlation = 0 . 75 ) ( Figure 2B ) and of reprogrammed neurons ( correlation = 0 . 82 ) ( Figure 2B ) . Importantly , when the DNAm age of each reprogrammed neuron was compared to the DNAm age of the corresponding fibroblast , there was a near-perfect correlation ( correlation = 0 . 91 ) , which suggests that the epigenetic clock is unperturbed during miRNA-based neuronal reprogramming ( Figure 2C ) and supports the notion of age maintenance during direct neuronal conversion . By contrast , it is known that iPSC generation resets the epigenetic clock to an embryonic state since iPSCs have a DNAm age that is negative or close to zero ( Horvath , 2013 ) . 10 . 7554/eLife . 18648 . 005Figure 2 . Conservation of the epigenetic clock of reprogrammed neurons from human fibroblasts . ( A ) Schematic diagram representing the hypothesis that the epigenetic clock of fibroblasts from different age groups is conserved in reprogrammed neurons after miR-mediated neuronal conversion . ( B ) Top: Predicted ages based on the methylation status ( DNAm age ) of fibroblasts plotted against the actual ages of fibroblasts ( correlation = 0 . 75 , p=9 . 1e-08 ) . Middle: Predicted DNAm ages of reprogrammed neurons against actual ages of starting fibroblast , correlation = 0 . 82 , p=1e-04 . Bottom: Combined plot of DNAm ages of fibroblasts and DNAm ages of reprogrammed neurons against actual ages ( correlation = 0 . 77 , p=1 . 6e-11 ) . ( C ) DNAm age of reprogrammed neurons plotted against the DNAm ages of the corresponding , starting fibroblast ages , correlation = 0 . 91 p=2 . 5e-06 . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 00510 . 7554/eLife . 18648 . 006Figure 2—source data 1 . Output for sample information and DNAm ages for fibroblasts and reprogrammed neurons compared to original age . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 006 Given the broad variability in gene expression with age in multiple cell types ( Berchtold et al . , 2008; Lu et al . , 2004; Fraser et al . , 2005; Glass et al . , 2013 ) and a recent demonstration of maintenance of age-associated transcriptomic changes in neurons directly converted by transcription factors ( Mertens et al . , 2015 ) , we sought to determine whether age-associated transcriptomic changes could be detected after miR-9/9-124-CDM-based neuronal conversion . The transcriptome and microRNA profiles of reprogrammed neurons from both young and old fibroblasts were analyzed alongside corresponding fibroblasts . Principal component analysis ( PCA ) of transcriptome revealed cell type-specific clustering of reprogrammed neurons versus fibroblasts , while age-associated segregation is observed in both fibroblasts and reprogrammed neurons ( Figure 3A ) . A cohort of upregulated and downregulated genes with aging was commonly observed in both fibroblasts and converted neurons ( Figure 3B ) , consistent with a previous report ( Mertens et al . , 2015 ) . Gene ontology ( GO ) analysis of differentially expressed genes with age in reprogrammed neurons is enriched for terms associated with age-related biological processes ( Figure 3—figure supplement 1 ) , including vesicle-mediated transport ( Wilmot et al . , 2008 ) , nervous system development ( Lu et al . , 2004 ) NF-kappaB transcription factor activity ( Tilstra et al . , 2011 ) , regulation of apoptosis and inflammatory response ( de Magalhães et al . , 2009 ) , and for genes previously identified to be associated with age in the human brain ( Lu et al . , 2004 ) . 10 . 7554/eLife . 18648 . 007Figure 3 . Age-associated changes in transcriptome and microRNA profiles in reprogrammed neurons . ( A ) Principle component analysis ( PCA ) of transcriptome profiling of reprogrammed neurons from young fibroblasts aged three days , five months one year ( green ) and from old fibroblasts aged 90 , 92 , and 92 years ( blue ) alongside corresponding young fibroblasts ( black ) and old fibroblasts ( red ) ( FDR < 0 . 05 ) . ( B ) Differentially expressed genes ( DEGs ) with age in fibroblasts ( x axis ) and in reprogrammed neurons ( y axis ) . Age-associated DEGs in reprogrammed neurons shown in green , DEGs in fibroblasts shown in red , and commonly shared DEGs with age in both fibroblasts and reprogrammed neurons shown in blue . ( C ) PCA of miRNA profile in reprogrammed neurons from young fibroblasts aged three days , five months one year ( green ) and from old fibroblasts aged 90 , 92 , and 92 years ( bue ) alongside the corresponding young fibroblasts ( black ) and old fibroblasts ( red ) ( FDR < 0 . 05 ) . ( D ) MicroRNAs that are differentially regulated with age in both fibroblasts ( red ) and reprogrammed neurons ( blue ) . Four microRNAs , miR-10a , miR-497 , miR-10b , and miR-195 , are upregulated with age , while 10 microRNAs are shown to be downregulated with age , p<0 . 05 . ( E ) Validation of expression of miRNA expression upregulated with aging ( miR-10a , miR-497 , miR-195 ) in reprogrammed neurons from old fibroblasts over reprogrammed neurons from young fibroblasts ( left ) . Validation of expression changes of miR-10a , miR-497 , miR-195 in human striatum slices ( middle ) and human cortex slices ( right ) from old individuals compared to young individuals . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 00710 . 7554/eLife . 18648 . 008Figure 3—source data 1 . Raw data for qPCR for microRNA expression analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 00810 . 7554/eLife . 18648 . 009Figure 3—source data 2 . Full GO terms for age-regulated genes in reprogrammed neurons and for predicted targets of miR-10a-5p and miR-497-5p . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 00910 . 7554/eLife . 18648 . 010Figure 3—figure supplement 1 . Gene ontology of DEGs in reprogrammed neurons . ( A ) GO analysis of age-downregulated transcripts ( top ) and upregulated transcipts ( bottom ) in reprogrammed neurons , log fc > 1 , FDR < 0 . 01 using BiNGO from Cytoscape version 3 . 3 . 0 . ( FDR 0 . 05 ) . GO terms related to aging-associated processes are represented here . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 01010 . 7554/eLife . 18648 . 011Figure 3—figure supplement 2 . Gene ontology of predicted targets of age-resulted microRNAs . ( A ) GO analysis of list of age-downregulated transcripts predicted to be targeted by upregulated microRNAs , miR-10a-5p ( top ) and miR-497-5p ( bottom ) . A list of original downregulated transcripts were filtered by fc<-1 . 5 , FDR < 0 . 05 . Target prediction analysis with Ingenuity Pathway Analysis , MicroRNA-mRNA Interactions . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 011 MicroRNA profiling similarly revealed distinct sample segregation both with age ( young versus old ) and cell type ( fibroblasts versus reprogrammed neurons ) ( Figure 3C ) . Interestingly , we detected fourteen microRNAs commonly regulated with age in both fibroblasts and reprogrammed neurons , including miR-10a , miR-497 , and miR-195 , whose expression increased with age ( Figure 3D ) . Because microRNAs have been implicated as global regulators of aging-associated cellular processes ( Liu et al . , 2012; Harries , 2014 ) through repression of existing target transcripts ( He and Hannon , 2004; Pasquinelli , 2012; Lewis et al . , 2005 ) , we reasoned that these age-upregulated microRNAs may target and repress classes of genes found to be downregulated in old reprogrammed neurons . Indeed , GO analysis of predicted targets amongst age-downregulated transcripts in reprogrammed neurons ( Figure 3B ) for miR-10a-5p and miR-497-5p revealed terms associated with age such as metabolism and cellular death and survival ( Figure 3—figure supplement 2 ) . Our results suggest the potential role of miR-10a and miR-497 in regulating genes involved in cell death and survival , metabolic pathways , and DNA repair . While the exact role of these microRNAs in aging is unknown , miR-10a and miR-497 have been previously implicated in aging-associated cellular processes including inflammation , senescence , metabolism and telomerase activity ( Kondo et al . , 2016; Qin et al . , 2012; Fang et al . , 2010 ) . Importantly , the increased expression of miR-10a , miR-497 and miR-195 detected in reprogrammed neurons was also validated by qPCR to be concordantly upregulated in human striatum and cortex samples from old individuals in comparison to young individuals ( Figure 3E ) . These results further support the validity of reprogrammed neurons for detecting age-associated changes in microRNA network , mirroring changes observed in human brain . Directly converted neurons were additionally assayed for cellular hallmarks of aging , including oxidative stress , DNA damage and telomere erosion ( López-Otín et al . , 2013 ) . Oxidative stress has been reported to increase with age , in part due to the accumulation of reactive oxygen species ( ROS ) ( Keating , 2008; Prigione et al . , 2010; Suhr et al . , 2010; Cui et al . , 2012 ) . FACS analysis of ROS levels using fluorescent marker MitoSOX ( Miller et al . , 2013 ) revealed that old reprogrammed neurons have increased ROS levels compared to young reprogrammed neurons , mirroring the observed age-associated differences of ROS levels in fibroblasts ( Figure 4A ) . Moreover , reprogrammed neurons were analyzed by Comet Assay , a single-cell gel electrophoresis technique that assesses DNA damage accumulation ( Singh et al . , 1990 ) . Old reprogrammed neurons were found to have longer comet tail lengths , an age-associated property also observed in fibroblasts , that reflects more extensive DNA damage accumulation compared to young cells ( Figure 4B ) . Additionally , neuronal conversion maintained the length of telomeres from starting fibroblast which is virtually unchanged ( Figure 4C ) , in contrast to the progressive increase in length commonly observed with iPSC reprogramming , where telomeres reach a plateau of around 12–14 kilobases after a few cellular passages ( Agarwal et al . , 2010; Batista et al . , 2011; Marion et al . , 2009 ) . Together , these cellular assays support the maintenance of aging marks in old reprogrammed neurons after direct conversion . 10 . 7554/eLife . 18648 . 012Figure 4 . Analysis of cellular biomarkers of age reveals conservation of neuron-specific aging in reprogrammed neurons . ( A ) ROS levels visualized by MitoSOX and analyzed by FACs . Representative dot plot of reprogrammed neuron from 1-year-old fibroblast ( left ) and from 91-year-old fibroblast ( right ) , y axis = FL2 channel ( MitoSOX ) , x axis = FSC . Quantification of percent of cells positive for MitoSOX reveals a significant difference in fibroblasts with age in addition to reprogrammed neurons with age . *p-value=0 . 0008; p-value=0 . 019 . ( B ) Representative images of comets indicating DNA damage from five month old fibroblasts ( top left ) versus old fibroblasts ( top right ) alongside reprogrammed neurons from 5-month-old fibroblasts ( bottom left ) and 92-year-old fibroblasts ( bottom right ) . Quantification of tail lengths in fibroblasts and reprogrammed neurons with age p-value=0 . 013 , p-value=0 . 004 . ( C ) Telomere length analyses of reprogrammed neurons are maintained from corresponding starting fibroblasts from one year old , 56 year old , and 86 year old donors . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 01210 . 7554/eLife . 18648 . 013Figure 4—source data 1 . Raw data for mitoSOX and comet assay . DOI: http://dx . doi . org/10 . 7554/eLife . 18648 . 013 Our in-depth analyses of multiple age signatures provide evidence that neuronal conversion from human somatic cells sampled at different ages generates neurons that emulate the donors’ ages . In addition to the demonstration of age-associated transcriptomic changes reported previously ( Mertens et al . , 2015 ) , our results provide novel insights into multiple key signatures associated with age— epigenetic , microRNA and cellular— that are consistently maintained in directly converted neurons . Because aging is a complex process affecting many hallmarks of a cell ( López-Otín et al . , 2013 ) , our assessment of a broad spectrum of age-related markers suggests that directly converted neurons may serve as an alternative model of neuronal aging to iPSC-derived neurons , whose erasure of multiple aging-associated signatures precludes it from adequately modeling , especially , late-onset diseases . Whereas , future studies may investigate whether additional aging marks are similarly conserved in reprogrammed neurons to model different facets of aging . While miR-9/9-124-based reprogramming can directly convert fibroblasts with similar efficiency to neurons , we note that the older fibroblasts have lower replicative potential in a culture . However , this does not impede in the conversion efficiency of old fibroblasts . MiRNA-mediated generation of aged neurons paves the road to direct conversion into specific neuronal subtypes to investigate the contribution of neuronal aging to late-onset neurodegenerative disorders . The following fibroblast cell lines ranging in age from three day old to 96 year old were obtained from the NIA Aging Cell Repository at the Coriell Institute for Medical Research , Coriell ID , RRID#: AG08498 , RRID:CVCL_1Y51; AG07095 , RRID:CVCL_0N66; AG11732 , RRID:CVCL_2E35; AG04060 , RRID:CVCL_2A45; AG04148 , RRID:CVCL_2A55; AG04349 , RRID:CVCL_2A62; AG04379 , RRID:CVCL_2A72; AG04056 , RRID:CVCL_2A43; AG04356 , RRID:CVCL_2A69; AG04057 , RRID:CVCL_2A44; AG04055 , RRID:CVCL_2A42; AG13349 , RRID:CVCL_2G05; AG13129 , RRID:CVCL_2F55; AG12788 , RRID:CVCL_L632; AG07725 , RRID:CVCL_2C46; AG04064 , RRID:CVCL_L624; AG04059 , RRID:CVCL_L623; AG09602 , RRID:CVCL_L607; AG16409 , RRID:CVCL_V978; AG06234 , RRID:CVCL_2B66; AG04062 , RRID:CVCL_2A47; AG08433 , RRID:CVCL_L625; AG16409 , RRID:CVCL_V978; GM00302 , RRID:CVCL_7277; AG01518 , RRID:CVCL_F696; AG06234 , RRID:CVCL_2B66 . We routinely check all our cell cultures and confirm it to be free of mycoplasma contamination . Authentication was completed by LINE and PCR-based techniques . The International Cell Line Authentication Committee ( ICLAC ) lists none of these primary cells are commonly misidentified cell lines . Fibroblast cell lines were cultured and expanded in DMEM media ( high glucose , Invitrogen ) supplemented with 10% or 15% fetal bovine serum ( Gibco ) , sodium pyruvate , non-essential amino acids , GlutaMAX ( Invitrogen ) , Pen/Strep solution , and Beta-mercaptoethanol . Fibroblast cell lines were expanded to a population doubling level ( PDL ) of ~19–21 . Formula to calculate PDL = 3 . 32log ( cells harvested/cells seeded ) + previous PDL . Cells were kept frozen at −150°C in the above culture medium with additional 40% FBS and 10% DMSO . Human fibroblasts ranging in age from 3 days to 96-year old were transduced with a lentiviral preparation of the Doxycline-inducible synthetic cluster of miR-9/9 and miR-124 ( miR-9/9-124 ) , alongside transcription factors CTIP2 , DLX1 , DLX2 , and MYT1L as previously described ( Richner et al . , 2015; Victor et al . , 2014 ) . Briefly , transduced fibroblasts were maintained in fibroblast media for two days with doxycycline prior to selection with Puromycin ( 3 μg/ml ) and Blasticidin ( 5 μg/ml ) at day three , then were plated onto poly-ornithine , fibronectin and laminin-coated coverslips at day five . Cells were subsequently maintained in Neuronal Media ( ScienCell , Carlsbad , CA ) supplemented with valproic acid ( 1 mM ) , dibutyryl cAMP ( 200 μM ) , BDNF ( 10 ng/ml ) , NT-3 ( 10 ng/ml ) , and RA ( 1 μM ) for 30–35 days before analysis . Reprogrammed neurons were fixed with 4% paraformaldehyde ( Electron Microscopy Sciences , Hatfield , PA ) for 20 min at room temperature ( RT ) , then permeabilized with 0 . 2% Triton X-100 for 10 min at room temperature . Cells were blocked with 1% goat serum , incubated with primary antibodies at 4°C overnight , then incubated with secondary antibodies for 1 hr at RT . Primary antibodies used for immunocytochemistry included chicken anti-MAP2 ( Abcam Cat# ab5392 RRID:AB_21381531; 1:10 , 000 dilution ) , mouse anti-β-III tubulin ( Covance Research Products Inc Cat# MMS-435P RRID:AB_2313773; 1:5000 ) , rabbit anti-β-III tubulin ( Covance Research Products Inc Cat# PRB-435P-100 RRID:AB_291637; 1:2000 ) , chicken anti-NeuN ( Aves Labs Cat# NUN RRID:AB_2313556; 1:500 ) , rabbit anti-GABA ( Sigma-Aldrich Cat# A2052 RRID:AB_477652; 1:2000 ) , mouse anti-GABA ( Sigma-Aldrich Cat# A0310 RRID:AB_476667 , 1:500 ) , and rabbit anti-DARPP32 ( Santa Cruz Biotechnology Cat# sc-11365 RRID:AB_639000; 1:400 ) . The secondary antibodies included goat anti-rabbit , mouse , or chicken IgG conjugated with Alexa-488 , −594 , or −647 ( Thermo Fisher Scientific , Waltham , MA ) . Images were captured using a Leica SP5X white light laser confocal system with Leica Application Suite ( LAS ) Advanced Fluorescence 2 . 7 . 3 . 9723 . Whole-cell patch-clamp recordings were performed at four weeks after transduction with miR-9/9-124-CDM . Intrinsic neuronal properties were studied using the following solutions ( in mM ) : Extracellular: 140 NaCl , 3 KCl , 10 Glucose , 10 HEPES , 2 CaCl2 and 1 MgCl2 ( pH adjusted to 7 . 25 with NaOH ) . Intracellular: 130 K-Gluconate , 4 NaCl , 2 MgCl2 , 1 EGTA , 10 HEPES , 2 Na2-ATP , 0 . 3 Na3-GTP , 5 Creatine phosphate ( pH adjusted to 7 . 5 with KOH ) . Membrane potentials were typically kept at −60 mV to −70 mV . In voltage-clamp mode , currents were recorded with voltage steps ranging from +10 mV to +80 mV . In current-clamp mode , action potentials were elicited by injection of step currents that modulated resting membrane potential from −20 mV to +80 mV . Local application of TTX ( Sigma-Aldrich#T8024 ) was achieved using a multibarrel perfusion system with a port placed within 0 . 5 mm of the patched cell . Reprogrammed neurons were harvested after 30 days of ectopic expression of miR-9/9*-124-CDM . DNA was extracted using phenol/chloroform/isoamyl alcohol followed by ethanol precipitation with a final concentration of 0 . 75M NaOAc and 2 μg of glycogen . DNA concentration was quantified using a standard curve with the Quant-iT dsDNA Assay Kit , broad range ( Thermo Fisher Scientific , Waltham , MA ) according to manufacturer’s instruction , while the DNA quality was determined by the ratio of absorbance of 260 nm and 280 nm at approximately 1 . 7–2 . 0 . The bisulfite conversion was performed for fibroblasts and reprogrammed neurons using the Zymo Research EZ-96 DNA Methylation-Gold Kit ( catalog #D5008 ) . DNA methylation data were generated on the HumanMethylation450k Bead Chip ( Illumina , San Diego , CA ) according to the manufacturer's protocols . Scanning was performed via Illumina’s iScan system in conjunction with the Illumina Autoloader 2 robotic arm . DNA methylation levels ( β values ) were established by calculating the ratio of intensities between methylated ( signal A ) and un-methylated ( signal B ) sites . The β value was calculated from the intensity of the methylated ( M corresponding to signal A ) and un-methylated ( U corresponding to signal B ) sites , as the ratio of fluorescent signals β = Max ( M , 0 ) /[Max ( M , 0 ) +Max ( U , 0 ) +100] . β values range from 0 ( completely un-methylated ) to 1 ( completely methylated ) . The data were normalized using the 'Noob' normalization method ( Triche et al . , 2013 ) . The epigenetic clock method is an accurate measurement of chronological age in human tissues ( Horvath , 2013 ) . Epigenetic age was estimated using the published software tools ( Horvath , 2013 ) . An online age calculator can be found at the webpage , https://dnamage . genetics . ucla . edu . The epigenetic clock has been shown to capture aspects of biological age: the epigenetic age is predictive of all-cause mortality even after adjusting for a variety of known risk factors ( Marioni et al . , 2015; Christiansen et al . , 2016; Horvath et al . , 2015a ) . The utility of the epigenetic clock method has been demonstrated in applications surrounding cognitive function ( Levine et al . , 2015 ) , obesity ( Horvath et al . , 2014 ) , Down syndrome ( Horvath et al . , 2015b ) , HIV infection ( Horvath and Levine , 2015 ) , and Parkinson's disease ( Horvath and Ritz , 2015 ) . Total RNA was extracted from reprogrammed neurons from young fibroblasts aged three days , five months , and one year and from old fibroblast aged 90 , 92 , and 92 years alongside corresponding starting fibroblast samples using TRIzol ( Thermo Fisher Scientific , Waltham , MA ) according to the manufacturer’s instruction and extracted using chloroform and ethanol precipitation . RNA quality was determined by the ratio of absorbance at 260 nm and 280 nm to be approximately 2 . 0 . Samples for RNA microarray were then standardly prepped and labeled with Illumina TotalPrep kits ( Thermo Fisher Scientific , Waltham , MA ) for Agilent Human 4x44Kv1 , while samples for microRNA microarray were prepared using Genisphere Flashtag labeling kits designed for Affymetrix miRNA 4 . 0 microarray . Standard hybridization and imagine scanning procedure were performed according to the manufacturer's protocol at Genome Technology Access Center at Washington University School of Medicine , St . Louis . The intensity of probes was imported into R environment and normalized by using package 'oligo' . Differentially expressed mRNA transcripts were identified by using package 'limma' with cut-off at adjusted p-value<0 . 01 and over logfc >1 fold expression change . For miRNA , the intensity of human-specific probes was isolated by using in-house python script , and were imported into R environment . Quantile normalization was performed by using 'preprocessCore' package , and differentially expressed miRNAs were identified by using package 'limma' with cut-off at adjusted p-value<0 . 01 and over one-fold expression change . cDNA was generated from 4 ng of RNA using specific primer probes from TaqMan MicroRNA Assays ( Thermo Fisher Scientific , Waltham , MA ) and subsequently analyzed on a StepOnePlus Real-Time PCR System ( AB Applied Biosystems , Foster City , CA ) . Expression data were normalized to RNU44 and analyzed using the 2−ΔΔCT relative quantification method . QPCR validation of miRNA expression was conducted in reprogrammed neurons from old fibroblasts aged 89 , 90 , 91 , 92 , 92 , 94 compared to reprogrammed neurons from young fibroblasts aged three days , five months , one , two , 12 years of age . QPCR experiments were conducted with human striatum and human cortex slices acquired from young individuals aged 9 , 11 , and 19 years compared to those from older individuals aged 83 , 85 , and 87 years . MitoSOX Red Mitochondrial superoxide indicator ( Thermo Fisher Scientific , Waltham , MA ) was diluted to 15 μM and incubated with cells for 15 min at 37°C . Cells were washed three times with PBS , dissociated with 0 . 25% Trypsin , then stained with DAPI . If FACs was not conducted on the same day , cells were fixed with 4% paraformaldehyde for 20 min at room temperature . Samples were compared to untreated ( unstained ) fibroblast and reprogrammed neurons . Cell sorting was performed on a FACSCalibur and LsrFortessa ( BD Biosciences ) , while quantification of the percent of the population of MitoSOX positive cells was performed using FlowJo X 10 . 0 . 7r2 . Each plot on the graph represents an individual experiment with multiple reprogrammed neurons . Unpaired t-test analysis of 3 sets of experiments of reprogrammed neurons from young fibroblasts compared to reprogrammed neurons from old fibroblasts . Young samples included reprogrammed neurons from fibroblasts aged three days , five month , one year and two years , while old samples were from donors aged 86 , 90 , 91 , 92A , and 92B years . Analyzed fibroblast samples include 1 , 2 , 91 , 72 , 74 , and 94-year-old samples . P-values were calculated with the student t-test . Cells were prepared and analyzed using the CometAssay Kit ( Trevigen ) according to manufacturer’s instruction . Cells were harvested after 30 days of neuronal reprogramming using 0 . 25% Trypsin , then whole cells were embedded in molten LMAgarose onto slides prior to overnight incubation in lysis buffer . Slides were then run in gel electrophoresis at 20 volts for 30 min , then stained with SYBR Green and visualized by epifluorescence microscopy . Tails lengths were measured by drawing a region of interest and p-values were calculated with student t-test for reprogrammed neurons from old fibroblasts aged 91 , 92A and 92B compared to reprogrammed neurons from young fibroblasts aged three days , five months , and one year old , while analyzed fibroblasts were aged five months , one year , 12 , 72 , 86 , and 92 years of age . Genomic DNA was collected from reprogrammed neurons from fibroblasts from one year , 56 , and 86 year old donors and the corresponding fibroblasts . The isolated genomic DNA was then digested with RsaI and HinfI and fractionated as described previously ( Tomlinson et al . , 2008 ) . Membranes were prepared by Southern transfer and hybridized to a radioactively end-labelled ( TTAGGG ) 4 oligonucleotide probe as described previously ( Batista et al . , 2011 ) .	As we age , so do our cells . When cells are used in the laboratory to study the biology of diseases , it is important that the age of the cells reflects the age at which the disease develops . This is particularly important for illnesses with symptoms that develop during old age , and where younger cells may appear to be relatively unaffected . Aging is a major risk factor in many brain disorders that affect elderly individuals . These late-onset disorders can be difficult to study because it is rarely possible to collect diseased cells from patients . Recent experimental advances , however , now mean that unrelated cell types – typically cells called fibroblasts , taken from a patient’s skin – can be converted directly into brain cells instead . These new brain cells will have the same genetic makeup as the patients they came from , but whether these converted cells would reflect the patient’s age too remained to be determined . By measuring a range of biological properties of the converted cells , Huh et al . now show that converted cells do indeed keep track of their age when they are changed from fibroblasts to brain cells . The age of the cells was tested by looking at age-linked markers attached to their DNA known as an “epigenetic clock” . In addition , Huh et al . measured the age of the cells by examining the expression of genes altered with aging . Other factors examined included the amount of damaged DNA and the size of DNA regions called telomeres , which become shorter with age . Together , all of these indicated that the converted brain cells retain the age of the fibroblasts that they were made from . So far this work has only been done using fibroblasts collected from healthy people . The same tests now need to be done using cells from people with late-onset illnesses like Huntington’s disease and Alzheimer’s disease . If the converted brain cells show signs of illness , it may provide new ways to study these illnesses using cells from specific patients , which may eventually lead to the development of new treatments .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "short", "report", "neuroscience" ]	2016	Maintenance of age in human neurons generated by microRNA-based neuronal conversion of fibroblasts
Under high light , oxygenic photosynthetic organisms avoid photodamage by thermally dissipating absorbed energy , which is called nonphotochemical quenching . In green algae , a chlorophyll and carotenoid-binding protein , light-harvesting complex stress-related ( LHCSR3 ) , detects excess energy via a pH drop and serves as a quenching site . Using a combined in vivo and in vitro approach , we investigated quenching within LHCSR3 from Chlamydomonas reinhardtii . In vitro two distinct quenching processes , individually controlled by pH and zeaxanthin , were identified within LHCSR3 . The pH-dependent quenching was removed within a mutant LHCSR3 that lacks the residues that are protonated to sense the pH drop . Observation of quenching in zeaxanthin-enriched LHCSR3 even at neutral pH demonstrated zeaxanthin-dependent quenching , which also occurs in other light-harvesting complexes . Either pH- or zeaxanthin-dependent quenching prevented the formation of damaging reactive oxygen species , and thus the two quenching processes may together provide different induction and recovery kinetics for photoprotection in a changing environment . Sunlight is the essential source of energy for most photosynthetic organisms , yet sunlight in excess of the organism’s photosynthetic capacity can generate reactive oxygen species ( ROS ) that lead to cellular damage . To avoid damage , plants respond to high light ( HL ) by activating photophysical pathways that safely convert excess energy to heat , which is known as nonphotochemical quenching ( NPQ ) ( Rochaix , 2014 ) . While NPQ allows for healthy growth , it also limits the overall photosynthetic efficiency under many conditions . If NPQ were optimized for biomass , yields would improve dramatically , potentially by up to 30% ( Kromdijk et al . , 2016; Zhu et al . , 2010 ) . However , critical information to guide optimization is still lacking , including the molecular origin of NPQ and the mechanism of regulation . Green algae is a sustainable alternative for biofuels and livestock feed ( Lum et al . , 2013; Wijffels and Barbosa , 2010 ) . In Chlamydomonas ( C . ) reinhardtii , the model organism for green algae , light-harvesting complex stress-related ( LHCSR ) is the key gene product for NPQ . LHCSR contains chlorophyll ( Chl ) and carotenoid ( Car ) held within its protein scaffold . Two isoforms of LHCSR , LHCSR1 and LHCSR3 , are active in NPQ , although LHCSR3 is accumulated at higher levels and so has the major role ( Dinc et al . , 2016; Maruyama et al . , 2014; Peers et al . , 2009; Tokutsu and Minagawa , 2013 ) . While the photophysical mechanism of quenching in light-harvesting complexes has not been determined , the primary proposals involve Chl-Car interactions ( Liao et al . , 2010; Ma et al . , 2003; Ruban et al . , 2007; Son et al . , 2020a; Son et al . , 2020b; de la Cruz Valbuena et al . , 2019 ) . NPQ is triggered by a proton gradient across the thylakoid membrane that forms through a drop in luminal pH ( Horton et al . , 1996 ) . Lumen acidification generally occurs when the light available causes an imbalance between the proton generation and the capacity of the photosynthetic apparatus to use protons for ATP production ( Joliot and Finazzi , 2010 ) . The C-terminus of LHCSR3 contains a number of luminal residues that are protonated upon the pH drop to trigger quenching ( Ballottari et al . , 2016; Liguori et al . , 2013 ) . The pH drop also activates the enzymatic conversion of the Car violaxanthin ( Vio ) to zeaxanthin ( Zea ) ( Eskling et al . , 1997 ) . Along with LHCSR , other homologous light-harvesting complexes are likely involved in quenching ( Nicol et al . , 2019 ) . In C . reinhardtii , the CP26 and CP29 subunits , which are minor antenna complexes of Photosystem II ( PSII ) , have been implicated in NPQ ( Cazzaniga et al . , 2020 ) . In higher plants , Zea has been reported to be involved in NPQ induction by driving light-harvesting complexes into a quenched state and/or by mediating interaction between light-harvesting complexes and PsbS , nonpigment binding subunits essential for NPQ induction in vascular plants ( Sacharz et al . , 2017; Ahn et al . , 2008; Jahns and Holzwarth , 2012 ) . Similarly , Zea binding to LHCSR1 in the moss Physcomitrella patens and LHCX1 ( a LHCSR homolog ) in the microalga Nannochloropsis oceanica has been shown to be essential for NPQ ( Pinnola et al . , 2013; Park et al . , 2019 ) . Finally , in C . reinhardtii , a reduction of NPQ in the absence of Zea has been reported ( Niyogi et al . , 1997 ) . In contrast , recent work has shown Zea to be unnecessary for NPQ induction or related to highly specific growth conditions ( Bonente et al . , 2011; Tian et al . , 2019; Vidal-Meireles et al . , 2020 ) . Thus , the contribution of Zea to quenching in green algae remains under debate . Because of the complexity of NPQ and the large number of homologous light-harvesting complexes , the individual contributions and mechanisms associated with LHCSR3 , pH , and Zea have been challenging to disentangle , including whether they activate quenching separately or collectively . With the power of mutagenesis , the contribution of LHCSR3 , and the dependence of this contribution on pH and Zea , can be determined . However , in vivo experiments leave the molecular mechanisms of LHCSR3 and its activation obscured . In vitro experiments , and particularly single-molecule fluorescence spectroscopy , are a powerful complement to identify protein conformational states ( Gwizdala et al . , 2016; Krüger et al . , 2010; Kondo et al . , 2017; Schlau-Cohen et al . , 2014; Schlau-Cohen et al . , 2015 ) . A recent method to analyze single-molecule data , 2D fluorescence correlation analysis ( 2D-FLC ) ( Ishii and Tahara , 2013a; Kondo et al . , 2019 ) quantifies the number of conformational states and their dynamics , including simultaneous , distinct processes . Thus , the conformational changes associated with NPQ can be resolved . Here , we apply a combined in vivo and in vitro approach to investigate NPQ in C . reinhardtii . We use mutagenesis , NPQ induction experiments , and fluorescence lifetime measurements on whole cells and single LHCSR3 complexes to show that pH and Zea function in parallel and that either parameter can activate full quenching and prevent ROS accumulation . The pH-dependent quenching in LHCSR3 is controlled by the protonation of residues in the C-terminus as shown by mutagenesis to remove these residues . The Zea-dependent quenching is activated even at neutral pH both in vitro and in vivo . Based on the in vitro results , we find two likely quenching sites , that is Chl-Car pairs within LHCSR3 , one regulated by pH and the other by Zea . The two quenching processes act in combination to provide different time scales of activation and deactivation of photoprotection , allowing survival under variable light conditions . The single-molecule fluorescence intensities are time averages , and so we also analyzed the fluorescence emission from single LHCSR3 through a photon-by-photon method , 2D fluorescence lifetime correlation ( 2D-FLC ) analysis . This method uses the associated lifetime data , and is more appropriate to analyze this data as the lifetime decays exhibit complex kinetics ( de la Cruz Valbuena et al . , 2019 ) . Applying the 2D-FLC analysis to single-molecule data identifies fluorescence lifetime states , which correspond to protein conformations with different extents of quenching , and rates of transitions between states , which correspond to switches between the protein conformations ( Kondo et al . , 2019 ) . To determine the number of lifetime states , the distributions of lifetime values were constructed ( Figure 2A–D ) . In a lifetime distribution , lifetime states appear as peaks with varying profiles . Traditional lifetime fitting requires an a priori assumption of the number of exponential terms required to fit a decay curve . In contrast , construction of a lifetime distribution does not require prior assignment of the number of decay timescales , which is particularly important when there are multiple different lifetimes as is the case for LHCSR3 ( de la Cruz Valbuena et al . , 2019 ) . The lifetime distribution also allows analysis of multi-exponential lifetimes , even for the low signal-to-background regime of single-molecule data . The initial lifetime distribution for each sample was calculated by first performing an inverse Laplace transform of all the lifetime data ( time between excitation and emission ) , which was recorded on a photon-by-photon basis ( Figure 2—figure supplement 1 ) . Photon pairs separated by a series of delay times were identified , and a 2D inverse Laplace transform was performed for the photon pairs associated with each delay time ( see Materials and methods for details ) . The final lifetime distributions ( Figure 2A–D ) were determined by fitting the data in order to optimize the lifetime distributions and to generate the correlation functions , which are discussed in more detail below . For each of the LHCSR3 samples , two lifetime states were observed in the distributions , an unquenched state ( ~2 . 5 ns ) and a quenched state ( ~0 . 5 ns ) . The dynamics of the lifetime states were investigated through the auto- and cross-correlation functions for the lifetime states of each sample ( Figure 2—figure supplement 2 ) . The correlation function is a normalized measure of how similar the photon emission time , that is , the lifetime , is as time increases ( Nitzan , 2006 ) . Therefore , an auto-correlation function for a given lifetime state contains the timescales for transitions out of the state and a cross-correlation function contains the timescales for transitions between the states ( anti-correlation ) and similar behavior of the states ( correlation ) due to processes throughout LHCSR3 , such as photobleaching . The auto- and cross-correlations were globally fit to estimate the parameters in the correlation model function given by Equation 1 in the Materials and methods , which includes the number of lifetime states , the brightness of each state , the population of each state , the rates of transitions between states , and the number of separate processes that transition between states , referred to as dynamic components . The parameters extracted from the fits are given in Figure 2—source data 1 for all samples . The best fits to the data included three dynamic components , where each component arises from distinct emissive states with separate conformational dynamics within single LHCSR3 ( Figure 2—figure supplement 2 , Figure 2—source data 1 ) . Correlation-based analysis of the photon fluctuations is a well-established tool to identify the number of independent emissive processes ( Schwille and Haustein , 2002; Mets , 2001 ) , and was adapted to determine the number of dynamic components ( Kondo et al . , 2019 ) . The cross-correlation for all LHCSR3 samples begins above zero ( Figure 2—figure supplement 2 ) , which appears in the presence of multiple dynamic components ( Kondo et al . , 2019 ) . The Chl a have the lowest energy levels , and , due to their significantly lower energy than the Chl b energy levels , primarily give rise to the emissive states . Because three components were observed within single LHCSR3 , they indicate multiple Chl a emissive sites within each LHCSR3 , consistent with previous models of LHCs ( Mascoli et al . , 2019; Mascoli et al . , 2020; Krüger et al . , 2010; Krüger et al . , 2011 ) . Thus , the dynamic components reflect conformational dynamics that switch between unquenched and quenched lifetime states at different places within LHCSR3 . The rate constants for the transitions between the lifetime states within each component were also extracted from the fit , primarily based on the dynamics of the cross-correlation functions ( Figure 2—source data 1 ) . Two of the components exhibit rapid dynamics , which arise from conformational changes that vary the extent of quenching of the Chl a emitters . The timescales of the transitions for one component are tens of microseconds and those for the other are hundreds of microseconds , which are both timescales that would be hidden in traditional single-molecule analyses . The third dynamic component is static at <0 . 01 s . Due to the lack of dynamics , we assigned the component to emitters far from , and thus unaffected by , quenchers for the unquenched state and partially photobleached complexes for the quenched state . Finally , the relative populations of the lifetime states for each component were also determined within the model . Assuming a Boltzmann distribution ( see Materials and methods ) , the relative rate constants were used to determine the equilibrium free-energy difference between the states for each component ( Figure 2—source data 1 ) . The free-energy barrier for a transition between states is related to the rate of the transition , which was used to approximate the barrier height ( Kondo et al . , 2019 ) . These free-energy differences and barrier heights were then combined to construct illustrative free-energy landscapes , which are shown in Figure 2 for the two dynamic components . We examined the dependence of the two dynamic components on pH , Zea and the C-terminal tail , which contains the pH-sensing residues . Figure 2E and F show the pH-dependence of the free-energy landscapes for the slower ( hundreds of microseconds ) dynamic component in LHCSR3-Vio and LHCSR3-Zea , respectively . In both cases , a decrease in pH from 7 . 5 to 5 . 0 stabilizes the quenched state . In LHCSR3-Vio , the quenched state is stabilized by a decrease in the transition rate from the quenched to the unquenched state , corresponding to a higher barrier in the free-energy landscape . In LHCSR3-Zea , the decrease in the transition rate from the quenched to the unquenched state is also accompanied by an increase in the transition rate from the unquenched to the quenched state , further stabilizing the quenched state relative to the unquenched one . In stop-LHCSR3-Vio , however , no change in the population of the quenched state is observed upon a decrease in pH ( Figure 2G ) , reflecting the expected pH-independence of the sample . Figure 2I and J show the Zea-dependence of the free-energy landscapes of LHCSR3 for the faster ( tens of microseconds ) dynamic component at pH 7 . 5 and pH 5 . 0 , respectively . At both pH levels , conversion from Vio to Zea stabilizes the quenched state via a decrease in the transition rate from the quenched to unquenched state . At pH 5 . 0 , the transition rate to the quenched state increases as illustrated by the lower barrier , which would enable rapid equilibration of population into the quenched state . The Zea-dependent behavior is maintained for stop-LHCSR3 ( Figure 2K ) , where the quenched state is still stabilized in the presence of Zea . Quenching mechanisms were further investigated in vivo by measuring fluorescence emission lifetimes at 77K of whole cells acclimated to LL or HL , as traditional NPQ measurements can be affected by artifacts ( Tietz et al . , 2017 ) . Under these conditions , the photochemical activity is blocked and by following the emission at 680 nm it is possible to specifically investigate the kinetics of PSII , the main target of NPQ . Cells were either grown in LL or HL , which determines the level of LHCSR protein ( Figure 1—figure supplements 1 and 2 ) and the fluorescence lifetime was recorded before and after exposure to 60 min of HL , which induces ΔpH and determines the level of Zea . These light conditions , combined with the genotypes generated , enabled studies that partially or fully separated the contributions of the different components of NPQ . Whole cell fluorescence lifetime traces show that LHCSR is necessary for the primary light-dependent component of NPQ in C . reinhardtii trigged by lumen acidification , in agreement with previous findings ( Peers et al . , 2009; Ballottari et al . , 2016 ) . WT cells and npq1 cells , which lack Zea , acclimated to HL show a faster fluorescence decay , or an increase in quenching , after exposure to 60 min of HL ( Figure 3A , gray bars , fluorescence decays and fits to data shown in SI ) . For npq4 lhcsr1 cells , which lack LHCSR , similar fluorescence decay kinetics were measured regardless of light treatment ( Figure 3A , purple ) , which is comparable to the unquenched kinetics of WT cells . WT and npq1 cells grown in control light ( low LHCSR content ) remain unquenched , even after exposure to 60 min of HL ( Figure 3—figure supplements 1 and 2 ) . These results are consistent with the NPQ measurements shown in Figure 1A . Similar to WT , npq1 cells grown in HL show a faster fluorescence decay after exposure to 60 min of HL ( Figure 3A , blue bars ) . While the results from WT show a role for pH and/or Zea in light-induced quenching in LHCSR3 , the results from the npq1 strain show that quenching can occur without Zea , that is , induced by the pH drop alone . The zep mutant , which constitutively accumulates Zea , presented a similar decay among all samples , regardless of HL or LL acclimation or light treatment , that was much faster , or more quenched , compared to the decay of dark-adapted WT ( Figure 3A , red , Figure 3—figure supplements 1 and 2 ) . These results indicate quenching upon Zea accumulation alone , consistent with the reduced Fv/Fm observed in this mutant ( Figure 1—figure supplement 7 ) . This result is also consistent with the pH-independent quenching observed through the single-molecule fluorescence shown in Figure 1D and Figure 2E and F . However , the quenching observed in the zep mutant was essentially unchanged in LL vs . HL acclimated zep cells suggesting that the Zea-dependent quenching observed in zep mutants is a more general process as opposed to one that occurs solely in LHCSR3 as quenching is observed even in the cells acclimated to LL that lack LHCSR3 . To investigate the generality of this quenching , monomeric or trimeric light-harvesting complexes were isolated from the zep mutant after exposure to 60 min of HL , which induces maximum Zea accumulation . These complexes had a twofold higher content of Zea compared to the same fraction isolated from WT ( CC4349 ) under the same conditions ( Figure 3—figure supplement 5 ) . The light-harvesting complexes isolated from the zep mutant also showed a 10% decrease in the fluorescence lifetime , suggesting that Zea-dependent quenching is at least somewhat shared with other light-harvesting complexes ( Figure 3—figure supplement 6 , Figure 3—figure supplement 5—source data 1 , and Figure 3—source data 3 ) . In contrast , no major differences in quenching properties were found in monomeric and trimeric LHC complexes isolated from WT cells before or after exposure to 60 min of HL , consistent with previous findings from higher plants and other green algae ( Xu et al . , 2015; Girolomoni et al . , 2020 ) . The main function of quenching the Chl singlet excited states is to thermally dissipate the fraction of absorbed excitation energy in excess of the capacity of the photosynthetic apparatus . Unquenched Chl singlet excited states may cause ROS formation and subsequent photoinhibition of their primary target , PSII ( Niyogi , 1999 ) . Singlet oxygen is the main ROS species formed at the level of PSII . In order to correlate the NPQ levels and quenching measured with ROS formation , singlet oxygen production was followed in the different genotypes herein investigated by using the fluorescent probe Singlet Oxygen Sensor Green ( SOSG ) ( Flors et al . , 2006; Figure 3B , Figure 3—figure supplements 3 and 4 ) . SOSG fluorescence can be used as a probe to follow singlet oxygen formation , although measuring the true production rates would require a different analytic method . Moreover , SOSG has been reported to produce singlet oxygen itself upon prolonged illumination , and thus requires the use of light filters in order to avoid direct excitation of the dye during HL treatment ( Kim et al . , 2013 ) . As expected from the low level of NPQ induction , npq4 lhcsr1 , which lacks LHCSR , demonstrated the highest level of singlet oxygen production , regardless of light treatment . Interestingly , the effect of Zea was almost negligible in HL acclimated samples ( with a very high NPQ induction ) . Notably , the amount of singlet oxygen production was correlated with average lifetime ( Figure 3A ) , that is , inversely correlated with quenching , confirming that the quenching of Chl singlet excited states investigated here plays a role in photoprotection . Two dynamic components were identified through the 2D-FLC analysis that suggest two distinct photoprotective processes , one pH-dependent and one Zea-dependent , operating simultaneously within LHCSR3 . Each component likely arises from a Chl-Car pair , where the Car can quench the emissive Chl . The two components both have greater population in the quenched state than in the unquenched state ( Figure 2—source data 1 ) , consistent with previous work where a quenching component was found to be present in LHCSR3 , even at pH 7 . 5 ( de la Cruz Valbuena et al . , 2019 ) . By considering the single-molecule results along with previous structural , spectroscopic and theoretical work , we speculate as to the likely sites associated with each component . Although no high-resolution structure of LHCSR3 has been determined , we illustrate possible quenching sites ( Figure 2H and L ) within a working structural model of LHCSR3 ( Bonente et al . , 2011 ) . As shown in Figure 2—source data 2 , LHCSR3 purified from C . reinhardtii contains eight Chl molecules ( 7–8 Chl a and 0–1 Chl b molecules ) and two Cars ( one lutein [Lut] and one Vio or Zea ) . Based on sequence comparison with LHCII and CP29 , the conserved Chl a binding sites are the following: Chls a 602 , 603 , 609 , 610 , 612 , and 613 , with Chls a604 , 608 , and 611 proposed as well ( Bonente et al . , 2011; Liguori et al . , 2016 ) . Previous spectroscopic analysis of LHCSR3 from C . reinhardtii has identified the likely binding sites of Lut and Vio/Zea within the structural model ( Bonente et al . , 2011 ) . Given that there are two Cars bound at the internal sites of LHCSR3 , it is likely that each Car and its neighboring Chl is the major contributor for one of the two dynamic components . The pH-dependent component ( Figure 2E–H ) likely involves Lut and the neighboring Chl a 613 . Both Chl a 612 ( coupled to Chl as 610 and 611 ) and Chl a 613 have previously been proposed as quenching sites given their physical proximity to the Lut ( Liguori et al . , 2016; Ruban et al . , 2007 ) . The Chl a 610–612 site contains the lowest energy Chl a , which has been shown to be a major energy sink and thus the primary emitter ( Müh et al . , 2010; Schlau-Cohen et al . , 2009; Novoderezhkin et al . , 2011 ) . Additionally , computational results have shown that the interaction between the Lut site and Chl a 612 has large fluctuations ( Liguori et al . , 2015 ) . This agrees with the slower dynamics found for this component . However , recent in vivo and in vitro analyses found that the removal of Chl a 613 results in a reduction in LHCSR specific quenching , while removal of Chl a 612 only affected which Chl was the final emitter of the complex ( Perozeni et al . , 2019 ) . While either of these sites are potential quenching sites , it is likely that Chl a 613 plays the major role in pH-dependent quenching in LHCSR3 in C . reinhardtii . With a decrease in pH from 7 . 5 to 5 . 0 , the equilibrium free-energy differences for the pH-dependent component , which were calculated using the relative rate constants from the global fit , were shifted toward the quenched state by over 200 cm−1 in LHCSR3-Vio and over 500 cm−1 in LHCSR3-Zea ( Figure 2—source data 1 ) . The specific conformational change upon protonation that leads to this stabilization remains undetermined . However , proposals in the literature include reduced electrostatic repulsion in the lumen-exposed domain causes a change in the distance and/or orientation between the helices ( Ballottari et al . , 2016 ) and an increase in protein-protein interactions ( de la Cruz Valbuena et al . , 2019 ) . These conformational changes could produce a displacement of Lut toward Chl a 613 . Analysis of stop-LHCSR3 , which lacks the pH-sensing residues in the C terminus , showed that the C terminus controls quenching activity by pH-induced stabilization of the quenched conformation of LHCSR3 . The negligible ( <30 cm−1 ) change in the equilibrium free-energy difference for this mutant ( Figure 2G , Figure 2—source data 1 ) upon a pH drop demonstrates the functional role of the C-terminal tail in the conformational change into the quenched state . The Zea-dependent component ( Figure 2J–K ) likely involves Vio/Zea and the neighboring Chl as 602–603 ( Bonente et al . , 2011; Di Valentin et al . , 2009; Lampoura et al . , 2002 ) . With conversion from Vio to Zea , the free-energy landscape changes significantly , and thus is likely to involve the region of LHCSR3 that surrounds Vio/Zea . In addition , MD simulations have shown this Car site to be highly flexible , sampling many configurations ( Liguori et al . , 2017 ) , which is consistent with the faster dynamics observed here . Upon substitution of Zea for Vio , the equilibrium free-energy difference becomes further biased toward the quenched state by over 550 cm−1 at pH 7 . 5 and over 300 cm−1 at pH 5 . 0 , where the difference was calculated from the populations of the lifetime states determined within the model . This result is consistent with a role of Zea in quenching of LHCSR3 that does not require a decrease in pH and therefore is distinct from the major pH-dependent component of NPQ observed in vivo in npq1 , which almost completely recovered in the dark ( Figure 1A ) . In the stop-LHCSR3 , the equilibrium free-energy differences for the Zea-dependent component is similar to the wild type samples ( Figure 2K ) . This is consistent with the Vio/Zea-Chl a 602–603 site as the major contributor for this component . Although qualitatively similar , there is a small decrease ( <200 cm−1 ) in the stabilization of the quenched state upon Zea incorporation . Thus , the C-terminal tail affects the states associated with both dynamic components , which arise from different emissive sites within LHCSR3 , and so likely has an allosteric effect throughout the protein . The static component , which is assigned to emitters far from the quenching site in the unquenched state , has a large contribution to the correlation profiles ( Figure 2—source data 1 ) . The large amplitude is consistent with the low number of Cars available to interact with the Chls and thus the presence of several unquenched emissive Chl a . Given the structural arrangement of the Cars and Chls , the unquenched state within the static component is likely due to Chls 604 , 608 , and 609 , which sit far from the Cars . The quenched state within the static component is likely due to partial photobleaching , which can lower the fluorescence intensity ( Kondo et al . , 2019 ) . Zea-dependent quenching is observed both in vivo and in vitro even at neutral pH . While the mechanism is described at the molecular level in the case of LHCSR3 , it is likely shared with other light-harvesting complexes . A strong reduction of fluorescence lifetime was observed in whole cells in the case of zep mutant , even in LL acclimated cells where the amount of LHCSR3 is minimal ( Figure 3—figure supplements 1 and 2 ) . This indicated that LHCSR subunits are not the sole quenching site where Zea-dependent quenching occurs , as seen in previous work implicating the minor antenna complexes ( Cazzaniga et al . , 2020 ) . Consistently , Zea-dependent quenching was measured in other light-harvesting complexes isolated from the zep mutant , but it was not sufficient to fully explain the strong quenching observed in whole cells . In the case of the zep mutant , not only does Zea completely substitute Vio ( de-epoxidation index is 1 , Figure 1—figure supplement 3 ) , but also the Zea/Chl ratio is much higher ( ~10 fold ) compared to the ratio observed in WT or npq4 lhcsr1 . This suggests an alternative possibility where the strong quenching observed in the zep mutant could be related to accumulation of Zea in the thylakoid membrane changing the environment where the photosystems and light-harvesting complexes are embedded , inducing the latter to a strong quenched state . Indeed , Zea has been previously reported to influence the assembly and organization of light-harvesting complexes in the thylakoid membranes of higher plants , affecting their quenching properties ( Sacharz et al . , 2017 , Shukla et al . , 2020 ) . While both possibilities allow for quenching in the presence of Zea even at neutral pH , it is the pH-independent quenching itself that is potentially the origin of the seemingly conflicting results in the literature , where Zea has been found to both reduce NPQ ( Niyogi et al . , 1997 ) and be unnecessary for its induction ( Bonente et al . , 2011; Baek et al . , 2016 ) . Our in vitro results point to pH and Zea controlling separate quenching processes within LHCSR3 and that either parameter can provide efficient induction of LHCSR3 to a quenched state for photoprotection . The 2D-FLC analysis on single LHCSR3 quantified two parallel dynamic components , or distinct quenching processes , one of which is pH-dependent and the other Zea-dependent . Likewise , in vivo full light-induced quenching upon lumen acidification was observed in the npq1 strain , which lacks Zea , and full quenching even at neutral pH was observed in the zep strain , which is Zea-enriched , suggesting two quenching and induction processes . The 2D-FLC analysis of the stop-LHCSR3 mutant shows that removal of the C-terminal tail removes pH-dependent quenching , while leaving Zea-dependent quenching nearly unaffected . Analogously , the WT LL grown strains , with reduced LHCSR accumulation , also present a significantly lower NPQ induction , supporting the critical role of the protonation of the C terminus residues unique to LHCSR in activating quenching in C . reinhardtii . Taken together , the in vivo and in vitro results indicated that either pH- or Zea-dependent quenching provides efficient photoprotection . While in vivo measurements suggest that pH-dependent quenching is often dominant over Zea-dependent quenching , and correspondingly more efficient in photoprotection , the conformational states and pigment pairs likely responsible exhibit spectroscopic signatures that suggest both quenching processes have similar conformational dynamics . In vivo measurements can be influenced by multiple variables , which are , in some cases , unpredictable , such as pleiotropic effects and acclimation responses . Thus , pH- and Zea-dependent quenching may both contribute to all quenching in the WT , while being alternatively triggered in the mutants through a compensatory effect . Under natural conditions , these processes combine to protect the system and there is likely interplay between them through compensatory acclimation or changes to the protein organization within the thylakoid . However , the timescales and induction associated with each quenching process are distinct; responsive pH-dependent quenching works in combination with the guaranteed safety valve of Zea-dependent quenching , potentially to protect against a rapid return to HL conditions . C . reinhardtii WT ( 4A+and CC4349 ) and mutant strains npq1 ( Niyogi et al . , 1997 ) and npq4 lhcsr1 ( Ballottari et al . , 2016 ) in the 4A+background and zep ( Baek et al . , 2016 ) in the CC4349 background were grown at 24°C under continuous illumination with white LED light at 80 µmol photons m−2 s−1 ( LL ) in high salts ( HS ) medium ( Harris , 2008 ) on a rotary shaker in Erlenmeyer flasks . 4A+ and CC4349 strains were obtained from the Chlamydomonas Resource Center ( https://www . chlamycollection . org/ ) and the npq1 strain ( Niyogi et al . , 1997 ) was kindly donated by Prof . Giovanni Finazzi ( CEA-Grenoble ) . HL acclimation was induced by growing cells for 2 weeks at 500 µmol photons m−2 s−1 in HS . As acclimation may result in complex single adaptation processes , we do not investigate these processes but instead focus our studies on the effect of acclimation on photoprotective mechanisms . SDS–PAGE analysis was performed using the Tris-Tricine buffer system ( Schägger and von Jagow , 1987 ) . Immunoblotting analysis was performed using αCP43 ( AS11 1787 ) , αPSAA ( AS06 172 ) , αLHCBM5 ( AS09 408 ) , and αLHCSR3 ( AS14 2766 ) antibodies purchased from Agrisera ( Sweden ) . The antibody αLHCBM5 was previously reported to also recognize LHCBM1-9 subunits and was thus used as αLHCII antibody ( Girolomoni et al . , 2017 ) . Violaxanthin de-epoxidation kinetics were performed by illuminating the different genotypes with red light at 1500 µmol photons m−2 s−1 up to 60 min . Pigments were extracted 80% acetone and analysed by HPLC as described in Lagarde et al . , 2000 . NPQ induction curves were measured on 60 min dark-adapted intact cells with a DUAL-PAM-100 fluorimeter ( Heinz-Walz ) at room temperature in a 1 × 1 cm cuvette mixed by magnetic stirring . Dark adaptation was performed in flasks under strong agitation with a shaker in order to avoid the onset of anaerobic conditions . Red saturating light of 4000 µmol photons m−2 s−1 and red actinic light of 1500 µmol photons m−2 s−1 were used to measure Fm and Fm’ , respectively , at the different time points . Samples were exposed for 60 min to actinic light followed by 20 min of dark recovery . Fm was measured on dark adapted cells , while Fm’ was measured at 10 min intervals . NPQ was then calculated as Fm/Fm’−1 . Proton motive force upon exposure to different light intensities was measured by Electrochromic Shift ( ECS ) with MultispeQ v2 . 0 ( PhotosynQ ) according to Kuhlgert , S . et al . MultispeQ Beta: A tool for large-scale plant phenotyping connected to the open photosynQ network ( Kuhlgert et al . , 2016 ) . pETmHis containing LHCSR3 CDS previously cloned as reported in Perozeni et al . , 2019 served as template to produce stop-LHCSR3 using Agilent QuikChange Lightning Site-Directed Mutagenesis Kit . Primer TGGCTCTGCGCTTCTAGAAGGAGGCCATTCT and primer GAATGGCCTCCTTCTAGAAGCGCAGAGCCA were used to insert a premature stop codon to replace residue E231 , generating a protein lacking 13 c-terminal residues ( stop-LHCSR3 ) . LHCSR3 WT and stop-LHCSR3 protein were overexpressed in BL21 E . coli and refolded in vitro in presence of pigments as previously reported ( Bonente et al . , 2011 ) . Pigments used for refolding were extracted from spinach thylakoids . Vio or Zea-binding versions of LHCSR3 were obtained using Vio or Zea containing pigment extracts in the refolding procedure . Zea-containing pigments were obtained by in vitro de-epoxidation ( de la Cruz Valbuena et al . , 2019; Pinnola et al . , 2017 ) Fluorescence emission at 300K with excitation at 440 nm , 475 nm and 500 nm was used to evaluate correct folding as previously reported ( Ballottari et al . , 2010 ) . Monomeric and trimeric light-harvesting complexes were isolated from solubilized thylakoids by ultracentrifugation in sucrose gradient as described in Tokutsu et al . , 2012 . Singlet oxygen production were estimated by using the fluorescent probe Singlet Oxygen Sensor Green ( SOSG ) ( Flors et al . , 2006 ) . SOSG fluorescence was measured in samples treated with red strong light ( 2000 µmol photons m−2 s−1 ) as described in Stella et al . , 2018 . While singlet oxygen estimation by SOSG is widely used , prolonged irradiation can lead to the formation of singlet oxygen by photodegradation of the fluorescent probe ( Kim et al . , 2013 ) . To prevent this artefact , direct excitation of the probe was prevented by insertion of a red filter ( >630 nm ) . Solutions of 12 µM purified LHCSR3 complexes were stored at −80°C . Immediately prior to experiments , LHCSR3 samples were thawed over ice and diluted to 50 pM using buffer containing 0 . 05% n-dodecyl-α-D-maltoside and either 20 mM HEPES-KOH ( pH 7 . 5 ) or 40 mM MES-NaOH ( pH 5 . 0 ) . The following enzymatic oxygen-scavenging systems were also used: ( 1 ) 25 nM protocatechuate-3 , 4-dioxygenase and 2 . 5 mM protocatechuic acid for pH 7 . 5 and ( 2 ) 50 nM pyranose oxidase , 100 nM catalase and 5 mM glucose for pH 5 . 0 . ( Aitken et al . , 2008; Swoboda et al . , 2012 ) Diluted solutions were incubated for 15 min on Ni-NTA-coated coverslips ( Ni_01 , Microsurfaces ) fitted with a Viton spacer to allow LHCSR3 complexes to attach to the surface via their His-tag . The sample was rinsed several times to remove unbound complexes and sealed with another coverslip . Single-molecule fluorescence measurements were performed in a home-built confocal microscope . A fiber laser ( FemtoFiber pro , Toptica; 130 fs pulse duration , 80 MHz repetition rate ) was tuned to 610 nm and set to an excitation power of 350 nW ( 2560 nJ/cm2 at the sample plane , assuming a Gaussian beam ) . Sample excitation and fluorescence collection were accomplished by the same oil-immersion objective ( UPLSAPO100XO , Olympus , NA 1 . 4 ) . The fluorescence signal was isolated using two bandpass filters ( ET690/120x and ET700/75 m , Chroma ) . The signal was detected using an avalanche photodiode ( SPCM-AQRH-15 , Excelitas ) and photon arrival times were recorded using a time-correlated single photon counting module ( Picoharp 300 , Picoquant ) . The instrument response function was measured from scattered light to be 380 ps ( full width at half maximum ) . Fluorescence intensity was analyzed as described previously using a change-point-finding algorithm ( Watkins and Yang , 2005 ) . Fluorescence emission was recorded until photobleaching for the following number of LHCSR3 in each sample: 132 LHCSR3-Vio at pH 7 . 5 ( 1 . 6•107 photons ) ; 173 LHCSR3-Vio at pH 5 . 5 ( 1 . 3•107 photons ) ; 95 LHCSR3-Zea at pH 7 . 5 ( 1 . 4•107 photons ) ; 216 LHCSR3-Zea at pH 5 . 5 ( 9 . 0•106 photons ) ; 125 stop-LHCSR3-Vio at pH 7 . 5 ( 2 . 5•107 photons ) ; 130 stop-LHCSR3-Vio at pH 5 . 5 ( 7 . 9•106 photons ) ; 148 stop-LHCSR3-Zea at pH 7 . 5 ( 1 . 3•107 photons ) ; 116 stop-LHCSR3-Zea at pH 5 . 5 ( 9 . 9•106 photons ) . Experiments were performed at room temperature . Each data set was collected over two or three days for technical replicates and the distribution generated each day was evaluated for consistency . 2D fluorescence lifetime correlation analysis was performed as detailed previously ( Kondo et al . , 2019 ) . Briefly , we performed the following analysis . First , the total number of states exhibiting distinct fluorescence lifetimes was estimated from the 1D lifetime distribution . The lifetime distribution is determined using the maximum entropy method ( MEM ) to perform a 1D inverse Laplace transform ( 1D-ILT ) of the 1D fluorescence lifetime decay ( Ishii and Tahara , 2013a ) . Next , a 2D fluorescence decay ( 2D-FD ) matrix was constructed by plotting pairs of photons separated by ∆T values ranging from 10−4 to 10 s . The 2D-FD matrix was transformed from t-space to the 2D fluorescence lifetime correlation ( 2D-FLC ) matrix in τ-space using a 2D-ILT by MEM fitting ( Ishii and Tahara , 2012; Ishii and Tahara , 2013a; Ishii and Tahara , 2013b ) . The 2D-FLC matrix is made up of two functions: the fluorescence lifetime distribution , A , and the correlation function , G . In practice , the initial fluorescence lifetime distribution , A0 , was estimated from the 2D-MEM fitting of the 2D-FD at the shortest ∆T ( 10−4 s ) . Then the correlation matrix , G0 , was estimated at all ∆T values with A0 as a constant . A0 and G0 , along with the state lifetime values determined from the 1D analysis , were used as initial parameters for the global fitting of the 2D-FDs at all ∆T values . A was treated as a global variable and G was treated as a local variable at each ∆T ( now G ( ∆T ) ) . The resulting fit provides the correlation function , G ( ∆T ) . The correlation function was normalized with respect to the total photon number in each state . Each set of correlation curves ( auto- and cross-correlation for one sample ) were globally fit using the following model function: ( 1 ) Gijs∆T=q2J2I∙∑x∑y≠xEy∙Φy∙Ry∞+Ex∙Φy∙RxΔT∙Ex∙Φx∙Cx This equation accounts for multiple , independent emitters within one protein ( multiple components ) . Here , x and y indicate the component number , i and j indicate the state ( auto correlation for i=j , cross correlation for i ≠ j ) , q accounts for experimental factors such as the detection efficiency , filter transmittance , gain of the detector , etc . , J is the laser power , and I is the total photon number proportional to the total measurement time . E , Φ , and C are diagonal matrices composed of the optical extinction coefficient , fluorescence quantum yield , and state population , respectively . R is a matrix element that is related to the rate matrix . The rate constants determined from the 2D-FLC analysis were used to calculate the free energies for each protein state shown in Figure 2E–F and H–J . The rate constants for transitions between the quenched and unquenched states are related to the free energies associated with both states through the Arrhenius equation: ( 2 ) kQ→U=A exp-EQ→UkBT ( 3 ) kU→Q=A exp-EU→QkBT Here , kQ→U and EQ→U* ( kU→Q and EU→Q* ) are the rate constant and activation energy , respectively , for the transition from the quenched ( Q ) to the unquenched ( U ) state . A is a constant , kB is the Boltzmann constant , and T is the temperature . Upon equilibration of the Q and A states , the free-energy difference , ∆E* , is given by the following equation: ( 4 ) kQ→UkU→Q=exp-∆EkBT Using the dynamic rates determined from the fits to the correlation function , we calculated ∆E at T = 300 K . The free-energy differences between the quenched and unquenched states are shown as the energetic differences between the minima in the energy landscapes shown in Figure 2 . The potential barriers were scaled by assuming the constant A in Equations 1 and 2 to be 1000 , which was shown previously to be a reasonable estimate for LHCSR1 ( Kondo et al . , 2019 ) . Low-temperature quenching measures were performed according to Perozeni et al . , 2019 . Cells were frozen in liquid nitrogen after being dark adapted or after 60 min of illumination at 1500 µmol photons m−2 s−1 of red light . Fluorescence decay kinetics were then recorded by using Chronos BH ISS Photon Counting instrument with picosecond laser excitation at 447 nm operating at 50 MHz . Fluorescence emissions were recorded at 680 nm in with 4 nm bandwidth . Laser power was kept below 0 . 1μW .	Green plants and algae rely on sunlight to transform light energy into chemical energy in a process known as photosynthesis . However , too much light can damage plants . Green plants prevent this by converting the extra absorbed light into heat . Both the absorption and the dissipation of sunlight into heat occur within so called light harvesting complexes . These are protein structures that contain pigments such as chlorophyll and carotenoids . The process of photoprotection starts when the excess of absorbed light generates protons ( elementary particles with a positive charge ) faster than they can be used . This causes a change in the pH ( a measure of the concentration of protons in a solution ) , which in turn , modifies the shape of proteins and the chemical identity of the carotenoids . However , it is still unclear what the exact mechanisms are . To clarify this , Troiano , Perozeni et al . engineered the light harvesting complex LHCSR3 of the green algae Chlamydomonas reinhardtii to create mutants that either could not sense changes in the pH or contained the carotenoid zeaxanthin . Zeaxanthin is one of the main carotenoids accumulated by plants and algae upon high light stress . Measurements showed that both pH detection and zeaxanthin were able to provide photoprotection independently . Troiano , Perozeni et al . further found that pH and carotenoids controlled changes to the organisation of the pigment at two separate locations within the LHCSR3 , which influenced whether the protein was able to prevent photodamage . When algae were unable to change pH or carotenoids , dissipation was less effective . Instead , specific molecules were produced that damage the cellular machinery . The results shed light onto how green algae protect themselves from too much light exposure . These findings could pave the way for optimising dissipation , which could increase yields of green algae by up to 30% . This could lead to green algae becoming a viable alternative for food , biofuels and feedstock .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2021	Identification of distinct pH- and zeaxanthin-dependent quenching in LHCSR3 from Chlamydomonas reinhardtii
In this study , we report a new protein involved in the homeostatic regulation of sleep in Drosophila . We conducted a forward genetic screen of chemically mutagenized flies to identify short-sleeping mutants and found one , redeye ( rye ) that shows a severe reduction of sleep length . Cloning of rye reveals that it encodes a nicotinic acetylcholine receptor α subunit required for Drosophila sleep . Levels of RYE oscillate in light–dark cycles and peak at times of daily sleep . Cycling of RYE is independent of a functional circadian clock , but rather depends upon the sleep homeostat , as protein levels are up-regulated in short-sleeping mutants and also in wild type animals following sleep deprivation . We propose that the homeostatic drive to sleep increases levels of RYE , which responds to this drive by promoting sleep . Sleep is a common and prominent behavior in almost all vertebrate animals and also in most invertebrates ( Cirelli , 2009; Sehgal and Mignot , 2011 ) . The function of sleep is a mystery , but it is surely of great importance to animals , as prolonged sleep deprivation can lead to death . Anatomical studies in mammals and birds have revealed brain structures and neurotransmitters that regulate sleep and wakefulness ( Saper et al . , 2010 ) . However , our understanding of the molecular mechanisms that drive the need to sleep is still in its infancy , partially due to the challenge of performing genetic experiments with mammalian models . In the past decade , several premier genetic organisms have been introduced into the sleep field , including fruit flies , worms and zebra fish . Mammalian counterparts of some sleep components identified in these model animals also regulate sleep ( Joho et al . , 2006 ) , which argues that ( i ) behavioral genetics in lower organisms provides an efficient tool to identify sleep components , ( ii ) at least some of the mechanisms underlying sleep are conserved through evolution . A two-process model for sleep regulation has been widely accepted by the sleep field . Process C ( the circadian clock ) controls the timing , in other words the onset and offset of sleep , whereas process S ( the sleep homeostat ) regulates sleep duration based on the sleep pressure built up during prior wakefulness ( Borbely , 1982 ) . This simple model explains sleep related phenomena , including sleep rebound after sleep deprivation . Molecular mechanisms of circadian control have been well characterized ( Zheng and Sehgal , 2012 ) , but , as noted above , relatively little is known about process S . Forward genetic screens in Drosophila have led to the identification of several mutants with altered sleep length , but while the genes implicated by these mutants are required for implementation of sleep drive , they have not yet been directly linked to this drive ( Sehgal and Mignot , 2011 ) . Using a forward genetic screen , we identified a new sleep mutant we termed redeye ( rye ) , which is directly controlled by the homeostatic drive to sleep . The rye mutation maps to a nicotinic acetylcholine receptor ( nAChR ) , which interacts with a previously identified sleep-regulating protein , SLEEPLESS ( SSS ) . Levels of RYE are expressed cyclically , in conjunction with the sleep state , and reflect sleep need such that they are up-regulated after sleep deprivation and in short-sleeping mutants . We conclude that homeostatic sleep drive promotes sleep at least in part by increasing expression of RYE . As previous screens have identified sleep components on the X chromosome and the second chromosome ( Cirelli et al . , 2005; Koh et al . , 2008; Stavropoulos and Young , 2011 ) , we sought to identify sleep-altering mutations on the third chromosome in Drosophila . To this end , we generated a stock of iso31 ( Ryder et al . , 2004 ) flies carrying a newly isogenized third chromosome and treated males of this stock with 10–25 mM ethyl-methanesulfonate ( EMS ) . Individual male progeny were bred to achieve homozygosity of the third chromosome , and females of the F3 generation were tested for daily sleep patterns in the presence of light:dark cycles ( Koh et al . , 2008 ) . As seen in previous screens ( Koh et al . , 2008 ) , sleep amounts were normally distributed in the 1857 lines assayed ( Figure 1A , Figure 1—source data 1 ) . Two homozygous mutant lines showed a severe reduction of sleep length , and we focused on one of these that we named redeye ( rye ) . rye homozygotes show a >50% reduction in both daytime and night-time sleep ( Figure 1C ) . The activity ( average number of beam crossings during periods of activity ) of rye mutants is comparable with that of wild type controls , suggesting it is not a hyperactive mutant ( Figure 1B ) . rye heterozygotes have slightly less sleep than wild type , suggesting that the rye mutation is partially dominant ( Figure 1B , C ) . The dramatic reduction of baseline sleep results largely from a shortening of the average sleep episode duration ( Figure 1D ) , which is indicative of a defect in sleep maintenance . The average number of sleep episodes at night is actually increased significantly in rye mutants , perhaps because the sleep homeostat senses sleep loss and compensates by initiating more bouts . We also assayed rye mutants in sleep deprivation assays and found that they did not lose much additional sleep in response to mechanical stimulation ( data not shown ) . We surmise that an inability to sustain sleep bouts leads to the buildup of high sleep need in rye mutants , which makes them resist any further sleep loss . 10 . 7554/eLife . 01473 . 003Figure 1 . Identification of a short sleep mutant , redeye ( rye ) , through a chemical mutagenesis ( EMS ) screen . ( A ) Histogram depicting average sleep levels in females from homozygous EMS-mutagenized lines ( n = 1857 ) . The mean sleep value for each mutant was calculated by assaying sleep in 4–8 individual flies in the presence of 12 hr L-12 hr D cycles . Average sleep for all lines is indicated on the X axis in increments of 25 min . The Y axis depicts the number of mutant lines within each group . The dashed lines mark the sleep values that correspond to plus/minus 3× standard deviation of the mean . The arrowhead indicates the redeye mutant . ( B ) Top: average activity pattern of females from control lines and from rye mutants recorded in 12hr L-D cycles ( n = 14–16 in 5 days ) . Bottom: sleep profiles with standard error bars . Black: lines isogenized on the 3rd chromosome for chemical mutagenesis; Blue: flies heterozygous for rye and the isogenized control chromosome; Orange: rye homozygotes . ( C ) Mean values of total sleep , daytime sleep and night-time sleep for rye mutants ( Figure 1—source data 1 ) . For each genotype 14–16 flies were assayed over a 5 day period . Bars represent standard error . One-way ANOVA was performed followed by Tukey post hoc analysis . * represents p<0 . 05 , p<0 . 01 and p<0 . 001 . ( D ) Sleep quality of rye mutants: Daytime and night-time sleep episode length and episode number are plotted in the box-and-whisker diagram ( Figure 1—source data 1 ) . The middle line represents the median value; Bottom and top line of each box represent 25% and 75% respectively; Bottom bar and top bar represent 5% and 95% respectively . ns: not significant , *p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 00310 . 7554/eLife . 01473 . 004Figure 1—Source data 1 . Measurement of sleep duration for all EMS mutants and sleep analysis of rye mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 004 As the circadian clock regulates the timing of sleep and some clock mutants show reduced sleep ( Hendricks et al . , 2003 ) , we also assayed rye flies for defects in circadian rhythms . These endogenously generated rhythms are best monitored under constant conditions , in the absence of cyclic environmental cues , so we monitored behavior in constant darkness ( DD ) . rye mutants display robust rest:activity rhythms in DD ( ∼62% rhythmicity ) despite showing prolonged duration of activity ( Figure 2A; Table 1 ) . Of the ‘arrhythmic’ rye homozygotes detected , half ( 6/11 ) actually displayed a rhythm of ∼12 hr period length , which likely resulted from persistence of bimodal behavior ( typically seen in light:dark ) in DD . Thus , most rye homozygotes are rhythmic . Consistent with the intact circadian behavior , circadian cycling of the clock protein , PERIOD ( PER ) , is normal in central clock cells of rye mutants . The peaks and troughs of expression , as well as the subcellular localization , at different times of day are similar to those in iso controls ( Figure 2B ) . Thus , PER is largely nuclear at ZT20 , entirely nuclear at ZT2 , expressed at low levels at ZT8 and undetectable at ZT14 . We conclude that only the homeostatic regulation of sleep is affected in rye mutants . 10 . 7554/eLife . 01473 . 005Figure 2 . rye homozygotes display normal circadian rhythms and clock protein cycling . ( A ) Double plotted activity records of iso controls and rye homozygotes in DD . Rhythmic , but prolonged cross-beam activity was recorded in rye homozygotes . ( B ) Immunostaining of PERIOD in ventral lateral neurons ( LNvs ) of iso controls and rye homozygotes at CT time points on the second day of DD following LD entrainment . PER oscillates in these clock neurons in iso as well as in rye homozygotes . PDF labels the small and large LNvs . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 00510 . 7554/eLife . 01473 . 006Table 1 . Rhythmic behavior of rye homozygotes in constant darknessDOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 006GenotypeRhythmicity% ( n ) Period ± SEM ( h ) †FFT ± SEM†W1118 ( iso ) 93 . 8% ( 30/32 ) 23 . 21 ± 0 . 040 . 051 ± 0 . 005rye62 . 1% ( 18/29 ) 23 . 66 ± 0 . 090 . 054 ± 0 . 005Flies were entrained to a light–dark cycle ( 12 hr/12 hr ) for 3 days before being moved into constant darkness ( DD ) . Behavior was analyzed from day 3 to day 11 in DD , and flies with a fast fourier transform value ( FFT ) above 0 . 01 were considered rhythmic . †Period lengths and FFT values of rhythmic flies are listed as average value plus minus the standard error of mean ( SEM ) . We identified the sleep-altering lesion in rye mutants through a combination of recombination mapping and deep-sequencing . Despite starting with a newly isogenized line , deep-sequencing identified many nucleotide changes in the rye background relative to the parental control and so considerable mapping was required to pinpoint the relevant mutation . Meiotic recombination analysis positioned rye between two genetic markers , thread and curled , in the centromeric region of the 3rd chromosome . Further mapping studies , using self identified single nucleotide polymorphism ( SNP ) markers , narrowed it down to ∼11 mega-bases ( Figure 3A , Figure 3—source data 1 ) . For finer localization , we utilized the genome-wide sequence data , which provided 40× and 10× genomes worth of nucleotide sequence , corresponding to 95% and 88% coverage , for the rye and iso31 genomes respectively . Following alignment with the published Drosophila genome , unique nucleotide polymorphisms present in rye mutants were pooled together as potential EMS-induced mutations ( n = 26 , 224 ) . Within the 11 Mb region identified through recombination mapping , there were 1457 potential mutations , but only nine that produced an amino-acid change in coding regions . Sanger sequencing allowed us to exclude seven of these . Two turned out to be deep-sequencing errors and five polymorphisms were also present in the wild type iso stock but not detected due to the low coverage of deep-sequencing in this stock . One polymorphism , in a gene annotated CG7320 , was at a site that was also polymorphic in iso , although the amino acid change was different in the two strains . Because of the change in iso , which does not produce a sleep phenotype , we considered CG7320 a less likely candidate for rye . The last of the nine changes identified by deep-sequencing was a C-T transition ( typical for EMS-induced mutations ) that leads to a threonine to methionine change in CG12414 and is present only in rye mutants ( Figure 3B ) . CG12414 encodes an α subunit of a nicotinic acetylcholine receptor ( nAchR ) ( Lansdell and Millar , 2000 ) . Five alternatively spliced variants produced by this gene are predicted to generate three possible open reading frames with one ( pC/pG ) containing an intact ligand bind domain ( LBD ) ( Figure 3C ) . The predicted protein is homologous to several nAChR subunits , including α3 , α2 , α6 , α4 and α7 in order of similarity . Alignment of CG12414 with α3 and α7 using the ClustalW2 algorithm shows that the rye mutation is at the junction between the LBD and the trans-membrane domain ( TM ) ( Figure 3D , E ) . Interestingly , this junction region is highly conserved across species and the specific threonine mutated in rye is conserved in all α7 subunits ( Figure 3D ) . 10 . 7554/eLife . 01473 . 007Figure 3 . Genetic mapping and genome-wide deep-sequencing reveal a missense mutation in a nicotinic acetylcholine receptor α subunit gene in rye mutants . ( A ) Through genetic mapping , using classical phenotypic markers and newly identified SNP markers , rye was mapped between two SNP markers ( SNP_L and SNP_R ) near the centromere on the 3rd chromosome . ( B ) Paired-end genomic deep-sequencing of rye mutants identified a missense mutation in CG12414 . We confirmed this through Sanger sequencing . This C-T/G-A transition that generated the ryeT227M allele is typical of EMS-induced mutations . ( C ) Schematic representation of the rye candidate gene , a nicotinic acetylcholine receptor ( CG12414: nAcRα-80B ) . The gene spans ∼100 kb with alternative splicing predicted to produce five transcripts ( RC-RG ) . The spliced forms are predicted to translate into three proteins isoforms ( pC-pG ) . pC ( pG ) produces the largest protein with the longest ligand binding domain ( LBD ) . ( D ) Alignment of the partial sequence of Drosophila RYE with nAChR α3 and α7 subunits in other animals using ClustalW2 . The region shown is at the boundary of the ligand binding domain ( LBD ) and transmembrane domain ( TM ) , and is evolutionarily conserved . T227 in RYE is marked in bold form . ‘’ identity; ‘:’ high similarity; ‘ . ’ similarity . ( E ) Protein sequence analysis predicts four transmembrane domains ( TM ) in RYE , which is typical of most nAChR proteins . RYE appears to contain a single ligand binding domain ( LBD ) in the extracellular region , and four TMs with three loop regions ( L1-L3 ) . The mutated threonine227 is in the LBD , close to the beginning of the TM . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 00710 . 7554/eLife . 01473 . 008Figure 3—Source data 1 . Recombination mapping of the rye mutation using a chromosome marked with visible markers h , th , cu , sr , e , and using SNP markers . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 008 The genetic mapping and deep-sequencing data combined implicated CG12414 as a candidate gene responsible for the rye sleep phenotype . To verify this function for CG12414 , we attempted to rescue rye mutants with a CG12414 transgene . To drive expression in areas that normally express this gene , we cloned the CG12414 promoter region ( 1 . 8 kb ) and generated a GAL4 construct . This GAL4 is expressed broadly in the Drosophila brain ( Figure 4A ) . We also cloned the CG12414 cDNA into a UAS construct and crossed the GAL4 and UAS transgenes into an iso31 stock to ensure a homogenous genetic background . These transgenes were then introduced into rye mutants . While either transgene alone did not alter the rye short sleep phenotype , the rye promoter driving RYE expression ( ryeP > uas-rye ) partially rescued the mutant phenotype and restored total sleep to levels seen in rye heterozygotes ( Figure 4B ) . This is expected , given that ryeT227M is partially dominant over wild type . In a wild type background , ryeP > uas-rye does not increase sleep length ( data not shown ) , suggesting that the increase in sleep is specific to the mutant . Thus , we identified the gene responsible for the short sleep phenotype in rye mutants , and henceforth refer to this gene CG12414 , previously called nAChR-80B , as redeye . 10 . 7554/eLife . 01473 . 009Figure 4 . An alpha subunit of the nicotinic acetylcholine receptor accounts for the rye mutant phenotype . ( A ) Expression pattern of ryeP-GAL4 . Rye-GAL4 was used to express nGFP , which is visualized with an anti-GFP antibody . The anti-nc82 staining marks the neuropil . ( B ) Expression of the alpha subunit , as described in Figure 3 , increases sleep duration in rye mutants ( Figure 4—source data 1 ) . Left: sleep profiles , with standard error , of rye mutants and mutants expressing a UAS construct of the putative rye cDNA under control of its own promoter ( ryeP-Gal4 ) in a 12:12 LD cycle . Right: quantification of sleep length . p<0 . 01 , p<0 . 001 . ( C ) Reduction of rye expression in rye neurons through the expression of an RNAi construct , together with Dicer2 , diminishes sleep length ( Figure 4—source data 1 ) . Left: sleep profile in a 12:12 LD cycle . Right: quantification of sleep length . p<0 . 05 . ( D ) Western blot analysis shows reduced expression of RYE when actin-GAL4 is used to drive rye RNAi ( VDRC#11392 ) with Uas-Dicer2 in female and male flies . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 00910 . 7554/eLife . 01473 . 010Figure 4—Source data 1 . Sleep behaviour of transgenically rescued rye mutants and rye RNAi lines . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 010 To further characterize rye function , we knocked down RYE expression through RNA interference . A rye RNAi line ( #11392 ) from the VDRC stock center in Vienna was crossed into the ryeP-GAL4 line along with a UAS-Dicer2 transgene , included to improve efficacy of the RNAi . Levels of RYE were reduced ( Figure 4D ) , as was the total sleep length ( Figure 4C ) . Since ryeT227M reduces sleep as well ( Figure 1 ) , these data suggest that the T227M mutant allele causes loss of RYE function . The sleep phenotype produced by rye knock-down is rather mild in comparison with the phenotype of the homozygous rye mutant , possibly due to inefficient knockdown in relevant cells or because ryeT227M is a dominant-negative mutation . The nicotinic acetylcholine receptor forms a pentameric structure , consisting of five homo-oligomeric or hetero-oligomeric subunits ( Miwa et al . , 2011 ) , so a dominant negative mutation in any one subunit may well interfere with activity of the complex . We next addressed if rye interacts with other known sleep mutants , in particular with a mutant we identified previously , sleepless ( sss ) , that has an extreme short-sleeping phenotype ( Koh et al . , 2008 ) . While transheterozygotes of rye and sss ( sss/+; rye/+ ) showed sleep duration comparable to that of rye/+ alone ( data not shown ) , rye heterozygotes carrying a genomic sss transgene , which increases SSS expression above that of wild type controls , showed a further reduction of sleep ( Figure 5A ) . The effect was observed with three independent insertions of the sss transgene , indicating that it was not due to insertion site effects ( Figure 5B ) . We note that this sss transgene does not have a significant effect on sleep in wild type flies although it completely rescues the sleep phenotype of sss mutants ( Koh et al . , 2008 ) . Thus , the effect is specific to rye mutants . 10 . 7554/eLife . 01473 . 011Figure 5 . Overexpression of sss exacerbates the rye phenotype by repressing RYE activity . ( A ) An extra copy of SSS reduces sleep in rye heterozygotes . Sleep profile of rye heterozygotes with or without a genomic sss transgenic line inserted on the third chromosome ( sssTG2 ) and of sss transgenics alone . ( B ) Quantification of sleep in the genotypes as indicated ( Figure 5—source data 1 ) . TG1 and TG3 are additional independent genomic sss transgenes inserted on the 3rd and X chromosomes respectively . Sleep reduction was observed in rye transheterozygotes with all three sss transgenic lines . p<0 . 05 . ( C ) Heterologous expression of SSS reduces current following ACh application in Xenopus oocytes expressing Drosophila RYE and human nAChR β2 ( Figure 5—source data 1 ) . p<0 . 05 . ( D ) Heterologous expression of SSS reduces current following ACh application in Xenopus oocytes expressing human nAChR α4β2 ( Figure 5—source data 1 ) . **p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 01110 . 7554/eLife . 01473 . 012Figure 5—Source data 1 . Sleep analysis of rye mutants overexpressing sss , and whole-cell membrane current recording of Xenopus oocytes . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 012 SSS is a GPI-anchored protein that regulates the Shaker potassium channel and its predicted structure resembles that of a specific class of toxins ( Wu et al . , 2010; Dean et al . , 2011 ) . It is most similar to the mammalian Lynx-1 protein , which is known to inhibit activity of the nicotinic acetylcholine receptor ( Miwa et al . , 2011 ) . The fact that SSS exacerbated the phenotype of flies heterozygous for rye , which have reduced activity of nAChR , suggested that SSS also regulates nAChRs . We tested a possible Lynx-1-like function of SSS in regulating nAChR activity by using a heterologous expression system , specifically Xenopus laveis oocytes . For functional expression of wild type RYE , we expressed it with a human β2 subunit , as insect β subunits typically do not work in heterologous systems ( Millar and Lansdell , 2010 ) . cRNAs encoding sss , wild type rye and human β2 were injected into oocytes , and 4–7 days later whole cell currents were measured following application of ACh ( Kuryatov and Lindstrom , 2011 ) . Heterologous expression of insect nAChRs has , in general , been challenging ( Millar , 2009 ) , which accounts for the low current values observed ( Figure 5C ) . However , we still detected a significant reduction of nAChR current upon co-expression with SSS ( Figure 5C ) . We also co-expressed in vitro transcribed sss and human α4 and β2 subunits and found that SSS also inhibits the robust current produced by the human receptor complex ( Figure 5D ) . Thus , in addition to its sleep-promoting role reported previously , SSS may also promote wake by repressing activity of nAChRs . As discussed below , loss of the sleep-promoting role has a dominant effect in sss mutants . We next asked if levels of rye vary over the course of the day . Groups of wild type flies were housed in a 12 hr-light and 12 hr-dark incubator and harvested at different Zeitgeber times ( ZT ) . Zeitgeber time defines time based on the environmental stimulus , so ZT0 corresponds to light-on and ZT12 to light-off . RNA was extracted and subjected to quantitative PCR analyses . While period ( per ) , a well-known circadian clock component , oscillates with a circadian rhythm ( Zheng and Sehgal , 2012 ) , rye mRNA levels were found to be constant ( Figure 6A ) . 10 . 7554/eLife . 01473 . 013Figure 6 . The RYE protein is expressed cyclically in association with the sleep state , but independently of the circadian clock . ( A ) Quantitative PCR analyses show oscillations of per mRNA ( left panel ) , but constant levels of rye mRNA ( right panel ) . actin was used as a normalization control ( Figure 6—source data 1 ) . ( B ) Left: a representative western blot of head extracts from wild type flies shows cyclic expression of RYE with two daily peaks , in the middle of the day ( ZT6-10 ) and in the middle of the night ( ZT18-22 ) , corresponding to the sleep state ( Figure 1B ) . PER , in contrast , shows only one daily peak . MAPK was used to control for loading . Right: densitometry quantification of western blots with error bars representing standard error ( n = 8 ) . RYE value at ZT0 is set as 1 ( Figure 6—source data 1 ) . p<0 . 05 , *p<0 . 01 . ( C ) RYE cycling is similar to wild type in Clkjrk flies , indicating that cycling per se does not require a functional clock . However , the phase is variable , so for quantification purposes five independent experiments were split into two groups of roughly similar phase ( Figure 6—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 01310 . 7554/eLife . 01473 . 014Figure 6—Source data 1 . qPCR analysis of per and rye expression during LD cycle and densitometry quantification of RYE expression . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 01410 . 7554/eLife . 01473 . 015Figure 6—figure supplement 1 . RYE expression in a LD cycle ( repeats ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 01510 . 7554/eLife . 01473 . 016Figure 6—figure supplement 2 . Supporting data for Figure 6C . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 016 To examine protein levels of RYE , we generated an antibody against the cytoplasmic L3 domain of RYE ( Figure 3E ) , which is the most variable region across all nAChR subunits ( Lindstrom , 2003 ) . On western blots , the antibody recognizes a band ( or sometimes a doublet ) of ∼60 Kd , which is the size predicted for the full-length isoform . Surprisingly , in contrast to the flat mRNA levels , levels of RYE protein show a robust rhythm with two peaks each day , one at around ZT6-10 and the other at ZT18-22 ( Figure 6B , Figure 6—figure supplement 1 ) . Interestingly , these correspond to the two daily times of sleep , the afternoon siesta and night-time sleep ( Figure 1B ) , suggesting that RYE is expressed in phase with sleep . The cycling of RYE persists in constant darkness , as do rhythms of sleep ( data not shown ) . However , RYE does not appear to be regulated like other cycling clock components , such as PER , which usually show only one peak ( Figure 6B ) . Furthermore , we found that RYE continues to cycle in Clkjrk flies that lack a circadian clock , although the phase of RYE expression tends to be more variable in these flies ( Figure 6C , Figure 6—figure supplement 2 ) . These data suggest that the RYE oscillation per se does not require process C of the two process model , but may depend upon process S ( sleep homeostasis ) . However , the circadian clock may function to provide more precise timing of RYE expression , perhaps through its control of sleep onset and offset . Increases in RYE expression could occur as a consequence of sleep , and thereby reflect the sleep state , but in that case levels should remain high throughout the sleep state . Given that RYE levels do not remain high throughout the night or through the afternoon siesta , another possibility is that elevations in RYE correspond to sleep drive and so occur at the time of sleep onset . To test this idea , we assayed RYE expression in short-sleeping mutants . These mutants have low sleep levels , but probably have high sleep drive that cannot be implemented . As shown in Figure 7 , for the most part RYE expression exhibits two peaks in these mutants , but the phase is more variable ( Figure 7—figure supplements 1–3 ) , indicating that defects in sleep behavior affect the RYE expression pattern . Interestingly , overall RYE values are higher in fumin ( Kume et al . , 2005; Ueno et al . , 2012 ) , sss ( Koh et al . , 2008 ) and insomniac ( Stavropoulos and Young , 2011 ) mutants than in wild type controls . As noted above , the sss mutation affects channel activity , fumin is a mutation in the dopamine transporter ( short sleep is thought to result from the high levels of dopamine in the synaptic cleft ) and insomniac is thought to affect protein turnover . The one feature these mutants have in common is short sleep , yet they all have elevated RYE expression during an LD cycle ( Figure 7A–C ) . Based upon these data , we suggest that RYE reflects sleep need . It is elevated at the time of sleep onset and is further increased in short-sleeping mutants because they have high sleep need . 10 . 7554/eLife . 01473 . 017Figure 7 . RYE levels are elevated in short-sleeping mutants . ( A ) Western analyses of RYE expression over the course of a day in iso controls and fumin mutants . While RYE still cycles in fumin , the mean value of all six daily time points from three independent experiments ( n = 3 ) shows that overall levels of RYE are higher than in wild type controls ( Figure 7—source data 1 ) . p<0 . 05 . ( B ) Western analyses of RYE expression over the course of a day in iso controls and sss mutants . As in A , the quantification reflects the mean across the day from three independent experiments ( n = 3 ) . p<0 . 05 . ( C ) Western analyses of RYE expression over the course of a day in iso controls and insomniac mutants . Quantification as above ( n = 3 ) . p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 01710 . 7554/eLife . 01473 . 018Figure 7—Source data 1 . Densitometry quantification of RYE expression in short sleep mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 01810 . 7554/eLife . 01473 . 019Figure 7—figure supplement 1 . Supporting data for Figure 7A . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 01910 . 7554/eLife . 01473 . 020Figure 7—figure supplement 2 . Supporting data for Figure 7B . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 02010 . 7554/eLife . 01473 . 021Figure 7—figure supplement 3 . Supporting data for Figure 7C . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 021 If the expression of RYE changes in response to the accumulation of sleep need , it should also be affected by acute sleep deprivation . We mechanically deprived flies of sleep in the second half night ( 6 hr ) in an LD cycle , and confirmed , through the presence of a substantial sleep rebound the following morning , that the flies were successfully deprived ( Figure 8A; Koh et al . , 2008 ) . RYE levels were assayed during the 5 hr sleep rebound window . In the non-deprived control , RYE levels were low at ZT0 and high at ZT5 , as noted above . Sleep-deprived flies had high levels of RYE immediately following deprivation ( ZT0 ) , but levels were low at ZT5 ( Figure 8A , Figure 8—figure supplement 1 ) . This fits the profile for sleep drive , which is expected to be high at the end of deprivation , but then dissipated over the course of rebound sleep . 10 . 7554/eLife . 01473 . 022Figure 8 . RYE expression is under control of the sleep homeostat . ( A ) Left: sleep profile of sleep deprived flies . Blue line: non-deprived controls , Grey line: sleep deprived flies . Arrowheads point to the 6 hr sleep deprivation ( SD ) window ( ZT18-24 ) and sleep rebound the following morning ( ZT0-5 ) . Right: a representative western blot of fly head samples at ZT0 and ZT5 in sleep-deprived ( SD ) animals and non-deprived ( ND ) controls . RYE levels are high immediately following sleep deprivation ( SD0 ) and dissipate after sleep rebound ( SD5 ) . In contrast , the circadian clock is not affected by SD , since PER cycling is comparable in the SD group and the non-deprived control ( ND ) . Densitometry quantification of the western data ( n = 5 ) is shown below with error bars showing standard error ( Figure 8—source data 1 ) . *p<0 . 05 . ( B ) Model for the role of RYE in the homeostatic regulation of sleep: We propose that RYE is required to maintain sleep , and posttranslational regulation of RYE reflects homeostatic sleep drive . Sleep need builds up during wakefulness and upregulates levels of RYE , which are essential to maintain sleep behavior . Homeostatic drive dissipates during sleep , and levels of RYE are reduced , leading to wakefulness . SSS represses RYE activity , thus acting as a wake-promoting factor in this particular context . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 02210 . 7554/eLife . 01473 . 023Figure 8—Source data 1 . Densitometry quantification of RYE expression after SD . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 02310 . 7554/eLife . 01473 . 024Figure 8—figure supplement 1 . Supporting data for Figure 8A . DOI: http://dx . doi . org/10 . 7554/eLife . 01473 . 024 We suggest that RYE represents a molecular correlate of delta power , a characteristic of an electroencephalogram ( EEG ) that reflects sleep drive . Recently , a few other molecules were reported to change with sleep drive , but the effects were at the level of the mRNA , the magnitude of the increase was less than we report here for RYE and loss of the molecules did not affect baseline sleep duration ( Seugnet et al . , 2006; Maret et al . , 2007; Naidoo et al . , 2007 ) . In addition , only one is expressed cyclically ( Nelson et al . , 2004 ) , indicating that others reflect sleep drive only under pathological conditions of sleep deprivation . RYE levels oscillate robustly in a daily cycle , although the phase is not as coherent as seen for circadian clock proteins . The timing of the peak varies within a temporal range , such that there is almost always a daytime peak and a night-time peak but not necessarily at the exact same time ( Figure 6—figure supplement 1 ) . We suggest that RYE cycles under control of the sleep homeostat , which may not time behavior as precisely as the circadian clock , perhaps because sleep can be influenced by many factors . The variability in RYE cycling is particularly pronounced in short-sleeping mutants and in the ClkJrk circadian clock mutant ( Figure 6—figure supplement 2 , Figure 7—figure supplements 1–3 ) , suggesting that the clock does influence RYE expression although it is not required for its cycling in an LD cycle . Interestingly , RYE cycles exclusively at the level of the protein , indicating translational or post-translational mechanisms . It is worth noting that a recently identified sleep regulator , Insomniac , is a component of specific protein degradation pathways in the cell ( Stavropoulos and Young , 2011 ) . Although our study indicates that RYE cycling does not require Insomniac ( Figure 7C ) , it is possible that it is regulated by other protein turn-over machinery . Thus , translational/posttranslational regulation appears to be part of the mechanism of sleep homeostasis . We show that RYE not only reflects sleep drive , but is also required for sleep maintenance ( Figures 1 , 4 and 8 ) . Given that RYE is induced by sleep deprivation and it promotes sleep , one might expect over-expression of the protein to increase sleep . However , transgenic expression of rye in a wild type background does not increase sleep , suggesting that while rye is necessary , it is not sufficient for sleep onset . We cannot exclude the possibility that RYE functions together with other signals as part of the sleep-inducing homeostatic drive . On the other hand , it is also possible that transgenic expression does not produce adequate amounts of RYE protein in relevant cells . This might be the case if RYE is tightly regulated at the level of protein stability . For the moment , though , we prefer the parsimonious explanation noted above , that RYE is not part of the homeostat , but immediately downstream of it . Acetylcholine signaling has been long proposed as an arousal factor , as the nAChR complex is a cation channel that normally promotes neuronal activity and ACh is released during wakefulness in mammals ( Platt and Riedel , 2011 ) . In contrast , our study indicates that at least one nAChR subunit ( RYE ) promotes sleep in the fly . There are more than 10 paralogs of nAChR subunits in the fly genome . One possibility is that RYE is expressed specifically in sleep promoting neurons , while other subunits of AChRs are in wake-promoting cells . An increase in ACh during wakefulness may contribute to the accumulation of sleep drive and to the increase in RYE ( Lindstrom , 2003 ) . Alternatively , sleep drive may increase RYE independently of ACh , but in either scenario , RYE then promotes sleep . The precise site of RYE action is currently not known . rye-gal4 driven GFP is expressed widely in the brain ( Figure 4A ) , but we cannot be sure that endogenous RYE is as widespread , as our antibody was not effective in immunohistochemistry experiments , and we find that GAL4 drivers are often quite promiscuous ( unpublished observations ) . SSS was previously identified as a sleep promoting factor , essential for maintaining baseline sleep and for homeostatic rebound ( Koh et al . , 2008 ) . An interaction between rye and sss is therefore not surprising . What is surprising is that overexpression of SSS promotes wakefulness in ryeT227M heterozygotes ( Figure 5A , B ) . SSS is a GPI-anchored protein that functions as a neuronal modulator . Previous studies indicate that SSS promotes activity of the voltage-gated potassium channel , Shaker ( Wu et al . , 2010 ) . In this study , we report that SSS acts like a brake on nAChR ( RYE ) activity ( Figure 5C , D ) , as does Lynx-1 , a SSS-like molecule , in mammals ( Miwa et al . , 2011 ) . Although the data we show for Drosophila receptors used only the RYE α subunit , it is likely that SSS also inhibits activity of other Drosophila nAChR receptors . As both sssP1 ( a null mutation ) and sssP2 ( a hypomorphic allele ) are short-sleeping mutants ( Koh et al . , 2008 ) , we propose that the overall effect of SSS is to promote sleep . The reduced sleep in sss mutants probably results from an increase of neuronal excitability , through inactivation of potassium channels ( Shaker ) ( Dean et al . , 2011 ) , or from hyperactivity of nAChR channels in wake-promoting neurons . Thus , typically the sleep-inhibiting effect of SSS , mediated through RYE , is masked by these other more dominant influences . However , in a sensitized background ( i . e . , rye/+ ) , this effect is evident . RYE promotes sleep , and so loss of RYE results in a decrease in sleep , which is further impacted by SSS overexpression ( Figure 8B ) . We note that there are some caveats to these data . For instance , the ryeT227M allele could confer a neomorphic function that accounts for the interaction with sss . Likewise , the effects in oocytes could be non-physiological , not necessarily reflecting what happens in the fly brain . However , given that we observe interactions in these two very different types of assays , and both assays indicate repression of nAchR function by SSS , which is the effect predicted from the role of the mammalian SSS-like protein , Lynx1 , we believe SSS does indeed regulate nAchRs such as RYE . Interactions between SSS and nicotinic acetylcholine receptors are also reported by recent work from another laboratory ( W Joiner , personal communication ) . It is interesting that genes identified through independent genetic screens in Drosophila are turning out to interact with one another . SSS and Shaker were isolated independently as sleep-regulating genes ( Cirelli et al . , 2005; Koh et al . , 2008 ) , and subsequently shown to interact , and now we find that RYE interacts with SSS . Given that each of these genes represents a relatively infrequent hit in an unbiased screen , the interactions suggest that genetic approaches are converging upon specific sleep-regulating pathways . Interestingly , a recent GWAS study for sleep-altering loci in humans identified significant effects of SNPs in an nAchR subunit as well as in a regulatory subunit of Shaker , suggesting that these mechanisms are also conserved across species ( Allebrandt et al . , 2013 ) . Wild type iso31 ( Ryder et al . , 2004 ) was crossed with TM2/TM6C , Sb ( Bloomington stock #5906 ) , and a single male progeny was selected and crossed with #5906 to balance the 3rd chromosome and generate a line isogenic for this chromosome . Flies were housed in Percival incubators ( Perry , IA ) and beam–break activity was recorded with the Trikinetics DAM system ( http://www . trikinetics . com/ ) . Pysolo ( Gilestro and Cirelli , 2009 ) software was used to analyze and plot sleep patterns . Sleep deprivation was achieved through repeated mechanical shaking ( 2 s randomized shaking in every 20-s interval ) . A 3rd chromosome marker line h , th , cu , sr , e ( Bloomington stock #576 ) was crossed to rye/TM6C , Sb , and heterozygote female progeny ( h , th , cu , sr , e/rye ) were further crossed to TM2/TM6C , Sb ( #5906 ) . A single recombinant male offspring was back-crossed to #576 to score the genotype of recessive markers and also back-crossed to rye for behavioral analysis to score the rye genotype . Overlap in the genotypes of recombinants narrowed down the location of the rye mutation to the region between markers th and cu . Genomic DNA of homozygous recombinants was subject to SNP analysis . The primer set 5′TGTTTAGTGGTGTTGTGTGAGC3′ and 5′GCCGAGTGTCATCGCCTTTG3′ for SNP_L and the primer set 5′AAAGGTCATCTTGCTTCGGAGTTG3′ and 5′GGAGTGGCTTCCTCGTCATC3′ for SNP_R were used for PCR amplification . Nucleotide polymorphisms were identified between iso and #576 . Thus , those recombinants were scored accordingly and rye was further narrowed down to a region between those two SNP markers . Fragmented genomic DNA ( ∼400 bp in size ) obtained through the Covaris instrument ( Woburn , MA ) was subjected to Illumina paired-end DNA library preparation . The libraries were amplified for 10 PCR cycles prior to Illumina Hi-Seq analyses . SNP calling analyses identified nucleotide polymorphisms in both rye mutants and iso flies . The primer set 5′GGCTCGATGTGCTTTCAAGAGTTC3′ and 5′CATGCCAGATGAGTGCGTTTC3′ was used to clone the rye promoter region from genomic DNA derived from the iso stock . The primer set 5′TTTAGGCTTAGTCCGCTACC 3′ and 5′AATGTCGTGGTTTGAAGTGC 3′ was used to clone the full length rye cDNA . The PhiC31 integration system was used for injection . The rye promoter construct was specifically inserted onto the 2nd chromosome and the Uas-rye construct was integrated on the 3rd chromosome . The primer set 5′ATGCATCATCACCATCACCATAGTATCTGCGTGACGGTTGTTG3′ and 5′GTGGTGGTGCACAACTGCCAACGTGAATATCC 3′ was used to clone the L3 epitope of RYE . A GST-RYEL3 fusion protein was expressed in Escherichia coli , excised from a PAGE gel and injected into rats for antibody generation . Drosophila genomic DNA: flies ( 3–15 ) were frozen and homogenized in DNA extraction buffer . After LiCl/KAc precipitation , supernatants were subject to isopropanol precipitation . DNA pellets were dissolved in TE for sequencing analyses and PCR amplification . RNA: adult flies were collected at indicated time points and fly heads ( 10∼20 ) were subject to Trizol extraction ( Ambion , Life Technologies , Grand Island , NY ) . cDNA libraries were made through high-capacity cDNA reverse transcription kits ( Applied Biosystems , Life Technologies , Grand Island , NY ) . Quantitative PCR analysis was performed using SYBR green reagents in the Applied Biosystems 7000 sequence detection system . The primer set 5′AGTTGAATGGAAGCCACCAGC3′ and 5′TGTTCATCCATGTGCCTCAG3′ was used for qPCR analysis of rye expression . Protein: flies were collected at indicated time points and fly heads ( ∼5 ) were prepared for protein extraction and for Western analyses as previously described ( Luo and Sehgal , 2012 ) . Adult fly brains were dissected in cold PBS buffers at indicated CT time points . Brains were fixed in 4% PFA for 30 m and incubated with 5% Normal Donkey Serum ( NDS ) for 1 hr at the blocking step . Incubation with primary rabbit anti-PER ( 1:1000 ) , mouse anti-PDF ( 1:1000 ) and rabbit anti-GFP , mouse anti nc82 was performed overnight in the cold room . After extensive washing , brains were incubated with the donkey anti-rabbit or anti-mouse secondary antibody ( 1:1000 ) for 2 hr . Images were taken under a Leica confocal microscope . Oocytes were removed surgically from Xenopus laevis and placed in an OR-2 solution containing 82 . 5 mM NaCl , 2 mM KCl , 1 mM MgCl2 , 5 mM HEPES , and pH 7 . 5 . They were defolliculated in this buffer containing 2 mg/ml collagenase type IA ( Sigma , St . Louis , MO ) for 1 . 5 hr . After defolliculation , oocytes were incubated at 18°C in 50% L15 ( Invitrogen , Carlsbad , CA ) 10 mM HEPES , pH 7 . 5 , 10 units/ml penicillin , and 10 g/ml streptomycin at 18°C . 3–6 days after injection , whole-cell membrane currents evoked by acetylcholine were recorded in oocytes at room temperature with a standard two-electrode voltage-clamp amplifier ( Oocyte Clamp OC-725; Warner Instrument , Hamden , CT ) . Recordings were performed at a holding potential of −50 mV . All perfusion solutions contained 0 . 5M atropine to block responses of endogenous muscarinic AChRs that might be present in oocytes . Acetylcholine was applied by means of a set of 2-mm glass tubes directed to the animal pole of the oocytes . Application was achieved by manual unclamping/clamping of a flexible tube connected to the glass tubes and to reservoirs with the test solutions . The recording chamber was perfused at a flow rate of 15–20 ml/min . cRNAs for human AChR subunits 4 and 2 and Drosophila sss were synthesized in vitro using SP6 RNA polymerase ( mMESSAGEmMACHINETM , Ambion ) . Oocytes were injected with 5 ng of cRNA of each of the subunits .	Almost all animals need to sleep , including most insects . In experiments in the 1980s , a group of rats that were completely deprived of sleep died within only a few weeks . Sleep has been implicated in processes including tissue repair , memory consolidation and , more recently , the removal of waste materials from the brain . However , a full understanding of why we sleep is still lacking . As anyone who has experienced jetlag can testify , the timing of the sleep/wake cycle is governed by the circadian clock , which leads us to feel sleepy at certain points of the day–night cycle and alert at others . The duration of sleep is regulated by a second process called sleep/wake homeostasis . The longer we remain awake , the more the body’s need for sleep—or ‘sleep drive’—increases , until it becomes almost impossible to stay awake any longer . Whereas many components of the circadian clock have been identified , relatively little is known about the molecular basis of this second process . Now , Shi et al . have identified a key component of the sleep/wake homeostatic system using the fruit fly and genetic model organism , Drosophila . Flies with a mutation in one particular gene , subsequently named redeye , were found to sleep only half as long as normal flies . While the insects were able to fall asleep , they would wake again only a few minutes later . Redeye encodes a subunit of a receptor that has previously been implicated in the control of wakefulness , known as the nicotinic acetylcholine receptor . Mutant flies had normal circadian rhythms , suggesting that their sleep problems were the result of disrupted sleep/wake homeostasis . Consistent with this , levels of redeye showed two daily peaks , one corresponding to night-time sleep and the second to the time at which flies would normally take an afternoon siesta . This suggests that redeye signals an acute need for sleep , and then helps to maintain sleep once it is underway . While redeye is not thought to be the factor that triggers sleep per se , it is directly under control of the sleep homeostat , and may be a useful biomarker for sleep deprivation . The fact that redeye was identified in fruit flies , a species whose genome has been fully sequenced , opens up the possibility of further studies to identify the genetic basis of sleep regulation .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Identification of Redeye, a new sleep-regulating protein whose expression is modulated by sleep amount
Postsynaptic cells can induce synaptic plasticity through the release of activity-dependent retrograde signals . We previously described a Ca2+-dependent retrograde signaling pathway mediated by postsynaptic Synaptotagmin 4 ( Syt4 ) . To identify proteins involved in postsynaptic exocytosis , we conducted a screen for candidates that disrupted trafficking of a pHluorin-tagged Syt4 at Drosophila neuromuscular junctions ( NMJs ) . Here we characterize one candidate , the postsynaptic t-SNARE Syntaxin 4 ( Syx4 ) . Analysis of Syx4 mutants reveals that Syx4 mediates retrograde signaling , modulating the membrane levels of Syt4 and the transsynaptic adhesion protein Neuroligin 1 ( Nlg1 ) . Syx4-dependent trafficking regulates synaptic development , including controlling synaptic bouton number and the ability to bud new varicosities in response to acute neuronal stimulation . Genetic interaction experiments demonstrate Syx4 , Syt4 , and Nlg1 regulate synaptic growth and plasticity through both shared and parallel signaling pathways . Our findings suggest a conserved postsynaptic SNARE machinery controls multiple aspects of retrograde signaling and cargo trafficking within the postsynaptic compartment . Synaptic connections form and mature through signaling events in both pre- and postsynaptic cells . The release of signaling molecules into the synaptic cleft depends on SNARE proteins that drive membrane fusion . This machinery is well understood for neurotransmitter release from the presynaptic cell: in response to an action potential , a v-SNARE in the synaptic vesicle membrane ( Synpatobrevin/VAMP ) engages t-SNARES in the presynaptic membrane ( Syx1 and SNAP-25 ) , forming a four-helix structure that brings the membranes into close proximity and initiates fusion ( Jahn and Scheller , 2006; Südhof and Rothman , 2009 ) . Although SNARE-dependent fusion drives membrane dynamics in all cell types , it is specialized in the presynaptic terminal to be Ca2+-dependent , employing Ca2+ sensors like Synaptotagmin 1 ( Syt1 ) to link synaptic vesicle fusion to Ca2+ influx following an action potential . The postsynaptic cell also exhibits activity-dependent exocytosis . Altering the composition of the postsynaptic membrane , including regulated trafficking of neurotransmitter receptors , is an important plastic response to neural activity ( Chater and Goda , 2014 ) . The postsynaptic cell also releases retrograde signals into the synaptic cleft to modulate synaptic growth and function . These retrograde messengers include lipid-derived molecules like endocannabinoids ( Ohno-Shosaku and Kano , 2014 ) , gases like nitric oxide ( Hardingham et al . , 2013 ) , neurotransmitters ( Koch and Magnusson , 2009; Regehr et al . , 2009 ) , neurotrophins ( Zweifel et al . , 2005 ) , and other signaling factors like TGF-β and Wnt ( Poon et al . , 2013; Salinas , 2005; Sanyal et al . , 2004; Speese and Budnik , 2007 ) . Adhesion complexes that provide direct contacts across the synaptic cleft also participate in retrograde signaling ( Futai et al . , 2007; Gottmann , 2008; Hu et al . , 2012; Mozer and Sandstrom , 2012; Peixoto et al . , 2012; Vitureira et al . , 2012 ) . Although retrograde signaling is a key modulator of synaptic function , little is known about how postsynaptic exocytosis is regulated and coordinated . Components of a postsynaptic SNARE complex have been recently identified in mammalian dendrites . The t-SNAREs Syntaxin 3 ( Stx3 ) and SNAP-47 are required for regulated AMPA receptor exocytosis during long term potentiation , while the v-SNARE synaptobrevin-2 regulates both activity-dependent and constitutive AMPAR trafficking ( Jurado et al . , 2013 ) . Stx4 has also been implicated in activity-dependent AMPAR exocytosis ( Kennedy et al . , 2010 ) . In Drosophila , a Ca2+-dependent retrograde signaling pathway relies on the postsynaptic Ca2+ sensor Syt4 . Syt4 vesicles fuse with the postsynaptic membrane in an activity-dependent fashion ( Yoshihara et al . , 2005 ) , and loss of Syt4 leads to abnormal development and function of the NMJ . Syt4 null animals have smaller synaptic arbors , indicating a defect in synaptic growth , and also fail to exhibit several forms of synaptic plasticity seen in control animals , including robust enhancement of presynaptic release in response to high frequency stimulation , and rapid budding of synaptic boutons in response to strong neuronal stimulation ( Barber et al . , 2009; Korkut et al . , 2013; Piccioli and Littleton , 2014; Yoshihara et al . , 2005 ) . However , a detailed understanding of how the postsynaptic cell regulates constitutive and activity-dependent signaling of multiple retrograde pathways is lacking . In addition to exocytosis , it is likely that many cellular processes including vesicle trafficking and polarized transport of protein and transcript are specialized to facilitate postsynaptic signaling . Identifying such regulatory mechanisms is crucial for understanding synaptic development and function . We conducted a candidate-based transgenic RNAi screen to identify regulators of postsynaptic exocytosis at the Drosophila NMJ , a model for studying glutamatergic synapse growth and plasticity ( Harris and Littleton , 2015 ) . Using a fluorescently tagged form of the postsynaptic Ca2+ sensor Syt4 , we screened for candidate gene products that disrupted the localization of Syt4 at the postsynaptic membrane . Here we describe our characterization of one candidate from this screen , Syntaxin 4 ( Syx4 ) . Drosophila Syx4 is the sole homolog of the mammalian Stx 3/4 family of plasma membrane t-SNAREs that also includes Syntaxin 1 ( Littleton , 2000 ) . The mammalian Stx3 and Stx4 homologs regulate activity-dependent AMPA receptor trafficking in mammalian neurons ( Jurado et al . , 2013; Kennedy et al . , 2010 ) , while Stx4 also participates in regulated secretory events in several other mammalian cell types , including insulin-stimulated delivery of the glucose transporter to the plasma membrane in adipocytes and glucose-stimulated insulin secretion from pancreatic beta cells ( reviewed by Jewell et al . , 2010 ) . Our results demonstrate that the Drosophila Syx4 homolog is essential for retrograde signaling , regulating the membrane delivery of both Syt4 and Neuroligin ( Nlg1 ) , a transsynaptic adhesion protein that plays important roles in synapse formation and function , and is linked to autism spectrum disorder ( ASD ) ( Bang and Owczarek , 2013; Bottos et al . , 2011; Südhof , 2008 ) . Through genetic interaction experiments , we define functions of the Syx4 , Syt4 , and Nlg1 pathway in regulating multiple aspects of synaptic growth and plasticity within the postsynaptic compartment . To identify regulators of Syt4 trafficking , we conducted a candidate-based RNAi screen at the NMJ . Our screening approach employed transgenic animals expressing Syt4 tagged with pHluorin , a pH-sensitive variant of GFP under the control of the UAS promoter ( UAS-Syt4-pH ) . When expressed with the muscle driver 24B-GAL4 , Syt4-pH protein decorates the postsynaptic membrane of the NMJ , overlapping with glutamate receptor ( GluR ) fields opposite active zones ( AZs ) ( Yoshihara et al . , 2005; Figure 1A ) . Syt4-pH is also found in numerous vesicular structures throughout the muscle , many of which overlap with the Golgi marker Lava lamp ( Lva ) ( Figure 1B ) . Postsynaptic expression of Syt4-pH rescues synaptic phenotypes previously reported in Syt4 null animals ( Figure 1—figure supplement 1 ) , including a decrease in the number of synaptic boutons and a decrease in the ability to grow new boutons ( “ghost boutons” , or GBs ) in response to strong neuronal stimulation ( Barber et al . , 2009; Korkut et al . , 2013; Piccioli and Littleton , 2014; Yoshihara et al . , 2005 ) . Thus , Syt4-pH is functional at the NMJ . 10 . 7554/eLife . 13881 . 003Figure 1 . A candidate RNAi screen for regulators of postsynaptic exocytosis . ( A , B ) Representative images of Syt4-pH expressed with the postsynaptic muscle driver 24B-GAL4 . Syt4-pH ( green ) accumulates in postsynaptic membrane that also contains domains of GluRIII ( magenta ) ( A ) . Syt4-pH also decorates numerous cytoplasmic puncta , many of which overlap with the Golgi marker Lva ( magenta ) , arrowheads ( B ) . ( C–F ) Examples of candidate RNAis affecting Syt4-pH localization: control ( C ) ; Syx4-RNAi reduces Syt4-pH at the membrane and causes a redistribution to prominent cytoplasmic puncta , arrowheads ( D ) ; Lasp-RNAi increases Syt4-pH at the membrane ( E ) ; and Syx7-RNAi causes a redistribution of Syt4-pH puncta around the NMJ without affecting the intensity at the membrane ( F ) . ( C′–F′ ) Close-ups of C–F . Scale bars = 7 μm ( A ) , 5 μm ( B–F ) , 2 μm ( C′–F′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 00310 . 7554/eLife . 13881 . 004Figure 1—figure supplement 1 . Both Syt4-GFP CRISPR knock-in and overexpression of Syt4-pH can replace endogenous Syt4 with respect to synaptic architecture and plasticity . ( A–D ) Representative images of NMJs stained with antibodies to HRP ( magenta ) and the postsynaptic marker Dlg ( green ) to highlight synaptic boutons . Acute budding of new varicosities ( “ghost boutons” ) was stimulated with spaced incubations in high K+ . Ghost boutons are identified as round HRP+ structures lacking Dlg signal ( arrowheads ) . Images are shown from the control genotype Syt4PRE ( A , A′ ) , a precise excision line that serves as a genetic background control for the Syt4BA1 allele ( B , B′ ) . Also shown are images from animals expressing Syt4-pH postsynaptically in the Syt4 null background ( Syt4BA1 24B>Syt4-pH; C , C′ ) , and the CRISPR GFP knock-in line Syt4GFP-2M ( D , D′ ) . ( E ) Quantification of bouton number normalized to yw , a genetic background control for Syt4GFP-2M . Blue line indicates the yw control mean . Data are presented as mean ± SEM . ( F ) Quantification of ghost bouton number per NMJ from animals without ( − ) or with ( + ) high K+ stimulation . Data are presented as mean ± SEM . Syt4BA1 24B>Syt4-pH animals have a normal number of boutons and exhibit normal budding of ghost boutons compared to the Syt4PRE control , and are significantly rescued compared to Syt4BA1 . Syt4GFP-2M animals have a normal number of boutons and exhibit normal budding of ghost boutons compared to the yw control line . Scale bars = 20 μm ( A–D ) , 6 . 7 μm ( A′–D′ ) . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in ( E ) and ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 00410 . 7554/eLife . 13881 . 005Figure 1—figure supplement 2 . pHluorin is quenched in live but not fixed preparations . Animals expressing Syt4-pH in the postsynaptic cell ( 24B>Syt4-pH ) were dissected and imaged live or following fixation in paraformaldehyde . The same animal was imaged first following incubation in pH 7 . 2 HL3 . 1 buffer and second following incubation in pH 5 . 0 HL3 . 1 buffer . Arrows indicate plasma membrane accumulations of Syt4-pH . Scale bars = 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 005 Candidate UAS-RNAi constructs were co-expressed with UAS-Syt4-pH in muscle , and animals were examined for changes in Syt4-pH distribution . We looked for changes in Syt4-pH intensity at the postsynaptic membrane ( defined as discreet Syt4-pH fields adjacent to the neuronal membrane ) , or other changes in the distribution , size or intensity of Syt4-pH-positive vesicular structures . Resolution of Syt4-pH localization was best achieved following tissue fixation , which is expected to interfere with the pH sensitivity of the pHluorin tag . Indeed , treatment of fixed samples with a low pH ( 5 . 5 ) adjusted buffer did not affect our detection of Syt4-pH in fixed tissue , compared to a dramatic quenching of fluorescence that was observed in a live preparation ( Figure 1—figure supplement 2 ) . Thus , we interpret the Syt4-pH localization pattern in our fixed-tissue assay as non-pH-dependent . We assembled a candidate list of gene products resident at synapses and/or involved in membrane trafficking ( Supplementary file 1 ) using the following criteria: 1 ) Drosophila orthologs of proteins identified in proteomics studies of mouse and rat brain synaptic membranes ( Abul-Husn et al . , 2009; Li et al . , 2007c ) ; 2 ) candidate genes identified in a Drosophila screen for transposon insertions affecting glutamate receptor expression or localization ( Liebl and Featherstone , 2005 ) ; and 3 ) known regulators of membrane trafficking ( eg Rabs , SNARE proteins , Vps proteins ) . Transgenic RNAi lines were obtained from the Transgenic RNAi Project ( TRiP ) at Harvard Medical School ( Perkins et al . , 2015 ) or the Vienna Drosophila RNAi Center ( Dietzl et al . , 2007 ) . For 190 candidates that had no RNAi stock already available , transgenic RNAi stocks were generated by the TRiP at Harvard Medical School; these stocks are currently available from the Bloomington stock center . Among the 442 lines screened , 15 candidates were identified with abnormal Syt4-pH distribution ( Table 1 ) . These candidates fell into three qualitatively distinct categories based on the intensity of Syt4-pH at the postsynaptic membrane: decreased intensity of Syt4-pH ( 7/15 , Figure 1D , D′ , eg . Syx4-RNAi ) , increased intensity of Syt4-pH ( 3/15 , Figure 1E , E′ , eg . Lasp-RNAi ) , and changes in the distribution of Syt4-pH-positive vesicles with otherwise normal intensity ( 5/15 , Figure 1F , F′ , eg . Syx7-RNAi ) . Two candidates had phenotypes consistent with previously published studies , supporting the efficacy of the screen: knockdown of Syx18/Gtaxin dramatically reduced Syt4 delivery , consistent with a role in postsynaptic membrane addition ( Gorczyca et al . , 2007 ) , and knockdown of β-spectrin resulted in changes in the spacing of Syt4-pH domains , consistent with a role for the spectrin cytoskeleton in AZ/GluR spacing ( Pielage et al . , 2006 ) . We chose to focus our analysis on one candidate , the plasma membrane t-SNARE Syx4 . Knockdown of Syx4 produced a decrease in the intensity of Syt4-pH at the postsynaptic membrane , along with large accumulations of Syt4-pH in the cytoplasm ( Figure 1D , D′ ) , suggesting that Syx4 may regulate membrane levels of Syt4 and modulate Syt4-dependent signaling mechanisms . 10 . 7554/eLife . 13881 . 006Table 1 . RNAis that alter the localization of Syt4-pH Candidate gene products are listed , along with the predicted gene function , the specific effect on Syt4-pH , and the RNAi constructs tested . RNAi lines were obtained from the Transgenic RNAi Project ( TRiP ) at Harvard Medical School ( Perkins et al . , 2015 ) a or the Vienna Drosophila RNAi Center ( Dietzl et al . , 2007 ) b . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 006Gene productCGFunctionSyt4-pH distributionRNAisSyntaxin 4CG2715t-SNAREReduced intensity at NMJ , large clusters in cytoplasmJF01714a V32413bSyntaxin 6CG7736t-SNAREReduced intensity at NMJ , large clusters in cytoplasmV1579b V1501bSyntaxin 18 ( Gtaxin ) CG13626t-SNAREReduced intensity at NMJ , large clusters in cytoplasmJF02263aMyoVCG2146Dilute class unconventional myosinReduced intensity at NMJ , large clusters in cytoplasmJF03035a V16902bActin-related protein 2/3 complex , subunit 3ACG4560Arp2/3 complex-mediated actin nucleationReduced intensity and size at NMJ , smaller cytoplasmic punctaJF02370aGdiCG4422Rab GDP-dissociation inhibitorReduced intensity at NMJ , large clusters in cytoplasmJF02617a V26537bRabex-5CG9139Rab5 guanyl-nucleotide exchange factor activityReduced intensity at NMJ , large clusters in cytoplasmJF02521aLaspCG3849Actin bindingIncreased intensity at NMJJF02075aNeuroglianCG1634Cell adhesion; axon guidance; synapse organizationIncreased intensity at NMJJF03151a V27201bContactinCG1084Cell adhesionIncreased intensity at NMJHM05134a HMS00186aSyntaxin 7CG5081t-SNARE , early endosomal regulationIntensity at NMJ normal , many small bright puncta cluster adjacent to NMJJF02436a V5413bDynamin associated protein 160CG1099Synaptic vesicle endocytosis; cell polarityIntensity at NMJ normal , many small bright puncta cluster adjacent to NMJJF01918a V16158bAdaptor Protein complex 2 , σ subunitCG6056EndocytosisIntensity at NMJ normal , many small bright puncta cluster adjacent to NMJJF02631aAdaptor Protein complex 2 , α subunitCG4260EndocytosisIntensity at NMJ normal , many small bright puncta cluster adjacent to NMJHMS00653aβ-spectrinCG5870Cytoskeleton; synapse organizationIrregular size and spacing at NMJHMS01746a V42053b To investigate the function of Syx4 , we created mutant alleles by mobilizing a transposable P-element located in the 5’-UTR of the Syx4 locus ( Figure 2A ) . Syx4 encodes a protein with a large N-terminal domain , a SNARE domain , and a C-terminal transmembrane domain ( Figure 2B ) . Two Syx4 proteins ( Syx4A and Syx4B ) are predicted from genome analysis , resulting in a longer ( A ) or shorter ( B ) N-terminus . RT-PCR analysis indicated that both of these isoforms are expressed in Drosophila larvae ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 13881 . 007Figure 2 . Syntaxin 4 is a postsynaptic plasma membrane SNARE . ( A ) Syx4 genomic region . Coding exons are indicated in green while non-coding exons are in blue . Two predicted start sites ( ATG ) are indicated in orange . The location of the P-element used for mutagenesis ( P ) is indicated in red . Three alleles of Syx4 were isolated . Deleted regions are indicated in red . Solid lines indicate regions known to be deleted from PCR analysis and sequencing , while dotted lines indicate regions within which breakpoints have been mapped . ( B ) Syx4 encodes a protein with an N-terminal domain , a SNARE domain and a C-terminal transmembrane domain . There are two predicted isoforms that differ in the size of the N-terminal domain . ( C , D ) Representative images of NMJs stained for Syx4 ( green ) and the neuronal membrane marker HRP ( magenta ) . Syx4 staining at the synapse in precise excision control animals ( C ) is absent in Syx473 mutant animals ( D ) . ( E ) Representative image from an animal stained for Syx4 ( green ) and expressing RFP-Syx4 ( magenta ) with 24B-GAL4 . ( F , G ) Representative images from animals expressing Syt4-pH with 24B-GAL4 in a control ( F ) or Syx473 ( G ) background . Syt4-pH is reduced at the postsynaptic membrane and redistributed to large cytoplasmic accumulations in Syx473 mutants . ( F′ , G′ ) Close-ups of F and G . ( H , I ) Representative images from Syt4GFP-2M knock-in animals in a control ( H ) or Syx473 ( I ) background . Synaptic localization of Syt4GFP-2M is reduced in Syx473 mutants . ( H′ , I′ ) Close-ups of H and I . Scale bars = 5 μm ( C–I ) , 2 μm ( F′ , G′ , H′ , I′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 00710 . 7554/eLife . 13881 . 008Figure 2—figure supplement 1 . RT-PCR analysis of Syntaxin 4 . Primers ( red arrows ) were designed to distinguish Syx4A and Syx4B transcripts by RT-PCR . F1 and R amplify a product from Syx4A transcript and F2 and R amplify a product from Syx4B transcript . Both transcripts are detected in control animals and both are absent from Syx473 nulls . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 008 We isolated three alleles that delete parts of the Syx4 coding region . Syx439 carries a deletion from the 5’-UTR to the first intron , removing the first exon and the start site for the Syx4A isoform . Syx448 carries a deletion from the 5’-UTR to the second intron , removing the first two exons and both predicted start sites . Finally , Syx473 carries a deletion from the 5’-UTR through the entire coding region of the gene . Several lines of evidence discussed below indicate Syx473 is a null allele . A precise excision with no deletion was also generated , and was used as a genetic background control . To examine the subcellular distribution of Syx4 , we generated polyclonal antisera against purified Syx4A protein . Syx4 is expressed in the muscle at the NMJ and is enriched postsynaptically , as revealed by co-staining with an antibody against HRP to highlight the presynaptic membrane ( Figure 2C ) . This staining is absent in Syx473 mutants ( Figure 2D ) , consistent with this allele removing the entire coding region of the gene . We also produced UAS-Syx4 and UAS-RFP-Syx4 constructs for both protein isoforms , in order to overexpress untagged or tagged Syx4 at the NMJ . Expression of RFP-Syx4A ( Figure 2E ) or RFP-Syx4B ( data not shown ) with the postsynaptic muscle driver 24B-GAL4 showed a similar distribution to the endogenous protein . Thus , Syx4 is expressed in the postsynaptic cell and accumulates at the synaptic membrane . We also tested the Syx473 allele for its effect on the distribution of Syt4-pH . Similar to our RNAi knockdown results , Syx4 mutants exhibited a decrease in Syt4-pH intensity at the NMJ , accompanied by a redistribution of Syt4-pH to large cytoplasmic accumulations ( Figure 2F , G , F′ , G′ ) . To test whether Syx4 regulates the localization of endogenous Syt4 , we used CRISPR/CAS9 to generate a C-terminal GFP knock-in line , Syt4GFP-2M . We produced a transgenic stock expressing custom guide RNAs ( gRNAs ) targeting the Syt4 locus . These animals were crossed to transgenic flies expressing germline-specific Cas9 , and embryos from the cross were injected with a donor plasmid for homology-directed repair to insert GFP into the Syt4 genomic locus ( Gokcezade et al . , 2014; Kondo and Ueda , 2013; Port et al . , 2014 ) . The Syt4GFP-2M line is homozygous viable and fertile , and does not exhibit synaptic defects that have been previously described in animals lacking Syt4 ( Figure 1—figure supplement 1 ) , indicating that Syt4GFP-2M protein is functional . Syt4GFP-2M shows synaptic localization at the NMJ ( Figure 2H , H′ ) , and this localization is lost in the Syx4 null mutant background ( Figure 2I , I′ ) . These findings indicate that the distribution of Syt4-pH reported in our screen is recapitulated by endogenous Syt4 protein . As we do not observe large cytoplasmic accumulations of Syt4GFP-2M in Syx4 mutants , this feature likely results from overexpression of Syt4-pH protein using the 24B-GAL4 driver . Because Syx4 mutants have a defect in Syt4 localization , we investigated whether Syx4 impacts synaptic development in a similar manner to Syt4 . Syt4 null mutants show abnormal development and function of the NMJ , including a decrease in the number of synaptic boutons , and a failure to express several forms of synaptic plasticity ( Barber et al . , 2009; Korkut et al . , 2013; Piccioli and Littleton , 2014; Yoshihara et al . , 2005 ) . We first quantified the number of boutons per NMJ at muscle 6/7 in hemisegment A3 . The null allele Syx473 exhibited a strong reduction in the number of synaptic boutons compared to control animals ( Figure 3A , B , G ) . When Syx473 was placed in trans with a large chromosomal deficiency that removed the entire Syx4 locus , a similar phenotype was observed compared to Syx473 mutants alone ( Figure 3G ) , consistent with Syx473 being a null allele . Syx448 animals also exhibited a decrease in bouton number compared to controls ( Figure 3C , G ) , which was less severe than the null mutant . The smallest deletion , Syx439 , had no change in bouton number compared to controls ( Figure 3D , G ) . These results indicate that Syx4 is required for normal synaptic bouton number . Furthermore , as the Syx439 allele lacks the start site for Syx4A and does not exhibit any phenotype , we hypothesize that expression of Syx4B from the second start site is sufficient for normal Syx4 function with respect to bouton number . 10 . 7554/eLife . 13881 . 009Figure 3 . Syntaxin 4 regulates synaptic growth at the NMJ . ( A–F ) Representative images of NMJs stained with antibodies to the postsynaptic marker Dlg ( green ) and the neuronal membrane marker HRP ( magenta ) to highlight the number of synaptic boutons; images are shown from precise excision control ( A ) , Syx473 ( B ) , Syx448 ( C ) , Syx439 ( D ) , Syx473 24B>Syx4A ( E ) , and Syx473 24B>Syx4B ( F ) animals . ( G ) Quantification of bouton number , normalized to controls . Blue line indicates the control mean . Red line indicates Syx473 null mean . Data are presented as mean ± SEM . ( H–J ) , Representative images of NMJs stained with antibodies to GluRIII ( green ) and the AZ marker Brp ( magenta ) ; images are shown from precise excision control ( H ) , Syx473 ( I ) , and Syx473 24B>Syx4A ( J ) animals . ( K ) , Quantification of AZ density , calculated as the number of AZs per volume HRP . Data are presented as mean ± SEM . ( L ) Quantification of GluRIII fluorescence per HRP fluorescence . Data are presented as mean ± SEM . Scale bars = 20 μm ( A–F ) , 5 μm ( H–J ) . Statistical comparisons are fully described in Figure 3—source data 1 , and are indicated here as *p<0 . 001 , p<0 . 01 , p<0 . 05 , ns = not significant; comparisons are with control unless indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 00910 . 7554/eLife . 13881 . 010Figure 3—source data 1 . Statistical data for Figure 3 . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in Figure 3G , K , and L . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 010 We next overexpressed UAS-Syx4A or UAS-Syx4B , in the presynaptic neuron ( with elav-GAL4 ) or the postsynaptic muscle cell ( with 24B-GAL4 ) . None of these overexpressions resulted in any change in synaptic bouton number compared to control animals ( Figure 3G ) . Thus , overexpression of Syx4 is not detrimental to synaptic growth . We next attempted to rescue Syx4 mutant defects by expressing either isoform in the null mutant background with either pre- or postsynaptic-specific drivers at the NMJ . When expressed in the postsynaptic cell , both Syx4A and Syx4B were able to rescue the Syx473 decrease in bouton number compared to Syx473 alone ( Figure 3E–G ) , restoring bouton number to control levels . In contrast , expressing either of these constructs presynaptically did not produce any rescue of the null mutant phenotype compared to Syx473 alone ( Figure 3G ) . These findings indicate that Syx4 is required postsynaptically to regulate bouton number . Furthermore , either isoform of Syx4 is sufficient for Syx4 function . We also examined the organization of neurotransmitter release sites in Syx4 mutants by staining for Bruchpilot ( Brp ) , a marker of the presynaptic AZ , and GluRIII , an obligate subunit of the postsynaptic glutamate receptor . By counting Brp+ puncta , we detected a significant decrease in the density of AZs per unit volume in Syx4 mutants compared to controls ( Figure 3H , I , K ) . The AZ density defect was rescued by postsynaptic overexpression of Syx4 ( Figure 3J , K ) . The amount of GluRIII fluorescence was unchanged in Syx4 mutants compared to controls , indicating normal amounts of GluRIII were present at the postsynaptic membrane ( Figure 3H–J , L ) . In addition , the apposition of Brp and GluRIII was unaffected ( Figure 3H–J ) . Thus , Syx4 mutants have a decrease in the density of AZs and a reduction in the total number of boutons , but no defects in the organization of individual release sites . The observation that postsynaptic Syx4 can regulate AZs in the presynaptic cell supports the hypothesis that Syx4 participates in retrograde signaling . Syx4 regulates the membrane localization of Syt4 , and loss of either gene leads to a reduction in the number of synaptic boutons at the larval NMJ ( Figure 3; Barber et al . , 2009 ) . To further investigate the relationship between Syt4 and Syx4 in synaptic development , we tested for genetic interactions between the null allele Syx473 and the null allele Syt4BA1 ( Adolfsen et al . , 2004 ) . Single heterozygotes of Syx4 ( Syx473/+ ) or Syt4 ( Syt4BA1/+ ) had no bouton number phenotype compared to control animals ( Figure 4A , B , F ) . Strikingly , double heterozygotes ( Syx473/+ Syt4BA1/+ ) had a strong decrease in bouton number compared to control animals and compared to single Syx473/+ or Syt4BA1/+ heterozygotes ( Figure 4C , F ) . This finding is consistent with Syx4 and Syt4 acting together to regulate bouton number and synaptic growth . 10 . 7554/eLife . 13881 . 011Figure 4 . Genetic interactions between Syntaxin 4 and Synaptotagmin 4 . ( A–E ) Representative images of NMJs stained with antibodies to the postsynaptic marker Dlg ( green ) and the neuronal membrane marker HRP ( magenta ) to highlight the number of synaptic boutons; images are shown from Syx473/+ ( A ) , Syt4BA1/+ ( B ) , Syx473/+ Syt4BA1/+ ( C ) , Syx473/+ Syt4BA1 ( D ) , and Syx473 Syt4BA1/+ ( E ) animals . ( F ) Quantification of bouton number , normalized to controls . Blue line indicates the control mean . Data are presented as mean ± SEM . L = lethal . Scale bars = 20 μm ( A–E ) . Statistical comparisons are fully described in Figure 4—source data 1 , and are indicated here as p<0 . 001 , p<0 . 01 , p<0 . 05 , ns = not significant; comparisons are with control unless indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 01110 . 7554/eLife . 13881 . 012Figure 4—source data 1 . Statistical data for Figure 4 . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in Figure 4F . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 012 To investigate the epistatic relationship between Syx4 and Syt4 , we produced animals that are hemizygous for Syx473 , which is on the X chromosome , and homozygous for Syt4BA1 ( Syx473 Syt4BA1 ) . These animals die during the 2nd larva instar , precluding analysis at the 3rd larval instar NMJ . As Syx473 hemizygotes die during the pupal stage , and Syt4BA1 homozygotes survive to adulthood , the early lethality observed in the double mutants reveals a synergistic interaction between the genes . This indicates that Syx4 and Syt4 act in parallel pathways , rather than a single epistatic pathway . We also produced animals that were heterozygous for one gene and homo/hemizygous for the other . All of these animals survived to the 3rd larval instar , allowing us to assess bouton number . Removing one copy of Syt4BA1 in the Syx473 hemizygous background did not modify the Syx473 hemizygous phenotype ( Figure 4E , F ) . In contrast , removing one copy of Syx473 in the Syt4BA1 homozygous background resulted in a further reduction in bouton number compared to Syt4BA1 homozygotes alone ( Figure 4D , F ) . Taken together with the observation that Syx4 regulates membrane localization of Syt4 , we conclude that Syx4 and Syt4 interact in one pathway , and also in parallel pathways , to regulate synapse development at the Drosophila NMJ . Based on the following observations , we hypothesize that Syx4 regulates the release of retrograde signals to control synaptic development: 1 ) Syx4 localizes to the postsynaptic membrane; 2 ) Syx4 regulates the membrane localization of Syt4; 3 ) Syx4 regulates bouton number both with and independently of Syt4; 4 ) postsynaptic Syx4 regulates presynaptic AZ density; and 5 ) Syx4 proteins have a conserved function as plasma membrane t-SNAREs . To identify retrograde signals potentially regulated by Syx4 , we tested for genetic interactions between Syx473 and components of characterized retrograde signaling pathways that affect bouton number at the NMJ . We failed to detect any dosage-dependent genetic interactions between Syx4 and components of the retrograde BMP signaling pathway that affects arbor size , neurotransmitter release , and synaptic plasticity ( Figure 5—figure supplement 1; Aberle et al . , 2002; Marqués et al . , 2002; McCabe et al . , 2003; Piccioli and Littleton , 2014; Rawson et al . , 2003 ) . In contrast , we detected strong genetic interactions between Syx4 and the genes encoding the adhesion proteins Neurexin 1 ( Nrx-1 ) and Neuroligin 1 ( Nlg1 ) ( Figure 5 ) . Neurexins and Neuroligins form transsynaptic adhesion complexes , with a Neurexin typically the presynaptic partner and a Neuroligin the postsynaptic partner . At the Drosophila NMJ , Nrx-1 and the three characterized Nlgs ( Nlg1-3 ) have been shown to play several roles in synaptic growth and organization , including regulation of bouton number , GluR subunit composition , and AZ size , spacing , and apposition ( Banovic et al . , 2010; Chen et al . , 2010 , 2012; Li et al . , 2007b; Sun et al . , 2011; Xing et al . , 2014 ) . 10 . 7554/eLife . 13881 . 013Figure 5 . Syntaxin 4 interacts with Neuroligin 1 and regulates its membrane localization . ( A–E ) , Representative images of NMJs stained with antibodies to the postsynaptic density marker Dlg ( green ) and the neuronal membrane marker HRP ( magenta ) to highlight the number of synaptic boutons; images are shown from Syx473/+ ( A ) , Nlg1ex3 . 1/+ ( B ) , Nrx-1273/+ ( C ) , Syx473/+ Nlg1ex3 . 1/+ ( D ) , and Syx473/+ Nrx-1273/+ ( E ) animals . ( F ) Quantification of bouton number , normalized to controls . Blue line indicates the control mean . Data are presented as mean ± SEM . ( G–H ) , Representative images of NMJs stained with antibodies against HRP ( magenta ) and expressing Nrx-1-GFP in a control ( G ) or Syx473 mutant ( H ) background . ( I–J ) Representative images of NMJs stained with antibodies against HRP ( magenta ) and expressing Nlg1-GFP in a control ( I ) or Syx473 mutant ( J ) background . ( K ) Quantification of GFP fluorescence per HRP fluorescence from animals expressing Nlg1-GFP in a control or Syx473 mutant background . Data are presented as mean ± SEM . ( L–M ) Representative images of NMJs stained with antibodies against HRP ( magenta ) and expressing Nlg1Δcyto-GFP in a control ( L ) or Syx473 mutant ( M ) background . ( N–Q ) Representative images of NMJs stained with antibodies against Brp ( magenta ) and GluRIII ( green ) , from precise excision control ( N ) , Nlg1ex3 . 1 ( O ) , Syx473/+ Nlg1ex3 . 1/+ ( P ) , and Syx473 Nlg1ex3 . 1 ( Q ) animals . Arrowheads indicate AZs lacking an apposed GluR field . Scale bars = 20 μm ( A–E ) , 5 μm ( I , J , L–Q ) . Statistical comparisons are fully described in Figure 5—source data 1 , and are indicated here as p<0 . 001 , p<0 . 01 , p<0 . 05 , ns = not significant; comparisons are with control unless indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 01310 . 7554/eLife . 13881 . 014Figure 5—source data 1 . Statistical data for Figure 5 . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in Figure 5F and K . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 01410 . 7554/eLife . 13881 . 015Figure 5—figure supplement 1 . Genetic interaction experiments between Syx4 and BMP pathway components . ( A ) No genetic interactions are detected between Syx473 and components of the BMP pathway: witA12 , witB11 , or gbb1 . Single and double heterozygous combinations are shown . Data are presented as mean ± SEM , ns = not significant . ( B ) Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 01510 . 7554/eLife . 13881 . 016Figure 5—figure supplement 2 . Genetic interaction experiments between single and double null mutants of Syx4 , Nlg1 , and Nrx-1 . ( A ) Double mutants combinations between Syx473 , Nlg1ex3 . 1 , and Nrx-1273 have severe synaptic growth defects . Data are presented as mean ± SEM . Statistical comparisons are indicated here as p<0 . 001 , p<0 . 01 , p<0 . 05 , ns = not significant; comparisons are with control unless indicated . ( B ) Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 016 We used the null alleles Nrx-1273 ( Li et al . , 2007b ) and Nlg1ex3 . 1 ( Banovic et al . , 2010 ) to test for interactions with Syx4 . Single heterozygotes of Nlg1 ( Nlg1ex3 . 1/+ ) , Nrx-1 ( Nrx-1273/+ ) , and Syx4 ( Syx473/+ ) all had a normal number of boutons compared to control animals ( Figure 5A–C , F ) . However , the double heterozygotes Syx473/+ Nlg1ex3 . 1/+ ( Figure 5D ) and Syx473/+ Nrx-1273/+ ( Figure 5E ) exhibited strong reductions in bouton number compared to controls and compared to each single heterozygote ( Figure 5F ) . Thus , Syx4 , Nrx-1 , and Nlg1 cooperate to regulate bouton number at the NMJ . We next tested whether the localization of Nrx-1 or Nlg1 was perturbed upon loss of Syx4 . We expressed GFP-tagged forms of Nrx-1 and Nlg1 ( Banovic et al . , 2010 ) at the synapse and measured fluorescence intensity in control and Syx4 mutant backgrounds . When we expressed Nrx-1-GFP in the presynaptic cell using elav-GAL4 , we did not detect any change in fluorescence intensity of GFP in Syx473 mutants compared to controls ( Figure 5G , H ) . However , when we expressed Nlg1-GFP in the postsynaptic cell using 24B-GAL4 , we detected a significant reduction in GFP signal at the synapse in the Syx473 null mutant background compared to controls ( Figure 5I–K ) . This result indicates that Syx4 regulates the levels of Nlg1 at the postsynaptic membrane . If the amount of Nlg1 at the synapse depends on Syx4 , it is possible that the cytoplasmic domain of Nlg1 is involved in its delivery or retention . To test this hypothesis , we examined the localization of a tagged Nlg1 construct lacking the cytoplasmic domain ( Nlg1Δcyto-GFP; Banovic et al . , 2010 ) . This construct was previously shown to localize to the NMJ , and to produce a dominant negative decrease in bouton growth ( Owald et al . , 2010 ) . Like the full-length construct , Nlg1Δcyto-GFP localizes to the postsynaptic membrane when expressed in a control background ( Figure 5L ) . Interestingly , this localization pattern is strikingly different when Nlg1Δcyto-GFP is expressed in a Syx473 mutant background . While some tagged protein is observed at the synapse , Nlg1Δcyto-GFP also appears in prominent bright clusters seen both near the NMJ ( Figure 5M ) and throughout the muscle ( data not shown ) . This finding indicates a strong effect of Syx4 on Nlg1 localization , which is enhanced when the cytoplasmic domain of the protein is absent . To further investigate the relationship between Syx4 , Nrx-1 , and Nlg1 , we produced Syx473 Nlg1ex3 . 1 and Syx473 Nrx-1273 double mutant animals . Analysis of the double mutants revealed a strong reduction in bouton number compared to controls , and double mutants were not significantly different from single Nlg-1 or Nrx-1 mutants ( Figure 5—figure supplement 2 ) . Thus , complete loss of Syx4 does not enhance the bouton formation defects seen in Nlg1 or Nrx-1 homozygous mutants . This observation suggests that Nlg1 and Nrx-1 are downstream of Syx4 with respect to bouton number . We next examined the organization of postsynaptic densities , which is perturbed in Nlg1 mutants ( Banovic et al . , 2010 ) , but not in Syx473 ( Figure 3 ) . Consistent with previous studies ( Banovic et al . , 2010 ) , we detected irregular and enlarged GluR fields in Nlg1 mutants , as well as defects in apposition between AZs and GluR fields , compared to controls ( Figure 5N , O , arrowheads ) . We then tested whether loss of Syx4 could modify the Nlg1 AZ/GluR phenotypes . We first tested animals that were double heterozygotes for Syx473 and Nlg1ex3 . 1 , and observed normal GluR field size and apposition ( Figure 5P ) . Thus , the dosage-dependent genetic interactions we detected with respect to bouton number are not seen in the case of AZ/GluR organization . Double mutant animals ( Syx473 Nlg1ex3 . 1 ) looked qualitatively similar to single Nlg1ex3 . 1 mutants ( Figure 5Q , arrowheads ) , indicating that loss of Syx4 did not modify this aspect of the Nlg1 phenotype . In summary , Syx4 and Nlg1 mutants have phenotypes that partially overlap ( bouton number ) , but Nlg1 mutants have additional defects not seen with loss of Syx4 ( organization of AZs/GluRs ) . Our genetic interaction data are consistent with Syx4 and Nlg1 cooperating to regulate bouton number , but not AZ/GluR organization . Taken together with the observation that loss of Syx4 leads to a partial reduction of Nlg1 at the membrane , one model is that minimal levels of Nlg1 are sufficient for AZ/GluR organization , but higher Syx4-dependent surface expression is required for regulating synaptic growth and bouton number . How does loss of Syx4 result in lower levels of Nlg1 at the postsynaptic membrane ? One possibility is that less Nlg1 is delivered , and another is that Nlg1 is less stable or more mobile once it gets to the membrane . To investigate these possibilities , we measured the mobility of Nlg1 in vivo by tagging it with a photoconvertible fluorophore , Dendra2 ( Adam et al . , 2009; Gurskaya et al . , 2006; Figure 6A ) . We added the Dendra2 tag in a juxta-membrane position , as previously described for Nlg1-GFP ( Banovic et al . , 2010 ) . When expressed in the postsynaptic compartment using the muscle driver 24B-GAL4 , Nlg1-Dendra2 localized to the synapse similarly to Nlg1-GFP ( Figure 6B ) . We then used a 405 nm laser to convert approximately 50% of the fluorescent signal in a single bouton ( Figure 6C , C′ ) , and followed the fluorescence intensity of the green ( non-photoconverted; NPC ) and red ( photoconverted; PC ) signals over a 10 min period ( Figure 6D and Video 1 ) . We measured mobility by calculating the relative change in fluorescence ( ΔF/F ) in both channels from immediately after PC ( t1 min ) to 9 min after PC ( t10 min ) , in the PC bouton ( ROI1 ) or adjacent NPC boutons ( ROI2 and ROI3 ) , after correcting for photobleaching ( Figure 6C′–E ) . If Nlg1-Dendra2 moved laterally in the membrane , we would expect to see a decrease in red/PC signal in ROI1 and an increase in red/PC signal in ROIs 2 and 3 . If Nlg1-Dendra2 was internalized from the membrane , we would expect to see a decrease in red/PC signal in ROI1 without any increase in red/PC signal in ROIs 2 and 3 . Interestingly , we measured extremely small ΔF/F values ( < 0 . 02% ) for the red/PC signal in all ROIs , reflecting very little change in fluorescence over the time course of the experiment ( Figure 6D , E ) . We performed the same experiment by expressing Nlg1-Dendra2 in the Syx473 mutant background and observed a similar effect , with small ΔF/F values in the red/PC channel , and no significant change compared to control values ( Figure 6E and Video 2 ) . Thus , over the time course measured , Nlg1 is immobile at the synapse , and this is not changed by loss of Syx4 . We also monitored the fluorescence of green/NPC molecules , and measured very small ΔF/F values ( < 0 . 02% ) in all ROIs for both controls and Syx473 mutants ( Figure 6D , E ) . It is more difficult to interpret the movement of NPC molecules in this experiment; however , given the conclusion from the red/PC channel data that Nlg1 does not move laterally in the membrane or get internalized from the membrane over the time course of the experiment , the stable fluorescence of NPC molecules allows us to infer that very little new unconverted Nlg1 is delivered to the synapse . Thus , our analysis of photoconvertible Nlg1 reveals that Nlg1 is stable at the synapse over a short time course and this stability is not compromised by loss of Syx4 . Based on these results , we hypothesize that the lower plasma membrane level of Nlg1 in Syx4 mutants is the result of changes in the delivery or removal of Nlg1 over a developmental time scale , or over a longer time course than our experimental paradigm . 10 . 7554/eLife . 13881 . 017Figure 6 . No change in mobility of Neuroligin 1 is observed in Syntaxin 4 mutants . ( A ) Nlg1-Dendra2 construct . The Dendra2 tag was placed between the transmembrane domain and the cytoplasmic tail . ( B–C ) Representative image from a single animal expressing Nlg1-Dendra2 in the postsynaptic cell . One bouton ( ROI1 ) was targeted with a 405 nm laser for photoconversion of the Dendra2 tag after 1 min . Non-photoconverted Nlg1-Dendra2 is shown in green and photoconverted Nlg1-Dendra2 is shown in magenta before ( B ) and immediately after ( C ) photoconversion . ( C′ ) Regions of interest: ROI1 , photoconverted region; ROIs 2 and 3 , adjacent regions . ( D ) Fluorescent intensity over time for photoconverted and non-photoconverted molecules in all three ROIs . Red arrows indicate time of photoconversion . ( E ) Quantification of ΔF/F of both photoconverted and non-photoconverted molecules , in all three ROIs , in both the control and Syx473 mutant backgrounds . Data are presented as mean ± SEM . Scale bars = 5 μm . Statistical comparisons are fully described in Figure 6—source data 1; no significant differences found ( ns = not significant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 01710 . 7554/eLife . 13881 . 018Figure 6—source data 1 . Statistical data for Figure 6 . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in Figure 6E . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 01810 . 7554/eLife . 13881 . 019Video 1 . Photoconversion of Nlg1-Dendra2 in control animals . Visualization of a synaptic arbor expressing postsynaptic Nlg1-Dendra2 at muscle 4 in a dissected third instar larva . One bouton is photoconverted after 20 sec , with about 50% of the green molecules converted to red ( shown here as magenta ) . Over the next 9 min of imaging , very little movement of photoconverted molecules is observed . Scale bar = 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 01910 . 7554/eLife . 13881 . 020Video 2 . Photoconversion of Nlg1-Dendra2 in Syx473 animals . Visualization of a synaptic arbor expressing postsynaptic Nlg1-Dendra2 at muscle 4 in a dissected third instar larva mutant for Syx4 . One bouton is photoconverted after 20 sec , with about 50% of the green molecules converted to red ( shown here as magenta ) . Over the next 9 min of imaging , very little movement of photoconverted molecules is observed . Scale bar = 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 020 In addition to synaptic growth during development , the Drosophila NMJ displays acute structural plasticity where new boutons bud rapidly in response to strong neuronal stimulation ( Ataman et al . , 2008 ) . Newly formed boutons , called ghost boutons ( GBs ) , are readily identifiable as round structures containing neuronal membrane , but without any postsynaptic apparatus . The activity-dependent budding of GBs requires retrograde BMP signaling ( Piccioli and Littleton , 2014 ) , as well as retrograde signaling mediated by Syt4 ( Korkut et al . , 2013; Piccioli and Littleton , 2014 ) . We investigated whether Syx4 regulates acute structural plasticity in vivo using a high K+ stimulation protocol ( Ataman et al . , 2008; Piccioli and Littleton , 2014 ) . As previously described , control animals exhibited robust GB budding in response to spaced incubations in high K+ over a 30 min period ( Figure 7A , A′ , K ) . However , budding was strongly suppressed in Syx4 null mutant animals compared to controls ( Figure 7E , E′ , K ) . Thus , like Syt4 , Syx4 regulates rapid activity-induced synaptic growth . Furthermore , Syx4 and Syt4 interact genetically with respect to GB budding , as budding was strongly suppressed in Syx473/+ Syt4BA1/+ double heterozygotes compared to normal robust budding in either single heterozygote ( Figure 7B , B′ , C , C′ , H , H′ , K ) . 10 . 7554/eLife . 13881 . 021Figure 7 . Syntaxin 4 , Synaptotagmin 4 , and Neuroligin 1 regulate acute structural plasticity at the NMJ . ( A–J ) Representative images of NMJs stained with antibodies to HRP ( magenta ) and the postsynaptic marker Dlg ( green ) to highlight synaptic boutons . Ghost bouton budding was stimulated with spaced incubations in high K+ . Ghost boutons are identified as round HRP+ structures lacking Dlg signal ( arrowheads ) ; images are shown from precise excision control ( A ) , Syx473/+ ( B ) , Syt4BA1/+ ( C ) , Nlg1ex3 . 1/+ ( D ) , Syx473 ( E ) , Syt4BA1 ( F ) , Nlg1ex3 . 1 ( G ) , Syx473/+ Syt4BA1/+ ( H ) , Syx473/+ Nlg1ex3 . 1/+ ( I ) , and Syt4BA1/+ Nlg1ex3 . 1/+ ( J ) animals . ( K ) Quantification of ghost bouton number per NMJ from animals without ( − ) or with ( + ) high K+ stimulation . Data are presented as mean ± SEM . Scale bars = 20 μm ( A–J ) , 6 . 7 μm ( A′–J′ ) . Statistical comparisons are fully described in Figure 7—source data 1 , and are indicated here as p<0 . 001 , p<0 . 01 , p<0 . 05 , ns = not significant; comparisons are with control unless indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 02110 . 7554/eLife . 13881 . 022Figure 7—source data 1 . Statistical data for Figure 5 . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in Figure 7K . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 02210 . 7554/eLife . 13881 . 023Figure 7—figure supplement 1 . Interaction experiments between Syt4 and Nlg1 . ( A ) No genetic interactions are detected between Syt4 and Nlg1 with respect to bouton number . Data are presented as mean ± SEM , ns = not significant , ANOVA . Control refers to a precise excision control line for the Syt4BA1 allele . ( B–C ) Representative NMJs expressing Nlg1-GFP in control ( B ) or Syt4BA1 ( C ) backgrounds . ( D ) Quantification of GFP fluorescence intensity per HRP fluorescence intensity , in animals expressing Nlg1-GFP in control or Syt4BA1 backgrounds . Data are presented as mean ± SEM , ns = not significant , Student’s t test . Scale bars = 5 μm . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented for the data in ( A ) and ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13881 . 023 We next tested whether GB budding is impaired in Nlg1 mutants , and whether Syx4 and Nlg1 interact in this context . Indeed , Nlg1ex3 . 1 mutants exhibited a strong suppression of GB budding compared to controls ( Figure 7G , G′ , K ) . Also , GB budding was strongly suppressed in Syx473/+ Nlg1ex3 . 1/+ double heterozygotes compared to single heterozygotes ( Figure 7C , C′ , D , D′ , I , I′ , K ) . These results indicate that Syx4 and Nlg1 interact to regulate activity-dependent formation of GBs . Because Syx4 regulates the levels of both Syt4 and Nlg1 at the postsynaptic membrane , and all of these proteins are involved in regulating bouton number and rapid activity-dependent bouton formation , we investigated whether Syt4 and Nlg1 interact with each other . Indeed , we observed dosage-dependent genetic interactions between Syt4 and Nlg1 with respect to GB formation , as Syt4BA1/+ Nlg1ex3 . 1/+ double heterozygotes had a strong reduction in GB budding compared to single heterozygotes ( Figure 7C , C′ , D , D′ , J , J′ , K ) . However , we were not able to detect an interaction between Syt4 and Nlg1 with respect to bouton number . The double heterozygotes Syt4BA1/+ Nlg1ex3 . 1/+ had a normal number of boutons compared to controls and compared to either single heterozygote ( Figure 7—figure supplement 1 ) , in contrast to the strong interactions we detected between Syx4/Syt4 ( Figure 4 ) and Syx4/Nlg1 ( Figure 5 ) . We also expressed both Nlg1-GFP and Nlg1Δcyto-GFP in the Syt4BA1 null background and did not observe any change in Nlg1 localization compared to controls ( Figure 7—figure supplement 1 ) . Thus , our data are consistent with Syx4 , Syt4 , and Nlg1 cooperating to regulate acute synaptic structural plasticity . With respect to bouton number , the data support Syx4 interacting with Nlg1 and Syt4 in separate pathways . Taken together , Syx4 acts postsynaptically to regulate multiple parameters of synaptic biology by interacting with Nlg1 and Syt4 and regulating their membrane localization . Our screen identified 15 candidate gene products that altered the localization of Syt4-pH . In addition to Syx4 , several other candidates motivate interesting hypotheses about regulatory pathways for postsynaptic exocytosis . MyoV is a Ca2+-sensitive unconventional myosin that regulates polarized traffic ( Krauss et al . , 2009; Li et al . , 2007a ) and the release of exosomes from motorneurons ( Koles et al . , 2012 ) . Thus , MyoV could play a role linking Ca2+ influx to vesicle delivery or release at the synapse . Indeed , MyoV homologs have been implicated in regulated AMPA trafficking in mammalian dendrites ( Correia et al . , 2008; Wang et al . , 2008 ) . Two Rab regulators ( Gdi and Rabex ) suggest that key vesicle trafficking steps en route to the synapse are modulated by Rab activation states . Also , two cell adhesion molecules ( Neuroglian and Contactin ) indicate potential transsynaptic mechanisms regulating retrograde signaling . Neuroglian has been shown to be required for synaptic stability ( Enneking et al . , 2013 ) and it is possible that Syt4-mediated retrograde signaling plays some role in this process . Syt4 has also been shown to be transferred transsynaptically from the presynaptic terminal to the postsynaptic terminal on exosomes ( Korkut et al . , 2013 ) . Thus , our approach of expressing Syt4-pH postsynaptically may not reveal components for the biosynthetic synthesis and transport of presynaptic Syt4 . Nevertheless , the requirement for Syt4 in the postsynaptic cell for retrograde signaling is clear ( Barber et al . , 2009; Korkut et al . , 2013; Piccioli and Littleton , 2014; Yoshihara et al . , 2005 ) , and the results of our screen highlight regulators of Syt4 trafficking to and from the postsynaptic membrane where Syt4 vesicles fuse in an activity-dependent manner ( Yoshihara et al . , 2005 ) . The observation that endogenously expressed Syt4-GFP ( Syt4GFP-2M ) shows a similar distribution to Syt4-pH supports the biological relevance of the screen data for identifying regulators of Syt4 trafficking in the postsynaptic cell . Our Syx4 null allele phenocopies the Syx4-RNAi knockdown , reducing the delivery of Syt4-pH to the postsynaptic membrane . Consistent with this finding , loss of Syx4 produces similar phenotypes to loss of Syt4 . Both null mutants exhibit a reduction in the total number of boutons at the NMJ , indicating a defect in synaptic growth . Moreover , genetic interaction experiments clearly indicate that Syx4 and Syt4 interact with respect to synaptic growth . A strong genetic interaction between Syx4 and Syt4 is also evident at the level of lethality , as double mutant animals are lethal at a much earlier stage than either single mutant alone . Thus , even though Syx4 affects the localization of Syt4 , suggesting they act in the same pathway , the genetic interaction data do not support a simple epistatic relationship . The difference in phenotypic severity , with the Syx4 bouton number defect being significantly stronger than the Syt4 defect , also points to Syt4 not being absolutely required for Syx4 signaling . A similar phenomenon is observed presynaptically where the t-SNARE Syx1 is indispensible for synaptic vesicle fusion , while fusion is only reduced in the absence of the synaptic vesicle Ca2+ sensor Syt1 . Taken together , we hypothesize that 1 ) Syx4 and Syt4 act together in a single pathway where Syx4 regulates the exocytosis of vesicles containing Syt4 , and 2 ) Syx4 and Syt4 also act in divergent pathways , where Syt4 cooperates with other t-SNARES , and Syx4 mediates the exocytosis of vesicles in a Syt4-independent manner . This model allows for multiple possible postsynaptic SNARE complexes , regulating distinct release events . Dissecting the other components of these fusion machineries , and distinguishing activity-dependent from constitutive release events , will be important to build our understanding of how retrograde signaling is regulated . In addition to affecting the localization of Syt4 , Syx4 mutants also exhibit a decrease in the amount of Nlg1 at the postsynaptic membrane . Nlg1 has several functions at the synapse , along with its presynaptic binding partner Nrx-1 . Together they regulate bouton number as well as the size and spacing of active zones and glutamate receptors ( Banovic et al . , 2010; Li et al . , 2007b; Owald et al . , 2010 ) , though some aspects of Nlg1 signaling appear to be independent of Nrx-1 ( Banovic et al . , 2010 ) . Mutations in Nrx and Nlg family genes are also linked to ASD , highlighting the importance of Nrx-Nlg signaling in neuronal development ( Bottos et al . , 2011; Südhof , 2008 ) . Consistent with a reduction of Nlg1 levels at the synapse , we observed strong genetic interactions between Syx4 , Nlg1 and Nrx-1 with respect to bouton number . However , the prominent AZ/GluR defects seen in Nlg1 and Nrx-1 mutants were not observed in Syx4 mutants , and heterozygous combinations did not produce these defects . It is likely that Syx4 mutants exhibit a partial loss of function of Nlg1 , and that bouton number is sensitive to this loss while AZ/GluR organization can be maintained with low levels of Nlg1 . A dramatic change in distribution of Nlg1Δcyto is observed in the Syx4 mutant background , providing further evidence that Syx4 regulates the localization of Nlg1 . The redistribution of Nlg1Δcyto to large accumulations is striking compared to full-length Nlg1 , which is simply reduced at the synapse in the Syx4 mutant background . This observation points to complex Syx4-dependent regulation of Nlg1 localization . One model is that trafficking of Nlg1 involves both a Syx4-dependent pathway and a second pathway that depends on an interaction with the Nlg1 C-terminus , which includes a PDZ-domain-binding motif . In this scenario , a severe Nlg1 trafficking defect is revealed only when both pathways are compromised . A second possibility is that in the absence of Syx4 , a portion of the Nlg1 content in the cell is degraded , but that this degradation step depends on the presence of the Nlg1 cytoplasmic tail , leading to the observed aggregation of Nlg1Δcyto in Syx4 mutants . Our analysis of Nlg1 trafficking in live animals reveals that Nlg1 is strikingly stable , in both control and Syx4 mutant backgrounds . Our motivation in performing these experiments was to test possible mechanisms underlying the decrease in Nlg1 levels in Syx4 mutants . It is possible that some Nlg1 mobility would be observed over a longer time course . Mammalian Nlg has been shown to undergo significant turnover at postsynaptic sites under LTP conditions in neuronal cell culture ( Schapitz et al . , 2010 ) . Also , synaptic activity has been shown to induce cleavage of Nlg and the subsequent destabilization of the Nrx-Nlg complex ( Peixoto et al . , 2012 ) . Thus , it remains a possibility that Nlg1 would be mobilized in response to activity in our preparation; however , we have not observed any increased mobility in response to high K+ incubations in preliminary tests ( data not shown ) . Our data are most consistent with Syx4 regulating Nlg1 over a developmental time course . A detailed examination of the relationship between Syx4 and Nlg1 dynamics will be crucial to understand how Syx4 contributes to this important pathway in synaptic development . We observed a strong suppression of acute structural plasticity in null mutants of Syx4 , Syt4 and Nlg1 . Double heterozygous combinations also indicated strong genetic interactions between all three of these genes with respect to plasticity . GB budding is regulated by both acute and developmental signaling . Because Syt4 postsynaptic vesicles fuse in an activity-dependent manner ( Yoshihara et al . , 2005 ) , it is possible that Syt4-dependent signaling releases an acute instructive cue for GB budding . Thus , one attractive model is that Nlg1 is delivered to the membrane in response to stimulation , depending on the Ca2+ sensitivity of Syt4 and the presence of the t-SNARE Syx4 at the membrane . It is also possible that Syx4-Syt4-Nlg1 signaling is required throughout development to potentiate the synapse to respond to strong neuronal stimulation . In conclusion , Syx4 , Syt4 , and Nlg1 interact to regulate several aspects of synaptic biology . Our data support multiple overlapping signaling pathways regulated by these proteins , reflecting a complex modulation of retrograde signaling to control synaptic growth and plasticity at the Drosophila NMJ . All Drosophila strains were cultured on standard media at 25°C . The following stocks were used: 24B-GAL4 ( BDSC 1767; Brand and Perrimon , 1993 ) ; elav-GAL4[2] ( BDSC 8765; Luo et al . , 1994 ) ; Df ( 1 ) ED6630 ( BDSC 8948; Ryder et al . , 2007 ) ; witA12 , witB11 ( Marqués et al . , 2002 ) ; UAS-Syt4-pHluorin ( Yoshihara et al . , 2005 ) ; Syt4BA1 ( Adolfsen et al . , 2004 ) ; gbb1 ( Wharton et al . , 1999 ) ; Nlg1ex3 . 1 , UAS-Nlg1-GFP , UAS-Nlg1Δcyto-GFP ( Banovic et al . , 2010 ) ; Nrx-1273 ( Li et al . , 2007b ) ; UAS-Syx4-RNAi ( TRiP JF01714; Perkins et al . , 2015 ) , UAS-Syx4-RNAi ( VDRC 32413; Dietzl et al . , 2007 ) . Full-length Syntaxin 4 was obtained from the Drosophila Genomics Resource Center ( DGRC RE02884; Stapleton et al . , 2002 ) . Three point mutations in the cDNA were corrected with a Quickchange Lightning Multi site-directed mutagenesis kit ( Agilent Technologies , Santa Clara , CA ) ( pos 71: G to A; pos 496: A to C; pos 693: T to A ) . The sequence listed in Flybase , and several other ESTs covering parts of the Syx4 sequence , agree that these changes reflect the correct sequence . UAS-Syx4 was produced by PCR-amplifying Syx4 using ExTaq ( ClonTech Laboratories , Mountain View , CA ) , and adding a 5’NdeI site and a 3’XbaI site . The PCR product was subsequently digested and subcloned into pValum ( Ni et al . , 2008 ) . The construct was injected into a third chromosome docking strain ( y1 w67c23;P{CaryP}attP2 ) by Best Gene Inc ( Chino Hills , CA ) . UAS-RFP-Syx4 was produced by PCR-amplifying Syx4 and subcloning into pENTR/D-TOPO ( Thermo Fisher Scientific , Waltham , MA ) . Syx4 was then moved into the destination vector pPRG using the Gateway system ( Thermo Fisher Scientific; Gateway vectors developed by T . Murphy , The Carnegie Institution of Washington , Baltimore , MD ) . The construct was injected into w1118 , along with a P-element helper plasmid , for random insertion by Best Gene Inc . Nlg1-Dendra2 was synthesized and subcloned into PBID-UASc ( Wang et al . , 2012 ) by Epoch Life Sciences ( Sugar Land , TX ) . The Dendra2 tag ( Adam et al . , 2009; Gurskaya et al . , 2006 ) was inserted between A865 and L866 , about 11 aa downstream of the TM domain . These 11 aa were then repeated at the end of Dendra2 , as previously described ( Banovic et al . , 2010 ) . The construct was injected into a second chromosome docking strain ( y1 w67c23; P{CaryP}attP40 ) by Best Gene Inc . Full length Syx4A was subcloned into pGEX-2T ( GE Healthcare , UK ) and GST-Syx4A protein was expressed and purified from OneShot BL21 cells ( Thermo Fisher Scientific ) as previously described ( Frangioni and Neel , 1993 ) . Rabbit immunosera were produced by SDIX ( Newark , DE ) . The P element line P{EPgy2}Syx4[EY0005] ( BDSC 14995; Bellen et al . , 2011 ) , carrying an insertion in the 5’-UTR of the Syx4 locus , was crossed to Tft/CyO , Δ2–3 ( BDSC 8201 ) to mobilize the insertion . Single mosaic male progeny were then crossed to 2–4 females from the 1st chromosome balancer stock Df ( 1 ) ED6630/FM7i ( BDSC 8948; Ryder et al . , 2007 ) . In the next generation , single white-eyed balanced females were crossed to 2–3 FM7i males . Approximately 150 lines were tested by PCR to detect deletions of the Syx4 locus . Three Syx4 alleles were identified and sequenced to determine the deletion breakpoints . Syx439: X:2743312 . . 2743832 deleted and >1 kb of P-element sequence inserted; Syx448: X:2742469 . . 2743832 deleted and ~690 bp of P-element sequence inserted; Syx473: X:2738999 . . [2750535–2752545] deleted . A precise excision was also identified and was used as a control line in all experiments unless otherwise indicated . Homology-directed repair ( HDR ) following CRISPR/Cas9-induced double strand break was used to generate C-terminally tagged Syt4 knock-in lines . To construct the HDR donor plasmid pDsRed-Attp-syt4-DNA-eGFP , 1 . 1 kb of genomic DNA downstream of the Syt4 stop codon was inserted at the BglII site of pDsRed-Attp ( Addgene #51019 , gift from Melissa Harrison , Kate O’Connor-Giles , & Jill Wildonger ) , producing the plasmid pDsRed-Attp-Syt4-p2 . Then 1 . 3 kb of genomic DNA upstream of the Syt4 stop codon was fused with eGFP coding sequence and inserted at the NheI site of pDsRed-Attp-syt4-p2 , producing pDsRed-Attp-Syt4-DNA-eGFP-pre . Finally , the gRNA binding sites in this plasmid were mutated , resulting the final donor plasmid . All cloning steps were performed using Gibson Assembly ( New England Biolabs , Ipswich , MA , #E5510 ) . Two gRNA sequences were designed according to Gokcezade et al . , 2014 and inserted into pCFD4-U6:1_U6:3tandemgRNAs ( Addgene #49411; Port et al . , 2014 ) . The plasmid was injected into y1 w67c23; P{CaryP}attP40 embryos by Best Gene Inc . to generate the Syt4-gRNA stock . To generate the GFP-tagged Syt4 flies , yw; nos-Cas9 flies ( Kondo and Ueda , 2013 ) were crossed with Syt4-gRNA flies , and the embryos from the cross were injected with the donor plasmid . Successful transformants were screened for the presence of 3XP3-DsRed in the flies . The nos-Cas9 and Syt4-gRNA expression cassettes were crossed out in the next generation . In the final stock , PCR and sequencing were performed to confirm the insertion and verify that no mutation was present . Several independent lines were generated and validated . One of the homozygous viable and fertile lines , Syt4GFP-2M , was used for all experiments . Homozygous animals were stained with antibodies against GFP to visualize Syt4GFP-2M protein . Larvae were reared at 25°C and dissected at the third wandering instar stage . Larvae were dissected in HL3 . 1 solution ( in mM , 70 NaCl , 5 KCl , 10 NaHCO3 , 4 MgCl2 , 5 trehalose , 115 sucrose , 5 HEPES , pH 7 . 2 ) and fixed in 4% paraformaldehyde or as otherwise indicated . Following washes in PBT ( PBS containing 0 . 3% Triton X-100 ) , larvae were blocked for one hour in PBT containing 2% normal goal serum , incubated overnight with primary antibody at 4°C , washed , incubated with secondary antibodies for 2 hr at room temperature , washed , and mounted in Vectashield ( Vector Laboratories , Burlingame , CA ) for imaging . For Syx4 stainings , Syx4 antibody was preabsorbed on Syx4 null mutant tissue to reduce background staining . Antibodies were as follows: mouse anti-Dlg , 1:1000 ( DSHB 4F3; Parnas et al . , 2001 ) ; anti-Brp , 1:500 ( DSHB nc82; Wagh et al . , 2006 ) ; anti-GluRIII , 1:500 ( Marrus et al . , 2004 ) ; anti-GluRIII-488 , 1:500 ( Blunk et al . , 2014; Marrus et al . , 2004 ) ; anti-Syx4 , 1:500; rabbit anti-Lva , 1:500 ( Sisson et al . , 2000 ) ; DyLight 649 conjugated anti-horseradish peroxidase , 1:1000 ( Jackson ImmunoResearch , West Grove , PA ) ; Alexa Fluor 488 goat anti-mouse , Alexa Fluor 488 goat anti-rabbit , and Alexa Fluor 546 goat anti-mouse , 1:400 ( Thermo Fisher Scientific ) . Images were acquired with a 40 × 1 . 3 NA oil-immersion objective ( Carl Zeiss , Germany ) . The recombinant stock UAS-Syt4-pHluorin , 24B-GAL4 was produced and used for the screen . Females from this line were crossed to UAS-RNAi males , and 3 progeny were dissected per RNAi line tested . Larvae were dissected in HL3 . 1 buffer , fixed in 4% paraformaldehyde , washed in PBT , incubated overnight at 4°C with antibodies against HRP , washed in PBT , and mounted in Vectashield ( Vector Laboratories ) . Syt4-pH distribution at the NMJ was analyzed in hemisegment A3 at muscle 4 . A control cross was included in every batch , where UAS-Syt4-pHluorin , 24B-GAL4 was crossed to a UAS line that had no effect on Syt4-pH distribution ( UAS-FLP; BDSC 4540; Duffy et al . , 1998 ) . A list of all RNAi stocks screened is found in Supplementary file 1 . The acute structural plasticity assay was performed as previously described ( Piccioli and Littleton , 2014 ) . Wandering third instar larvae were dissected in HL3 solution ( in mM , 70 NaCl , 5 KCl , 0 . 2 CaCl2 , 20 MgCl2 , 10 NaHCO3 , 5 trehalose , 115 sucrose , and 5 HEPES , pH7 . 2 ) . Dissecting pins were then moved inward to 60% of the original size for each larva . Relaxed fillets were subjected to three 2 min incubations in high K+ solution ( in mM , 40 NaCl , 90 KCl , 1 . 5 CaCl2 , 20 MgCl2 , 10 NaHCO3 , 5 trehalose , 5 sucrose , and 5 HEPES , pH 7 . 2 ) , spaced by 10 min in HL3 . After the third high K+ incubation , larvae were returned to HL3 solution for 2 min and then stretched to their original size and fixed . Analyses were conducted using Volocity ( version 6 . 3 ) or FIJI / ImageJ ( version 2 . 0 . 0-rc-32/1 . 49v; Schindelin et al . , 2012 ) . Ghost boutons were identified by the presence of a presynaptic bouton ( HRP–labeled ) that lacked Dlg staining in fixed preparations . Counting of boutons and GBs was conducted at hemisegment A3 at muscle 6/7 , and n refers to the number of NMJs analyzed , with no more than two NMJs analyzed per animal , and with animals derived from at least three independent experiments . Measurements of AZ density and GluR intensity were conducted on 12 1b boutons per animal , using 1 terminal bouton and 5 adjacent non-terminal boutons , on two different branches; n refers to the number of animals analyzed . AZ density was quantified manually by counting Brp–labeled puncta and dividing by the volume of HRP . GluR intensity was quantified by measuring the fluorescence intensity of GluRIII signal within an ROI defined by the HRP signal , and the average intensity within the ROI was divided by the average HRP intensity . All analyses were performed blind to genotype . Wandering third instar larvae expressing postsynaptic Nlg1-Dendra2 were dissected in HL3 . 1 saline at room temperature . Images were acquired with a Carl Zeiss LSM 700 with a 40× 0 . 8 NA water-immersion objective using Zen software ( Zeiss ) . A single confocal plane of a muscle 4 NMJ in hemisegment A3 was continuously imaged for 10 min in the red and green channels . After the first 20 s , a single bouton was targeted with pulses of the 405 nm laser at 100% power until ~50% of the green signal was converted . Images were stabilized using the Image stabilizer plug-in from FIJI / Image J ( K . Li , The image stabilizer plugin for ImageJ , http://www . cs . cmu . edu/~kangli/code/Image_Stabilizer . html , February , 2008 ) . Fluorescence intensity was measured in the photoconverted ROI , two adjacent ROIs , and a 4th distant ROI ( photobleaching ROI ) . Each intensity measurement was divided by the photobleaching ROI measurement at that time point to correct for photobleaching . ΔF/F was calculated on corrected measurements as ( FT10min-FT1min ) /FT1min100 . Statistical analyses were conducted using GraphPad Prism . Statistical significance in two-way comparisons was determined by a Student’s t-test , while ANOVA analysis was used when comparing more than two datasets . The P values associated with ANOVA tests were adjusted P values obtained from a Tukey’s post hoc test . In all figures , the data is presented as mean ± SEM; * p<0 . 001 , p<0 . 01 , * p<0 . 05 , n . s . not significant . Statistical comparisons are with control unless noted . Sample size ( n ) , mean , SEM , and pairwise statistical comparisons are presented in figure supplements . RNA was extracted from 5 larvae per sample using an RNease Mini kit ( Qiagen Sciences , Germantown , MD ) and treated with DNAse I ( Qiagen ) . RT-PCR was carried out using a SuperScript One-Step RT-PCR System with Platinum Taq ( Thermo Fisher Scientific ) . Forward primers were designed to bind to unique 5′-UTR sequences of the Syx4A and Syx4B transcripts .	Synapses are connections that allow a neuron to communicate with a neighboring cell ( often another neuron ) . When an electrical impulse traveling down the “presynaptic” neuron reaches the synapse , it causes the neuron to release molecules called neurotransmitters . These molecules then bind to receptors on the surface of the other “postsynaptic” cell and cause that cell to respond in a particular way . Communication between the two cells at the synapse can also go in the opposite direction , with the postsynaptic cell signaling to the presynaptic cell . Such “retrograde” signals typically regulate the properties of the synaptic connection , such as changing the strength or shape of the synapse , or altering which proteins are present there . While a lot is known about how a presynaptic neuron communicates with the postsynaptic cell , not as much is known about how retrograde signals are regulated . Harris et al . therefore set out to identify and characterize new factors that control retrograde signaling , and started by producing a list of likely candidate molecules . These candidates were then screened by removing them one at a time from the synapses between motor neurons and muscle cells in fruit flies and observing the effect this had on a molecule called Synaptotagmin 4 . Synaptotagmin 4 is normally found at the membrane of the postsynaptic cell . Harris et al . found that removing one candidate molecule , called Syntaxin 4 , from the postsynaptic cell reduced the amount of Synaptotagmin 4 at the membrane . Further investigation showed that Syntaxin 4 also helps to deliver a protein called Neuroligin 1 to the postsynaptic membrane , which is important for organizing the synapse . By identifying Syntaxin 4 as a new regulator of retrograde signaling , Harris et al . open up several avenues of investigation that could reveal more about the mechanisms that influence how synapses work .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "neuroscience" ]	2016	The postsynaptic t-SNARE Syntaxin 4 controls traffic of Neuroligin 1 and Synaptotagmin 4 to regulate retrograde signaling
Little is known about the capacity of lower vertebrates to experience itch . A screen of itch-inducing compounds ( pruritogens ) in zebrafish larvae yielded a single pruritogen , the TLR7 agonist imiquimod , that elicited a somatosensory neuron response . Imiquimod induced itch-like behaviors in zebrafish distinct from those induced by the noxious TRPA1 agonist , allyl isothiocyanate . In the zebrafish , imiquimod-evoked somatosensory neuronal responses and behaviors were entirely dependent upon TRPA1 , while in the mouse TRPA1 was required for the direct activation of somatosensory neurons and partially responsible for behaviors elicited by this pruritogen . Imiquimod was found to be a direct but weak TRPA1 agonist that activated a subset of TRPA1 expressing neurons . Imiquimod-responsive TRPA1 expressing neurons were significantly more sensitive to noxious stimuli than other TRPA1 expressing neurons . Together , these results suggest a model for selective itch via activation of a specialized subpopulation of somatosensory neurons with a heightened sensitivity to noxious stimuli . Itch is an unpleasant sensation that elicits a scratch behavior in terrestrial vertebrates . In mammals , chemically-induced itch is thought to be mediated by pruritic receptors on somatosensory neurons ( Bautista et al . , 2014; Hoon , 2015 ) . These receptors are typically G-protein coupled receptors ( GPCRs ) that , upon activation , prompt the opening of downstream transient receptor potential ( TRP ) channels , facilitating activation of the neuron ( Ross , 2011; Zhang , 2015 ) . This coupling of pruritic receptors to TRPA1 or TRPV1 is especially intriguing in that these TRP channels also serve as nociceptors , mediating responses to algogenic ( painful ) stimuli ( Laing and Dhaka , 2016; Ikoma et al . , 2006 ) . Zebrafish ( Danio rerio ) have proven to be a valuable tool in the study of nociception ( Gau et al . , 2013 ) . The zebrafish ortholog of Trpa1 , trpa1b , is required for nociceptive responses to aversive pungent chemicals ( Prober et al . , 2008 ) . Orthologs of genes involved in mammalian itch transduction are also present in the zebrafish ( Kaslin and Panula , 2001; Pei et al . , 2016; Xu et al . , 2011 ) . Studying how these itch genes operate in the somatosensory system of zebrafish could reveal conserved itch transduction mechanisms , providing insight into the evolution of itch . In an effort to determine if pruritic stimuli are capable of eliciting somatosensory activity in zebrafish , we screened five compounds known to both induce acute pruritus in mammals and act on receptors expressed by zebrafish ( Schön and Schön , 2007; Bell et al . , 2004; Lieu et al . , 2014; Yamaguchi et al . , 1999; Tsujii et al . , 2009 ) , excluding pruritogens that act on receptors that do not have a zebrafish ortholog , such as MRGPR agonists . Allyl isothiocyanate ( AITC ) , a known algogen and TRPA1 agonist ( Prober et al . , 2008; Jordt et al . , 2004 ) , was used as the positive control . We evaluated somatosensory neuronal responses to pruritic compounds using transgenic larvae that pan-neuronally express the neuronal activity indicator CaMPARI ( elavl3:CaMPARI ) , a fluorescent protein that permanently photoconverts from green to red in the presence of calcium and 405 nm light ( Fosque et al . , 2015 ) . Using this approach , we were able to view trigeminal neuronal activity in 3 day post fertilization ( dpf ) larval zebrafish following the application of each pruritogen ( Figure 1A–H ) . Of the pruritogens screened , only imiquimod ( IMQ ) significantly ( p<0 . 05 ) activated zebrafish trigeminal ganglia ( TG ) neurons ( Figure 1I ) . We have previously reported that noxious stimuli evoke locomotion in larval zebrafish ( Gau et al . , 2013 ) . When 5dpf larval zebrafish were exposed to individual pruritogens , only IMQ elevated baseline locomotion ( p<0 . 001 ) , producing a dose dependent increase ( Figure 1—figure supplement 1A ) . When coupled with our findings in CaMPARI transgenics , these results indicate that IMQ likely acts through somatosensory neurons to evoke behavioral responses . While 5-HT did produce a reduction in locomotion , it did not activate somatosensory neurons ( Figure 1I , J ) . Given the limited sensitivity of the locomotor assay we were unable to differentiate nocifensive behavior elicited by AITC from potentially pruritic behavior elicited by IMQ . To address this issue , we employed an adult zebrafish behavioral assay ( Correia et al . , 2011; Maximino , 2011 ) . Injection of IMQ into the lip elicited a lip-rubbing behavior that may constitute a form of itch-scratch response in zebrafish , a behavior that was absent in sham-injected control fish ( p<0 . 001 ) ( Figure 1K , L; Videos 1—3 ) and distinct from previously described zebrafish nocifensive , escape , or exploratory behaviors ( Colwill and Creton , 2011; Egan et al . , 2009; Levin et al . , 2007 ) . Consistent with studies of nocifensive behaviors , injection of AITC produced freezing behavior ( p<0 . 01 ) ( Figure 1—figure supplement 1C ) as well as a significant decrease ( p<0 . 05 ) in velocity not observed in control or IMQ injected fish ( Figure 1—figure supplement 1B ) . Such distinct behavioral responses imply that zebrafish are capable of experiencing , and responding differentially to , discrete stimuli analogous to itch and pain in mammals . The mechanism by which IMQ elicits itch in mammals is unclear . IMQ , a treatment for various skin disorders , acts through TLR7 to stimulate an immune response , with intense itching and painful burning commonly reported as side effects ( Chang et al . , 2005; Lebwohl et al . , 2004 ) . In mice , however , IMQ is reported to be itch selective , and only elicits scratching , but not nociceptive , behaviors ( Liu et al . , 2010; Kim et al . , 2011 ) . Furthermore , there is dispute surrounding TLR7’s role in IMQ-induced itch . One study found that Tlr7-/- mice showed deficits in IMQ evoked itch and proposed that Tlr7 expressed in dorsal root ganglion ( DRG ) neurons was mediating IMQ transduction ( Liu et al . , 2010 ) . TLR7 has also been reported to couple with TRPA1 in DRG neurons to evoke nociception in response to other TLR7 agonists ( Park et al . , 2014 ) . However , conflicting studies found that Tlr7-/- mice exhibited no deficits in IMQ-evoked itch , and RNAseq analysis of DRG neurons found no evidence for Tlr7 expression ( Kim et al . , 2011; Usoskin et al . , 2015; Li et al . , 2016 ) . We therefore investigated whether TLR7 and/or TRPA1 were involved in transducing the neural and behavioral responses to IMQ in zebrafish . In situ hybridization studies in zebrafish larvae revealed that tlr7 mRNA expression is restricted to known hematopoietic regions in larval zebrafish ( Bresciani , 2014; Du et al . , 2011 ) , and was notably absent in both TG and Rohon-Beard ( RB ) neurons ( Figure 2A; Figure 2—figure supplement 1I ) . As expected , trpa1b expression was observed in both TG and RB neurons ( Figure 2B; Figure 2—figure supplement 1C ) . Therefore , any role TLR7 could play in IMQ-evoked behavior would be via indirect mechanisms . To determine if trpa1b and/or tlr7 are required for IMQ induced behaviors , we introduced early nonsense mutations in the coding sequences of both genes ( Kimura et al . , 2014 ) . Tlr7-/- zebrafish larvae exhibited a significant ( p<0 . 001 ) increase in total locomotion when exposed to IMQ ( 100 µM ) that was indistinguishable from controls ( Figure 2C ) . However , trpa1b-/- larvae demonstrated no response to IMQ ( 100 µM ) , while their WT siblings displayed normal IMQ induced behaviors ( p<0 . 01 , Figure 2D ) . These data support a mechanism where trpa1b , but not tlr7 , is necessary for mediating behavioral responses to IMQ in larval zebrafish . As expected based on previous reports , behavioral responses to AITC were absent in trpa1b-/- fish ( Figure 2—figure supplement 1F ) ( Prober et al . , 2008 ) . Notably , trpa1b-/- fish demonstrated an equivalent increase in locomotor behavior as their WT siblings when exposed to increased temperatures ( Figure 2—figure supplement 1E ) . This indicates that the trpa1b mutation specifically affects Trpa1b-mediated sensations , rather than causing generalized sensory impairment . To test whether the presence of Trpa1b was necessary to mediate neuronal responses in larval zebrafish TG , we conducted in vivo calcium imaging using elavl3:GCaMP5g larvae ( Akerboom et al . , 2012 ) ( Figure 2G ) . In WT larvae , IMQ activation was seen exclusively in a subset of AITC responsive neurons ( 4/36 , n = 5 larvae ) . Trpa1b-/- fish , however , exhibited no response to either IMQ or AITC ( 0/89 total neurons , n = 5 larvae ) ( Figure 2H ) . Notably , the highly specific TLR7 agonist loxoribine did not evoke TG neuron activation , larval locomotion , or adult lip-rubbing behavior , further strengthening the finding that Tlr7 does not play a role in IMQ evoked behaviors in zebrafish ( Figure 1I , J; Figure 1—figure supplement 1D ) . This finding is similar to reports in the mouse demonstrating that loxoribine does not elicit pruritic behavioral responses ( Kim et al . , 2011 ) . To determine if TRPA1 directly interacts with IMQ to produce itch , we utilized the TLR7-deficient cell line HEK293T ( Hornung et al . , 2005 ) . HEK cells transfected with zebrafish , mouse , and human Trpa1 showed a dose-dependent increase in intracellular calcium following application of IMQ and AITC that was not observed in HEK cells alone ( Figure 3A–F; Figure 3—figure supplement 1A , C ) . Importantly , we found that loxoribine did not activate HEK cells transfected with zebrafish or mouse Trpa1 ( Figure 3—figure supplement 2A–D ) , indicating that TRPA1 is responsive to IMQ , and not to TLR7 agonists in general . While we observed no expression of tlr7 in the zebrafish TG ( Figure 2A ) , given the lack of consensus over its functional role we sought to determine whether TLR7 might serve as a pruritic co-receptor that potentiates the TRPA1 response . We co-transfected HEK cells with Trpa1 and Tlr7 and examined the calcium responses following treatment with IMQ and observed no discernible differences in the average peak responses to IMQ between Trpa1 and Trpa1 + Tlr7 conditions ( Figure 3A–C , Figure 3—figure supplement 1C ) . Whole cell electrophysiological experiments corroborated these findings . When stimulated with IMQ , voltage-clamped HEK cells transfected with mouse or zebrafish Trpa1 demonstrated a significant increase in current ( Figure 3G; Figure 3—figure supplement 1G ) . Co-transfecting mouse Tlr7 with mouse Trpa1 in HEK cells had no demonstrable effect on current influx ( Figure 3—figure supplement 1H ) . Additionally , no difference was found in the IMQ current density dose-response curves for cells transfected with zebrafish trpa1b , mouse Trpa1 , or mouse Trpa1 + mouse Tlr7 ( Figure 3H ) . In contrast to previous reports , we found no evidence that TLR7 coupled with TRPA1 in the presence of loxoribine as measured by ratiometric calcium imaging and whole cell electrophysiology in mouse and zebrafish ( Figure 3—figure supplement 2A–D ) ( Liu et al . , 2010 ) . Together , our data suggest that TLR7 plays no role in the direct activation of somatosensory neurons , and does not appear to potentiate the response of TRPA1 to IMQ . Following these results , we confirmed that mouse and human TLR7 were present and functional in our assays ( Figure 3—figure supplement 1I–K ) ( Mitchell and Sugden , 1995 ) . Intriguingly , zebrafish Tlr7 did not respond to either IMQ or loxoribine in a dual-luciferase assay , suggesting that zebrafish Tlr7 is not activated by mammalian TLR7 agonists ( Figure 3—figure supplement 2E ) . However , due to a lack of a Tlr7 positive control , we were unable to confirm that Tlr7 was functional in our heterologous expression system . With this caveat in mind , the lack of zebrafish Tlr7 response to these TLR7 agonists lends further credence to the conclusion that Tlr7 is not involved in somatosensory neuronal activation or behavior in this species . If TRPA1 does not couple with TLR7 , but is instead directly activated by both IMQ and AITC , how could IMQ be itch-selective in the mouse ? To address this question we assessed the EC50 and peak responses to IMQ and AITC in HEK cells transfected with Trpa1 from different species . The EC50 of IMQ for zebrafish , mouse and human TRPA1 was consistently higher than the AITC EC50 , demonstrating that IMQ is a weaker agonist than AITC ( Figure 3—figure supplement 1K ) . Notably , while the IMQ EC50 of zebrafish Trpa1 and human TRPA1 were only 2–3 fold greater than that of the EC50 of AITC , mouse TRPA1 demonstrated a ~40 fold difference between the EC50 of AITC and IMQ . Furthermore , only mouse TRPA1 elicited significantly lower maximum responses to IMQ than AITC ( Figure 3D–F ) . In similar electrophysiology experiments , HEK cells transfected with zebrafish trpa1b exhibited identical current density responses upon stimulation with the maximum dose of either AITC or IMQ , but cells transfected with mouse Trpa1 exhibited significantly higher current density responses following stimulation with AITC , relative to IMQ ( Figure 3I ) . Such physiological differences in TRPA1 function between species could provide a potential mechanism for the itch selectivity of IMQ in mice , implying that IMQ is not a strong enough mouse TRPA1 agonist to elicit nociception in this species . Conversely , the ability of both zebrafish and human TRPA1 to respond equally to AITC and IMQ at maximal doses potentially explains how IMQ can elicit both itch and pain sensations in humans , and suggests that the same may be observable in fish . In the preceding in vivo calcium imaging experiments , we observed that only a small proportion of zebrafish Trpa1+ neurons , identified by their responsiveness to AITC , were also responsive to the IMQ stimulus . To further explore this result , we used elval3:H2BGCaMP6 ( Chen et al . , 2013 ) transgenic zebrafish to record the response properties of larval TG neurons to IMQ and AITC . No IMQ+/AITC- neurons were found across 13 larvae . Among AITC+ neurons , 28% ( 31/111 ) were responsive to IMQ ( Figure 4A–C ) . The above data affirms that IMQ+ neurons are a subset within a larger population of Trpa1+ TG neurons , implying that a population coding strategy for pruritus might be at play . This does not itself answer the question of how such an itch-selective Trpa1+ subpopulation might be activated by a TRPA1 agonist without recruiting other Trpa1+ neurons that may code for nociceptive behaviors . Based on our finding that IMQ is a weaker TRPA1 agonist than AITC , one potential explanation is that such an itch-selective Trpa1+ population is more sensitive to TRPA1 agonists , and can be activated by weaker ( pruritic ) stimuli . In IMQ+/AITC+ neurons , we determined that the average maximum fluorescence intensity response to the IMQ stimulus was significantly lower than that of the AITC stimulus ( p<0 . 001 ) , implying that at the concentrations used , IMQ is indeed a weaker TRPA1 stimulus than AITC in vivo ( Figure 3—figure supplement 2A ) . Furthermore , we found that the maximum AITC response of IMQ+/AITC+ neurons was significantly greater than that of AITC+ only neurons ( Figure 4D ) . Likewise , IMQ+/AITC+ neurons displayed a significantly greater average AITC response than AITC+ only neurons ( p<0 . 05 , Figure 4—figure supplement 1B ) . These data suggest that IMQ+ TG neurons are primed to respond to TRPA1 agonists and support a model where relatively weak TRPA1 stimuli , such as IMQ at the concentration used , could selectively recruit a potential itch-coding subpopulation of Trpa1+ neurons . Higher intensity stimuli like AITC at the concentrations used , however , would activate the majority of Trpa1+ neurons to evoke nocifensive behaviors , positively correlating with findings that nociception takes precedence over itch sensation ( Roberson et al . , 2013; Ross et al . , 2010; Liu et al . , 2011 ) . To verify that IMQ activates a selective subset of Trpa1+ neurons as opposed to activating Trpa1+ neurons stochastically , we performed calcium imaging experiments in which 3dpf elavl3:H2BGCaMP6 were exposed to successive pulses of 100 μM IMQ . Of the IMQ+ neurons we identified across five fish , 92 . 3% ( 12/13 ) responded to both pulses of IMQ , whereas only 7 . 7% ( 1/13 ) responded only to the second pulse of IMQ ( Figure 4—figure supplement 2A–C ) . Additionally , it is possible the single neuron that responded only to the second pulse of IMQ may also be a dual-responder . While the GCaMP fluorescence change only crossed our response threshold during the second pulse , it is possible that the sloping baseline may have obscured a minimal response to the first pulse , especially considering the low amplitude of the second response . However , for purposes of completion we decided to include this trace in our final counts . Given the finding that only ~30% of AITC responsive neurons responded to one pulse of IMQ ( 100 μM ) , if one assumes that this is the probability that any given AITC responsive neuron would respond to IMQ ( 100 μM ) , and that responses to IMQ are stochastic in nature , one would expect only a small fraction of neurons to be double responders ( ~9% ) ( Figure 4D ) . The finding that nearly all IMQ-responsive neurons were dual responders argues that these neurons comprise a distinct population of Trpa1 expressing neurons , primed to respond to low intensity TRPA1 dependent stimuli . Although the IMQ dose-response curve in zebrafish trpa1b-transfected HEK cells was rightward shifted , it was eventually able to elicit the same amount of intracellular calcium flux that AITC evoked . This implies that at a sufficiently high concentration , IMQ might be able to recruit neurons outside of the IMQ+ subpopulation observed in the above larval calcium imaging experiments , thus eliciting neuronal and behavioral responses characteristic of AITC at nociceptive concentrations . Likewise , it is also possible that at low enough concentrations , AITC is capable of eliciting the pruritic neuronal and behavioral responses we observed following application of IMQ . In CaMPARI fish we observed that decreasing the concentration of applied AITC correlated with a reduction in the number of photoconverted neurons ( Figure 4E ) . Additionally , administering higher IMQ concentrations converts an equivalent number of neurons as high concentrations of AITC ( Figure 4E ) , suggesting that for TRPA1 agonists , eliciting pruritus or nociception is dependent more on stimulus intensity than identity . In vivo GCaMP imaging bolstered our CaMPARI findings that stimulus intensity affects which subpopulations of Trpa1+ neurons are activated . In these experiments , we observed that increasing the stimulus intensity activated more neurons . Only a subset of neurons that responded to a high concentration ( 50 μM ) of AITC responded to a lower concentration ( 10 μM ) of AITC ( 11/42 ) , and of those even fewer neurons responded to 100 μM IMQ ( 4/11 ) ( Figure 4—figure supplement 1C ) . Furthermore , increasing the IMQ concentration to 200 µM in adult behavioral experiments evoked nocifensive behaviors such as elevated freezing and significantly reduced velocity , and the itch-like lip rubbing behavior seen at 100 µM was notably absent ( Maximino , 2011; Sneddon , 2009 ) ( Figure 4F; Figure 1—figure supplement 1C , Figure 4—figure supplement 1D ) . Conversely , low concentrations of AITC ( 5 μM ) elicited both itch-like lip rubbing behavior and increased velocity ( Figure 4F; Figure 4—figure supplement 1C ) . These data indicate that the subpopulation of Trpa1+ neurons that drive itch behavior in the zebrafish are distinct in their sensitivity to TRPA1 agonists , but can be activated by either AITC or IMQ at the appropriate concentration to produce equivalent behaviors . We found that IMQ elicited responses in 9 . 6% ( 83/864 ) of cultured AITC+ DRG neurons ( Figure 5A ) from WT animals . To determine if TRPA1 mediates IMQ responses in mice , we examined IMQ-evoked responses in DRG neurons from both WT and Trpa1-/- animals using ratiometric calcium imaging . We found that both IMQ and AITC responses were completely abolished in DRG neurons obtained from Trpa1-/- animals , while neurons from WT siblings exhibited normal responsivity to both stimuli ( Figure 5B , C ) . Furthermore , application of loxoribine did not elicit calcium responses in mouse DRG neurons , providing evidence that TLR7 stimulation does not result in activation of somatosensory neurons ( Figure 5—figure supplement 1H ) . We next explored whether IMQ-evoked scratching behavior was also dependent on TRPA1 . High-dose IMQ ( 125 μg ) paw injections did not evoke nocifensive behaviors in WT mice ( n = 0/10 IMQ-injected ) , consistent with previous reports ( Kim et al . , 2011 ) . However , scratching bouts at a low concentration of IMQ ( 10 μg , nape injected ) were significantly attenuated in Trpa1-/- mice , demonstrating that TRPA1 is required for normal IMQ-induced scratching behavior ( Figure 5D ) . Interestingly , a higher dose of IMQ ( 50 μg ) evoked equivalent scratching behavior in both WT and Trpa1-/- mice ( Figure 5—figure supplement 1G ) . This result , taken together with our finding that isolated mouse DRG neuron responses to IMQ are TRPA1-dependent , suggests that IMQ can also evoke itch via indirect activation of somatosensory neurons , perhaps downstream of an immune response ( Bautista et al . , 2014; Hoon , 2015 ) . Given the itch selectivity of IMQ in the mouse , we sought to determine whether IMQ+ neurons were part of a population of DRG neurons that encode TRPA1-dependent pruritus . We therefore measured the overlap of IMQ+ neurons with DRG neurons that responded to a mixture of the TRPA1-dependent pruritogens deoxycholic acid ( DC ) and chloroquine ( CQ ) ( Figure 5E–F ) ( Tsujii et al . , 2009; Wilson et al . , 2011; Liu et al . , 2009 ) . We found that the vast majority of IMQ+ neurons ( 73% , 23/30 ) also responded to these pruritic stimuli ( Figure 5G ) , indicating that in the mouse , IMQ+ neurons belong to a subpopulation of itch-encoding neurons . Due to the parallels noted between zebrafish and mouse IMQ responses , we proceeded to investigate whether the correlation between stimulus intensity and neuronal activation that we observed in the zebrafish was conserved in the mouse . In the mouse , increasing the concentration of AITC activated more DRG neurons in a dose-dependent manner ( Figure 5—figure supplement 1I ) . Furthermore , within the population of neurons that responded to lower concentrations of AITC , the IMQ+ subpopulation was enriched ( Figure 5—figure supplement 1J ) . We also found that IMQ+ neurons in the mouse had a smaller peak responses to IMQ than AITC ( Figure 5—figure supplement 1A–C ) . Additionally , AITC peak responses within the IMQ+ population were significantly greater than in the AITC+ only population ( p<0 . 01 ) , demonstrating that the heightened sensitivity of IMQ+ neurons to TRPA1 agonists is conserved in the mouse ( Figure 5H ) . Subsequent experiments revealed that IMQ+ DRG neurons exhibited a significantly higher maximum response to the TRPV1 agonist capsaicin ( CAPS ) than CAPS+ only neurons ( Figure 5—figure supplement 1D–F ) , implying that the IMQ+ neurons may be intrinsically more sensitive to noxious stimuli , not exclusively TRPA1 agonists . Finally , in accordance with our zebrafish data , we observed that AITC stimulus intensity dictates whether mice exhibit pruritic or nocifensive behaviors . In order to discriminate between nocifensive and pruritic behaviors , we employed a ‘cheek model of itch’ assay in which compounds injected into the cheek may elicit scratching ( a pruritic response ) or wiping ( a nocifensive response ) ( Shimada and LaMotte , 2008; Akiyama et al . , 2010 ) . We observed that AITC ( 50 mM ) produces significant scratching behavior with no appreciable wiping behavior ( p<0 . 001 and N . S . respectively ) , while a higher dose of AITC ( 100 mM ) results in a significant attenuation of the observed scratching behavior , as well as a significant wiping behavior ( p<0 . 05 and p<0 . 001 respectively ) ( Figure 5I ) . Our work demonstrates that IMQ can directly activate TRPA1 to elicit pruritic behavioral responses in both the zebrafish and mouse . Furthermore , we have shown that the immune receptor TLR7 does not mediate somatosensory neuronal responses to IMQ . Our results imply that in both species a subset of highly sensitive TRPA1-expressing itch-encoding neurons can respond to weaker TRPA1 agonists to encode sensations of itch and elicit discrete itch behaviors . More intense stimuli , such as those that evoke nocifensive behaviors , appear to recruit this highly-sensitive subset as well as less-sensitive TRPA1-expressing neurons . Our finding that IMQ responsive neurons in the mouse are part of a TRPA1-expressing subpopulation that is activated by other TRPA1 dependent pruritogens provides further evidence that these more sensitive neurons indeed signal itch . Parallel observations between the zebrafish and mouse suggest that this relatively simple mechanism for conveying , and distinguishing between , pruritic and algogenic stimuli originated early in vertebrate evolution and appears to be preserved in mammals . In sum , our results support the existence of a population-coding based strategy through which differential activation of TRPA1-expressing somatosensory neurons with high or low sensitivities to TRPA1 agonists can relay the discrete sensations of itch and pain respectively . Adult Zebrafish ( Danio rerio ) were raised with constant filtration , temperature control ( 28 . 5 ± 2°C ) , illumination ( 14 hr:10 hr light-dark cycle , lights on at 9:00 AM ) , and feeding . All animals were maintained in these standard conditions and the Institutional Animal Care and Use Committee approved all experiments . Adult zebrafish not used in behavioral experiments were bred in spawning traps ( Thoren Caging Systems , Hazelton , PA ) from which embryos were collected . Embryonic and larval zebrafish were raised in petri dishes ( Fisher Scientific , Hampton , NH ) of E2 medium with no more than 50 embryos per dish at 28 . 5 ± 1°C in an incubator ( Sanyo ) . Embryos were staged essentially as described ( Kimmel et al . , 1995 ) and kept until 5dpf . Trpa1+/+ and congenic Trpa1−/− mice on the C57BL/6J background were described previously ( Cruz-Orengo et al . , 2008 ) . All mice were housed under a 12 hr light/dark cycle with food and water provided ad libitum . All behavioral tests were videotaped from a side angle , and behavioral assessments were done by observers blind to the treatments or genotypes of animals . All mice used for behavior tests were age , sex and body weight matched . All experiments were performed in accordance with the guidelines of the National Institutes of Health and the International Association for the Study of Pain , and were approved by the Animal Studies Committee at Washington University School of Medicine . HEK 293T cell stocks were initially purchased from ATCC , which authenticated their identity via STR profiling . Cells tested negative for mycoplasma contamination . Cells were cultured in DMEM ( Life Technologies , Carlsbad , CA ) supplemented with fetal bovine serum and antibiotics ( penicillin/streptomycin ) , and passaged every 2–3 days . Individual larvae were processed as previously described ( Meeker et al . , 2007 ) . Larvae were anesthetized with tricaine , and placed in individual PCR tubes with a small quantity of E2 media . An equivalent amount of a 2X base solution made from a 50x stock ( 1 . 25 M NaOH , 10 mM EDTA pH 12 ) was then added to each tube , and all tubes were incubated at 95°C for 30 min . Following this , 1x neutralization solution ( again made from from a 50x solution , 2M Tris-HCl pH 5 ) was added , and the resulting DNA solutions were stored at −20°C . Adult genomic DNA was extracted using similar methods , but with a few minor modifications . Individual fish were anesthetized with tricaine , and a small portion of the tail fin was removed with a scalpel and placed into an individual PCR tube . 1X base solution was then applied to the piece of tissue , which was incubated for 30 min at 95°C until an equivalent amount of 1X neutralization solution was added . Nonsense mutants for both trpa1b and tlr7 were generated essentially as previously described ( Shah et al . , 2016 ) . To synthesize the template DNA required for the in vitro transcription we employed a two oligo PCR method , one oligo contained the RNA loop structure required for recognition by the Cas9 enzyme and had the sequence 5'[gatccgcaccgactcggtgccactttttcaagttgataacggactagccttattttaacttgctatttctagctctaaaac]3' . The second , gene specific , oligo had the sequence 5'[aattaatacgactcactata ( N20 ) gttttagagctagaaatagc]3’ , where ( N20 ) refers to the 20 nucleotide oligo that binds the genome . In the case of the trpa1b nonsense mutants the N20 oligo was 5’[GGCGTATAAATACATGCCAC]3’ . In the case of the tlr7 nonsense mutant the N20 oligo was 5’[GGGGATGTAGGACAAGTTGT]3’ . A mixture of 400 μL of Cas9-encoding mRNA and 200 ng/μL of the proper sgRNA was injected into zebrafish embryos of the AB background at the one cell stage . Fish were then screened for mutations using the following primers , tlr7 5’ GGATGCGTTTATGCTGCTTGACAA , tlr7 3’ AATGTTGTTGTTGTACAGGTAGAGCTC , trpa1b 5’-CTCATACATTCATAAACCTGCCTGATAT , and trpa1b 3’ – TGGAGGGGCGTCAGACCCTTT , and Sanger Sequencing . We identified two nonsense mutants for trpa1b , one that possessed a 4 bp insertion and one that possessed a 7 bp deletion . We then outcrossed these founders ( F0 ) to WT fish of the same genetic background ( AB line ) and screened for germline transmission in the F1 generation . Members of the F1 generation were additionally backcrossed to WT fish to establish an F2 generation . Heterozygous F2 zebrafish were then crossed to each other to produce an F3 generation that was used for experiments; additionally F2 zebrafish were crossed to transgenics expressing calcium indicator proteins ( CaMPARI , GCaMP ) under neuronal promoters in order to perform functional imaging studies . In some instances , F3 zebrafish were backcrossed a fourth time to establish younger generations of fish . Animals that were homozygous for either the 4 bp insertion or 7 bp deletion possessed identical phenotypes ( i . e . , lack of behavioral response to AITC ) . Likewise , zebrafish with a 4 bp/7 bp phenotype possessed an identical phenotype as 4 bp/4 bp and 7 bp/7 bp homozygotes . We identified one nonsense mutant for tlr7 that possessed a 1 bp deletion . As with the generation of the trpa1b mutant line , this founder was outcrossed to a WT fish and the offspring were screened for germline transmission . Subsequent generations were backcrossed in the manner described above . In the trpa1b experiments , WT siblings of trpa1b-/- fish were used as the controls . In the tlr7 experiments , a pure tlr7-/- line was established and compared to age-matched WT fish , since we were unable to genotype the 1 bp mutation via conventional methods ( gel electrophoresis , HRMA ) and could only identify the mutation via sequencing . At 5dpf , larval zebrafish ( AB background ) were placed into individual wells on a 96-well mesh bottom plate ( Millipore , Burlington , MA ) resting in a bath of E2 medium . The 96-well plate was then transferred to a hot plate that was maintained at a constant temperature of 28 . 5°C . Then the 96-well plate was moved from the E2 medium bath to the experimental bath for four minutes , during which the behavioral response of the larval zebrafish was recorded with a HD camcorder ( Canon , Japan ) . Experiments were performed blindly and each larva’s total locomotive behavioral response was tracked using Ethovision ( Noldus , Netherlands ) . Statistical analysis was done using an analysis of variance ( Graphpad Prism 6 ) or Student’s t-test . All experimental compounds were purchased from Sigma Aldrich unless otherwise noted and were made up in 1% dimethyl sulfoxide ( DMSO , Sigma Aldrich , St . Louis , MO ) and E2 medium . In experiments involving nonsense mutants and their WT siblings , all larvae were genotyped following video capture of the behavioral response . Briefly , each larva was removed from its well and placed into a PCR tube in 25 μL of E2 media . gDNA was extracted using the base extraction technique described above . For experiments involving trpa1b-/- fish , all larvae were genotyped by HRMA ( CFX Connect , BioRad , Hercules , CA ) using the primers 5’-CTCATACATTCATAAACCTGCCTGATAT and 3’-TGGAGGGGCGTCAGACCCTTT . As mentioned previously , due to difficulties in genotyping the 1 bp deletion in tlr7 nonsense mutants , animals were identified by genomic sequencing , and a pure tlr7-/- was created for use in behavioral experiments and were compared to AB fish . elavl3:CaMPARI zebrafish in the Casper background were simultaneously exposed to chemical stimuli and a 405 nm light in order to permanently photoconvert active neurons ( Fosque et al . , 2015 ) . Briefly , 3dpf larval zebrafish were paralysed by injecting α-bungarotoxin protein ( Sigma ) into the chest cavity using microinjection needles pulled on a Flaming-Brown Micropipette Puller ( model P-87 , Sutter Instrument Co . , Novato , CA ) and a Picrosprizter II microinjection apparatus ( General Valve Corporation , Fairfield , NJ ) . Paralysed fish were then placed in small glass-bottomed dishes ( Wilco Wells , Netherlands ) filled with an individual chemical from the pruritic screen and allowed to incubate for 2 min . Following this incubation period , glass-bottomed dishes were placed on the stage of an inverted fluorescent microscope ( Olympus , Japan , model Ix81S1F-3 ) and the larvae were exposed to a 405 nm light for 40 s using MetaMorph software ( Molecular Devices , San Jose , CA ) . Post-exposure fish were removed from the chemical and placed in a petri dish filled with embryo media and tricaine to prevent any future activation of sensory neurons . Immediately prior to imaging , larvae were mounted on coverslips in 1 . 5% agarose +tricaine in EM . TG and surrounding neural tissue were imaged using a 20x lens on an LSM 880 confocal microscope ( Zeiss , Germany ) . Zen Black software was used to scan through the entire TG , acquiring a 1024 × 1024 pixel image slice at every ~5 µm that could then be stacked in the Z plane until the entire ganglion was imaged . Images were examined for photoconverted ( red-labeled ) neurons , and totals were established for each TG in each condition . When used , ANOVA statistical tests were done against control . Adult zebrafish were placed in traps ( Thoren Caging Systems ) and were transported to the experimental area , which was maintained at 28 . 5 ± 2°C , where they were left to acclimate for one hour . After completing acclimation fish were transported one at a time to the injection area . Each fish was anaesthetized by exposure to 12 . 0 ± 0 . 3°C system water . They were then immobilized and injected in the upper lip using a 33 gauge Hamilton needle and 20 μL Hamilton syringe . Fish were injected with 10 μL of experimental or control solution , all of which were made up in 1% DMSO , 1x PBS , and distilled water . After injection , fish were placed into a trap and transferred to the recording area . The behavioral response was recorded for five minutes using an HD camcorder ( Cannon ) . The velocity of the fish was then analyzed using Ethovision ( Noldus ) and all facial interactions were manually scored to prevent any bias in the data . All analysis was blinded . All statistical analysis were done with an analysis of variance ( Graphpad Prism 6 ) or Student’s t-test . In the case of nonsense mutant experiments , after the behavioral responses were captured , each fish was euthanized by tricaine overdose and fin clipped , and the fin section was placed in a PCR tube . Then , gDNA was extracted from the excised tissue using the previously described base extraction technique . Trpa1b genotype was determined using the same HRMA strategy as employed in the larval behavioral experiments . Since a homozygous tlr-/- line was employed for adult behavioral experiments , genotyping post-experiment was unnecessary . When used , ANOVA statistical tests were done against control . Whole-mount colorimetric in situ hybridization to determine trpa1b and tlr7 expression was performed on 3dpf larvae as described previously ( Gau et al . , 2013 ) . Pigment formation was inhibited by exposing larvae to 1-phenyl 2-thiourea ( PTU ) at 24hpf . Larvae were hybridized with DIG-labeled riboprobes for trpa1b or tlr7 overnight at 65°C . They then underwent a series of stringent washes , followed by incubation in α-DIG conjugated Fab fragments ( Roche , Switzerland , 1:10 , 000 ) and staining in NBT/BCIP solution . Larvae were washed with PBTw and stored in glycerol until imaging , whereupon they were mounted in 100% glycerol and photographed using an upright Axioplan2 microscope ( Zeiss ) . 3dpf zebrafish larvae from either elavl3:GCaMP5 ( Akerboom et al . , 2012 ) or elavl3:H2BGCaMP6 ( Chen et al . , 2013 ) transgenic line were paralysed as described above . After paralysis , larvae were mounted in 2% agarose in EM on coverslips , which were then placed into a perfusion chamber ( Warner Instruments ) . Once solidified , the agarose immediately surrounding the head was cut away with a scalpel to ensure maximal exposure to chemical stimuli . The perfusion chamber was placed onto the stage of an Olympus Fluoview FV-1000 multiphoton microscope equipped with an infrared laser controlled by Mai Tai software ( Spectra-Physics , Thermo Electron Corporation , Walthom , MA ) . Larvae were imaged under the following parameters: laser wavelength of 880 nm , resolution of 4 . 0 µs/pixel , frame rate of 1–3 s per frame , frame size of 512 × 512 pixels . Laser intensity , HV , and zoom were optimized for individual larva . For experiments comparing multiple stimuli , each stimulus was separated by an equivalent period of E2 media washout . All solutions were made in E2 media containing 2% DMSO . For calcium imaging experiments involving trpa1b-/- animals and their WT/trpa1b+/- siblings , elavl3:GCaMP5:trpa1b-/- larvae were employed . One day prior to imaging , larvae were anesthetized with tricaine , tail-clipped , genotyped via HRMA , and housed in individual wells within a 24-well plate until ready for use in experiments . 2-APB was used a positive control . DRGs were isolated from 6- to 12-week-old C57Bl/6J mice . All experiments were performed in compliance with institutional animal care and use committee standards and experiments were performed essentially as described ( Kimball et al . , 2015 ) . Dissociation and culturing of mouse DRG neurons were performed as described with the following modifications ( Story et al . , 2003 ) . Dissected DRGs were dissociated by incubation for 1 hr at 37°C in a solution of culture medium [Ham’s F12/Dulbecco’s modified Eagle’s medium ( DMEM ) with 10% horse serum , 1% penicillin-streptomycin ( Life Technologies , Carlsbad , CA ) ] containing 0 . 125% collagenase ( Worthington Biochemicals , Lakewood , NJ ) , followed by a 30 min incubation in 10 ml of culture media plus 1 . 25 units of papain . Calcium imaging was performed essentially as described previously ( Story et al . , 2003 ) . Growth media was supplemented with 100 ng/ml nerve growth factor . For experiments involving heterologous expression , human embryonic kidney ( HEK ) 293T cells were transiently transfected with one or two of the following plasmid constructs: zebrafish trpa1b , zebrafish tlr7 , mouse Trpa1 , mouse Tlr7 , human TRPA1 , and/or human TLR7 . All constructs except for the one encoding zebrafish tlr7 were also co-transfected with pIRES-eGFP plasmid in order to estimate transfection efficiency . ( For zebrafish tlr7 , this step was unnecessary because the construct was already in the pIRES-eGFP vector . ) The buffer solution for all experiments was 10 mM HEPES in 1X Hanks’ balanced salt solution ( HBSS ) ( Invitrogen , Carlsbad , CA ) . The threshold for activation was defined as 30% above baseline for both DRG and heterologous expression experiments . Student’s t-test was used for all statistical calculations . All averaged traces represent mean ± s . e . m . All reported fluorescence values of each cell were normalized to the fluorescence of that cell during the initial baseline wash period . Maximum response values of each cell were calculated as the difference between the maximum and minimum fluorescence values of the cell during a stimulus application period . To verify the functionality of our transfected Tlr7 constructs , we determined levels of NF-kB induction following stimulation with the TLR7 agonist loxoribine using a Dual-Luciferase Reporter Assay System ( Promega , Madison , WI ) . Briefly , HEK 293T cells were seeded at ~80% confluency in individual wells of a 24-well plate ( N ≈ 4 × 105 cells per well ) and transiently transfected with the same zebrafish , mouse , or human Tlr7 constructs as employed in calcium imaging following a standard lipofectamine protocol . All cells were also co-transfected with a nF-kB Firefly luciferase reporter plasmid ( p1242 3x-KB-L , Addgene [Mitchell and Sugden , 1995] ) , a Renilla luciferase control plasmid ( p207-CMV-Renilla , gift from Tom Reh ) , and pIRES-eGFP ( if necessary ) for estimating transfection efficiency . As a negative control , another set of HEK 293T cells were transfected only with pIRES-eGFP . Twenty-four hours following transfection , the culture media was removed from all cells and replaced with normal serum-free media or with serum-free media containing 200 μM loxoribine . Following a 24 hr treatment period , culture media was removed , plated cells were rinsed briefly with DPBS , then lysed with passive lysis buffer ( Promega ) and gentle agitation on a multi-purpose rotator ( Barnstead , Hampton , NH ) . Lysates were analyzed on a Viktor3 1420 Multilabel Counter ( PerkinElmer , Waltham , MA ) , which generated luminescence values in CPS ( counts per second ) for both Firefly and Renilla luciferase activity . Assays were also performed on 1X PLB samples to estimate background luminescence . Background luminescence was subtracted from each measurement , and the Firefly/Renilla CPS ratio was calculated for each condition . To verify that TLR7 was being expressed by HEK 293T cells used in our experiments , we transfected cells on coverslips with either mouse or human Tlr7 constructs and pIRES-eGFP; some cells were only transfected with pIRES-eGFP to serve as a negative control . 48 hr following transfection , cell culture media was removed , and coverslips were washed briefly with 1X DPBS and fixed for 10 min at room temperature in 4% paraformaldehyde in 1X PBS ( Electron Microscopy Sciences , Hatfield , PA ) . Coverslips were again rinsed briefly in DBPS and then blocked in 10% goat serum in 1X PBST ( PBS with 0 . 1% Tween-20 ) for 1 hr at room temperature . Primary antibodies against TLR7 ( rabbit anti-TLR7 , Boster , Pleasanton , CA , 1:250 ) and GFP ( chick anti-GFP , 1:1000 , Invitrogen ) were made in PBST with 10% goat serum and applied to the coverslips , which were incubated overnight at 4°C . Coverslips were then rinsed 3X in PBST to remove primary antibodies , treated with secondary antibodies ( AlexaFluor goat anti-chicken 488 and AlexaFluor goat anti-rabbit 568 , both at 1:1000 , Life Technologies and Invitrogen , respectively ) for approximately 2 hr at room temperature , washed in PBST , and mounted on slides with DAPI-containing Vectashield medium ( Vector Laboratories , Inc . , Burlingame , CA ) . Confocal imaging of mounted cells was performed using a Zeiss microscope and Zen Black acquisition software . Whole-cell patch-clamp recordings were performed at room temperature ( 22–24°C ) using an Axon 700B amplifier ( Molecular Devices , Sunnyvale , CA ) on the stage of an inverted phase-contrast microscope equipped with a filter set for GFP visualization ( Nikon Instruments Inc . , Melville , NY , USA ) ( Feng et al . , 2017 ) . Pipettes pulled from borosilicate glass ( BF 150-86-10; Sutter Instrument , Novato , CA ) with a Sutter P-1000 pipette puller had resistances of 2–4 for whole-cell patch-clamp recordings when filled with pipette solution containing 140 mM CsCl , 2 mM EGTA , and 10 mM HEPES with pH 7 . 3 and 315 mOsm/l osmolarity . Cells were perfused with extracellular solution containing 140 mM NaCl , 5 mM KCl , 0 . 5 mM EGTA , 1 mM MgCl2 , 2 mM CaCl2 , 10 mM glucose , and 10 mM HEPES ( pH was adjusted to 7 . 4 with NaOH , and the osmolarity was adjusted to ≈ 340 mOsm/l with sucrose ) . The whole-cell membrane currents were recorded using voltage ramps from −100 to +100 mV for 500 ms at holding potential of 0 mV . Data were acquired using Clampex 10 . 4 software ( Molecular Devices , Novato , CA ) . Currents were filtered at 2 kHz and digitized at 10 kHz . Data were analyzed and plotted using Clampfit 10 ( Molecular Devices , Novato , CA ) . The concentration-response curve was fitted with the logistic equation: Y = Ymin + ( Ymax − Ymin ) / ( 1 + 10^[ ( logEC50 − X ) ×Hill slope] ) , where Y is the response at a given concentration , Ymax and Ymin are the maximum and minimum responses , X is the logarithmic value of the concentration and Hill slope is the slope factor of the curve . EC50 is the concentration that gives a response halfway between Ymax and Ymin . All data are presented as mean ± s . e . m . Mice were shaved on the nape of the neck or on the face two days before the assay . On the day of experiment , mice were acclimated for 1 hr by placing each of them individually in the recording chamber followed by intradermal injection of 10 μg of IMQ to the nape of the neck . Immediately after the injection , mice were videotaped for 30 min without any person in the recording room . After the recording , the videotapes were played back and the number of scratching bouts towards the injection site was counted by an investigator blinded to the treatment . Cheek injection of AITC was performed as described ( Shimada and LaMotte , 2008 ) . Briefly , during anesthesia with isoflurane ( 2% in 100% oxygen ) , the right cheek ( approx . 5 × 8 mm area ) was shaved . Mice were acclimated in the recording chambers at least two days before experiments began . AITC was dissolved in DMSO to make a 5 M stock solution and diluted in saline to make the working solution . Every mouse received an injection of 10 ul volume at the shaved area . Immediately after the injection , mice were videotaped for 30 min without any person in the recording room . After the recording , the videotapes were played back and the number of scratching and wiping bouts towards the injection site was counted by an investigator blinded to the treatment . Footpad injections were performed as described previously ( Liu et al . , 2016 ) . Briefly , either a saline control or IMQ ( 125 μg ) was injected into the footpad of the hindpaw . Animals were filmed , and nocifensive behaviors ( licking , biting ) in the recorded videos were scored by an investigator blind to the treatment .	Itch is a common and uncomfortable sensation that creates a strong desire to scratch . This mechanism may have evolved so animals can remove harmful parasites or substances from themselves . Feelings like touch , pain , and itch arise when stimuli such as mechanical pressure , temperature , or chemicals activate groups of specialized neurons in the skin . This response takes place when certain proteins – or receptors – at the surface of the neurons are stimulated . For instance , TRP ion channels such as TRPA1 play an important role in both the itch and pain responses . In mammals , directly activating these channels elicits pain . Itch is felt when itch responsive receptors are activated on a distinct set of neurons , which in turn activate TRP receptors . Although these processes have been well-studied in mammals , little is known about the existence of itch sensation in other animals . To explore this , Esancy , Condon , Feng et al . exposed zebrafish to chemicals that induce itch in mammals , and found that imiquimod , a medicine used to treat certain skin conditions , can elicit itch in fish . When this chemical was injected into the lips of a fish , the animal rubbed them against the walls of its tank , akin to scratching an itch . Further experiments showed that imiquimod directly activated the pain-sensing ion channel TRPA1 . In fact , this receptor was essential to the ‘scratching’ behavior: fish genetically engineered to lack TRPA1 did not react to the drug . Fluorescent proteins were then used to track when the neurons that carry TRPA1 were activated . This revealed that , in the skin of zebrafish , there are at least two functionally distinct populations neurons that express TRPA1 . One population , whose activation is associated with the animal ‘scratching’ , responds even when TRPA1 receives a low level of stimulation . The other population is less sensitive: it responds only to high-intensity stimuli and is associated with a pain response such as freezing and slower movements . Further experiments in the mouse suggest that this mechanism is present in mammals as well . This coding strategy explains how pain and itch can be experienced when the same receptors are being activated . Studying how animals like fish experience itch gives an insight into how detecting these sensations could have evolved . In turn , understanding this mechanism at the molecular and cellular levels may help find new ways to design better treatments for itch and pain disorders .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2018	A zebrafish and mouse model for selective pruritus via direct activation of TRPA1
As neural circuits form , growing processes select the correct synaptic partners through interactions between cell surface proteins . The presence of such proteins on two neuronal processes may lead to either adhesion or repulsion; however , the consequences of mismatched expression have rarely been explored . Here , we show that the Drosophila CUB-LDL protein Lost and found ( Loaf ) is required in the UV-sensitive R7 photoreceptor for normal axon targeting only when Loaf is also present in its synaptic partners . Although targeting occurs normally in loaf mutant animals , removing loaf from photoreceptors or expressing it in their postsynaptic neurons Tm5a/b or Dm9 in a loaf mutant causes mistargeting of R7 axons . Loaf localizes primarily to intracellular vesicles including endosomes . We propose that Loaf regulates the trafficking or function of one or more cell surface proteins , and an excess of these proteins on the synaptic partners of R7 prevents the formation of stable connections . During nervous system development , growing axons must navigate through a complex environment and select the correct synaptic partners from numerous potential choices . Recognition of cell surface molecules plays an important role in axon guidance and targeting and the establishment of specific synaptic connections ( Yogev and Shen , 2014 ) . Interactions between cell surface molecules can lead to either adhesion or repulsion , and their relative levels on different cells are important for appropriate connections to form . For instance , gradients of ephrins and their Eph receptors enable retinal axons to form a topographic map in visual areas of the brain because Eph levels determine the sensitivity to ephrins ( Triplett and Feldheim , 2012 ) . In the Drosophila olfactory system , olfactory receptor neurons preferentially connect to projection neurons that express matching levels of the adhesion molecule Teneurin ( Hong et al . , 2012 ) . As defects in synaptic adhesion molecules can lead to autism and other neurodevelopmental disorders ( Van Battum et al . , 2015; Gilbert and Man , 2017 ) , identifying mechanisms that regulate synaptic partner choice is likely to enhance our understanding of such human diseases . The Drosophila visual system has been a fruitful model for investigations of circuit assembly and synaptic specificity ( Plazaola-Sasieta et al . , 2017 ) . The two color photoreceptors in the fly retina , R7 and R8 , project to distinct layers in the medulla , M6 and M3 respectively . The R7 growth cone first actively targets a temporary layer , and then passively reaches its final layer due to the growth of other neuronal processes ( Ting et al . , 2005; Özel et al . , 2015 ) . Early stabilization of the R7 and R8 growth cones in different layers depends on differences in their relative levels of the transcription factor Sequoia ( Seq ) ; the adhesion molecule N-cadherin ( Ncad ) is thought to be the relevant target of Seq in these cells ( Petrovic and Hummel , 2008; Kulkarni et al . , 2016 ) . Both Ncad and the receptor protein tyrosine phosphatase ( RPTP ) Lar are required to stabilize R7 terminals in the M6 layer . In the absence of either protein they remain in the M3 layer , although defects are observed earlier in development in Ncad mutants than in Lar mutants ( Clandinin et al . , 2001; Lee et al . , 2001; Maurel-Zaffran et al . , 2001; Ting et al . , 2005; Özel et al . , 2015; Özel et al . , 2019 ) . Another RPTP , Ptp69D , is partially redundant with Lar , and the depth of R7 axon termination correlates with the total level of RPTP activity ( Newsome et al . , 2000; Hofmeyer and Treisman , 2009; Hakeda-Suzuki et al . , 2017 ) . Stabilization of R7 contacts also requires the presynaptic proteins Liprin-α and Syd-1 that act downstream of Lar ( Choe et al . , 2006; Hofmeyer et al . , 2006; Holbrook et al . , 2012; Özel et al . , 2019 ) . The primary synaptic targets of R7 that are responsible for its function in driving the spectral preference for ultraviolet light are the Dm8 medulla interneurons ( Gao et al . , 2008; Takemura et al . , 2013; Karuppudurai et al . , 2014; Ting et al . , 2014 ) . These cells fall into two subclasses , yellow ( y ) and pale ( p ) , and their survival depends on their correct pairing with the appropriate R7 cell subtype , expressing either Rh4 ( yR7 ) or Rh3 ( pR7 ) ( Courgeon and Desplan , 2019; Menon et al . , 2019 ) . The synapses R7 cells form on Dm8 cells often include the projection neurons Tm5a ( for yR7s ) or Tm5b ( for pR7s ) as a second postsynaptic element ( Gao et al . , 2008; Takemura et al . , 2013; Menon et al . , 2019 ) . Another interneuron , Dm9 , is both pre- and postsynaptic to R7 and R8 and mediates inhibitory interactions between ommatidia ( Takemura et al . , 2013; Takemura et al . , 2015; Heath et al . , 2020 ) . It is not known which , if any , of these cell types provide Ncad or RPTP ligands that stabilize filopodia from the R7 growth cone ( Yonekura et al . , 2007; Hofmeyer and Treisman , 2009; Hakeda-Suzuki et al . , 2017; Özel et al . , 2019 ) . Glia are also involved in establishing the pattern of R7 synaptogenesis , as they prevent excessive synapse formation through the adhesion protein Klingon ( Klg ) and its partner cDIP ( Shimozono et al . , 2019 ) . Here we identify a novel CUB-LDL domain transmembrane protein , Lost and found ( Loaf ) , that acts in photoreceptors to promote the formation of stable R7 contacts in the M6 layer . R7 mistargeting to the M3 layer is observed when loaf function is lost from photoreceptors , but not in a fully loaf mutant animal . Similar defects can be induced in loaf mutants by expressing Loaf in neurons that include Tm5a , Tm5b , Dm9 , and Dm8 , suggesting that R7 targeting is disrupted when Loaf is absent from R7 but present in its postsynaptic partners . Loaf does not itself promote cell adhesion and localizes primarily to endosomes . We propose that Loaf controls the trafficking or function of cell surface molecules that are used to match R7 to the correct postsynaptic neurons . A microarray-based screen for genes with enriched expression in the R7 and R8 photoreceptors relative to R1-R6 identified CG6024 , which encodes an uncharacterized transmembrane protein ( Pappu et al . , 2011 ) . CG6024 is also a predicted target of Glass ( Naval-Sanchez et al . , 2013 ) , a transcription factor required for photoreceptor differentiation and axon guidance ( Moses et al . , 1989; Selleck and Steller , 1991 ) . To test whether CG6024 has a function in axon targeting by R7 or R8 , we expressed RNAi transgenes targeting CG6024 with two different drivers: GMR-GAL4 drives expression in all differentiating cell types in the eye ( Freeman , 1996 ) , and removing a stop cassette from Actin>CD2>GAL4 with the eye-specific recombinase ey3 . 5-FLP ( Bazigou et al . , 2007 ) leads to RNAi expression in the entire eye disc . In both cases , R8 targeting was unaffected , but we observed a loss of R7 terminals from the M6 layer of the medulla ( Figure 1A–C ) ; 30–60% of R7 axons were mistargeted to the M3 layer ( Figure 1D–F ) . This phenotype appears to arise during the second stage of R7 targeting , when filopodia are stabilized to form synapses ( Ting et al . , 2005; Özel et al . , 2019 ) . R7 axons targeted correctly to their temporary layer at 40 hr after puparium formation ( APF ) when CG6024 was knocked down , but many terminals did not reach or were not stabilized in their permanent target layer , M6 , at 60 hr APF ( Figure 1G–K ) . We named the gene lost and found ( loaf ) based on the failure of R7 axons lacking loaf to find the right target layer and on the rescue of this phenotype discussed below . The Loaf protein contains extracellular CUB and LDLa domains and a predicted transmembrane domain ( Figure 2A ) , making it a candidate to directly mediate target recognition by R7 . In the experiments above , we used two independently generated RNAi lines targeting the same region of the gene to knock down loaf ( Figure 2A ) , both of which produced similar R7 mistargeting phenotypes ( Figure 1D ) . To confirm that this phenotype was due to loss of loaf rather than an off-target effect of the RNAi , we used the CRISPR-Cas9 system to generate deletion alleles that removed the LDLa , transmembrane and cytoplasmic domains of the protein ( Figure 2A , I ) . The sgRNAs were directed against a region of the gene distinct from the RNAi target sequence , and using them to delete the loaf gene in the eye by somatic CRISPR reproduced the R7 targeting defect ( Figure 2A , B , H ) . Surprisingly , germline removal of loaf resulted in homozygous mutant flies that were viable and showed largely normal R7 targeting ( Figure 2C , D , H ) , indicating that global loss of loaf does not affect this process . Expressing loaf RNAi had no effect in this loaf mutant background ( Figure 2E , H ) , confirming that the RNAi phenotype was due to its effect on loaf rather than another gene . Together , these results indicate that the phenotype caused by removing loaf from the eye is dependent on the presence of loaf in the optic lobes . R7 targeting may therefore depend on the amount of Loaf in R7 relative to other cells rather than its absolute presence or absence . To test this hypothesis , we generated clones of cells in the eye that were homozygous for loaf deletion alleles in an otherwise heterozygous background . As predicted , these showed mistargeting of R7 axons to the M3 layer ( Figure 2F , H , Figure 2—figure supplement 1A ) . The mistargeting was significantly rescued by expressing either HA-tagged or untagged Loaf within the mutant clones ( Figure 2G , H , Figure 2—figure supplement 1B–D ) , confirming that it is due to loss of loaf from photoreceptors . These results could be explained if correct targeting depends on the relative levels of Loaf in R7 and another cell type . Loss of Loaf in R7 when it is present in the other cell type would cause mistargeting . When Loaf is absent from all cells , redundant mechanisms may be sufficient to maintain R7 terminals in the correct layer . The medulla interneuron Dm8 , which mediates the preference for ultraviolet over visible light , was reported to be the major postsynaptic target of R7 ( Gao et al . , 2008; Takemura et al . , 2013; Ting et al . , 2014 ) . We therefore considered the hypothesis that R7 and its postsynaptic partner Dm8 must both express Loaf to form a stable connection ( Figure 3F ) . We first determined the effect of removing loaf function from Dm8 . Expressing loaf RNAi or Cas9 and loaf sgRNAs in neurons that include Dm8 cells with DIP-γ-GAL4 or traffic jam ( tj ) -GAL4 ( Carrillo et al . , 2015; Courgeon and Desplan , 2019 ) did not cause any R7 targeting phenotype ( Figure 3—figure supplement 1A , B ) . As it was difficult to assess the reduction in Loaf levels caused by these manipulations , we generated loaf mutant clones in the brain and labeled the mutant Dm8 cells with ortc2b-GAL4 ( Ting et al . , 2014 ) . R7 axons that contacted the dendrites of mutant Dm8 cells correctly reached the M6 layer , and there was no obvious defect in the position or morphology of the mutant Dm8 dendrites ( Figure 3A , B ) . We predicted that expressing Loaf in Dm8 cells in a loaf mutant background would result in a mismatch between R7 and Dm8 that would be similar to removing loaf from R7 in a wild-type background ( Figure 3F ) . We tested this by expressing UAS-LoafHA in loaf mutant flies with the Dm8 drivers DIP-γ-GAL4 , tj-GAL4 and drifter ( drf ) -GAL4 ( Hasegawa et al . , 2011; Carrillo et al . , 2015; Courgeon and Desplan , 2019 ) , as well as a combination of tj-GAL4 and DIP-γ-GAL4 . However , we did not observe significant levels of R7 mistargeting ( Figure 3C–E , Figure 3—figure supplement 1C , D ) , arguing against a requirement for matching Loaf levels in R7 and Dm8 . Since the presence or absence of Loaf in Dm8 did not appear to affect R7 targeting , we searched for other Loaf-expressing cells that might interact with R7 . We used several methods to examine the location of Loaf expression in the brain . RNA-Seq analysis of sorted cell types in the adult brain revealed widespread expression of loaf , although at varying levels ( Konstantinides et al . , 2018; Davis et al . , 2020 ) . However , Loaf translation in photoreceptors reaches its maximum at mid-pupal stages , when R7 axons are targeting the M6 layer ( Ting et al . , 2005; Zhang et al . , 2016 ) , so adult expression levels in other cells might not be reflective of this developmental stage . At pupal stages , we observed that a GFP protein trap insertion in loaf was expressed in many cells in the medulla ( Figure 2A , Figure 4—figure supplement 1A ) . Finally , we generated an antibody that recognizes the cytoplasmic domain of Loaf ( Figure 2I ) . As this antibody cross-reacted with another protein present in the cell bodies of medulla neurons ( Figure 4C ) , we could only evaluate Loaf expression within the neuropil . In pupal brains , Loaf was enriched in specific layers of the medulla neuropil and also in R7 axons and terminals ( Figure 4A; Figure 4—figure supplement 1C ) . This staining was absent in loaf mutants ( Figure 4C ) , and the enrichment in R7 processes was specifically lost when loaf RNAi was expressed with GMR-GAL4 ( Figure 4B; Figure 4—figure supplement 1D ) . The Loaf protein trap was primarily present in cell bodies and was not visibly enriched in R7 axons; we believe that this insertion disrupts the normal localization of the protein , as clones homozygous for the insertion showed R7 mistargeting ( Figure 4—figure supplement 1A , B ) . When misexpressed in photoreceptors , LoafHA was efficiently transported to R7 axons and terminals ( Figure 4—figure supplement 1F ) . Consistent with our findings , recent single-cell RNA-Seq data from dissociated optic lobes show that significant loaf expression is present in almost every cluster throughout the pupal stage , although its levels are generally lower in clusters identified as glia . loaf expression in photoreceptors is highest at P40 and P50 , but declines at later stages ( Kurmangaliyev et al . , 2020; Özel et al . , 2021 ) . Because these data did not identify a specific cell type that would be most likely to interact with R7 using Loaf , we tested whether R7 mistargeting could be induced by expressing Loaf in broad categories of cells in a loaf mutant background . We observed no phenotype when Loaf was expressed in glia with repo-Gal4 , or in neuronal populations that expressed homothorax ( hth ) -GAL4 , brain-specific homeobox ( bsh ) -GAL4 , or Visual system homeobox ( Vsx ) -GAL4 ( Hasegawa et al . , 2011; Li et al . , 2013; Erclik et al . , 2017; Figure 4D ) . Expressing Loaf in photoreceptors with ey3 . 5-FLP , Act>CD2>GAL4 in a loaf mutant background likewise had no effect on R7 ( Figure 4D ) , indicating that the presence of Loaf in R7 when it is absent in other cells did not impede its targeting . However , we did observe a significant level of R7 mistargeting when Loaf was expressed in neurons that expressed apterous ( ap ) -GAL4 , which is active from the third larval instar ( Morante et al . , 2011; Li et al . , 2013 ) or in cholinergic neurons with Choline acetyltransferase ( ChAT ) -GAL4 , which is active from mid-pupal stages ( Meissner et al . , 2019 ) , in a loaf mutant background ( Figure 4D–F ) . ap is expressed in the majority of cholinergic neurons in the medulla ( Konstantinides et al . , 2018 ) , supporting the idea that cells in this population use Loaf to interact with R7 . R7 targeting defects also occurred when Loaf was expressed in glutamatergic neurons in a loaf mutant background with Vesicular glutamate transporter ( VGlut ) -GAL4 , which is active from early pupal stages ( Meissner et al . , 2019; Figure 4D; Figure 4—figure supplement 1E ) , indicating that more than one type of neuron interacts with R7 through Loaf . The populations of cholinergic and glutamatergic neurons include the major synaptic targets of R7 , suggesting the possibility that Loaf acts in these cells to influence their interactions with R7 . The synapses that R7 forms with Dm8 also include the cholinergic output neurons Tm5a and Tm5b ( Gao et al . , 2008; Karuppudurai et al . , 2014; Menon et al . , 2019; Davis et al . , 2020 ) . To test the importance of Tm5a/b neurons we used GMR9D03-GAL4 , which is expressed in a subset of these cells from early in development ( Han et al . , 2011; Figure 5—figure supplement 1A , B ) to express Loaf in a loaf mutant background . This again produced significant R7 mistargeting ( Figure 5A , H ) , consistent with the hypothesis that Loaf levels in Tm5a/b influence R7 . Although GMR9D03-GAL4 is also expressed in lamina neurons L2 and L3 ( Akin et al . , 2019 ) , restoring Loaf only in lamina neurons with GH146-GAL4 ( Schwabe et al . , 2014 ) did not affect R7 ( Figure 5H ) . Importantly , loaf mutant Tm5a/b cells did not have obvious morphological defects or cause R7 mistargeting ( Figure 5D , E ) . Among glutamatergic neurons , both Dm8 and Dm9 are synaptic partners of R7 . Dm9 is a multicolumnar neuron that tracks R7 axons closely and mediates inhibition between neighboring ommatidia ( Nern et al . , 2015; Heath et al . , 2020 ) . The transcription factors Vestigial ( Vg ) and Defective proventriculus ( Dve ) are strongly enriched in Dm9 cells ( Davis et al . , 2020 ) , and dve-GAL4 drives expression in Dm9 ( Figure 5—figure supplement 1C ) , although vg-GAL4 expression was not detectable in the adult brain . When used to express Loaf in a loaf mutant background , neither driver alone significantly affected R7 targeting , but the combination had a significant effect ( Figure 5B , H ) , making Dm9 a candidate to provide Loaf that affects R7 targeting . Again , loaf mutant Dm9 cells and their presynaptic R7 axons appeared normal ( Figure 5F , G ) . Finally , we tested whether a contribution of Dm8 might be detectable in combination with other R7 synaptic target cells by restoring Loaf to loaf mutants with both ap-GAL4 and tj-GAL4 . This produced significantly more R7 mistargeting than ap-GAL4 alone ( Figure 5C , H ) , suggesting that Dm8 or another tj-GAL4 expressing neuron such as Dm11 ( Courgeon and Desplan , 2019 ) , which also projects to the M6 layer ( Nern et al . , 2015; Figure 5—figure supplement 1D ) , may contribute to the pool of Loaf that influences R7 targeting . However , Tm5a/b and Dm9 appear to play a more significant role ( Figure 5I ) . Overexpressing Loaf with the GMR9D03-GAL4 , dve-GAL4 and vg-GAL4 , or ap-GAL4 and tj-GAL4 drivers in a wild-type background did not cause R7 mistargeting ( Figure 5—figure supplement 2 ) ; because Loaf is normally enriched in R7 terminals , it is possible that the Loaf levels produced in the processes of synaptic partner cells in these overexpression experiments did not exceed those present in R7 . To determine whether Loaf could function as a homophilic cell adhesion molecule , we transfected HA-tagged Loaf into S2 cells and conducted cell aggregation assays ( Ting et al . , 2005; Astigarraga et al . , 2018 ) . We did not observe significant aggregation of the transfected cells , although the positive control Sidekick ( Sdk ) ( Astigarraga et al . , 2018 ) induced aggregation under the same conditions ( Figure 6A , B , Figure 6—figure supplement 1A ) . Unlike Sdk , neither tagged nor untagged Loaf showed strong localization to the plasma membrane; most Loaf was present in punctate structures inside the cells ( Figure 6C–E ) . These structures showed partial colocalization with Hepatocyte growth-factor-regulated tyrosine kinase substrate ( Hrs ) and Rab7 ( Figure 6D , E ) , two markers of late endosomes , but did not colocalize with the recycling endosome marker Rab11-GFP ( Figure 6—figure supplement 1C ) . When expressed in the retina in vivo , LoafHA also partially colocalized with Rab7 and Hrs , but not with the lysosomal markers ADP-ribosylation factor-like 8 ( Arl8 ) or Vacuolar H+-ATPase 55kD subunit ( Vha55 ) ( Figure 6F , G , J ) , and untagged Loaf again showed a similar localization ( Figure 6—figure supplement 1D ) . In clones of cells mutant for the ESCRT complex component Tumor susceptibility gene 101 ( TSG101 ) , endocytosed proteins such as Notch accumulate in late endosomes ( Moberg et al . , 2005 ) , and we found that LoafHA colocalized with Notch ( Figure 6—figure supplement 1E ) , confirming its presence in the endocytic pathway . GFP-tagged endogenous Loaf appeared to localize to the cytoplasm of all photoreceptors , but unlike overexpressed Loaf , it was primarily found close to the plasma membrane rather than in late endosomes ( Figure 6—figure supplement 1B ) ; as noted above , this tag disrupts the function of the Loaf protein . As a more stringent test of whether Loaf ever reaches the plasma membrane , we transfected S2 cells with a form of Loaf tagged at its extracellular N-terminus with the V5 epitope , and incubated live cells with antibodies to V5 . No staining was observed in these conditions ( Figure 6H ) . As controls for this experiment , V5 staining was detected in cells that were fixed and permeabilized , and antibodies to HA detected cotransfected HASdk on the surface of live cells as well as in vesicles internalized during the incubation ( Figure 6H , I ) . These results suggest that Loaf is not itself a cell surface adhesion molecule , but could regulate the trafficking or cell surface localization of proteins involved in cell adhesion or synapse formation . We next searched for candidate proteins that might be regulated by Loaf . One possibility we considered was Lar , an RPTP that is required for normal R7 targeting ( Clandinin et al . , 2001; Maurel-Zaffran et al . , 2001; Hofmeyer and Treisman , 2009 ) . Lar acts in R7 and not the target region ( Clandinin et al . , 2001; Maurel-Zaffran et al . , 2001 ) , so its ligand , which remains unknown , would also have to be regulated by Loaf to account for the effect of Loaf in synaptic partners of R7 . To test for a genetic interaction between loaf and Lar , we knocked down these genes using the photoreceptor driver long GMR-GAL4 ( lGMR-GAL4 ) ( Wernet et al . , 2003 ) . Expression of either loaf RNAi or Lar RNAi with this driver affected only a subset of R7 axons , but simultaneous expression of both RNAi lines had a synergistic effect , causing almost all R7 axons to terminate in the M3 layer ( Figure 7A–D ) . This suggests that Loaf and Lar are involved in the same process . Similarly , loaf knockdown enhanced the mistargeting phenotype of mutations in the downstream gene Liprin-α , although this effect could simply be additive ( Figure 7—figure supplement 1G–J ) . Overexpression of Lar in photoreceptors , either alone or together with Loaf , did not cause any significant defects ( Figure 7—figure supplement 1A , B , D ) . However , overexpression of Lar in loaf mutant photoreceptors could rescue R7 targeting ( Figure 7—figure supplement 1C , D ) , indicating that Lar can compensate for the lack of loaf and is thus unlikely to be its primary effector . Consistent with this conclusion , we found that loaf was not required for HA-tagged Lar to be transported into photoreceptor axons ( Figure 7—figure supplement 1E , F ) . We also investigated LDL receptor related protein 4 ( Lrp4 ) , based on its role as a presynaptic organizer in the olfactory system ( Mosca et al . , 2017 ) , its postsynaptic signaling function at the vertebrate neuromuscular junction ( Yumoto et al . , 2012 ) , and the requirement for chaperones to promote the trafficking of other LDL family members ( Culi et al . , 2004; Wagner et al . , 2013 ) . We found evidence that the level of Lrp4 can affect R7 targeting and that its effect on R7 is regulated by Loaf . Overexpressing Lrp4 in photoreceptors caused R7 axons to contact each other or hyperfasciculate either in the M3 or M6 layers of the medulla ( Figure 7E , H ) . These defects were more severe , and included overshooting of the M6 layer by some R7 axons , when Lrp4 was coexpressed with Loaf ( Figure 7F , H ) , but were almost absent when Lrp4 was expressed in loaf mutant photoreceptors ( Figure 7G , H ) . Lrp4 overexpression also resulted in abnormal numbers and arrangements of cone and pigment cells in the retina ( Figure 7—figure supplement 2A ) . Again , these defects were more severe when Lrp4 was coexpressed with Loaf , and were not observed when Lrp4 was expressed in loaf mutant cells ( Figure 7—figure supplement 2B , C ) . Although Lrp4HA had a more granular appearance in loaf mutant than in wild-type photoreceptor cell bodies ( Figure 7—figure supplement 2D , F ) , it was still transported into their axons ( Figure 7—figure supplement 2E , G ) , and its level of expression appeared unaffected ( Figure 7—figure supplement 2H ) . Despite the effect of Loaf on Lrp4 function , Lrp4 is unlikely to fully explain the effects of loaf on R7 targeting , as R7 axons projected normally in Lrp4 mutant clones ( Figure 7—figure supplement 2I ) . Moreover , expressing loaf RNAi in photoreceptors resulted in R7 mistargeting even in an Lrp4 null mutant background ( Figure 7—figure supplement 2J , K ) . These results show that Loaf can affect the function of cell surface proteins , and suggest that it could act by regulating Lrp4 and/or other cell surface molecules that act as a readout of its levels to control the interactions between R7 and its postsynaptic partners . The layered arrangement of neuronal processes in the medulla makes R7 axon targeting a sensitive model system in which to elucidate how growth cones select the correct postsynaptic partners . However , it has not been clear which cells are responsible for retaining R7 axons in the M6 layer . The RPTP Lar , which forms a hub for the assembly of presynaptic structures through the adaptor protein Liprin-α ( Choe et al . , 2006; Hofmeyer et al . , 2006; Takahashi and Craig , 2013; Bomkamp et al . , 2019 ) , acts in R7 to stabilize filopodia in the M6 layer by promoting synapse formation ( Clandinin et al . , 2001; Maurel-Zaffran et al . , 2001; Özel et al . , 2019 ) . Our findings that Lar and loaf show a strong genetic interaction and that Lar overexpression can rescue the loss of loaf suggest that like Lar , Loaf stabilizes synaptic contacts . Although Lar family RPTPs can recognize a variety of ligands ( Han et al . , 2016 ) , the ligand involved in R7 targeting and its cellular source remain unknown ( Hofmeyer and Treisman , 2009 ) . One candidate is Ncad , which is required at an early stage of development in both R7 and medulla neurons , but its widespread expression has made it difficult to determine in which neurons it acts to promote R7 synapse stabilization ( Lee et al . , 2001; Ting et al . , 2005; Yonekura et al . , 2007; Özel et al . , 2015 ) . R7 cells form numerous synapses with Dm8 interneurons , which are essential for ultraviolet spectral preference ( Gao et al . , 2008; Takemura et al . , 2013 ) and fall into two classes that are postsynaptic to either yR7 or pR7 cells ( Carrillo et al . , 2015 ) . Each R7 subtype promotes the survival of the class of Dm8 cells ( y or pDm8 ) with which it synapses ( Courgeon and Desplan , 2019; Menon et al . , 2019 ) . The Dm8 dendrites that remain in the absence of R7 cells still project to the M6 layer ( Courgeon and Desplan , 2019 ) , but it is not known whether R7 relies on Dm8 for targeting or survival information . Many synapses between R7 and Dm8 also include the projection neurons Tm5a ( yR7 ) or Tm5b ( pR7 ) as a second postsynaptic element ( Gao et al . , 2008; Takemura et al . , 2015; Menon et al . , 2019 ) . In addition , Dm9 interneurons are both pre- and postsynaptic to R7 and mediate center-surround inhibition , similarly to horizontal cells in the mammalian retina ( Takemura et al . , 2013; Takemura et al . , 2015; Heath et al . , 2020 ) . Our data indicate that the level of Loaf in Tm5a/b and Dm9 is more important for R7 targeting than its level in Dm8 , suggesting that these cells may determine the stability of R7 contacts in the M6 layer . However , we cannot rule out the possibility that the drivers we used to express Loaf in Dm8 did not cause a phenotype because the level or timing of expression was not optimal . Our observation that the absence of Loaf from R7 only causes a phenotype when Loaf is present in its postsynaptic partners implies that Loaf is not essential for R7 targeting . In loaf mutants , redundant mechanisms must stabilize R7 terminals in the M6 layer; cell surface protein interactions often only specify a preference for one synaptic partner over another ( Xu et al . , 2019 ) . Synaptic connections may not form entirely normally in these conditions , as loaf mutants show a reduced sensitivity to ultraviolet light when compared to isogenic controls ( C . -H . Lee , pers . comm . ) . Importantly , R7 mistargeting is much more striking when loaf is absent from photoreceptors , but present in the brain . We were able to reproduce this mistargeting by expressing loaf only in subsets of neurons in the brain that include the major postsynaptic partners of R7 . The most parsimonious explanation for these phenotypes is that a mismatch in Loaf expression between R7 and its partners results in mistargeting . A similar phenomenon was observed for the homophilic cell adhesion molecule Klingon , which affects synapse formation when removed from either R7 or glial cells , but not when removed from both simultaneously ( Shimozono et al . , 2019 ) . Matching pre- and postsynaptic levels are also important for the Drosophila Teneurin proteins to promote synapse formation ( Hong et al . , 2012; Mosca et al . , 2012 ) . This type of level matching , in which the presence of a protein in only one of the two partners is more deleterious than its absence from both , is well suited to refining synaptic specificity by eliminating inappropriate connections . Interestingly , Loaf matching seems to be asymmetric; R7 mistargeting results if Loaf is absent in R7 and present in the postsynaptic cell , but not if it is absent in the postsynaptic cell and present in R7 ( Figure 5I ) . It is possible that matching levels in some way neutralize the activity of Loaf , or of a cell surface molecule regulated by Loaf . An excess of this molecule on the postsynaptic cell might prevent it from initiating or stabilizing synapses with R7 , or drive it to preferentially connect with other neurons . However , the asymmetry could also reflect the presence of Loaf in multiple postsynaptic cells; loss of loaf from only one cell type may not be sufficient to disrupt R7 targeting . Our results suggest that Loaf does not itself act as a cell surface adhesion molecule . When epitope-tagged or untagged forms of Loaf are overexpressed in photoreceptors or cultured cells , they localize to intracellular vesicles that include endosomes and do not appear to reach the cell surface . In addition , they do not induce cell aggregation , further arguing against a homophilic adhesion function . CUB domains are present in a variety of functionally distinct proteins , and are thought to bind protein ligands , sometimes in combination with calcium ions ( Gaboriaud et al . , 2011 ) . Some CUB domain proteins are involved in endocytosis of other molecules ( Moestrup and Verroust , 2001; Xu and Wang , 2016 ) , while members of the Neuropilin and Tolloid-like ( Neto ) family of CUB-LDL proteins are required for the normal localization and activity of glutamate receptors and other postsynaptic proteins ( Zheng et al . , 2004; Kim et al . , 2012; Ramos et al . , 2015; Sheng et al . , 2015 ) . It is thus possible that Loaf controls the level of other proteins on the cell surface by mediating their trafficking or endocytosis . The endosomal protein Commissureless functions in this manner , by trafficking the Roundabout axon guidance receptor directly from the Golgi to endosomes so that it does not reach the cell surface ( Keleman et al . , 2002 ) . In another example , Rab6 and its activator Rich traffic Ncad to the cell surface , facilitating R7 targeting ( Tong et al . , 2011 ) . Differences in Neurexin levels between axons and dendrites are also dependent on endocytosis and sorting ( Ribeiro et al . , 2019 ) , and trafficking of synaptic adhesion molecules in general is highly regulated ( Ribeiro et al . , 2018 ) . Consistent with this model , we found that loss or gain of Loaf affects the function of coexpressed Lrp4 , a presynaptic organizer in the olfactory system that has postsynaptic functions at mammalian neuromuscular junctions ( Yumoto et al . , 2012; Mosca et al . , 2017 ) . However , Lrp4 alone cannot explain the effects of Loaf , as removing loaf from photoreceptors still affects R7 targeting in an Lrp4 null mutant . Loaf may act through a protein similar to Lrp4 , or through a combination of proteins . Alternatively , it is possible that under some conditions , perhaps in the presence of other interacting proteins , Loaf itself can reach the cell surface and function there . Some synaptic organizing molecules are transported to axons in lysosome-related vesicles and secreted in a regulated manner ( Arantes and Andrews , 2006; Vukoja et al . , 2018; Ibata et al . , 2019 ) . Further study of the mechanism of Loaf action will provide insight into the cellular mechanisms that enable synaptic connections to be stabilized only on the appropriate cells as neural circuits develop . Further studies by the authors have revealed that GMR9D03-GAL4 is also expressed in other transmedullary neurons such as Tm15 and Tm25 , raising the possibility that Loaf in these neurons could contribute to R7 targeting . Fly stocks used were Rh5-GFP ( Bloomington Drosophila Stock Center [BDSC] #8600 ) ; Rh6-GFP ( BDSC #7461 ) ; gl-lacZ ( Moses and Rubin , 1991 ) , R22E09-LexA , LexAop-myrTomato; GMR-GAL4 ( Pecot et al . , 2013 ) ; ey3 . 5-FLP , Act>CD2>GAL4 ( BDSC #35542 and #4780 ) ; lGMR-GAL4 ( BDSC #8605 ) ; UAS-loaf RNAiBL P{TRiP . JF03040}attP2 ( BDSC #28625 ) ; UAS-loaf RNAiKK P{KK112220}VIE-260B ( Vienna Drosophila Resource Center [VDRC] #102704 ) ; UAS-Lar RNAi P{KK100581}VIE-260B ( VDRC #107996 ) ; UAS-dcr2 ( BDSC #24650 ) ; panR7-lacZ ( Hofmeyer et al . , 2006 ) ; nos-Cas9 ( BDSC #54591 ) ; UAS-Cas9-P2 ( BDSC #58986 ) ; DIP-γ-GAL4 ( Carrillo et al . , 2015 ) ; tj-GAL4NP1624 ( Kyoto Stock Center #104055 ) ; drf-GAL4 ( Brody et al . , 2012 ) ; Mi{PT-GFSTF . 1}CG6024MI00316-GFSTF . 1 ( BDSC #64464 ) ; apmd544-GAL4 ( BDSC #3041 ) ; ChAT-GAL4 ( BDSC #6798 ) ; repo-GAL4 ( BDSC #7415 ) ; hth-GAL4 ( Wernet et al . , 2003 ) ; bsh-GAL4 ( Hasegawa et al . , 2011 ) ; Vsx-GAL4 ( BDSC #29031 ) ; VGlut-GAL4 ( BDSC #26160 ) ; GMR9D03-GAL4 ( BDSC #40726 ) ; GH146-GAL4 ( BDSC #30026 ) ; dveNP3428-GAL4 ( Kyoto Stock Center #113273 ) ; vg-GAL4 ( BDSC #6819 ) ; ortC1a-GAL4 ( BDSC #56519 ) ; ortC2b-GAL4 ( Ting et al . , 2014 ) ; GMR64H01-GAL4 ( BDSC #39322 ) ; UAS-LarHA ( Hofmeyer and Treisman , 2009 ) ; UAS-Lrp4HA; Lrp4dalek ( Mosca et al . , 2017 ) ; Liprin-αoos ( Hofmeyer et al . , 2006 ) ; GMR9D03-DBD ( BDSC #68766 ) ; GMR38H04-AD ( BDSC #75758 ) ; MCFO-1 ( Nern et al . , 2015 ) , and TSG1012 ( Moberg et al . , 2005 ) . loaf mutant clones and loaf mutant clones overexpressing other proteins were generated using ey3 . 5-FLP , UAS-CD8GFP; lGMR-GAL4; FRT80 , tub-GAL80 . loafMiMIC-GFSTF clones were generated using eyFLP; lGMR-GAL4 , UAS-myr-tdTomato; FRT80 , tub-GAL80 . Clones in which specific cell types were labeled were generated by crossing ortC2b-GAL4 ( or other GAL4 lines ) ; FRT80 ( or FRT80 , loafΔ33 ) to hs-FLP122 , UAS-CD8GFP; FRT80 , tub-GAL80 . Overexpression clones were generated by crossing UAS-LoafHA ( or UAS-Loaf , UAS-LarHA or UAS-Lrp4HA ) ; FRT82 to ey3 . 5-FLP , UAS-CD8GFP; lGMR-GAL4; FRT82 , tub-GAL80 . To obtain sparse labeling of Tm5a/b/c neurons , flies with the genotype hsflp2PEST; UAS>stop>CD4-tdGFP/CyO; GMR9D03-GAL4 were heat-shocked for 7 min at late L3 stage and dissected in the adult . loaf mutant clones in a background of Lrp4 overexpression were generated by crossing UAS-Lrp4HA; lGMR-GAL4 , FRT80 , loafΔ33 to eyFLP; FRT80 , ubi-RFP . Lrp4 mutant clones were generated using ey-FLP , tub-GAL80 , FRT19; lGMR-GAL4 , UAS-CD8-GFP . To restore Loaf to specific cell types in a loaf mutant background , tj-GAL4 ( or other GAL4 lines ) ; loafΔ33/SM6-TM6B was crossed to UAS-LoafHA , UAS-myrTomato; loafΔ33/SM6-TM6B or to UAS-LoafHA; panR7-lacZ , loafΔ33/SM6-TM6B . GAL4 lines on the third chromosome were recombined with loafΔ33 and recombinants carrying loaf were identified by PCR , except for GAL4 lines inserted at the attP2 site , which is very close to loaf . In these cases , new loaf alleles were directly introduced by CRISPR onto the GAL4 chromosome using nos-Cas9 and our transgenic loaf sgRNA flies , and identified by PCR . Adult heads were dissected in cold 0 . 1 M sodium phosphate buffer ( PB ) pH 7 . 4 , fixed in 4% formaldehyde in PB for 4 hr at 4°C and washed in PB . Heads were then submerged in a sucrose gradient ( 5% , 10% , 20% ) and left in 25% sucrose overnight at 4°C for cryoprotection . Heads were embedded in OCT tissue freezing medium and frozen in dry ice/ethanol , and 12 μm sections were cut on a cryostat . Sections were post-fixed in 0 . 5% formaldehyde in PB for 30 min at room temperature and washed three times in PB with 0 . 1% Triton ( PBT ) before incubation with primary antibodies overnight at 4°C . Sections were washed four times for 20 min with PBT and incubated with secondary antibodies for 2 hr at room temperature . Sections were washed again four times for 20 min before mounting in Fluoromount-G ( Southern Biotech ) . Pupal and adult whole brains were fixed in 4% paraformaldehyde in PBS for 30 min at room temperature , washed 3 times for 10 min in PBST ( PBS + 0 . 4% Triton-X 100 ) and blocked in PBST +10% donkey serum prior to incubation with primary antibodies overnight at 4°C . For Loaf staining of pupal brains , this incubation was extended to 4 days . Samples were washed in PBST three times for at least 1 hr each and incubated with secondary antibodies for 2 . 5 hr at room temperature . Samples were washed three times for 20 min in PBST and once in PBS before mounting in SlowFade Gold AntiFade reagent ( Life Technologies ) on bridge slides . Pupal retinas were fixed in 4% paraformaldehyde in PBS for 30 min on ice , washed for 15 min in PBT and incubated with primary antibodies overnight at 4°C . Retinas were washed three times for 5 min with PBT , incubated with secondary antibodies for 2 hr at 4°C and washed again three times for 5 min before mounting in 80% glycerol in PBS . Confocal images were collected with Leica SP8 and Zeiss LSM510 confocal microscopes . The primary antibodies used were mouse anti-Chp ( 1:50; Developmental Studies Hybridoma Bank [DSHB] 24B10 ) , chicken anti-GFP ( 1:400; Life Technologies ) , rat anti-HA ( 1:50; Roche 3F10 ) , rabbit anti-β galactosidase ( 1:100 , Fisher ) , rat anti-Ncad ( 1:50; DSHB ) , rabbit anti-dsRed ( 1:500; Takara Bio ) , guinea pig anti-Loaf ( 1:400 , Proteintech ) , mouse anti-Cnx99A ( 1:10 , DSHB 6-2-1 ) , mouse anti-Hrs ( 1:10 , DSHB 27–4 ) , mouse anti-Rab7 ( 1:10 , DSHB ) , rabbit anti-ATP6V1B1 ( Vha55; 1:200 , Abgent ) , rabbit anti-Arl8 ( 1:200; DSHB ) , rat anti-Elav ( 1:100 , DSHB ) , mouse anti-Notch ( 1:10; DSHB C17 . 9C6 ) , mouse anti-Arm ( 1:10; DSHB N2 7A1 ) , sheep anti-GFP ( 1:200 , Bio-Rad #4745–1051 ) , rabbit anti-RFP ( 1:500; MBL International #PM005 ) , rabbit anti-V5 ( 1:1000; Abcam ab9116 ) , mouse anti-FLAG ( 1:500 , Sigma F3165 ) , mouse anti-Dac ( 1:40; DSHB mAbdac2-3 ) , rabbit anti-Bsh ( 1:1800 ) ( Özel et al . , 2021 ) , and guinea pig anti-Runt ( 1:600; GenScript ) . Rhodamine-phalloidin ( Invitrogen R415 ) was used at 1:20 . The Loaf polyclonal antibody was made by Proteintech using the cytoplasmic domain ( aa 292–378 ) as an antigen . Guinea pig anti-serum was affinity purified . Guinea pig anti-Runt was made by GenScript using the full-length protein as an antigen . Secondary antibodies ( Jackson Immunoresearch and Life Technologies ) were coupled to the fluorochromes Cy3 , AlexaFluor 488 , or AlexaFluor 647 . To quantify the R7 targeting defect , fluorescent image stacks of 12 μm adult head sections labeled for gl-lacZ , Rh3/4-lacZ , or anti-24B10 were gathered in 0 . 5 μm steps . Maximum intensity projections were obtained and termini projecting beyond the R8 layer were counted as ‘R7 correctly targeted’ and those stopping in the R8 layer were counted as ‘R7 incorrectly targeted . ’ Termini in the R8 layer were counted as total cartridge number per section . The percentage of mistargeting R7s was calculated for each section , except that when the phenotype was scored in photoreceptor clones , the percentage was calculated from all mistargeting R7 axons within clones from all sections . To quantify defects in UAS-LRP4 clones , GFP-labeled R7 termini that contacted each other in the M3 layer were counted as ‘M3 clumping’ , while termini hyperfasciculating in the M6 layer were counted as ‘M6 clumping . ’ Three termini contacting each other were counted as two instances of M3 or M6 clumping depending on in which layer the clumps resided . GFP-labeled R7 termini that extended past the M6 layer were counted as ‘overshooting . ’ A terminus that extended past the M6 layer and turned to contact another clone was counted both as ‘M6 clumping’ and ‘overshooting . ’ The percentage of R7s belonging to each of these groups was calculated from all the R7 clones within each section . To quantify cell aggregates in S2 cell culture experiments , fluorescent image stacks from fixed cells that had been labeled for GFP and HA were gathered in 0 . 5 μm steps . Each image was examined for GFP positive ( control ) or GFP and HA-positive ( Sdk or Loaf ) cells that contacted each other as aggregates . The number of cells in each aggregate was counted for each image . To measure intracellular colocalization , single confocal slices were processed with a median filter with neighborhood of 1 in ImageJ and each channel was linear contrast enhanced to spread values evenly from 0 to 255 . A rectangular ROI was drawn around the central region of a single ommatidium and ImageJ was used to calculate Pearson’s correlation coefficient on each region with pixel intensity above a threshold of 16 out of a range of 255 , to eliminate background . To extract proteins , adult heads were dissected and frozen on dry ice , and then homogenized in Laemmli buffer ( 4% SDS , 20% glycerol , 120 mM Tris-Cl pH 6 . 8 , 0 . 02% bromophenol blue , 10% beta-mercaptoethanol ) . Samples were heated at 95°C for 5 min and loaded onto a SDS-PAGE gel . Gels were run first at 80 volts for 20 min , then 100 volts for the remainder of the time and transferred onto nitrocellulose membranes ( Bio-Rad ) for one hour at 100 volts . Membranes were washed for 5 min in TBST ( 20 mM Tris ( pH 7 . 6 ) , 136 mM NaCl , 0 . 2% Tween-20 ) , and blocked in 5% low-fat milk in TBST solution for one hour . Membranes were incubated overnight with primary antibody in TBST with 5% milk at 4°C , washed three times for 10 min in TBST and incubated in horseradish peroxidase-conjugated secondary antibodies ( 1:10 , 000; Jackson ImmunoResearch ) at room temperature in TBST with 5% milk for 2 hr . Membranes were washed three times for 10 min in TBST and once for 10 min in TBS . Enhanced chemiluminescence ( Thermo SuperSignal WestPico ) was used to develop the blots . Primary antibodies used were guinea pig anti-Loaf ( 1:1000 , Proteintech ) and mouse anti ß-tubulin ( 1:10 , 000; Sigma , T4026 ) . UAS-Loaf-FLAG-HA is clone UFO07678 ( Drosophila Genomics Resource Center ) . UAS-Loaf was cloned by inserting an Nhe I/Xba I fragment of clone UFO07678 into the Xba I site of pUAST-attB . Both constructs were integrated into the VK1 PhiC31 site at position 59D3 . The loaf sgRNA sequences GCTGGTGATTACGTCGGTGA ( loaf gRNA 1 ) and TGCGGGACCATCCGGGTACC ( loaf gRNA 2 ) identified on http://www . flyrnai . org/crispr2 were made with gene synthesis in pUC57 ( GenScript ) and cloned into pCFD4 ( Port et al . , 2014 ) by Gibson assembly . The construct was integrated into the attP40 site at 25C6 . These flies were crossed to nos-Cas9 flies to make germline mosaic flies . The progeny of these flies were crossed to balancer flies and screened by PCR using primers outside the expected deletion ( CGCACGAACTTTGTGACACT and CTCAAGTCAATCGGTCCTTCC ) . In loafΔ20 , the deletion extends from TACGTCGGTGA in gRNA1 through TGCGGG in sgRNA2 , creating a frameshift and a stop codon after 30 novel amino acids . loafΔ33 has the final CGGTGA of sgRNA replaced by GATT , and then deletes through TGCGGG in sgRNA2 , creating a stop codon immediately following Thr 208 at the end of the CUB domain . Injections and screening of transgenic flies were carried out by Genetivision . A V5-Loaf construct in which the V5 epitope tag ( GKPIPNPLLGLDST ) was inserted following H90 , four residues after the predicted signal peptide cleavage site , was synthesized by GenScript and cloned into pUASTattB using the EcoRI and XbaI sites . S2 cells were grown in Schneider’s Drosophila Medium ( GIBCO Invitrogen ) with 10% heat inactivated fetal bovine serum and 50 units/ml penicillin-50 g/ml streptomycin ( GIBCO Invitrogen ) at 25°C . Cells were spun down and resuspended in PBS . Poly-L-lysine-treated slides were covered with 0 . 1–0 . 2 ml of the cell suspension . Cells were fixed for 10 min at room temperature with 4% paraformaldehyde , permeabilized for 15 min with 0 . 2% PBT , then blocked with 10% normal donkey serum . Slides were incubated with primary antibodies overnight at 4°C in a humid chamber , washed four times with PBS , and incubated with secondary antibody at room temperature for 1–2 hr . Samples were washed three times with PBS before mounting with Vectashield ( Vector Labs ) . To stain cell surface proteins , cells were incubated with primary antibody in PBS for 2 hr at room temperature prior to fixation . Pictures were collected on a Leica SP8 confocal microscope . For aggregation assays , S2 cells were pelleted 48 hr after transient transfection using Effectene Transfection Reagent ( Qiagen ) and washed in fresh medium . A total of 2 . 5 ml of cells at a concentration of 4 × 106 cells/ml were rocked at 50 rpm for at least 3 hr . Plates were then analyzed for the presence of cell aggregates . Pictures were collected on a Zeiss AxioZoom microscope .	New nerve cells in a developing organism face a difficult challenge: finding the right partners to connect with in order to form the complex neural networks characteristic of a fully formed brain . Each cell encounters many potential matches but it chooses to connect to only a few , partly based on the proteins that decorate the surface of both cells . Still , too many cell types exist for each to have its own unique protein label , suggesting that nerve cells may also use the amount of each protein to identify suitable partners . Douthit , Hairston et al . explored this possibility in developing fruit flies , focusing on how R7 photoreceptor cells – present in the eye to detect UV light – connect to nerve cells in a specific brain layer . It is easy to spot when the process goes awry , as the incorrect connections will be in a different layer . Experiments allowed Douthit , Hairston et al . to identify a protein baptized ‘Lost and found’ – ‘Loaf’ for short – which R7 photoreceptors use to find their partners . Removing Loaf from the photoreceptors prevented them from connecting with their normal partners . Surprisingly though , removing Loaf from both the eye and the brain solved this problem – the cells , once again , formed the right connections . This suggests that R7 photoreceptors identify their partners by looking for cells that have less Loaf than they do: removing Loaf only from the photoreceptors disrupts this balance , leaving the cells unable to find their match . Another unexpected discovery was that Loaf is not present on the surface of cells , but instead occupies internal structures involved in protein transport . It may therefore work indirectly by controlling the movement of proteins to the cell surface . These findings provide a new way of thinking about how nerve cells connect . In the future , this may help to understand the origins of conditions in which the brain is wired differently , such as schizophrenia and autism .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "neuroscience" ]	2021	R7 photoreceptor axon targeting depends on the relative levels of lost and found expression in R7 and its synaptic partners
Molecular bistables are strong candidates for long-term information storage , for example , in synaptic plasticity . Calcium/calmodulin-dependent protein Kinase II ( CaMKII ) is a highly expressed synaptic protein which has been proposed to form a molecular bistable switch capable of maintaining its state for years despite protein turnover and stochastic noise . It has recently been shown that CaMKII holoenzymes exchange subunits among themselves . Here , we used computational methods to analyze the effect of subunit exchange on the CaMKII pathway in the presence of diffusion in two different micro-environments , the post synaptic density ( PSD ) and spine cytosol . We show that CaMKII exhibits multiple timescales of activity due to subunit exchange . Further , subunit exchange enhances information retention by CaMKII both by improving the stability of its switching in the PSD , and by slowing the decay of its activity in the spine cytosol . The existence of diverse timescales in the synapse has important theoretical implications for memory storage in networks . Memories are believed to be stored in synapses , encoded as changes in synaptic strength ( Hebb , 2005; Takeuchi et al . , 2014; Choi et al . , 2018 ) . Long-term potentiation ( LTP ) , an activity-dependent change in synaptic strength , is considered to be the primary post-synaptic memory mechanism ( Bliss and Collingridge , 2013; Mayford et al . , 2012 ) . Various behavioral experiments strongly suggest a critical role for CaMKII in induction of LTP ( Lucchesi et al . , 2011; Giese et al . , 1998 ) . In the CA1 region of the hippocampus , blocking CaMKII activity blocks the induction of LTP ( Chang et al . , 2017 ) . After LTP induction , several other pathways including protein synthesis ( Aslam et al . , 2009 ) , clustering of receptors ( Shouval , 2005 ) , receptor translocation ( Hayer and Bhalla , 2005 ) and PKM-ζ activation ( Sacktor , 2012 ) , have been suggested as mechanisms for long-term maintenance of synaptic state . Recent evidence from behavioral assays suggests that CaMKII may also be involved in long-term maintenance of memory ( Rossetti et al . , 2017; but see Chang et al . , 2017 ) . Any putative molecular mechanism involved in long-term maintenance of memory must be able to maintain its state despite the potent resetting mechanisms of chemical noise and protein turnover . In the small volume of the synapse ( ∼0 . 02 µm3 [Bartol et al . , 2015] ) , the number of molecules involved in biochemical processes range from single digits to a few hundred , thereby increasing the effect of chemical noise . John Lisman proposed that a kinase and its phosphatase could form a bistable molecular switch able to maintain its state for a very long time despite turnover ( Lisman , 1985 ) . It has been shown by various mathematical models that CaMKII and its phosphatase protein phosphatase 1 ( PP1 ) may form a bistable switch ( Zhabotinsky , 2000 ) which can retain its state for years despite stochastic chemical noise and protein turnover ( Miller et al . , 2005; Hayer and Bhalla , 2005 ) . Although there is experimental evidence that CaMKII/PP1 is bistable in in vitro settings ( Bradshaw et al . , 2003; Urakubo et al . , 2014 ) , experimental evidence for in vivo bistability is lacking . In spine cytosol , CaMKII has been shown not to act like a bistable switch but rather a leaky integrator of calcium activity ( Chang et al . , 2017 ) . However , CaMKII may be bistable in special micro-environments such as the ‘core’ PSD where it attaches to the NMDA receptor ( Dosemeci et al . , 2016; Petersen et al . , 2003 ) . From a computational perspective , the CaMKII/PP1 bistable system is an attractive candidate for memory storage ( Koch , 2004 ) . Bistability provides a plausible solution to the problem of state maintenance . Previous modeling work has shown that the CaMKII/PP1 system may form a very stable switch despite protein turnover and stochastic noise in the small volume of the synapse . The stability increases exponentially with the number of holoenzymes ( Miller et al . , 2005 ) . It is important to note that this model exhibits bistable behavior only in a narrow range of PP1 concentrations in the PSD . This strict restriction may be met because phosphorylated CaMKII is protected from phosphatases in PSD except PP1 ( Strack et al . , 1997a ) , which is tightly regulated in the PSD ( Bollen et al . , 2010 ) . CaMKII has another remarkable property which was hypothesized by Lisman ( Lisman , 1994 ) but discovered only recently , namely , subunit exchange . In this process , two CaMKII holoenzymes can exchange active subunits leading to spread of CaMKII activation ( Stratton et al . , 2014 ) . In this paper , we adapt the Miller and Zhabotinksy ( MZ ) model ( Miller et al . , 2005 ) to include subunit exchange and diffusion , and quantify the effects of subunit exchange on the properties of the CaMKII-PP1 system in two adjacent neuronal micro-environments: PSD and spine cytosol . In the PSD , PP1 is tightly regulated and CaMKII is protected from other phosphatases . But in the spine cytosol , CaMKII is accessible to other phosphatases along with PP1 . We examined how state switching lifetimes in the PSD are affected by subunit exchange in different contexts of PP1 levels , turnover , and clustering of CaMKII . In the spine cytosol , we show how the integration of calcium stimuli generates two time-courses of CaMKII activity as a result of subunit exchange ( Chang et al . , 2017 ) . The basic computational units in our model are individual CaMKII subunits , and the CaMKII ring consisting of six or seven CaMKII subunits . We treat the CaMKII ring as a proxy for the CaMKII holoenzyme , which consists of two such rings stacked over each other ( Woodgett et al . , 1983; Hoelz et al . , 2003; Chao et al . , 2011 ) . We define Active CaMKII as a holoenzyme ( ring of six or seven subunits ) in which at least two subunits are phosphorylated at Thr286 . In our model , CaMKII exists in 15 possible states compared to two in the MZ model ( see Materials and methods ) . This leads to many more reactions than the MZ model . We also replaced the Michaelis-Menten approximation in the MZ model by equivalent mass-action kinetics ( e . g . Equation 2 ) . Since analytical comparison of the two models was not possible , we first compared numerical results from our model without diffusion and without subunit exchange with the MZ model ( Figure 1 ) . Our model exhibited all the key properties of the MZ model: ( 1 ) In the PSD , under basal calcium ( Ca2+ ) stimulus conditions , CaMKII/PP1 formed a bistable switch ( Figure 1C , D ) , ( 2 ) The stability of the switch increased exponentially with system size ( Figure 1E ) , ( 3 ) Increased number of PP1 molecules ( NPP1 ) shut off the switch ( Figure 2 ) , and ( 4 ) Bistability was robust to slow turnover of CaMKII ( Figure 3 ) . Thus , our baseline model exhibited all the key properties that had previously been predicted for the bistable CaMKII switch . However , subunit exchange and diffusion introduced several interesting additional properties , which we examine below . We first analyzed the switch sensitivity to PP1 . In our model as well in the MZ model , the number of PP1 molecules ( NPP1 ) has an upper limit for the switch to exhibit bistability . This constraint arises because PP1 must saturate in the ON state of the switch , that is the maximal enzymatic turnover of PP1 must be smaller than the rate of activation of CaMKII subunits . However , unlike the MZ model where the addition of one extra PP1 molecule changed the residence time of the ON state by roughly 90% ( Figure 2C in Miller et al . , 2005 ) , we did not find the residence time of the ON state to be this sensitive to PP1 . In our model , on average it required half the number of holoenzymes ( i . e . 0 . 5× NCaMKII ) extra PP1 molecules to cause a similar 90% change in the residence time of the ON state . This number is roughly equal to the maximum number of CaMKII subunits ( released from CaMKII holoenzymes during subunit exchange Equation 3 ) that can exist at any given time in our model . We conjecture that this reduced sensitivity to PP1 is due to the fact that PP1 participates in many more reactions in our model . We found that a system consisting of NCaMKII = 12 holoenzymes remained bistable for NPP1 = 3× to 8× NCaMKII without subunit exchange , and for NPP1 = 12× to 16× NCaMKII with subunit exchange for Dsub = 0 . 1 , and DPP1 = 0 . 5 µm2 s−1 ( Figure 2B ) . Thus , subunit exchange shifted the middle of the bistable range to higher values of PP1 . The width of the range of PP1 over which bistability was present saw a moderate increase in the presence of subunit exchange ( blue and red sigmoidal fit in Figure 2B ) . A similar trend was observed for other values of DPP1 and Dsub ( data not shown ) . In the presence of subunit exchange , the ON state of the switch has a tighter distribution ( blue vs . red histogram , Figure 2A ) , that is , there are fewer holoenzymes that are completely de-phosphorylated by PP1 . We interpret this as follows: In the presence of subunit exchange , any subunit in a holoenzyme de-phosphorylated by the PP1 is likely to be rapidly re-phosphorylated . This is because , when the switch is in ON state , most diffusing subunits present in the PSD are in the phosphorylated state . Hence , in addition to auto-phosphorylation , the exchange reactions ( Equation 3 ) turn unphosphorylated holoenzymes to phosphorylated holoenzymes with a significant rate . Taken together , subunit exchange acts as a compensatory mechanism for dephosphorylation by PP1 in the ON state of the switch . Subunit exchange also had a strong effect on time spent by the switch in transition from one stable state to another ( relaxation time ) . When subunit exchange was enabled , the relaxation time was reduced ( red vs . blue dashed line in Figure 2B ) and also became independent of NPP1 . As mentioned previously , due to subunit exchange , the ON state has a tighter distribution ( blue vs . red histogram in Figure 2A ) . This means that there were fewer ineffective transitions from the ON to the OFF state . As expected , the standard deviation of the relaxation time was also greatly reduced in the presence of subunit exchange ( red and blue curve , Figure 2C ) . Thus , subunit exchange makes the switch’s ON state less noisy and more robust to dephosphorylation by PP1 . Parallel results were obtained for the effect of subunit exchange on CaMKII switch robustness in the context of protein turnover . Turnover acts at a constant rate to replace any active CaMKII holoenzyme with an inactive holoenzyme ( Equation 6 ) , thus decreasing the stability of the ON state . Without subunit exchange , switch stability as measured by residence time of the ON state decreased exponentially with increasing turnover rate . With subunit exchange , however , residence time of the ON state remained roughly constant upto a ∼10 fold increase in turnover ( Figure 3B ) , after which subunit exchange could not phosphorylate all the inactive holoenzymes produced by turnover . At this point , the switch started to show a similar steep decay of stability as was seen without subunit exchange . As expected , turnover increased the number of switching events in the regime of bistability in both cases . Thus , subunit exchange broadens the zone of bistability of the switch , both with respect to the range of NPP1 , and the turnover rate over which the switch remains bistable . It also reduces fluctuations in the ON state of the switch . As suggested in Stratton et al . , 2014 , we found that subunit exchange facilitated the spread of CaMKII activation ( Figure 4 ) . When subunits were allowed to diffuse , an active subunit could be picked by a neighboring inactive CaMKII holoenzyme , making it partially phosphorylated . This process overcomes the first slow step of CaMKII phosphorylation ( Equation 1 ) , especially when subunit exchange makes many phosphorylated subunits available , thereby facilitating the spread of activation . We simulated NCaMKII = 18 inactive holoenzymes in a cylindrical arena with a volume of 0 . 0275 µm3 and a length of 540 nm representing the PSD . The cylinder was divided into 18 voxels ( one holoenzyme in each voxel ) . Each voxel was separated by 30 nm , which is the average nearest-neighbour distance for CaMKII holoenzymes ( Feng et al . , 2011 ) . Each voxel was considered to be a well-mixed environment that is diffusion was instantaneous within the voxel . Between voxels , diffusion was implemented as cross-voxel jump reactions ( see Materials and methods ) . We did not try 2D/3D diffusion because of its simulation complexity and because it would be expected to be qualitatively similar ( Fange et al . , 2010 ) . We fixed the diffusion coefficient of PP1 ( DPP1 ) to quantify the effect of varying the diffusion coefficient of subunits ( Dsub ) and basal calcium levels . We used NPP1 = 0 . 5 µm2 s−1 which is the observed value of the diffusion coefficient of Ras , a similar sized protein ( Harvey et al . , 2008 ) . We ran simulations for 4 h at basal calcium concentration [Ca2+] = 80 nM+ϵ ( where ϵ is the fluctuation in basal calcium levels , see Figure 1C ) , and without subunit exchange ( i . e . Dsub = 0 ) . We set NPP1 = 15× NCaMKII to make sure the system showed no significant CaMKII activity ( Figure 4B , red curve ) . This served as the baseline to quantify the effect of subunit exchange . When we enabled subunit exchange by setting Dsub to a non-zero value , CaMKII activity rose to a maximum within 4 h even for a low value of Dsub = 0 . 001 µm2 s−1 ( Figure 4C , black curve ) . The first step of CaMKII phosphorylation ( Equation 1 ) is slow since it requires binding of two calcium/calmodulin complex ( Ca2+/CaM ) simultaneously ( at basal [Ca2+] = 80 nM+ϵ , v1 = 1 . 27 × 10-5 s-1 ) . However , subunit exchange can also phosphorylate a subunit in a holoenzyme by adding an available phosphorylated subunit to it ( Equation 3 ) . Note that a Dsub value as low as 0 . 001 µm2 s−1 is good enough for subunit exchange to be effective . With this value of Dsub , it takes roughly 0 . 9 s for the subunit to reach another holoenzyme which is , on average , 30 nm away . Under these conditions , the rate of picking up available active subunits ( given in Equation 3 ) is faster than v1 . Expectedly , for larger Dsub values ( e . g . , 0 . 001 and 0 . 1 µm2 s−1 ) , subunit exchange becomes more effective ( compare red and blue with the rest in Figure 4D ) . As expected , at higher basal Ca2+ levels ( 120 nM ) , the system showed higher CaMKII activity for all values of Dsub ( Figure 4D ) . Increasing Dsub increased the effect of subunit exchange , as measured by the decreased rise time of CaMKII activity from 10% to 90% ( Figure 4D ) . However , the time of onset of CaMKII activation as measured by rise time from 0% to 10% was dependent only on basal Ca2+ levels but not on Dsub ( Figure 4E ) . Thus , subunit exchange facilitates the spread of kinase activity following CaMKII activation but does not affect the onset of CaMKII activation . Next , we probed the effect of subunit exchange between spatially separated CaMKII clusters . We considered NCaMKII holoenzymes organized into three clusters of size NCaMKII/3 , each separated by a distance d . This configuration corresponds to cases where receptors and CaMKII holoenzymes are clustered at the synapse . When there is no subunit exchange across voxels ( Dsub = 0 ) , these switches are expected to switch independently like multiple coins flipped together , resulting in a binomial distribution of activity . The clustered system had three relatively stable bistable systems ( long residence time , Figure 1E ) . As expected , without subunit exchange , activity in this system had a binomial distribution ( Figure 5B , red plot ) . Then we allowed PP1 and CaMKII subunits to undergo linear diffusion . We set DPP1 = 0 . 5 µm2 s−1 as before and varied Dsub to quantify effect of subunit exchange . Subunit exchange led to synchronization of switching activity . The population of clustered CaMKII acted as a single bistable switch ( Figure 5B , blue plot ) . This effect was strong and robust to variation in Dsub . Even for a very small value of Dsub = 0 . 01 µm2 s−1 , we observed strong synchronization ( Figure 5D ) . The synchronization disappeared completely for Dsub less than 0 . 0001 µm2 s−1 , and for d greater than 100 nm ( Figure 5D ) . Thus , for most physiologically plausible values of diffusion coefficient Dsub , subunit exchange causes synchronization of switching activity of clustered CaMKII . Finally , we asked if subunit exchange might account for the complex time-course of CaMKII dynamics in spine as observed in recent experiments ( Chang et al . , 2017 ) . We designed a simulation to replicate an experiment where CaMKII was inhibited by a genetically encoded photoactivable inhibitory peptide after activating CaMKII by glutamate uncaging ( Murakoshi et al . , 2017 ) . In the spine , CaMKII is more accessible to phosphatases than in the PSD , where our previous calculations had been located . To model the increased availability of phosphatases , we increased the concentration of PP1 by an order of magnitude , and increased the volume of the compartment to match the volume of a typical spine head that is 0 . 02 µm3 ( Bartol et al . , 2015 ) . We found that CaMKII acted as a leaky integrator of the calcium activity with a typical exponential decay dynamics ( Figure 6A ) . We then enabled the diffusion of CaMKII subunits ( Dsub = 1 µm2 s−1 ) and PP1 ( DPP1 = 0 . 5 µm2 s−1 ) . These conditions decreased the rate of dephosphorylation roughly by a factor of 5 ( 41 . 65 s vs . 200 . 82 s ) ( Figure 6B ) . We expected that subunit exchange should have a strong effect on the time-course of decay of activity of clustered CaMKII in spine cytosol ( e . g . CaMKII bound to actin ) because the proximity of holoenzymes would lead to rapid exchange . Thus , if there are populations of clustered as well as non-clustered CaMKII in the spine , we expected that they would exhibit long and short time-courses of activity decay , respectively . Therefore a mixed population of clustered and non-clustered CaMKII should decay with two time-constants . Our simulations supported this prediction . In Chang et al . ( 2017 ) , the decay kinetics of CaMKII were obtained by curve fitting of experimental data . It was given by a double-exponential function: F⁢ ( t ) =Pf⁢a⁢s⁢t⁢e-t/τf⁢a⁢s⁢t+Ps⁢l⁢o⁢w⁢e-t/τs⁢l⁢o⁢w where Pfast=0 . 74 , Pslow=0 . 26 , τfast=6 . 4±0 . 7s , τslow=92 . 6±50 . 7s ( Figure 6C , magenta ) . We used their estimate of Pf⁢a⁢s⁢t and Ps⁢l⁢o⁢w to construct a mixed population of slow and fast decaying CaMKII ( Figure 6A , black ) , and simulated the decay kinetics of CaMKII for this system . We fit the resulting decay curve with a double-exponential function ( Figure 6C , dashed red ) . The time-constants obtained ( 8 . 4 s , 86 . 2 s ) matched well with experimentally estimated time-constants of ( 6 . 4 s ± 0 . 7 , 92 . 6 s ± 50 . 7 ) . Thus , we suggest that subunit exchange may be a mechanism that leads to CaMKIIα activity decaying with two time-courses in spine cytosol . Here , we have shown that subunit exchange strongly affects the properties of the CaMKII/PP1 pathway , both in its role as a bistable switch in the PSD and as a leaky integrator of Ca2+ activity in spine cytosol . In the PSD , where the model was tuned to elicit bistable dynamics from clustered CaMKII , subunit exchange improved the stability of the CaMKII/PP1 switch by synchronizing the kinase activity across the PSD ( Figure 6 ) . It also improved active CaMKII tolerance of PP1 , and of turnover rate ( Figure 2 and Figure 3 ) . In the case where CaMKII was uniformly distributed in PSD , subunit exchange facilitated more rapid activation of CaMKII ( Figure 4B–D ) ( Stratton et al . , 2014 ) . These simulation results predict that a CaMKII mutant lacking subunit exchange would be deficient in switch stability and slower to be activated by Ca2+ , thereby resulting in degraded memory retention and deficient learning in memory-related behavioral experiments , respectively . In the spine head , subunit exchange facilitated integration by prolonging the decay time-course of kinase activity ( Figure 6 ) . The fact that CaMKII dynamics changed from an integrator to bistable switch as we moved from spine cytosol ( a phosphatase rich environment ) to the PSD ( where PP1 is tightly controlled ) suggests an interesting sub-compartmentalization of functions in these microdomains . Furthermore , we observed that the clustering of CaMKII had important implications for its sustained activity . Subunit exchange is unlikely to have any impact on neighbouring spines . The mean escape time of a single CaMKII subunit from a typical spine is between 8 s to 33 s ( Holcman and Schuss , 2011 ) . Any phosphorylated subunit is almost certain to be de-phosphorylated by PP1 during this time . We therefore predict that the effects of synchronization are local to each PSD , where PP1 is known to be tightly controlled . Subunit exchange loses its potency in the phosphatase rich region of the bulk spine head or dendrite . We therefore consider it unlikely that CaMKII subunit exchange plays any role in intra-spine information exchange such as synaptic tagging . CaMKII is non-uniformly distributed in the PSD where it is mostly concentrated in a small region of 16 nm to 36 nm below the synaptic cleft ( Petersen et al . , 2003 ) . In the PSD , CaMKII may exist in large clusters given that the PSD is rich in CaMKII binding partners . Our study predicts that subunit exchange may lead to synchronization when CaMKII is clustered , or more rapid activation by Ca2+ when it is uniformly distributed . Given that CaMKII can form clusters with N-methyl-D-asparate ( NMDA ) receptors , it would be interesting to study the mixed case where some CaMKII is clustered and the rest is uniformly distributed . This would require a detailed 3D simulation and is beyond the scope of this study . Finally , we suggest that the existence of diverse time-scales of CaMKII activity – bistable and highly stable synchronized bistable in PSD , slow and fast decaying leaky integrator in spine head ( Table 1 ) – has important theoretical implications . A very plastic synapse is good at registering activity dependent changes ( learning ) but poor at retaining old memories . On the other hand , a rigid synapse is good at retaining old memories but is not efficient at learning . A theoretical meta-model which sought to strike a balance between these two competing demands requires that a diversity of timescales must exist at the synapse ( Benna and Fusi , 2016 ) for optimum performance . In this model , complex synapses with state variables with diverse time-scales are shown to form a memory network in which storage capacity scales linearly with the number of synapses , and memory decay follows 1/t — a power-law supported by psychological studies ( Wixted and Ebbesen , 1991 ) . This model requires the memory trace to be first stored in a fast variable and then progressively and efficiently transferred to slower variables . Our study suggests a concrete mechanism for such a process . Here , the Ca2+ concentration in the PSD can be mapped to the fastest variable . The CaMKII integrator in the cytosol could represent the second slower variable to which the trace is transferred from Ca2+ . Further , the state information is transferred to the third slower CaMKII bistable switch . The dynamics of CaMKII in the PSD forms an even slower bistable variable for longer retention of the memory trace . It is possible that memory is transferred from here to even slower variables , such as sustained receptor insertion ( Hayer and Bhalla , 2005 ) , PKM-ζ activation ( Sacktor , 2012 ) , or local protein synthesis ( Aslam et al . , 2009 ) . We assumed the resting Ca2+ level in spine to be 80 nM ( Berridge , 1998 ) . In the MZ model , Miller et al . assumed that Ca2+ entry through NMDA receptors can be approximated by a Poisson train with an average rate of 0 . 5 Hz . Since , on average , ∼0 . 5 NMDA receptors open ( Nimchinsky et al . , 2004 ) upon pre-synaptic stimulation , we reduced the frequency of NMDA opening events to 0 . 25 Hz . We used a periodic pulse with a time-period of 4 s and duty cycle of 50% . To model NMDA activity in the 2 s long ON period of our 4 s long periodic pulse , we sampled from a uniform distribution with median of 120 nM ( 50% change , on average ) and range of 40 nM ( Figure 1C ) . This distribution is informed by Figure 2B , C from ( Nimchinsky et al . , 2004 ) . We did not consider decay dynamics of Ca2+ influx through the NMDA channel since the timescale of decay ( roughly 100 ms ) is much shorter than our simulation runtimes of days , and including this detail would have made the simulations very slow . The effect of ignoring decay dynamics are expected to be negligible given that the time-scale of CaMKII activation is much larger than the time course of Ca2+ decay dynamics . Furthermore , we did not consider contributions to background Ca2+ fluctuations by other channels . This background activity is represented by ϵ in the figures and text . The activation of CaMKII in our study follows the same dynamics as in the MZ model ( Equation 1 ) . In our paper , by phosphorylation/activation of a CaMKII subunit or a holoenzyme , we mean phosphorylation at Thr286 . The first step in CaMKII activation requires simultaneous binding of two ( Ca2+/CaM ) to the two adjacent subunits of CaMKII . Once a subunit is phosphorylated , it catalyzes phosphorylation of its neighbour ( auto-phosphorylation ) which requires binding of only one ( Ca2+/CaM ) . Therefore , further phosphorylation proceeds at much faster rate . The phosphorylation of CaMKII is given by Equation 1 ( Bradshaw et al . , 2003; Miller et al . , 2005 ) . ( 1 ) xayn−a→v1xa−1yn−a+1→v2xa−2yn−a+2v1=k1[H31+H3]2 , v2=k1H31+H3 , whereH=Ca2+KH1where n = 6 or 7 for 1≤a≤n;k1=1 . 5s−1 ( Hanson et al . , 1994 ) , and kH1=0 . 7µM ( De Koninck and Schulman , 1998; Miller et al . , 2005 ) . At resting Ca2+ concentration of 100 nM , v1=1 . 27×10−5s−1 and v2=4 . 36×10−3s−1 ( i . e . , v2/v1≈343 ) . The rate constant v1 above defines the initial rate of phosphorylation . Furthermore , addition of phosphorylated subunits can happen through subunit exchange ( Equation 3 ) . We treat these as independent variables . The phosphorylation rates v1 and v2 are relatively well constrained by the experimental literature . The subunit exchange rates were estimated ( Materials and methods ) to be in the range of 1 s-1 . Once fully phosphorylated , CaMKII moves to the PSD where it binds to the NMDA receptor . Upon binding , it is no longer accessible other phosphatases except PP1 . The dephosphorylation of the CaMKII ring , and the subunit are given by Equation 2 . ( 2 ) PP1+xayn−a⇌k−k+PP1 . xayn−a→k2PP1+xa+1yn−a−1PP1+x⇌k−k+PP1 . x→k2PP1+ywhere n = 6 or 7 , and 1≤a≤n . Following ( Miller et al . , 2005 ) , we also assumed k-=0 . This gave us k+=k2kM = 1/µM/s . We could not find any experimental estimate of kM in recent literature , therefore we used the same value of kM as in the MZ model ( Miller et al . , 2005 ) . Since CaMKII ring consists of either 6 or 7 subunits in our model , any ring with six subunits cannot lose a subunit , and a ring with seven subunits cannot gain a subunit . The reactions which result in either gain or loss of a subunit are given by Equation 3 where 0≤a≤6or7 . ( 3 ) xay7−a+x⇌kx−kx+xa+1y6−axay6−a+y⇌ky−ky+xay7−a We were not able to find values for kx+ , kx- , ky+ , and ky- in the literature . We used the data in Stratton et al . ( 2014 ) to estimate the possible timescale of subunit exchange rate . ( Bhattacharyya et al . , 2016 ) speculate that upon activation , the hub of the holoenzyme becomes less stable and more likely to open up and lose a subunit that is an active holoenzyme loses subunits at a faster rate . Therefore , we maintained the following ratio kx−≈10kx+NCaMKII and ky−≈10ky+NCaMKII in all simulations where NCaMKII is the number of holoenzymes in the system . In the PSD , PP1 is the primary – and perhaps only – phosphatase known to dephosphorylate CaMKII ( Strack et al . , 1997b ) . We followed the MZ model for Equation 5 where inhibitor-1 ( I1 ) inactivates PP1 . Phosphorylated inhibitor-1 ( I1P ) renders PP1 inactive by forming I1P-PP1 complex ( I1P . PP1 ) . ( 5 ) PP1+I1P⇌k4k3I1P . PP1I1P=I1vPKAvCaN1+ ( CakH2 ) 3 ( CakH2 ) 3where k3 = 100 /µM/s , k4 = 0 . 1 s-1 ( Endo et al . , 1996 ) , and vPKA/vCaN = 1 ( Miller et al . , 2005 ) . The turnover of CaMKII is a continuous process given by Equation 6 with rate vt=30⁢h-1 ( Ehlers , 2003 ) . ( 6 ) xay6−a→vtx6y0for6≥a≥1xay7−a→vtx7y0for7≥a≥1 Diffusion is implemented as a cross voxel jump reaction . Diffusion of a species X with diffusion-coefficient DX between voxel A and B separated by distance h is modelled as a reaction XA⇌kkXB where k=DX/h2 , and [XA]=[XB]=[X]/2 ( Erban et al . , 2007 ) . Based on our own numerical results ( Appendix 1—figure 2 ) and other studies ( Isaacson , 2009; Erban and Chapman , 2009 ) , we are confident that h≥10⁢hc⁢r⁢i⁢t where hc⁢r⁢i⁢t=k+DP⁢P⁢1+Ds⁢u⁢b is a good value . We have hcrit ≤ 3 . 2 nm whenever DPP1 + Dsub ≥ 0 . 5 µm2 s−1 . For all simulations presented in main text , we maintained h≥hc⁢r⁢i⁢t . For a few illustrative examples where h is smaller than hc⁢r⁢i⁢t , see Figure 4—figure supplement 1D , E . All simulations were performed using the stochastic solver based on the Gillespie method , in the MOOSE simulator ( https://moose . ncbs . res . in , version 3 . 1 . 4; Ray and Bhalla , 2008 ) . This model is available at https://github . com/dilawar/SinghAndBhalla_CaMKII_SubunitExchange_2018 ( copy archived at https://github . com/elifesciences-publications/SinghAndBhalla_CaMKII_SubunitExchange_2018 ) . The table of parameters is in SI ( Table 2 ) . To validate our implementation of diffusion , we compared trajectories of two systems: one in a single well-mixed cylinder with parameters tuned to elicit bistable behavior ( henceforth , we call it the reference bistable ) , and a spatial system implemented as a discretized cylinder as described above . We expect the later to converge to reference bistable system when the diffusion constants become large such that the molecules are effectively well-mixed . We put six CaMKII holoenzymes in a cylinder of length 180 nm discretized into six voxels , separated by a distance of 30 nm . The long-term behavior of discretized system was most sensitive to DPP1 ( Figure 8B ) and almost independent of Dsub ( Figure 8A ) . The discretized system converges to reference bistable for DPP1 ≥ 0 . 5µm2s-1 ( Figure 8C ) . Table 2 summarizes the parameters of our model .	The brain stores memories by changing the strength of synapses , the connections between neurons . Synapses that change their strength easily can quickly encode new information . But such synapses are also unstable . They tend to revert back to their original state and so struggle to retain information . By contrast , synapses that are slow to change their strength are slow to learn , but are good at remembering . The difference is a little like that between writing a message in wet sand versus carving it into stone . It is quick and easy to write on sand , but the resulting marks are temporary . Writing on stone is slow and difficult , but the results last far longer . The brain must strike a balance between how fast synapses can learn and how well they can retain that information . One molecule that helps with this is a synaptic protein called CaMKII . Each CaMKII molecule consists of multiple subunits and exists in either an active or inactive state . Experiments have shown that CaMKII molecules can swap subunits . But how does this affect memory ? Singh and Bhalla used a computer model to simulate subunit exchange between CaMKII molecules . The results revealed that when active CaMKII molecules swap subunits , synapses become better at retaining information . However , when inactive CaMKII molecules swap subunits , synapses do not become better at encoding information . Subunit exchange by CaMKII thus helps synapses stabilize existing memories , rather than form new ones . This makes it easier for the brain to retain stored information despite threats to stability such as the turnover of proteins . A better knowledge of how the brain balances quick learning and slow forgetting may help us to better understand brain disorders , such as Alzheimer’s disease ( in which patients struggle to remember ) , and post-traumatic stress disorder ( in which patients struggle to forget ) . Biological memory networks can also inspire artificial memory systems . Damaging a few components of a computer memory can erase all the stored information . By contrast , the brain loses many neurons every day without suffering the same catastrophic failure . Mimicking such fault tolerance in an artificial system would be highly valuable for storing critical memories .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "computational", "and", "systems", "biology", "neuroscience" ]	2018	Subunit exchange enhances information retention by CaMKII in dendritic spines
Endometriosis is a chronic painful disease highly prevalent in women that is defined by growth of endometrial tissue outside the uterine cavity and lacks adequate treatment . Medical use of cannabis derivatives is a current hot topic and it is unknown whether phytocannabinoids may modify endometriosis symptoms and development . Here we evaluate the effects of repeated exposure to Δ9-tetrahydrocannabinol ( THC ) in a mouse model of surgically-induced endometriosis . In this model , female mice develop mechanical hypersensitivity in the caudal abdomen , mild anxiety-like behavior and substantial memory deficits associated with the presence of extrauterine endometrial cysts . Interestingly , daily treatments with THC ( 2 mg/kg ) alleviate mechanical hypersensitivity and pain unpleasantness , modify uterine innervation and restore cognitive function without altering the anxiogenic phenotype . Strikingly , THC also inhibits the development of endometrial cysts . These data highlight the interest of scheduled clinical trials designed to investigate possible benefits of THC for women with endometriosis . Endometriosis is a chronic inflammatory disease that affects 1 in 10 women of childbearing age ( Zondervan et al . , 2019 ) . It is characterized by the growth of endometrium in extrauterine locations , chronic pain in the pelvis and the lower abdomen , infertility , emotional distress and loss of working ability ( Fourquet et al . , 2011; Márki et al . , 2017; Zondervan et al . , 2019 ) . Current clinical management provides unsatisfactory outcomes . On the one hand , hormonal therapy has unwanted effects including contraception and emotional disturbances ( Ross and Kaiser , 2017; Skovlund et al . , 2016 ) , whereas surgical excision of the growths is associated with high-recurrence rates and post-surgical pain ( Garry , 2004 ) . Hence , clinical treatments are limited and women often unsatisfactorily self-manage their pain ( Armour et al . , 2019 ) . In this context , marijuana legalization for medical purposes in American and European states has led to increased availability of phytocannabinoids ( Abuhasira et al . , 2018 ) . While cannabis may provide pain relief in certain conditions ( Campbell et al . , 2001 ) , it is unclear whether it may modify endometriosis symptoms or development . Δ9-tetrahydrocannabinol ( THC ) is the main psychoactive constituent of the Cannabis sativa plant , and multiple animal and clinical studies suggest its efficacy relieving chronic pain ( De Vry et al . , 2004; Harris et al . , 2016; King et al . , 2017; Ueberall et al . , 2019; Williams et al . , 2008 ) , although controversial results have been obtained in human clinical trials ( Stockings et al . , 2018 ) . However , THC has important side effects including cognitive deficits and anxiety ( Célérier et al . , 2006; Kasten et al . , 2017; Puighermanal et al . , 2013 ) . This work investigates the effects of natural THC in a mouse model of endometriosis that reproduces the ectopic endometrial growths and some of the behavioral alterations of clinical endometriosis . Our data show that THC is effective inhibiting hypersensitivity in the caudal abdominal area without inducing tolerance , as well as reducing the pain unpleasantness associated with endometriosis . Notably , THC also prevents the cognitive impairment observed in mice with ectopic endometrium without modifying anxiety-like behavior at this particular dose . Interestingly , THC shows efficacy limiting the development of ectopic endometrium , revealing disease-modifying effects of this natural cannabinoid . Our first aim was to characterize a novel experimental procedure to evaluate at the same time nociceptive , cognitive and emotional manifestations of endometriosis pain in female mice . Mice were subjected to a surgical implantation of endometrial tissue in the peritoneal wall of the abdominal compartment or to a sham procedure . Mice receiving ectopic endometrial implants developed persistent mechanical hypersensitivity in the caudal abdominal area , whereas sham mice recovered their baseline sensitivity and showed significant differences in comparison to endometriosis mice since the second week of implantation ( Figure 1a and Figure 1—figure supplement 1 ) . To test whether mechanical hypersensitivity of endometriosis mice was specific to this abdominal region , nociceptive responses were also measured in the hind paw . In this distant area , mechanical sensitivity remained unaltered , indicating that pain sensitization did not generalize to other sites ( Figure 1b and Figure 1—figure supplement 2 ) . To discern whether increased nociception was accompanied by a component of negative affect , a measure of pain unpleasantness was taken on day 14 after the surgeries ( Figure 1c ) . Endometriosis mice showed increased nocifensive behaviors to mechanical stimuli when compared with sham mice . Similarly , endometriosis mice exhibited enhanced anxiety-like behavior reflected in lower percentages of time and entries to the open arms of the elevated plus maze ( Figure 1d ) . Total arm entries were similar in both groups ( Figure 1d ) . In line with these findings , previous rodent models of endometriosis found increased mechanosensitivity in the lower abdomen ( Arosh et al . , 2015; Greaves et al . , 2017 ) and affective-like disturbances ( Filho et al . , 2019; Li et al . , 2018 ) . Previous works associate nociceptive and emotional distress in chronic pain settings with cognitive decline ( Bushnell et al . , 2015; La Porta et al . , 2015; You et al . , 2018 ) , although this cognitive impairment has not yet been revealed in rodent models of endometriosis . We found in our model a dramatic impairment of long-term memory in endometriosis mice ( Figure 1e ) . While mnemonic effects of this pathology have not been thoroughly evaluated , a cognitive impairment may contribute to the loss of working ability consistently reported in women with endometriosis ( Hansen et al . , 2013; Sperschneider et al . , 2019 ) . Hence , mice with ectopic endometrium recapitulate in our model some of the symptomatology observed in the clinics , although manifestations of spontaneous pain could not be evaluated in this work . Mice receiving endometrial implants developed 3 to 5 endometrial cysts in the peritoneal wall of the abdominal compartment . Cysts were of 2 . 59 ± 0 . 34 mm diameter , filled with fluid , with glandular epithelium and stroma and innervated by beta-III tubulin positive fibers ( Figure 1f ) , as shown in women ( Tokushige et al . , 2006; Wang et al . , 2009 ) and other rodent models ( Arosh et al . , 2015; Berkley et al . , 2004 ) . Interestingly , we also found increased expression of the neuronal marker beta-III tubulin in the uteri of endometriosis mice ( Figure 1—figure supplement 3 ) , mimicking not only some of the symptoms but also the histological phenotype observed in women with endometriosis ( Miller and Fraser , 2015; Tokushige et al . , 2006 ) . Our second objective was to assess the effects of THC exposure on the endometriosis model to select an appropriate dose for a chronic treatment . Acute doses of THC were first tested in endometriosis and sham mice at a time point in which endometriotic lesions and hypersensitivity in the caudal abdomen were fully developed . Acute THC administration produced a dose-dependent reduction of abdominal mechanical hypersensitivity ( Figure 2 ) . The acute ED50 of THC 1 . 916 mg/kg ( ≈2 mg/kg ) was chosen for the repeated administration . Repeated exposure to THC 2 mg/kg , once daily for 28 days , provided a sustained alleviation of mechanical hypersensitivity during the whole treatment period ( Figure 3a and Figure 3—figure supplement 1 ) . Repeated THC starting on day 1 could have exerted a preventive effect at endometriosis stages in which pain sensitivity may have not been fully developed . To discern whether the absence in loss of efficacy was due to an inhibition of endometriosis development or to an actual lack of tolerance , we assessed the persistence of THC efficacy once pain was already present . THC given for the first time on day 14 was as effective as THC given on the same day after a daily treatment starting on day 8 ( 7 days long , Figure 3b and Figure 3—figure supplement 2 ) . Therefore , THC did not lose its efficacy when repeated administration started once painful symptomatology was established . The absence of tolerance to THC-induced antinociception is in contrast with the tolerance described at higher THC doses in other pain models ( Greene et al . , 2018; LaFleur et al . , 2018; Wakley et al . , 2014 ) . As expected , no effects of endometriosis or THC treatments were found in mechanical sensitivity of distant areas ( Hind paw , Figure 3—figure supplement 3 ) . Endometriosis mice treated with vehicle showed an increase in nocifensive behaviors compared with sham mice ( Figure 3c ) . Interestingly , the 7 day treatment with THC inhibited this component of negative affect , while the effects of an acute administration of THC were highly variable . This variable response could be associated to aversive effects associated with a first exposure to THC , an event described in humans ( MacCallum and Russo , 2018 ) and mice ( Kubilius et al . , 2018 ) . Additional experiments were conducted to assess the effects of THC on the anxiety-like behavior induced by endometriosis pain ( Figure 3d ) . As in previous experiments , endometriosis mice showed a lower percentage of time in the open arms of the elevated plus maze ( Figure 3d ) , revealing increased anxiety-like behavior . However , the percentage of entries to open arms was similar in endometriosis and sham mice . Therefore , the anxiogenic-like effect of ectopic endometrium in these experimental conditions was mild and the present model was not optimal to reveal the emotional component of this painful situation . No significant effects of repeated THC 2 mg/kg were observed on the percentages of time and entries , although THC-treated mice showed a subtle increase in anxiety-like behavior ( Figure 3d , percentage of time in open arms ) . Previous studies described anxiogenic-like effects of slightly higher doses ( 3 mg/kg ) in naïve male mice ( Viñals et al . , 2015 ) , and anxiolytic-like effects when using lower doses ( 0 . 3 mg/kg , Puighermanal et al . , 2013; Viñals et al . , 2015 ) . Thus , possible effects of THC alleviating pain-related anxiety-like behavior in endometriosis mice could be hindered by intrinsic anxiogenic effects of this THC dose . Therefore , doses with less pain-relieving efficacy could potentially be effective promoting anxiolytic-like effects considering the intrinsic effects of THC on emotional-like behavior . Alternatively , the absence of clear effects of THC on anxiety-like behavior may be associated to the evaluation time point , which was 6 hr after administration to study the impact of pain relief on anxiety-like behavior , rather than to assess direct drug effects . Total arm entries were similar among groups ( Figure 3d ) . Memory performance was also assessed the third week after starting the THC treatment . As expected , mice exposed to the chronic nociceptive manifestations of endometriosis showed a pronounced cognitive impairment , as well as sham mice exposed to THC , in accordance with previous reports in naïve males ( Kasten et al . , 2017; Puighermanal et al . , 2013 ) . Surprisingly , endometriosis mice repeatedly treated with natural THC showed intact discrimination indices ( Figure 3e ) suggesting protective effects of THC in this chronic inflammatory condition . In agreement , recent studies have shown cognitive improvements after THC exposure in old male and female mice ( Bilkei-Gorzo et al . , 2017; Sarne et al . , 2018 ) . Exogenous and endogenous cannabinoids have shown modulatory effects on the female reproductive system ( Walker et al . , 2019 ) . Thus , we analyzed the effects of THC on the ectopic and eutopic endometrium and on ovarian follicle maturation . Interestingly , endometriosis mice receiving THC 2 mg/kg for 32 days showed an evident inhibition of the development of endometrial cysts ( cyst diameter and area of endometrial tissue , Figure 4a ) without significant effects on cyst innervation ( Figure 4—figure supplement 1a ) . In agreement , a previous study showed antiproliferative effects of WIN 55212–2 , a synthetic cannabinoid agonist , on endometrial cell cultures and in ectopic endometrium implanted in immunodepressed mice ( Leconte et al . , 2010 ) . The assessment of the uterine diameter and the area of eutopic endometrium ( Figure 4—figure supplement 1b ) showed no effects of the THC treatment , suggesting that the antiproliferative activity of THC on endometrial cells is restricted to ectopic sites . However , possible effects of THC on established endometriosis lesions were not evaluated . Repeated THC increased the expression of neuronal markers in the uteri of sham mice , similar to the increase provoked by the ectopic endometrium ( Figure 4b ) . Interestingly , THC prevented this increase in endometriosis mice ( Figure 4b ) indicating again that THC exposure may have different consequences under chronic inflammatory conditions . In agreement , recent studies showed differential effects of THC on the nervous system of rodents with and without chronic inflammation ( Bilkei-Gorzo et al . , 2017; Sarne et al . , 2018 ) . To investigate a possible estrogenic influence on these histological findings , we analyzed 17 β-estradiol plasma levels . As expected , 17 β-estradiol plasma levels depended on the phase of the estrous cycle: mice in proestrus had the highest concentration followed by mice in diestrus , and mice in estrus showed the lowest levels ( Figure 4c , left graph ) . We found that 17 β-estradiol was similar in all experimental groups ( Figure 4c , right graph ) , although the levels of this estrogen were positively correlated with cyst diameter ( Figure 4d , left ) , proving the estrogenic influence on ectopic endometrial lesions . 17 β-estradiol levels were not correlated with endometrial area of the cysts ( Figure 4d , middle ) , or uterine innervation ( Figure 4d , right ) , suggesting independent THC effects on these histological changes . We also assessed possible effects of THC on ovarian functioning , since previous works have suggested inhibitory effects of THC on folliculogenesis and ovulation ( Adashi et al . , 1983; El-Talatini et al . , 2009 ) . Numbers of preantral follicles , antral follicles and corpora lutea were similar in all groups in our experimental conditions ( Figure 4—figure supplement 1c ) . These data suggest that endometriosis and THC were void of overt effects on ovarian follicle maturation and luteinization , however , other effects of endometriosis or THC on fertility cannot be excluded in our model . Similarly , the presence of prominent symptoms of endometriosis such as dysmenorrhea or dyspareunia could not be evaluated . Here we show for the first time that chronic administration of a moderate dose of the phytocannabinoid THC relieves mechanical hypersensitivity of caudal abdominal area , pain unpleasantness and cognitive impairment associated with the presence of ectopic endometrial cysts . These behavioral manifestations correlate with a decrease in the size of ectopic endometrium in THC-exposed mice . However , the pain-relieving effects of this particular dose of THC were not accompanied by a modification of anxiety-like behavior associated with endometriosis and effects on spontaneous pain were not evaluated in this work . Interestingly , THC produced opposite cognitive effects in sham and endometriosis mice . THC also induced an increase in markers of uterine innervation in sham animals , but prevented such changes in endometriosis mice , suggesting again different effects of THC under chronic inflammatory conditions . Importantly , THC also inhibited the growth of ectopic endometrium without apparent consequences on the eutopic endometrium and ovarian tissues . Altogether , the present data obtained in a preclinical model of endometriosis underline the interest in conducting clinical research to assess the effects of moderate doses of THC on endometriosis patients . Based on our results , we ( clinicaltrials . gov , #NCT03875261 ) and others ( gynica . com ) have planned the initiation of clinical trials to provide evidence on the translatability of these results to women with endometriosis . These novel clinical trials will evaluate this new possible endometriosis treatment under pathological human conditions . However , cannabis has a large number of potential side effects , as well as a high potential for abuse liability ( Curran et al . , 2016 ) , that have to be considered by physicians and patients . Therefore , the use of cannabis in unregulated scenarios should be discouraged taking into account these serious side effects . Female C57Bl/6J mice ( Charles Rivers , Lyon , France ) were used in all the experiments . Mice were 8 weeks old at the beginning of the experiments and were housed in cages of 4 to 5 mice with ad libitum access to water and food . The housing conditions were maintained at 21 ± 1°C and 55 ± 10% relative humidity in controlled light/dark cycle ( light on between 8 AM and 8 PM ) . Animals were habituated to housing conditions and handled for 1 week before the start of the experiments . All animal procedures were conducted in accordance with standard ethical guidelines ( European Communities Directive 2010/63/EU and NIH Guide for Care and Use of Laboratory Animals , 8th Edition ) and approved by autonomic ( Generalitat de Catalunya , Departament de Territori i Sostenibilitat ) and local ( Comitè Ètic d'Experimentació Animal , CEEA-PRBB ) ethical committees . Mice were randomly assigned to treatment groups and all experiments were performed blinded for pharmacological and surgical conditions . THC was purchased from THC-Pharm-GmbH ( Frankfurt , Germany ) as natural THC with 98 . 8% purity . This source of natural THC has been widely used in multiple research studies ( Busquets-Garcia et al . , 2018; Busquets-Garcia et al . , 2011; Cutando et al . , 2013; Flores et al . , 2014; Forsberg , 1970; Gunasekaran et al . , 2009; Lopez-Rodriguez et al . , 2014; Morrison et al . , 2011; Puighermanal et al . , 2013 ) . To corroborate the purity of the THC samples , High Performance Liquid Cromatography – Ultraviolet ( HPLC-UV ) was used for cannabinoid analysis and Gas Chromatography and Flame Ionization Detection ( GC-FID ) for terpenes ( Canna Foundation , Paterna , Spain ) . These analyses revealed no detectable amounts of other cannabinoids or terpenes ( Source Data Files 2 , 3 and 4 ) . THC was diluted in a vehicle composed of 2 . 5% ethanol , 5% Cremophor EL ( C5135 , Sigma-Aldrich St . Louis , MO , USA ) , and 92 . 5% saline , and was administered subcutaneously in a volume of 5 ml/kg . The phase of the estrous cycle was assessed by histological examination of cells extracted by vaginal lavage ( Byers et al . , 2012 ) the day of the surgeries and the day of euthanasia . Briefly , mice were gently restrained and 20 μl of saline were flushed 5 times into the vagina . The resulting fluid was placed onto gelatinized slides , stained with methylene blue and observed at 40X magnification under a light microscope ( DM6000 B , Leica Biosystems , Nussloch , Germany ) . Endometriotic lesions were surgically-induced as previously described ( Somigliana et al . , 1999 ) , with some modifications . Briefly , uterine horns from donor mice at diestrus were excised , opened longitudinally and biopsied into four pieces ( 2 × 2 mm ) . Recipient mice were anesthetized with vaporized isoflurane in oxygen ( 4% V/V for induction; 2 . 5% V/V for maintenance ) and a midline incision of 1 cm was made to expose the abdominal compartment . Endometriosis mice had four uterine fragments sutured to the parietal peritoneum , whereas sham-operated mice received four similar-sized fragments of abdominal fat . Transplanted tissues and abdominal muscle and skin were stitched using 6–0 black silk ( 8065195601 , Alcon Cusi S . A . , Barcelona , Spain ) . The nociceptive , affective and cognitive manifestations associated with the presence of ectopic endometrium were determined in a first experiment . After the measurement of baseline mechanical sensitivity ( day −1 ) , endometriosis or sham surgery was performed ( day 0 ) , and nociceptive responses were assessed again 7 , 14 , 21 and 28 days after surgery . Anxiety-like behavior and cognitive performance were evaluated on days 23 and 27 , respectively . At the end of the experimental sequence ( day 32 ) , mice were euthanized by cervical dislocation for sample collection . A second experiment was conducted to investigate the presence of generalized nociceptive sensitization . Nociceptive responses to hind paw mechanical stimulation were assessed before ( day −4 ) and 16 days after surgery . In parallel , mechanical sensitivity of the caudal abdominal area was evaluated on days −2 and 14 after surgery . An additional evaluation of nocifensive behaviors to abdominal mechanical stimulation was performed on day 14 . A third experiment was conducted to obtain the ED50 of acute THC administration for the alleviation of mechanical hypersensitivity . Endometriosis and sham mice were tested in the von Frey assay after administration of different doses of THC ( 1 . 25 , 2 , 2 . 5 and 5 mg/kg ) or vehicle . Measurements were done 45 min after subcutaneous administration of THC or vehicle at time points in which endometriotic lesions and hypersensitivity in the caudal abdomen were fully developed ( days 33–41 ) . The effects of chronic THC or vehicle were evaluated in endometriosis and sham mice in a fourth experiment . Chronic treatment with THC ( 2 mg/kg ) or vehicle administered once a day ( 9 AM ) started on day 1 after surgery and lasted until day 32 . Behavioral measures were conducted as in the first experiment . Mice were tested on the nociceptive paradigm 45 min after drug or vehicle administration and on the anxiety-like and memory tests 6 hr after administration . Mice were euthanized on day 32 by cervical dislocation for sample collection . A fifth experiment with 4 sets of mice was conducted to investigate THC tolerance development once the pain symptomatology was established . One of the groups underwent a sham surgery and the other three received endometrial implants . The sham group and one of the endometriosis groups received vehicle from day 1 to 16; one of the endometriosis groups received vehicle for 13 days and on day 15 , and acute doses of THC ( 2 mg/kg ) on days 14 and 16; the last endometriosis group received a repeated treatment with a daily administration of THC ( 2 mg/kg ) from day 7 to 16 . All mice were tested for mechanical sensitivity in the caudal abdominal area and the hind paw 45 min after drug or vehicle administration on days −2 , 7 and 14 ( caudal abdomen ) , and −4 and 16 ( hind paw ) , respectively . The effects of THC on nocifensive behavior were measured on day 14 . Mechanical sensitivity was quantified by measuring the responses to von Frey filament stimulation of the caudal abdominal area or the right hind paw . Von Frey filaments ( 1 . 65 , 2 . 36 , 2 . 44 , 2 . 83 , 3 . 22 and 3 . 61 corresponding to 0 . 008 , 0 . 02 , 0 . 04 , 0 . 07 , 0 . 16 and 0 . 4 g; Bioseb , Pinellas Park , FL , USA ) were applied in increasing order of force , 10 times each , for 1–2 s , with an inter-stimulus interval of 5–10 s . Abrupt retraction of abdomen , immediate licking , jumping and scratching of the site of application were considered positive responses in the evaluation of abdominal mechanical sensitivity . Paw withdrawal , shaking or licking was considered a positive response in the evaluation of paw mechanical sensitivity . The area under the curve ( AUC ) was calculated by applying the linear trapezoidal rule to the plots representing the frequency of response versus the numbers of von Frey filaments , which represent the logarithm of the filament force expressed in mg x 10 . Unpleasantness of pain in response to a mechanical stimulus was measured as previously described ( Corder et al . , 2019; Corder et al . , 2017 ) with minor modifications . Briefly , this parameter was evaluated using a single application of the von Frey filament 4 . 08 ( corresponding to 1 g ) against the caudal abdominal area shown in Figure 1—figure supplement 1b . The time spent protecting the area by guarding or seeking escape during the following 30 s was considered nocifensive behavior . The elevated plus maze test was used to evaluate anxiety-like behavior in a black Plexiglas apparatus consisting of 4 arms ( 29 cm long x 5 cm wide ) , 2 open and 2 closed , set in cross from a neutral central square ( 5 × 5 cm ) elevated 40 cm above the floor . Light intensity in the open and closed arms was 45 and 5 lux , respectively . Mice were placed in the central square facing one of the open arms and tested for 5 min . The percentages of time and entries to the open arms were determined as 100 x ( time or entries to open arms ) / ( time or entries to open arms + time or entries to closed arms ) as a measure of anxiety-like behavior . The novel object recognition task was assayed in a V-shaped maze to measure cognitive performance ( Puighermanal et al . , 2009 ) . On the first day , mice were habituated for 9 min to the maze . On the second day , mice were placed again in the maze for 9 min and two identical objects were presented at the ends of the arms of the maze . Twenty-four h later , one of the familiar objects was replaced with a novel one and mice were placed back in the maze for 9 min . The time spent exploring each object ( novel and familiar ) was recorded and a discrimination index ( DI ) was calculated as the difference between the time spent exploring the novel and the familiar object , divided by the total time exploring the two objects . A threshold of 10 s of total interaction with the objects was set to discard low levels of general activity . Endometriotic lesions , uterine horns and ovaries were harvested from each mouse and fixed in 4% paraformaldehyde in phosphate buffered saline ( PBS ) for 4 hr and cryoprotected in 30% sucrose with 0 . 1% sodium azide for 6 days at 4°C . Samples were then embedded in molds filled with optimal cutting temperature compound ( 4583 , Sakura Finetek Europe B . V . , Alphen aan den Rijn , The Netherlands ) and stored at −80°C until use . Endometriotic lesions and uteri were serially sectioned at 20 μm with a cryostat ( CM3050 , Leica Biosystems , Nussloch , Germany ) , mounted onto gelatinized slides and stored at −20°C until use . Sections of endometriotic lesions and uteri were stained with hematoxylin and eosin and observed under a Macro Zoom Fluorescence Microscope ( MVX10 , Olympus , Tokyo , Japan ) for assessment of diameter and histological features . Cyst sections were blocked and permeabilized with 3% normal donkey serum in PBS with 0 . 3% Triton X-100 for 2 hr and incubated overnight with rabbit anti-beta-III tubulin antibody ( ab18207 , 1:2000 , Abcam , Cambridge , United Kingdom ) in 3% normal donkey serum in PBS with 0 . 3% Triton X-100 at 4°C . After washing with PBS , sections were incubated for 1 hr at room temperature with anti-rabbit Alexa Fluor A488 antibody ( A21206 , 1:1000 , Thermo Fisher Scientific , Waltham , MA , USA ) . Slides were washed with PBS and coverslipped with DAPI Fluoromount-G ( 0100–20 , SouthernBiotech , Birmingham , AL , USA ) mounting media . Uterine sections were blocked and permeabilized with 5% normal goat serum in PBS with 0 . 3% Triton X-100 for 2 hr and incubated overnight with rabbit anti-beta-III tubulin antibody ( ab18207 , 1:2000 , Abcam ) in 5% normal goat serum in PBS with 0 . 3% Triton X-100 at 4°C . After washing with PBS , sections were incubated for 1 hr at room temperature with anti-rabbit Alexa Fluor A555 antibody ( ab150078 , 1:1000 , Abcam , Cambridge , United Kingdom ) . Slides were washed with PBS and coverslipped with DAPI Fluoromount-G ( 0100–20 , SouthernBiotech ) . Images of immunostained sections of cysts and uteri were captured with the X2 objective of a Macro Zoom Fluorescence Microscope ( MVX10 , Olympus , Shinjuku , Tokyo , Japan ) and processed and quantified using the NIH Image J software . An observer who was blinded to treatment group assignment converted from 4 to 8 images per animal into negative black-and-white images and the threshold was manually adjusted . Images were then dilated , skeletonized and the mean percentage of immunoreactive area was obtained by running the ‘Analyze particles’ function . Plasma samples were collected the day of euthanasia in tubes containing calcium EDTA . 17β-estradiol levels were determined with an enzyme-linked immunosorbent assay - ELISA ( ES180S-100 , Calbiotech , El Cajon , CA , USA ) according to manufacturer instructions . Sections of ovaries were stained with hematoxylin and eosin and observed under an upright microscope ( DM6000 B , Leica Biosystems ) . The number of pre-antral and antral follicles was determined in every nine sections . Only follicles containing an oocyte were counted and the total number of follicles was estimated by multiplying the raw counts by nine according to published criteria ( Myers et al . , 2004 ) . The number of corpora lutea was determined by direct counting of every 18 sections according to the average corpus luteum diameter ( Numazawa and Kawashima , 1982 ) . Data obtained with the nociception model were analyzed using one-way repeated measures ANOVA ( surgery as between‐subject factor ) , two-way repeated measures ANOVA ( surgery and treatment as between‐subject factors ) or mixed models ( surgery and treatment as between‐subject factors and time as within-subject factor ) whenever appropriate . Dose-response curve was fitted and ED50 determined using GraphPad Prism 8 ( San Diego , CA , USA ) . Data obtained with the elevated plus maze test , novel object recognition task , histology , immunostaining and ovarian follicle counting were analyzed using a Student t-test ( surgery ) or a two‐way ANOVA ( surgery and treatment ) . Post hoc Bonferroni analysis was performed after ANOVA when appropriate . The nonparametric Kruskal-Wallis test was used whenever data did not have a normal distribution or equal variances , followed by Mann Whitney U when appropriate . Correlation between variables was determined using the Pearson correlation coefficient . Data are expressed as individual data points and mean ± SEM , and statistical analyses were performed using IBM SPSS 23 software ( Chicago , IL , USA ) . The differences were considered statistically significant when the p value was below 0 . 05 .	Endometriosis is a common disease in women caused by tissue that lines the uterus growing outside the uterine cavity on to other organs in the pelvis . This can cause a variety of symptoms including chronic pelvic pain , infertility , and pain during menstruation or sexual intercourse . These symptoms may contribute to anxiety , depression , loss of working ability and a reduced quality of life . Currently available treatments for endometriosis , including hormonal therapy and surgery , have a limited effect and can produce unwanted side effects . For example , women who undergo surgery to remove the growths may experience post-surgical pain or a recurrence . As a result , women with endometriosis often rely on self-management strategies like dietary changes or exercise . Although cannabis consumption has a large number of potential side effects and can lead to substance abuse , it has been shown to provide pain relief in some conditions . But it is unknown whether it could be useful for treating endometriosis . Now , Escudero-Lara et al . have created a mouse model that mimics some of the conditions of human endometriosis: pelvic pain , anxiety and memory impairments . The mice were treated with moderate doses of Δ9-tetrahydrocannabinol ( THC ) , which is the main pain-relieving component of cannabis . The THC reduced pelvic pain and cognitive impairments in the mice with the endometriosis-like condition , but it had no effect on their anxious behavior . Escudero-Lara et al . also noticed that endometrial growths were also smaller in the treated mice indicating that THC may also inhibit endometriosis development . These experiments suggest that THC may be a useful treatment for patients with endometriosis . Clinical trials are already ongoing to test whether these findings translate to patients with the condition . Although THC and cannabis are readily available in some areas , Escudero-Lara et al . discourage using unregulated cannabis products due to the potential risks .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "short", "report", "medicine", "neuroscience" ]	2020	Disease-modifying effects of natural Δ9-tetrahydrocannabinol in endometriosis-associated pain
Many learned motor behaviors are acquired by comparing ongoing behavior with an internal representation of correct performance , rather than using an explicit external reward . For example , juvenile songbirds learn to sing by comparing their song with the memory of a tutor song . At present , the brain regions subserving song evaluation are not known . In this study , we report several findings suggesting that song evaluation involves an avian 'cortical' area previously shown to project to the dopaminergic midbrain and other downstream targets . We find that this ventral portion of the intermediate arcopallium ( AIV ) receives inputs from auditory cortical areas , and that lesions of AIV result in significant deficits in vocal learning . Additionally , AIV neurons exhibit fast responses to disruptive auditory feedback presented during singing , but not during nonsinging periods . Our findings suggest that auditory cortical areas may guide learning by transmitting song evaluation signals to the dopaminergic midbrain and/or other subcortical targets . Most human behaviors , such as speech , music , or athletic performance , are learned through a gradual process of trial and error . In all of these behaviors , the motor actions are shaped during learning by an internal model of good performance ( Wolpert et al . , 1995 ) . While a great deal has been learned about the neural mechanisms by which external rewards , such as food or juice drops , are represented in the brain and might shape future behavior ( Schultz et al . , 1997; Hikosaka , 2007 ) , little is known about how the brain evaluates its own performance , where such self-evaluation is computed and how these signals might shape future behavior . Here we use the songbird as a model system to understand how complex behaviors can be learned by self-evaluation of motor performance , rather than relying on explicit external rewards such as food or social reinforcement . Songbirds learn their vocalizations by storing a memory , or ‘template’ , of their tutor's song ( Doupe and Kuhl , 1999 ) . By listening to themselves sing and comparing their own song to this template ( Konishi , 1965 ) , they gradually converge to what can be a nearly exact copy of the tutor's song . Vocal learning requires a basal-ganglia forebrain circuit called the anterior forebrain pathway ( AFP ) , the output of which projects to cortical premotor vocal areas ( Nottebohm et al . , 1976; Bottjer et al . , 1984; Scharff and Nottebohm , 1991 ) . The AFP is crucial for driving learned changes in song and shaping plasticity in the song motor pathway ( Brainard and Doupe , 2000; Andalman and Fee , 2009; Warren et al . , 2011 ) . A key component of this vocal learning circuit is Area X , a basal ganglia homologue containing both striatal and pallidal components ( Person et al . , 2008 ) . Of critical importance in understanding the mechanisms of vocal sensorimotor learning in the songbird is to determine where song is evaluated and how the resulting evaluation signals are transmitted to the AFP to shape learning . Some hypotheses have emphasized the possible role of HVC ( used as a proper name ) , a premotor cortical circuit that receives auditory inputs and also projects to Area X ( Troyer and Doupe , 2000; Mooney , 2006; Prather et al . , 2008; Roberts et al . , 2012 ) . While the involvement of HVC in evaluating ongoing song remains an open question , several lines of evidence suggest that HVC does not transmit error-related signals to the AFP ( Hessler and Doupe , 1999; Kozhevnikov and Fee , 2007 ) , or receive auditory inputs ( Hamaguchi et al . , 2014 ) during singing . Another possibility is suggested by the large projection to Area X from dopaminergic nuclei in the midbrain—the ventral tegmental area ( VTA ) and substantia nigra pars compacta ( SNc ) ( Person et al . , 2008; Figure 1A ) . Inspired by the recently hypothesized role of dopaminergic signaling in reinforcement learning ( Houk et al . , 1994; Schultz et al . , 1997; Bayer and Glimcher , 2005; Tsai et al . , 2009 ) , we and others have proposed that this dopaminergic input to Area X may play a role in song learning by signaling vocal errors ( Hara et al . , 2007; Gale et al . , 2008; Gale and Perkel , 2010; Fee and Goldberg , 2011 ) . Notably , a recent anatomical study in songbirds has described a projection to the dopaminergic midbrain from the arcopallium ( Gale et al . , 2008 ) , an avian cortical region containing subtelencephalically projecting neurons analogous to those in deep layer-5 of mammalian cortex ( Dugas-Ford et al . , 2012; Karten , 2013 ) . 10 . 7554/eLife . 02152 . 003Figure 1 . Characterization of avian ‘cortical’ areas projecting to the dopaminergic midbrain . ( A ) Left panel: schematic of the songbird brain ( in sagittal view ) showing classical song control brain areas . Nuclei HVC and RA form the descending song motor pathway , necessary for song production . Nucleus RA is in the arcopallium , a cortical-like region homologous to layer-5 neurons of the mammalian cortex . Also shown is the anterior forebrain pathway ( AFP ) , a circuit necessary for vocal learning but not vocal production . The AFP consists of a basal ganglia homologue Area X , thalamic nucleus DLM , and cortical-like nucleus LMAN , which projects to RA . Right panel: schematic showing a set of pathways that are the focus of this paper , including a previously described projection to Area X from neurons in the dopaminergic midbrain nuclei VTA and SNc ( Gale et al . , 2008 ) . Also shown is AIV , a part of the intermediate arcopallium found to project to VTA and SNc . Projections to AIV from auditory cortical areas L1 , CM , and HVC-shelf are elucidated here . ( B ) Sagittal section showing the injection site of a retrograde tracer ( CTB , white color , arrow ) within VTA/SNc ( TH-stained neurons , red ) . ( C ) Three sagittal sections through the arcopallium showing retrogradely labeled neurons in AIV ( fluorescence , white ) and relation to RA ( dark field image , purple ) . Panel at left is most lateral; sections 200 µm apart . ( D ) Upper panel: image of injection site of a GFP-expressing virus ( HSV , fluorescence white ) in the anterior part of AIV , showing relation to RA ( dark field image ) . Bottom panel: sagittal section of VTA/SNc showing anterogradely labeled fibers from AIV ( green ) and TH-stained neurons ( red ) . ( E ) Same as D , with an injection site in AIV ventral to RA . All scale bars 200 µm unless indicated otherwise . In panels A–E , anterior is left; dorsal is up . ( F ) Coronal section showing axons in RA and Ad ( green ) , anterogradely labeled from LMAN and LMAN-shell , respectively ( Dextran 10 MW Alexa 488 ) . AIV neurons ( white ) labeled by injection of a retrograde tracer ( CTB ) in VTA and SNc ( medial , left; dorsal , up ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 00310 . 7554/eLife . 02152 . 004Figure 1—figure supplement 1 . Retrograde labeling of AIV neurons from VTA and SNc . ( A ) Site of CTB injection ( white , 13 . 8 nl total injected volume ) into VTA ( tyrosine hydroxylase immunostain , red ) . ( B ) Serial sagittal sections through the arcopallium showing retrogradely labeled neurons ( white ) in the vicinity of RA . ( Most lateral image is #1 , 100 µm between slices ) . ( C and D ) Site of CTB injection into SNc and retrograde label ( white ) in the arcopallium . ( E–H ) In each pair , the image on the left shows the injection site in VTA/SNc . The image on the right shows a single sagittal section through the arcopallium near the lateral third of RA ( corresponding to sections 2 or 3 in panels B and D ) . We refer collectively to the regions containing neurons retrogradely labeled from the dopaminergic midbrain ( VTA/SNc ) as the ventral intermediate arcopallium ( AIV ) . Scale bars for all images are 200 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 004 Here we set out to examine the role of the cortical area projecting to VTA/SNc in song learning . First , we characterized the spatial extent of the arcopallial neurons projecting to VTA or SNc , and characterized the inputs to these neurons from auditory cortical fields . We then carried out lesions targeted to this area to examine the effect on tutor imitation and on vocal plasticity that occurs after deafening . Finally , we recorded from VTA/SNc-projecting neurons in the arcopallium of singing birds and observed fast transient responses to disruptive auditory feedback during singing . Altogether , our findings suggest that descending auditory cortical projections may play a role in song evaluation via projections to the dopaminergic midbrain and/or other subcortical targets . To elucidate the spatial pattern of cortical neurons projecting to the dopaminergic midbrain in songbirds , we made small injections of a retrograde tracer ( cholera toxin subunit β , CTB ) into VTA and SNc ( n = 12 zebra finches , Figure 1B , see ‘Materials and methods’ ) . Consistent with an earlier report ( Gale et al . , 2008 ) , we observed a distinct mass of retrogradely labeled cell bodies within the ventral-most extent of the intermediate arcopallium . Labeled neurons were distributed in a complex pattern that was highly consistent across birds ( Figure 1C , Figure 1—figure supplement 1 ) . The main cluster of labeled neurons was arranged in a stem-like column ventral to RA ( robust nucleus of the arcopallium ) , while a distinct ‘stripe’ of labeled neurons was observed 300–400 µm anterior to the main cluster . A smaller number of labeled neurons formed a thin shell around the dorsal surface of RA , as previously described ( Gale et al . , 2008 ) . We will refer specifically and collectively to the arcopallial areas containing neurons retrogradely labeled from the dopaminergic midbrain , described above , as the ventral intermediate arcopallium ( AIV ) . Injections of anterograde tracer ( HSV viral vector expressing green fluorescent protein , GFP ) within the anterior and posterior clusters of AIV neurons each revealed axonal arborization intermingled with TH-positive neurons in both VTA and SNc ( Figure 1D , E , n = 6 birds ) . It is important to elucidate the anatomical relation between the area we identify as AIV and the dorsal arcopallium ( Ad ) , which has previously been implicated in vocal learning ( Bottjer and Altenau , 2010 ) , and also in locomotory control ( Feenders et al . , 2008; Jarvis et al . , 2013 ) ( called lateral intermediate arcopallium , LAI , in the latter studies ) . Ad was labeled by injections of anterograde tracer ( Alexa 488 conjugated dextran 10 MW ) into the nidopallium lateral to LMAN ( LMAN-shell ) , and AIV was retrogradely labeled in the same birds by injection of CTB into VTA and SNc ( n = 8 hemispheres ) . In coronally sectioned tissue , Ad was seen as a band of labeled fibers extending lateral to RA ( Figure 1F ) , as previously described . Neurons retrogradely labeled from VTA/SNc were visible as a wedge of cell bodies ventrolateral to RA and ventromedial to Ad . Few labeled cell bodies were observed within the region of labeled fibers in Ad , although some retrogradely labeled neurons were seen in the thin ‘neck’ of labeled fibers connecting RA and Ad . Injections of retrograde tracer ( CTB ) into the anterior or posterior parts of AIV ( n = 5 ) revealed no retrogradely labeled neurons in the nidopallium adjacent to LMAN ( LMAN-shell ) . Thus , AIV and Ad appeared to be largely distinct non-overlapping regions . To further elucidate the possible cortical afferents to AIV , we made injections of retrograde tracer ( CTB ) into different locations anterior and ventral to RA ( n = 10 birds ) . Following injections into the anterior parts of AIV ( n = 5 birds ) , retrogradely labeled neurons were reliably observed in auditory cortical areas HVC-shelf , caudal mesopallium ( CM ) and L1 ( Figure 2A–F , Figure 2—figure supplement 1A–C ) ( Kelley and Nottebohm , 1979; Vates et al . , 1996; Mello et al . , 1998 ) . We note that , while these previous studies have identified inputs to ventral arcopallium ( RA-cup ) from L1 and HVC-shelf , the projection described here from CM is novel . Following injections into the more posterior parts of AIV ( n = 5 birds ) , retrogradely labeled neurons were observed in the caudal nidopallium ( NC ) posterior to HVC-shelf ( Figure 2G , H ) . 10 . 7554/eLife . 02152 . 005Figure 2 . Cortical inputs to ventral intermediate arcopallium ( AIV ) : retrograde and anterograde tracing . ( A ) Injection of a retrograde tracer ( CTB ) anterior to RA in the vicinity of AIV . ( B ) Retrograde label in cortical auditory areas L1 and CM resulting from the injection shown above . ( C ) Labeled neurons in HVC-shelf of the same bird . ( D–F ) Same as panels A–C . Injection of CTB was targeted to the anterior ‘stripe’ part of AIV . ( G ) Injection of CTB into the posterior part of AIV , directly ventral to RA . ( H ) Retrogradely labeled neurons in caudal nidopallium ( NC ) resulting from the injection shown above . ( I–L ) Anterograde tracing from auditory cortical areas L1 , CM , and HVC-shelf . ( I ) Upper panel—injection of a GFP-expressing virus ( HSV , green ) into auditory field L1 . Middle panel shows labeled axons in the vicinity of AIV neurons retrogradely labeled ( purple ) from VTA/SNc . Color in RA is due to auto fluorescence , not label . Bottom panel shows higher magnification view from image above . ( J ) Same as ( I ) , but the injection of HSV-GFP was made into the auditory caudal mesopallium ( CM ) . ( K ) Same as ( I ) , but the injection of HSV-GFP was made into HVC-shelf . Scale bars: 100 µm for lower panels of I–L; 200 µm for all other panels . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 00510 . 7554/eLife . 02152 . 006Figure 2—figure supplement 1 . Cortical inputs to ventral intermediate arcopallium ( AIV ) . ( A ) Injection site of CTB in anterior AIV ( see Figure 1—figure supplement 1B , D panels 2–4 ) . ( B ) Retrogradely labeled cell bodies in L1 and CM resulting from the injection shown in panel A . Retrograde label was also observed in L3 following injections of tracer anterior to RA . But the presence of labeled neurons in L3 was less reliable than in L1 and CM . Further studies will be required to resolve the connectivity between L3 and AIV . ( C ) Retrogradely labeled cell bodies in posterior HVC-shelf resulting from the injection shown in panel A . ( D–H ) Upper panels—injection of a GFP-expressing virus ( HSV , green ) into L1 ( D ) , CM ( E ) , HVC-shelf ( F ) and NC ( G and H ) . Middle panels show labeled fibers arborizing in AIV in the vicinity of neurons retrogradely labeled from VTA/SNc ( purple ) . Bottom panels show a higher-magnification view of the sections shown in the middle panels . Scale bars: 100 µm for the bottom images in panels D–I; 200 µm for all other images . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 006 To confirm that neurons in these four areas terminate within AIV , we carried out anterograde tracing using a GFP-expressing virus ( HSV-GFP ) . Injections of anterograde tracer into the caudal nidopallium ( NC ) posterior to HVC-shelf revealed axonal arborization primarily in the posterior ‘stem’ region of AIV ( Figure 2—figure supplement 1G–H , n = 8 birds ) . Injections of anterograde tracer into auditory cortical areas L1 , CM , and HVC-shelf all revealed axonal arborization overlapping with neurons retrogradely labeled from VTA/SNc , particularly within the anterior ‘stripe’ region of AIV ( Figure 2I–L , Figure 2—figure supplement 1D–F; L1 and CM , n = 4 birds each; HVC-shelf , n = 10 birds ) . In these experiments , axonal arbors were not restricted to the region of retrogradely labeled neurons in AIV , but were also seen anterior to the AIV ‘stripe’ and in the area directly anterior to RA . The functional connectivity of projections from the auditory areas L1 , CM , and HVC-shelf was further confirmed by electrically stimulating these areas and demonstrating short latency activation of multi-unit recording sites in AIV and of antidromically identified VTA/SNc-projecting AIV neurons ( Figure 3 ) . Together these findings suggest the existence of multiple descending pathways by which auditory information could reach the dopaminergic midbrain and other downstream auditory targets of the arcopallium ( Kelley and Nottebohm , 1979; Vates et al . , 1996; Mello et al . , 1998 ) . 10 . 7554/eLife . 02152 . 007Figure 3 . Electrophysiological verification of functional connectivity . Electrical stimulation in auditory cortical areas L1 , CM , and HVC-shelf drives spiking in VTA/SNC-projecting neurons in AIV . ( A ) Schematic at left illustrates the location of the stimulating electrodes ( red ) in L1 and the location of the recording electrode ( blue ) in AIV . Middle traces: responses of 3 antidromically identified AIV neurons to L1 stimulation ( overlayed responses from ten trials ) . Right traces: antidromic response of the same neurons to electrical stimulation in VTA/SNc . ( B and C ) The panels are analogous to those shown in ( A ) , except the stimulation electrode is placed in caudal mesopallium ( CM ) or in the posterior part of HVC-shelf . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 007 To test the role of AIV in vocal learning and imitation , we carried out excitotoxic lesions targeted to AIV in young zebra finches , and examined the subsequent effect on song imitation . Birds were lesioned after being tutored in their home cage but prior to evidence of vocal imitation ( Figure 4A , B , 48–51 days post hatch , dph ) . After the lesion surgery ( ‘Materials and methods’ ) , birds were maintained in isolation and their songs were recorded until adulthood ( 90 dph ) . Control siblings underwent the same tutoring protocol but did not receive AIV lesions . A quantification of song imitation ( by analysis of pupil-tutor similarity , see ‘Materials and methods’ ) revealed that the AIV-lesioned birds exhibited deficits in vocal learning ( Figure 4C–F ) . Lesioned birds had a significantly lower similarity to their tutor ( imitation score ) than their unlesioned control siblings ( Figure 4—figure supplement 1A , paired t test , p=0 . 0048 , rank-sum test between all controls and all AIV-lesioned birds p=0 . 0024 ) . The overall distribution of acoustic features ( Tchernichovski et al . , 2000 ) was not significantly different between AIV-lesioned and control groups , suggesting that these lesions did not create gross abnormalities in song production ( Figure 4—figure supplement 1B–D ) . 10 . 7554/eLife . 02152 . 008Figure 4 . Bilateral lesions of AIV impair tutor imitation . ( A ) Excitotoxic lesion targeted to AIV by injection of NMA . Lesion borders indicated by arrowheads ( sagittal section; anterior left , dorsal up ) . ( B ) Schematic timeline of experimental protocol . ( C ) Song spectrograms of an adult bird ( 90 dph ) that underwent bilateral lesion of AIV as a juvenile , compared to an unlesioned control sibling . Top: spectrogram of tutor song . Middle: two example song spectrograms of the lesioned bird ( 45% lesion , 0 . 205 imitation score ) . Bottom: two example song spectrograms of the control sibling ( 0 . 235 imitation score ) . ( D ) Same as panel C , but for a different pair of birds ( lesioned bird , 66% lesion , 0 . 1566 imitation score; control sibling , 0 . 27 imitation score ) . ( E and F ) Song spectrograms of two additional adult birds that underwent lesions of AIV as juveniles . These birds did not have control siblings . ( Bird in panel E , 61% lesion , 0 . 097 imitation score; bird in panel F , 77% lesion , 0 . 114 imitation score . ) DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 00810 . 7554/eLife . 02152 . 009Figure 4—figure supplement 1 . Effect of AIV lesions on song motor production and imitation . ( A ) AIV-lesioned birds showed significantly reduced tutor imitation score compared to unlesioned controls . Lines connect data from siblings . ( B ) Average distribution of song spectral features ( FM , Weiner entropy , pitch goodness , and pitch ) of AIV lesioned birds ( black ) and unlesioned controls ( blue ) at 90 dph . ( C and D ) AIV lesions produce no immediate effect on juvenile songs . ( C ) Song example of a juvenile bird just prior to AIV-lesion ( 50 dph; top two spectrograms ) and in the first day of singing after AIV lesion ( 53 dph; bottom two spectrograms ) . This bird had a 55% AIV-lesion and ultimately exhibited a severe deficit in imitation ( tutor imitation score 0 . 094 at 90 dph ) . ( D ) Distributions of several acoustic features prior the AIV lesion ( blue ) and after AIV lesion ( red ) , for the bird shown in panel C . Numbers in each subplot show the cross correlation between the feature distributions before and after lesion . Across all AIV-lesioned birds ( n = 17 birds ) the correlation coefficients between pre- and post-lesion distributions were 0 . 9837 ± 0 . 03 , 0 . 944 ± 0 . 06 , 0 . 988 ± 0 . 13 , and 0 . 96 ± 0 . 03 for FM , Weiner entropy , pitch goodness , and pitch , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 009 We investigated the possibility that the observed learning deficits after AIV lesions might be caused by the loss of neurons in dorsal arcopallium ( Ad ) , which is dorsolaterally adjacent to AIV . In a subset of juvenile birds ( n = 7 ) , lesions were targeted to the portion of Ad most likely affected by our AIV lesion procedure ( Figure 5A , ‘Materials and methods’ ) . These birds underwent the same tutor exposure and learning protocol as did the AIV-lesioned birds described above . The Ad-lesioned birds exhibited significantly higher song imitation scores compared to AIV-lesioned birds ( Wilcoxon rank-sum test p=0 . 012 ) , and no significant deficit in song imitation compared to non-lesioned controls ( Figure 5B , Wilcoxon rank-sum test p=0 . 74 ) . These birds exhibited no overt motor or locomotor deficits . 10 . 7554/eLife . 02152 . 010Figure 5 . AIV lesions , but not Ad lesions , impair tutor imitation . ( A ) Left: coronal section showing the relation between AIV neurons ( retrogradely labeled from VTA/SNc , blue ) and nucleus Ad ( anterograde labeling from LMAN-shell , green; medial , left; dorsal , up ) Right: coronal section showing excitotoxic lesion of Ad , revealed by loss of NeuN staining ( white , note that the lesion affected some of the overlying nidopallium ) . ( B ) Tutor imitation score plotted as a function of the extent of AIV lesion for each bird ( hollow circles , n = 17 AIV-lesioned birds ) . Ad-lesioned birds ( n = 7 ) are shown as 0% lesion ( hollow gray diamonds ) since Ad lesions had minimal impact on AIV . No significant impairment of song-imitation was observed in Ad-lesioned birds as compared to unlesioned controls ( blue diamonds , n = 19 birds ) . Solid line denotes least square fit to Ad-lesioned and AIV lesioned data points . Red dashed horizontal line indicates mean of all similarity comparisons between 20 unrelated adult birds ( red shaded area indicates SEM ) ( C ) Boxplot of the distribution of imitation scores of all control birds ( cyan , unlesioned , and Ad-lesioned controls ) and birds with large AIV-lesions ( black , >50% lesion ) . Also shown is a boxplot of the distribution of similarity scores of all pairwise comparisons between 20 unrelated adult birds ( red ) . Whiskers denote 10–90 percentile . Asterisk in each boxplot denotes the mean , heavy line denotes median . ( D ) Distribution of syllable self-similarity in AIV-lesioned birds ( top ) and in control birds ( bottom , unlesioned , and Ad-lesioned controls combined ) . Dashed lines denote mean of distributions . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 01010 . 7554/eLife . 02152 . 011Figure 5—figure supplement 1 . Effect of AIV lesions on song imitation , development of song stereotypy , and rate of singing . The tutor imitation scores shown in main Figure 5B are based on a product of the acoustic similarity score the sequence similarity score ( Mandelblat-Cerf and Fee , 2014 ) . Here the contribution of acoustic and sequence similarity are shown separately in panels A and B respectively . ( C ) Plot of maturity index ( a measure of song stereotypy; see ‘Note on methods’ below ) as a function of AIV lesion size . Larger AIV lesions in juvenile birds resulted in lower stereotypy in the final adult song . In panels A–C , dashed line denotes averaged scores of each metric for comparisons between unrelated tutors . Solid lines denote linear regression using least square . R-values are 0 . 69 , 0 . 46 , and 0 . 50 respectively . F-statistic p-values are noted . The song maturity index of AIV-lesioned birds is significantly smaller than that of control birds ( unlesioned and Ad-lesioned birds; Wilcoxon rank-sum test , p=0 . 01 and p=0 . 025 , respectively ) . ( D–E ) To test whether the deficits in learning ( tutor imitation scores ) after AIV lesion are due to a reduced rate of singing ( i . e . , less practice ) , we quantified the amount of singing in AIV-lesioned birds and Ad-lesioned controls . ( D ) Total seconds of singing per day for Ad-lesioned birds ( average of 7 birds , red ) and AIV-lesioned birds ( average of 17 birds , black ) . There was no significant difference in the accumulated amount of singing ( t test , p=0 . 52 ) , nor in the daily singing rate , between these populations , as tested every 5 days ( t tests , p-values of all comparisons , 0 . 8>p>0 . 22 ) . ( E ) Tutor imitation scores are not correlated with the amount of singing both for AIV-lesioned ( upper panel , black ) , Ad-lesioned ( upper panel , red ) , and unlesioned control birds ( bottom panel , blue ) ( F-statistics all comparisons , p>0 . 4 ) . Note on methods: our song similarity algorithm compares many renditions of the pupil song with the tutor motif . This comparison results in an acoustic similarity score and a sequence similarity score . Acoustic similarity estimates the similarity of each of the tutor's syllables to the best match in the pupil's song . The sequence similarity score represents the acoustic similarity of song segments preceding and following these closest-match syllables . The overall tutor imitation score is the product of the acoustic and sequence similarity scores . Song stereotypy ( panel C ) was quantified by measuring the peak of the spectrogram cross-correlation between different bouts of the birds song with itself ( Aronov et al . , 2008 ) . This measure was shown to gradually increase through song development , and therefore we refer to it as the maturity index . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 011 The lack of effect of Ad lesions , on either vocal learning or other aspects of motor behavior , led us to wonder whether complete lesions of Ad might have an observable effect . Larger lesions of Ad were carried out in three additional birds ( the lesions were extended 0 . 2 mm more laterally , but within Ad ) . All three birds exhibited severe akinesia and immobility requiring the termination of the experiment . Because these birds did not sing , we were not able to assess the effects of the larger Ad lesions on vocal learning . We wondered if the song imitation deficits produced by the AIV lesions described above were correlated with the size of the lesion . Several different lesion protocols were tested in this study , resulting in different extents of AIV lesion across the data set ( n = 17 birds total ) . In all lesioned birds , the extent to which AIV was lesioned was quantified histologically at the age 90 dph ( ‘Materials and methods’ ) . Because Ad lesions minimally impacted AIV , Ad-lesioned birds ( n = 7 ) were included in this analysis as sham lesions ( 0% AIV lesion ) . Song imitation scores were negatively correlated with the fraction of AIV that was lesioned ( Figure 5B , linear regression using least square , R = 0 . 63 , F-statistic p=0 . 0008 , Figure 5—figure supplement 1A , B ) . Birds for which we estimated that more than half of AIV was lesioned ( n = 8 birds , referred to as a large AIV lesion ) produced poor imitations of the tutor song compared to controls ( Figure 5C , rank-sum test , p=0 . 00016; Ad-lesioned and unlesioned control groups were pooled in this comparison ) . On average , birds with large AIV lesion exhibited an 87 ± 9% loss of imitation capacity compared to the controls ( ‘Materials and methods’ ) . The distribution of tutor imitation scores for birds with large lesions was not significantly different from the distribution of pairwise similarity scores between unrelated adult birds in our colony ( Figure 5C , Wilcoxon rank-sum test , p=0 . 31 ) . Altogether , our findings suggest that the songs of birds with large AIV lesions were nearly as dissimilar to their own tutors as unrelated adult birds in our colony are to each other . Importantly , the impaired vocal imitation in AIV-lesioned birds cannot be attributed to the amount of vocal practice . We compared the amount of singing in the period from surgery up to 90 dph for AIV-lesioned birds vs Ad-lesioned controls . No difference was observed between these groups ( Figure 5—figure supplement 1D , E , t test , p=0 . 56 ) . Not surprisingly , both Ad- and AIV-lesioned birds sang less than the unlesioned control group ( ranksum p=0 . 003 and 0 . 016 respectively ) . However , within any experimental or control group , there was no correlation between the amount of singing and the degree of song imitation ( F-statistics all comparisons , p>0 . 4 ) . In addition to impaired vocal imitation , AIV-lesioned birds developed adult songs that exhibited a significantly reduced overall song stereotypy ( p=0 . 0045 , rank-sum test , Figure 5—figure supplement 1C ) ( Aronov et al . , 2008 ) and syllable stereotypy ( p=0 . 01 , rank-sum test , Figure 5D ) , compared to the control groups ( Ad-lesioned and unlesioned control groups were individually significant , but were pooled in the comparisons stated above ) . Another common feature of the songs of AIV-lesioned birds was the presence of a large number of ‘atypical’ syllables that were uncharacteristically long in duration or contained patterns of acoustic modulation not usually observed in zebra finch song . Auditory feedback is necessary for vocal learning in juvenile birds , as revealed by experiments showing that birds deafened as juveniles are unable to learn normal songs ( Konishi , 1965 ) . Because AIV is a downstream target of auditory brain areas , we wondered if the effects of AIV lesions might be similar to deafening . We took advantage of the fact that deafening results in a dramatic degradation of song in young adults that have largely finished learning their songs ( Nordeen and Nordeen , 1992; Lombardino and Nottebohm , 2000 ) . For example , deafening at the age of 75–80 dph resulted in a near complete loss of song structure , typically within 1 week ( n = 14 birds , Figure 6A , D , see ‘Materials and methods’ for quantification ) . In contrast to the effect of deafening , bilateral lesions targeted to AIV ( n = 6 birds , 75–80 dph ) produced no significant song degradation within a 2-week period following the lesion ( t test p=0 . 28 , Figure 6B , D ) . 10 . 7554/eLife . 02152 . 012Figure 6 . Effect of AIV lesion on adult song and on song degradation after deafening . ( A ) Examples of song spectrograms from a bird deafened at the age 80 dph . Shown from top to bottom; song before surgery ( deafening ) , first song post-surgery , song 1 week post-surgery , and song 2 weeks post-surgery . Note rapid degradation of song structure within 1 week after deafening . ( B ) Examples of song spectrograms from a bird that underwent complete bilateral lesion of AIV at the age 80 dph . Note the lack of song degradation . ( C ) Song spectrograms from a bird that underwent both bilateral lesion of AIV and deafening at the age 80 dph . ( D ) Plot of song self-similarity , normalized to the average self-similarity of pre-surgery song . Note that lesioned birds exhibited a reduced rate of song degradation after deafening , compared to deafening alone , 1 week and 2 weeks post-surgery ( t test , p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 01210 . 7554/eLife . 02152 . 013Figure 6—figure supplement 1 . Immediate effects on song of excitotoxic and electrolytic lesions in AIV . ( A ) Excitotoxic lesions in AIV do not cause a decrease in song variability , as would be expected if the lesions impacted LMAN axons entering RA . Example song spectrograms prior to AIV lesion surgery ( top ) and the first day of singing after the surgery ( bottom ) . Data are from a bird that received bilateral lesion of AIV at the age 77 dph . ( B ) Quantification of song variability before and after AIV lesion in older juvenile birds ( 75–80 dph ) . Data are from the same birds shown in the main Figure 6 . Song variability was assessed by computing the variance of song self-similarity scores . ( C–F ) Partial electrolytic lesion anterior to RA produces an immediate effect on song acoustic structure , presumably due to the destruction of axons in the RA output tract . Such immediate effects on song structure were never observed after excitotoxic lesions in AIV , which also extended anterior to RA . ( C ) Sagittal section ( NeuN , white ) showing RA and the electrolytic lesion anterior to it ( red arrow ) . ( D ) Song spectograms prior to the electrolytic lesion ( upper panels ) and the first day of singing post lesion ( lower panels ) , for the same bird in panel C . ( E ) Plot of song self-similarity post-surgery , normalized to the average self-similarity of pre-surgery song ( 75–80dph ) . Post-electrolytic-lesion similarity scores ( n = 5 birds , red ) were significantly smaller than pre-lesion similarity scores ( ranksum test , p=0 . 008 ) , and significantly smaller than the normalized self-similarity scores of AIV-lesioned birds ( n = 6 birds , gray ) . ( ranksum test , p=0 . 0043 ) . ( F ) Plot of song self-similarity of AIV-lesioned birds ( gray ) and electrolytically lesioned birds for 2 weeks after surgery . Self-similarity of electrolytically lesioned birds showed an immediate drop after lesion that remained significantly lower than pre-surgery , and lower than AIV-lesioned birds , for at least 2 weeks ( all comparisons , p<0 . 01 ) . Note on methods: bilateral electrolytic lesions were carried out by passing cathodal current of 70 µA for 45 s at multiple sites in the arcopallium anteroventral to RA . The head was oriented stereotaxically with the flat anterior portion of the skull rotated forward at a pitch of 45° . RA was completely mapped based on its characteristic electrophysiological signature ( regular spiking interspersed with bursts ) . Once all of the borders of RA were localized , these were used to calculate the coordinates of the lesion sites . Two penetrations were made 200 µm anterior to the anterior edge of RA , spaced 300 µm from each other on the medial–lateral axis . For each penetration , two lesions were made in two depths: 200 µm above and below the bottom edge of RA . Songs were acquired for 2 weeks ( n = 3 birds ) or for 2 days ( n = 2 birds ) , after which birds were sacrificed and the brain analyzed histologically . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 013 Notably , AIV lesions had no immediate effect on song structure or song variability ( paired t test p=0 . 52 , Figure 6—figure supplement 1A , B ) , suggesting that AIV plays no direct role in song production or in the generation of vocal variability . These results also suggest that our AIV lesions did not significantly disrupt fibers of passage , either afferent fibers entering RA laterally from LMAN ( lateral magnocellular nucleus of the anterior nidopallium , an area which conveys variability into the motor output during singing [Kao et al . , 2005; Ölveczky et al . , 2005] ) , or efferent fibers exiting RA rostrally ( Figure 6—figure supplement 1C–F ) . It has been shown that lesions of the basal ganglia–forebrain learning circuit , AFP , largely prevent the degradation of song that occurs after deafening , leading to the view that such degradation is an active process requiring vocal learning circuitry ( Brainard and Doupe , 2000; Nordeen and Nordeen , 2010 ) . Since we have hypothesized that AIV is involved in vocal learning , we wondered whether lesions targeted to this area might slow the degradation of song after deafening . To test this idea , we carried out bilateral cochlear removal in combination with bilateral AIV lesions ( n = 15 , 75–80 dph ) . Songs were recorded for 2 weeks following this procedure and were analyzed for similarity to the pre-surgery song . Song degradation after deafening was significantly slower in AIV-lesioned birds than in birds with intact AIV ( Figure 6C , D , comparison at 7 days and 14 days post-lesion , t test p=0 . 0074 , 0 . 0096 respectively ) . The reduced rate of deafening-induced song degradation after AIV lesion suggests a role for descending auditory cortical pathways , including possibly those projecting to the dopaminergic midbrain , in mediating the plastic changes in the AFP and motor pathways that occurs after deafening . We next wanted to test if neurons in AIV—specifically those projecting to VTA/SNc—exhibit auditory responses during singing consistent with their hypothesized role in vocal learning . To address this question we used a motorized microdrive to record neural activity in AIV of singing zebra finches ( n = 7 birds , 90-120 dph ) . Recordings were targeted based on electrophysiological mapping of RA borders ( ‘Materials and methods’ ) . Recordings were made from 37 single units , of which 17 were antidromically identified and collision-tested VTA/SNc projectors ( Figure 7A , B ) . We also recorded 13 multiunit sites exhibiting robust antidromic response to VTA/SNc stimulation ( ‘Materials and methods’ , antidromic responses to VTA/SNc stimulation were restricted to the borders of AIV as defined by retrograde labeling , Figure 7—figure supplement 1 ) . AIV single-units discharged at low rates during singing ( 1–10 Hz , Figure 7C ) , and exhibited only small changes in average firing rate compared to non-singing ( Figure 7D , significant firing rate change in 15/37 neurons , paired t test p<0 . 05 ) . Only a small fraction of AIV neurons exhibited significant firing rate modulations locked to the song motif ( n = 4/37 ) . These modulations did not appear to be premotor in nature , and were substantially weaker than those reported in premotor song control nuclei ( Leonardo and Fee , 2005 ) , or auditory cortical areas ( Sen et al . , 2001 ) . 10 . 7554/eLife . 02152 . 014Figure 7 . AIV neurons projecting to VTA/SNc exhibit error-related auditory responses during singing . ( A ) Schematic diagram of recording setup showing stimulating electrode placed in VTA/SNc for antidromic identification of AIV neurons . ( B ) Voltage traces showing an antidromically evoked spike of a VTA/SNc-projecting AIV neuron ( red spikes ) . Collision test shown in black traces . ( C ) Distribution of firing rates of AIV neurons during singing . ( D ) Firing rates of AIV neurons during singing vs non-singing ( antidromically identified neurons , hollow circles; non-identified AIV neurons , filled circles ) . ( E ) Recording of an antidromically identified AIV neuron during singing ( song spectrogram , top; extracellular voltage trace , bottom ) with presentation of noise bursts ( red arrows ) . ( F ) Motif-aligned spike raster plot of the AIV neuron from panel E during singing with no noise bursts presented . Each row in the raster plot corresponds to a rendition of the song motif . Each dot corresponds to a spike . Along each row , gray areas denote syllables and white areas denote silent gaps . Two vertical red lines denote motif onset and offset . Shaded area along the histogram denotes SEM . Note lack of singing-related firing rate modulations . ( G ) Spike raster plot and spike histogram aligned to noise bursts during singing . Red vertical line denotes noise onset . Other notations are as in F . Note the brief response to the noise burst during singing . ( H–J ) Another example of an antidromically identified AIV neuron ( larger spike waveform ) recorded during singing . ( I ) Motif-aligned raster plot of the AIV neuron from panel H . ( J ) Spike raster plot and histogram aligned to noise bursts during singing . Rasters are sorted by the duration of the syllable in which the noise burst occurred . Note that the response to noise burst occurs during different syllable types . ( K ) Average peri-stimulus histogram for all antidromically identified AIV neurons , aligned to noise burst onset during singing . ( L ) Distribution of response latency after noise burst ( left ) and response duration ( right ) , for AIV neurons that showed a significant response to noise bursts . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 01410 . 7554/eLife . 02152 . 015Figure 7—figure supplement 1 . Simultaneous mapping of VTA/SNc-projecting neurons in AIV by antidromic activation and retrograde tracing . Does electrical stimulation in VTA/SNc antidromically activate only AIV neurons , or does it also activate other descending arcopallial pathways ? To assess this question , we carried out mapping of antidromically stimulated activity in ventral arcopallium and compared this , in the same animal , to the pattern of AIV neurons retrogradely labeled from VTA/SNc . ( A ) Sagittal section showing neurons in AIV retrogradely labeled by an injection of a CTB into VTA/SNc . Antidromic responses were measured along the electrode tracks indicated ( diagonal gray lines ) . Yellow circles indicate sites where statistically significant antidromic activation from electrical stimulation in VTA/SNc was observed . White circles indicate where no significant response was observed . Dashed line indicates the penetration along which reference electrolytic burn marks were made . ( B–D ) Example responses to VTA/SNc stimulation are shown at selected recording sites along each of three electrode tracks . Significant antidromic responses were observed only at locations where retrogradely labeled neurons were subsequently found . Specifically , there was no antidromic response seen in the gap between the main posterior mass of retrogradely neurons and the more anterior stripe . Nor were antidromic responses observed antero-ventral to the anterior stripe , within the borders of RA , nor dorsal to the thin rim of retrogradely labeled neurons along the dorsal border of RA . Note on methods: Retrograde tracer ( CTB ) was injected into VTA and SNc , and bipolar stimulating electrodes were implanted into the same region . The head was then oriented with the flat anterior portion of the skull rotated forward at a pitch of 70° . RA was completely mapped based on its characteristic electrophysiological signature ( regular spiking interspersed with bursts ) . Electrode penetrations ( Carbostar , Kation Scientific ) were made in a plane 200 µm medial from the lateral edge of RA and penetrations were made at several locations along the AP axis in this plane . For each penetration , recordings were made along a 1–1 . 5 mm range of DV coordinates spanning through the arcopallium . For each recording site the existence of antidromic responses was tested with single 0 . 2 ms monopolar pulses , current up to 350 µA and with both polarities of stimulation current . If no antidromic response was observed visually , responses were recorded using the maximum current ( 350 µA ) , otherwise , responses were recorded at 130% of threshold current . In each recording penetration , we noted the most ventral coordinate at which the characteristic spontaneous activity of RA was found . After recordings were complete , another electrode was used to make an electrolytic burn at known locations relative to the recording sites . Alignment of antidromic and retrograde maps . The brains were sectioned and stained for NeuN , and retrogradely labeled neurons were visualized . The AP coordinate of every recording site was determined with respect to the electrolytic burns . The DV coordinate of every recording site was determined by aligning the ventral border of RA as determined histologically ( NeuN stain ) and electrophysiologically ( spontaneous activity ) . Significance of an antidromic response: we analyzed the significance of antidromic responses as follows . For each recording site we computed the frequency spectrum of the averaged response to the stimulus ( 15 ms window , spanning 2–17 ms after stimulus onset ) . We then computed the power between 200 Hz and 2 KHz . A t test was used to test the hypothesis that the power in this frequency band was significantly higher than the power obtained from recordings with no antidromic response ( recorded at 1–2 mm dorsal to RA and inspected visually to verify they did not show antidromic response ) . Note that results were consistent for different frequency bands . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 015 While natural song learning proceeds slowly ( Tchernichovski et al . , 2001 ) , playback of disruptive auditory feedback , such as brief bursts of broadband noise , can induce rapid learned changes in song structure over the course of a few hours ( Tumer and Brainard , 2007; Andalman and Fee , 2009 ) . Such learning has been interpreted as evidence that distorted auditory feedback can be used to introduce experimentally controlled ‘errors’ in song performance ( Tumer and Brainard , 2007; Andalman and Fee , 2009; Fee and Goldberg , 2011 ) . More than 40% of the VTA/SNc-projecting neurons exhibited a significant neuronal response to distorted auditory feedback ( 50 ms noise bursts ) presented during singing ( Figure 7E–K , p<0 . 02 for n = 7/17 neurons , comparison of spike count in a 150 ms window before and after noise burst onset for each neuron; paired t test , p<10−5 for average response , bootstrap analysis , ‘Materials and methods’ ) . Significant responses were also observed at 54% of the multiunit recording sites ( n = 7/13 ) , and in 30% of AIV neurons that did not meet the criteria for antidromic identification ( n = 6/20 neurons ) . The response of AIV neurons to noise bursts was brief , occurring with an average latency of 23 ± 12 ms from noise onset and having an average duration of 90 ± 43 ms ( ‘Materials and methods’ , Figure 7L ) . There was no significant difference between the responses of antidromically identified neurons and those that were not antidromically identified . Thus , these data are pooled in the statistics on latency and duration . Are AIV neurons responsive to noise bursts only during singing , or can similar auditory responses be evoked during non-singing ? We recorded the responses of AIV neurons to noise bursts presented during playback of the bird's own song ( BOS ) in 25 identified VTA/SNc-projecting AIV neurons under anesthesia ( n = 3 birds ) , and another 28 such neurons in freely behaving birds ( n = 4 birds ) . Under these conditions , noise bursts elicited no significant response in any of these neurons ( Figure 8A , C , E , ‘Materials and methods’ ) . We also examined whether AIV neurons respond to noise bursts presented in a silent background ( non-singing , no BOS ) . In this case , auditory responses were observed in a small fraction of neurons in both awake and anesthetized birds: Altogether , 12/53 neurons exhibited a significant response within a 1 s window after noise onset ( bootstrap , p<0 . 02 , ‘Materials and methods’ ) . However , these responses were significantly slower and exhibited longer latency than during singing ( p<0 . 001 , Figure 8B , D , E , only one neuron responded significantly within 150 ms of noise onset ) . In comparison , all the neurons that were found to be noise-responsive during singing ( significant activity within a 1 s window ) , exhibited this response within 150 ms of the noise burst onset . Altogether , we have found that the auditory responsiveness of AIV neurons to disruptive auditory stimuli is highly state-dependent , exhibiting fast and robust firing-rate changes only during singing . 10 . 7554/eLife . 02152 . 016Figure 8 . Response of AIV neurons to noise bursts during non-singing . ( A ) Activity of an antidromically identified AIV neuron recorded during presentation of noise bursts during playback of birds own song ( BOS ) . Top to bottom panels: song spectrogram and simultaneous recording of neuronal activity; raster plot and spike histogram aligned to BOS onset ( black line ) . Time of noise bursts indicated by vertical red lines . Yellow band denotes period of BOS playback . Note lack of response to noise bursts . ( B ) Response of the same neuron in panel A to presentation of isolated noise bursts ( no BOS , no singing ) . Note slow timecourse of the response . ( C and D ) Recording of another antidromically identified AIV neuron ( notation same as in panels A and B ) . ( E ) Averaged PSTH for all VTA/SNc-projecting AIV neurons that exhibited a significant response to isolated noise bursts within 1 s after noise onset . Average response of these same neurons to noise bursts presented during BOS playback ( blue ) . In comparison , averaged response is shown for all VTA/SNc-projecting AIV neurons that responded significantly to noise burst during singing ( black trace ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02152 . 016 Motivated by the hypothesized role of dopaminergic signaling in reinforcement learning ( Houk et al . , 1994; Schultz et al . , 1997; Bayer and Glimcher , 2005; Tsai et al . , 2009 ) , we set out to examine the role in vocal learning of a recently discovered songbird cortical area ( Gale et al . , 2008 ) that projects to VTA and SNc . We have characterized the spatial extent of neurons within the intermediate arcopallium that project to VTA and SNc , and refer to the region of arcopallium retrogradely labeled from these midbrain dopaminergic areas as AIV . Using a combination of anatomical and electrophysiological techniques , we have elucidated the afferent inputs to AIV neurons from other cortical areas . We examined the effect on vocal learning of lesions targeted to AIV , and carried out electrophysiological recordings of AIV neurons . Our findings are broadly consistent with the hypothesis that an arcopallial region with descending auditory cortical projections , both to the dopaminergic midbrain and to midbrain and brainstem auditory centers , plays a role in vocal learning . The area we have identified as AIV is adjacent to other structures that have been hypothesized to play a role in vocal learning ( Bottjer and Altenau , 2010 ) . The dorsal arcopallium ( also referred to as the lateral intermediate arcopallium , LAI , by Jarvis et al . , 2013 ) , receives inputs from the nidopallium adjacent to LMAN ( LMAN-shell ) , and projects to the basal ganglia , motor thalamus , tectum , and the reticular formation ( Bottjer et al . , 2000 ) . Our anatomical findings suggest that AIV , while adjacent to Ad , is part of a distinct , largely non-overlapping , anatomical circuit . In contrast to Ad , AIV does not receive a projection from LMAN-shell; rather it receives inputs from auditory cortical areas ( CM , L1 , and HVC-shelf ) and caudal nidopallium . Furthermore , our findings show that AIV , but not Ad , projects to VTA/SNc . Given its anatomical overlap with RAcup , parts of AIV also likely project to MLD or to the shell of the thalamic nucleus ovoidalis ( Mello et al . , 1998 ) . Earlier studies have presented conflicting views on the function of Ad/LAI , particularly in regard to its role in vocal learning ( Bottjer et al . , 2000; Achiro and Bottjer , 2013 ) . Bottjer and Altenau ( 2010 ) describe evidence that lesions of Ad produce deficits in vocal learning , not unlike those we report here for lesions of AIV . In contrast , Feenders et al . showed immediate early gene activation in Ad/LAI and in LMAN-shell ( which projects to Ad/LAI ) during locomotory activity ( i . e . , hopping ) , but find no evidence for activation in these areas during singing . This led them to suggest that Ad/LAI and RA form the output of a general avian cortical motor circuit , of which RA is a component specialized for vocal production ( Feenders et al . , 2008 ) . Our findings are consistent with the view that AIV , but not Ad , play a role in vocal motor learning . We found that lesions of AIV caused deficits in vocal imitation , while lesions of Ad did not cause any such deficit , suggesting that the loss of song imitation following lesions targeted to AIV was not a secondary consequence of an unintended lesion in Ad . Our finding that larger lesions of Ad produced severe akinesia and immobility is consistent with the hypothesis that Ad plays a role in locomotion and other motor behaviors , rather than vocal learning ( Feenders et al . , 2008 ) . There are several potential explanations for the discrepancy between our findings on the effects of Ad lesions and those of Bottjer and Altenau ( 2010 ) . Clearly , the Ad lesion protocol of Bottjer and Altenau impacted a subset of neurons important for vocal learning , while our Ad lesion protocol did not . This subset of neurons , in principle , could be within Ad , within AIV , or could be in another , as yet undescribed , adjacent region important for vocal learning . In our view , the most parsimonious explanation for the detrimental effects on vocal learning is that the Ad lesions of Bottjer and Altenau had an unintended impact on neurons in AIV . More detailed electrophysiological studies of the pathways into and out of Ad will be needed to completely resolve these different views of its function . The effects of lesions directed to AIV in juvenile birds bear some resemblance to the effects of lesions in the AFP , a basal ganglia–thalamocortical circuit necessary for vocal learning . Lesions of Area X , the basal ganglia component of the AFP , in young birds have a substantial detrimental effect on song imitation and result in persistent song variability into adulthood ( Scharff and Nottebohm , 1991 ) . In contrast , lesions of Area X have relatively little immediate effect on song performance in adult and juvenile birds ( Goldberg and Fee , 2011; Kojima et al . , 2013 ) . Similarly , we found that birds in which AIV was lesioned at an early age produced a poor imitation of the tutor song and developed a less stereotyped song than did control birds , and AIV lesions in adult or juvenile birds have little immediate effect on vocal performance . Lesions of Area X also largely block the degradation of song following deafening ( Kojima et al . , 2013 ) , and similarly , but to a lesser extent , we found that lesions of AIV slow song degradation in deafened birds . The similar patterns of findings in AIV-lesioned birds and Area X-lesioned birds is consistent with the possibility that the AIV , or some components of it , may interact with the AFP , perhaps by transmitting to it a signal important for its function . In principle , there are several ways the loss of a descending auditory cortical projections might affect vocal learning . We first address the possible role of the projection from AIV to the dopaminergic midbrain in vocal learning . In the songbird , dopaminergic inputs to Area X have been hypothesized to carry signals related to motivational aspects of singing and to regulate vocal variability , in particular while switching between less variable song directed to a female bird to more variable song produced in social isolation ( Yanagihara and Hessler , 2006; Hara et al . , 2007; Leblois et al . , 2010 ) . Notably , we find that juvenile and young adult birds exhibit normal song variability after AIV lesions , suggesting that AIV may not be involved in the regulation of vocal variability , and that the deficits in vocal learning do not result from a loss of exploratory song variability . Alternatively , reduced glutamatergic drive after AIV lesion could result in reduced tonic dopaminergic input to Area X . In mammals , reduced dopaminergic input to the striatum can have deleterious effects on BG function , including reduced spontaneous activity and the loss of spines in the medium spiny neurons ( Kreitzer and Malenka , 2008 ) . Songbirds exhibit patterns of dopamine receptor expression in cortex and Area X ( Kubikova et al . , 2010 ) , and responses to dopamine stimulation ( Ding and Perkel , 2002 ) that parallel those found in mammals . Finally , another possibility is that impoverished vocal imitation after lesions of AIV results from the loss of a reinforcement signal that carries information about recent song performance . The short-latency response of AIV neurons to disrupted auditory feedback is consistent with the possibility that AIV may play a role in computing or transmitting a fast online signal to VTA/SNc . Previous studies have shown that neurons in the vicinity of RA , and potentially within AIV , have several sub-telencephalic targets other than VTA/SNc , including a higher-order auditory thalamic nucleus Ov-shell and parts of the auditory midbrain ( Vates et al . , 1996 ) . Thus , deficits in vocal learning following lesions targeted to AIV could potentially arise from loss of signaling in these other pathways . Further studies will be required to elucidate the relation between neurons projecting to these different downstream targets , and to determine the role of these different pathways in vocal learning . One key to understanding the function of AIV is to characterize its cortical afferents , and several candidates have been identified . Güntürkün et al . have described a projection in pigeons from the lateral caudal nidopallium ( NCL ) to the ventral arcopallium ( Kroner and Gunturkun , 1999 ) . The projection we describe here in the zebra finch—from the caudal nidopallium ( NC ) posterior to HVC—terminates primarily in the posterior ‘stem’ part of AIV . This projection may be related to the previously described nidopallium caudolaterale ( NCL ) projection , although NCL appears to be more lateral than NC . NCL was recently found to be involved in error and reward processing ( Starosta et al . , 2013 ) , and based on several lines of evidence , it has been suggested that NCL may be analogous to mammalian prefrontal cortex ( Kroner and Gunturkun , 1999 ) . Notably , in mammals the major cortical projection to VTA arises from prefrontal cortex , a key structure for cognitive control and goal-directed actions ( Murase et al . , 1993 ) . Previous studies in the zebra finch have also identified inputs to the intermediate arcopallium from auditory cortical regions , including primary auditory cortical field L1 and HVC-shelf in the nidopallium ( Kelley and Nottebohm , 1979; Vates et al . , 1996; Mello et al . , 1998 ) . These projections were found to terminate near RA , in a region termed RA-cup , and are likely homologous to projections in the pigeon from the dorsal nidopallium ( Nd ) to the ventromedial intermediate arcopallium ( AIvm ) ( Wild et al . , 1993 ) . We have confirmed the existence of these projections , and using anatomical tracing and electrophysiological techniques , we have demonstrated that these inputs directly or indirectly innervate AIV neurons projecting to the dopaminergic midbrain . Inputs from L1 and anterior parts of HVC shelf appear localized to the parts of AIV anteroventral to RA . Inputs from posterior HVC shelf appear to terminate in both the anterior and posterior parts of AIV , and show significant overlap with inputs from the more posterior caudal nidopallium described above . Our anatomical studies have also revealed a novel projection to AIV from the caudal mesopallium ( CM ) , an avian auditory area previously identified as homologous to higher-order auditory cortex ( Bauer et al . , 2008 ) or to upper layers of primary auditory cortex ( Karten , 2013 ) . The projection from CM terminates within AIV , and electrical stimulation in CM causes spiking activity in VTA/SNc-projecting AIV neurons , likely through orthodromic activation . The existence of a projection to AIV from CM and L1 may be important for two reasons . The cluster of neurons in CM retrogradely labeled from AIV was overlapped with the region of CM , also known as Avalanche ( Av ) , that forms bidirectional synaptic interactions with both HVC and NIf ( Bauer et al . , 2008; Akutagawa and Konishi , 2010 ) , two song control nuclei also involved in vocal learning ( Roberts et al . , 2012 ) . Furthermore , a small subpopulation of neurons in CM and L1 have been shown to respond to noise bursts presented during singing , but not to noise bursts presented during playback of the birds own song ( Keller and Hahnloser , 2009 ) . These responses , highly reminiscent of responses in AIV , have led to the suggestion that one computational function of CM and L1 may be to compare actual auditory feedback with the song template ( Keller and Hahnloser , 2009 ) . Further studies are required to determine if the hypothesized error-related responses of AIV neurons derive from activity in this population of CM and L1 neurons . One can speculate about the potential role for AIV in computing or transmitting a fast online signal to VTA/SNc that potentially carries information about recent song performance . Most VTA/SNc-projecting AIV neurons responded to disruptive auditory feedback with a latency of less than 25 ms and response duration of less than 100 ms . This response may have a sufficiently fast temporal resolution to mediate vocal learning , assuming that the synaptic learning rules in Area X and the motor pathway employ a short-term synaptic memory such as an eligibility trace ( Sutton and Barto , 1998; Fee and Goldberg , 2011; Redondo and Morris , 2011 ) . Such a fast dopaminergic reinforcement signal could , in principle , be used in Area X to correlate vocal performance with an efference copy of vocal motor commands ( Fee and Goldberg , 2011 ) and to drive corticostriatal plasticity at HVC inputs to medium spiny neurons ( MSNs ) . The MSNs could then act through their downstream pallidothalamic pathway to ( 1 ) bias the motor system in favor of vocal commands that previously led to better song performance and ( 2 ) drive plasticity in the motor pathway such that the juvenile song gradually converges to the desired vocal output ( Andalman and Fee , 2009; Fee and Goldberg , 2011; Warren et al . , 2011; Charlesworth et al . , 2012 ) . Additional studies would be required to establish the nature of auditory reinforcement signals in the pathway we have described here , or possibly through other downstream targets , and to establish a causal role for these circuits in vocal learning . Animal subjects were male zebra finches ( n = 110 ) ( 45–120 days post hatch , dph ) . Birds were obtained from the Massachusetts Institute of Technology zebra finch breeding facility ( Cambridge , Massachusetts ) . The care and experimental manipulation of the animals were carried out in accordance with guidelines of the National Institutes of Health and were reviewed and approved by the Massachusetts Institute of Technology Committee on Animal Care . Retrograde labeling of neurons in AIV was obtained by injecting fluorescently labeled cholera toxin β subunit ( Molecular Probes ) into the ventral tegmental area ( VTA ) and substantia nigra pars compacta ( SNc ) . VTA/SNc were localized stereotaxically with reference to the principal thalamic auditory relay nucleus Ovoidalis , as described below . We have developed a technique that allows us to target VTA and SNc with high reliability , based on the mapped location of the thalamic auditory relay nucleus Ovoidalis ( Ov ) . Specifically , the head was oriented with the flat anterior portion of the skull rotated forward at a pitch of 50° . Using an extracellular electrode ( Carbostar-1 , Kation Scientific , Minneapolis , MN ) , tilted in the LM direction 2° towards the midline , Ov was located and mapped based on its robust auditory responses . Injections were made into VTA relative to Ov as follows: 300 µm anterior from the center of Ov , 200 µm medial from the medial edge of Ov , and 1800 µm ventral to the middle of Ov in the DV direction . Injections into SNc were targeted 350 µm posterior and 350 µm lateral to the VTA coordinates . For electrophysiology experiments requiring antidromic stimulation in VTA/SNc , bipolar stimulating electrodes were placed such that , to the extent possible , one wire was in VTA and the other wire was in SNc . Bilateral lesions of AIV were carried out by injecting 2% N-Methyl-DL aspartic acid ( NMA , Sigma , St Louis , MO ) at multiple injection sites in ventral arcopallium around RA chosen to maximally lesion VTA/SNc-projecting neurons . All AIV lesions were carried out stereotaxically . The head was oriented with the flat anterior portion of the skull rotated forward at a pitch of 70° . Because a large portion of AIV is located ventral to RA , access to AIV was achieved by a laterally rotated penetration . The head was rotated 45° ( roll ) to the left for lesions of AIV in the right hemisphere , and rotated to the right for lesions of AIV in the left hemisphere . Using a carbon fiber electrode ( Mod #E1011-20; Carbostar-1 , Kation Scientific . ) RA was then completely mapped based on its characteristic electrophysiological signature ( regular spiking interspersed with bursts ) . Once all of the edges of RA were localized in the rotated frame , these were used to calculate the coordinates of the injection sites . In the injection coordinates given below ( ML , AP , DV ) , dimensions are in µm; positive numbers represent more lateral , more anterior , and more ventral displacements , respectively . The coordinates are specified relative to the lateral edge of RA in the medial–lateral direction , relative to the center of RA in the anterior–posterior direction , and relative to the most ventral edge of RA in the dorsal–ventral direction . Injections were made through a glass pipette using a digitally controlled injection system ( Nanoject , Drummond Scientific , Broomall , PA ) . Injections at each site were made in multiple boluses of 13 . 8 nl , which were spaced at 5-min intervals to avoid backflow along the pipette . AIV lesions ( n = 17 birds ) were created by up to seven different injection sites ( in five different penetrations ) at the coordinates specified below . The last number in each coordinate indicates the number of injection boluses made at that site: ( +200 , −400 , +200; 4 ) ; ( +300 , +0 , +200; 4 ) ; ( +200 , +300 , +0; 4 ) ; ( +0 , +700 , +0; 2 ) ; ( +0 , +700 , +300; 2 ) ; ( −300 , +900 , +0; 2 ) ; ( −300 , +900 , +300; 2 ) . Note that , given the complex spatial structure of AIV , it was difficult to achieve a complete lesion of VTA/SNc projecting neurons . For example , we did not attempt to lesion the thin shell of AIV neurons dorsal to RA . Furthermore , our lesions may have affected areas of the intermediate arcopallium adjacent to AIV . We targeted the portion of Ad most likely affected by our AIV lesion procedure ( n = 7 birds ) as follows: Bilateral lesions of Ad were carried out stereotaxically by injecting 1% NMA at multiple injection sites . The stereotaxic procedure was similar to that described above for AIV lesions . RA was mapped electrophysiologically to determine the lateral border and center ( in the AP direction ) of RA . Ad was also mapped electophysiologically by its characteristic bursting patterns , which distinguished it from the overlying arcopallium . Injection sites are listed below: the ( ML , AP ) coordinates are given in the same notation given above for the AIV lesion . However , the injections were targeted to the center of Ad in the dorso–ventral dimension as determined from electrophysiological mapping of Ad . Injections at each site were made in 3 boluses of 13 . 8 nl , which were spaced at 5-min intervals to avoid backflow along the pipette . Injections were made at the following locations: ( +250 , −100 ) ; ( +250 , +300 ) ; ( +650 , −100 ) ; ( +650 , +300 ) ; ( +1000 , −100 ) ; ( +1000 , +300 ) . This lesion protocol typically spared the lateral-most ∼200 µm of Ad ( see Figure 5A ) . To determine the effects of complete lesions of Ad on vocal learning , we also carried out ( n = 3 birds ) an Ad lesion protocol with two additional lateral injection sites . Injections were made at the following locations: ( +250 , −100 ) ; ( +250 , +300 ) ; ( +650 , −100 ) ; ( +650 , +300 ) ; ( +950 , −100 ) ; ( +950 , +300 ) ; ( +1200 , −100 ) ; ( +1200 , +300 ) . These birds exhibited severe immobility after surgery , and failed to eat or drink , requiring early termination of the experiment . Male juvenile birds were maintained in the aviary in their home cages with both parents until the age 44–45 dph , at which point they were transferred to sound isolation/recording chambers until they began to sing and baseline song was acquired . For experimental birds , bilateral AIV or Ad lesions were carried out as described above . After surgery , birds typically began to sing within 5 days ( range 2–5 days ) . Birds were maintained in isolation and recorded continuously until they reached the age of 90 dph . Once it was confirmed that post-90 dph , songs had been recorded , we histologically verified the extent of AIV lesions , as described below . Unmanipulated control birds were isolated at 44–45 dph ( as for experimental birds ) and were maintained in sound isolation/recording cages until they reached the age of 90 dph , or until post-90 dph song had been acquired . Assignment of birds to unmanipulated control or experimental groups was initially made randomly , but subsequent juveniles from a given breeding cage ( with a given tutor ) were placed in control or experimental groups to provide sibling matches between these conditions . AIV lesions were examined histologically at the end of each experiment . Retrograde tracer ( fluorescently labeled cholera toxin β subunit , Molecular Probes ) was injected into VTA/SNc and 4 days later the bird was sacrificed and perfused transcardially with saline and then 4% paraformaldehyde . The brain was removed and post-fixed in 4% paraformaldehyde overnight . The brain was sectioned , stained for NeuN , and imaged under a fluorescence microscope ( Zeiss Axioplan , Germany ) . The extent of AIV lesions was analyzed on the basis of loss of NeuN staining and the density of retrogradely labeled neurons in these regions . Quantification was based on the size of the lesion in each parasagittal section extending from the medial edge of RA to 200 µm past the lateral edge , and yielded two numbers indicating the fraction of AIV lesioned on the left and right side . The extent of any damage to RA was also monitored and birds with more than 5% RA lesion were excluded from the analysis ( n = 2 ) . The histological assessment of lesion size was conducted blind to the behavioral effects on song , and the quantification of the behavioral effects was done blind to the histological analysis of lesion size . To test the functional connectivity of projections from L1 , CM , and HVC-shelf to VTA/SNc-projecting neurons in AIV , we implanted bipolar stimulating electrodes in each of these auditory cortical areas ( L1 , n = 2 birds; CM , n = 3 birds; HVC-shelf , n = 3 birds ) . A bipolar stimulating electrode was also placed in VTA/SNc to antidromically identify neurons in AIV . Guided by the results of the anterograde tracing experiments ( Figure 2 , Figure 2—figure supplement 1 ) , recordings were made in the regions of AIV ventral and anterior to RA using a Carbostar electrode ( 1 MOhm , Kation Scientific ) . Within this region , single-unit and multi-unit activity was broadly responsive to stimulation in both VTA/SNc and in L1 , CM , or HVC-shelf . Threshold currents required to elicit spiking responses were between 50 and 200 µA ( single 0 . 2-ms unipolar pulses ) . Location of the stimulating electrodes was verified histologically . We modified the previously published Song Analysis Pro ( SAP ) algorithm ( Tchernichovski et al . , 2000 ) and developed an automated procedure for comparison of pupil songs with the tutor motif ( Software is available at Mandelblat-Cerf and Fee , 2014 ) . Following SAP , songs were represented by acoustic features . Similarity between time points in the two songs is based on the Euclidean distance in the feature space between these points . The procedure builds a similarity matrix for all possible pairs of time points between the two song samples . To quantify tutor–pupil similarity the tutor motif was segmented into separate syllables . While syllable segmentation is highly reliable in tutor song , it is less reliable in the more variable pupil song , which was thus segmented into equally sized sections , each twice the duration of the tutor motif . For each tutor syllable our procedure uses the similarity matrix described above to find the sections of the pupil song bout that have the highest similarity and assigns an overall similarity score . For each match between a tutor's syllable and a pupil's song section , a sequencing score is evaluated by computing the match of the next syllable in the tutor song with the next section of the pupil song . The algorithm then computes a composite ‘tutor imitation score’ as the product of song similarity and sequence similarity . For each bird we examined song spectrograms and computed the imitation scores for the first day of singing after isolation , and before any further manipulation . Three birds exhibited clear evidence of tutor imitation and were excluded from further experiments . To determine the fractional extent to which AIV-lesions reduced the capacity of birds to imitate their tutor song , imitation scores of each bird were placed on a linear scale ranging from the average imitation score of control birds 0 . 197 ± 0 . 011 to the average similarity score from all pairwise comparisons of 20 unrelated adult birds in our colony 0 . 104 ± 0 . 0027 , yielding a total dynamic range of 0 . 093 [=0 . 197–0 . 104] ± 0 . 00636 . For birds in which we estimate that more than half of AIV was lesioned , the average imitation score was 0 . 116 ± 0 . 008 . Therefore , we can compute the similarity of AIV-lesioned birds to their tutor songs relative to the song similarity of unrelated birds , as a fraction of the total dynamic range , as 12 . 9% [= ( 0 . 116–0 . 104 ) /0 . 093] ± 9 . 2% . Thus , AIV-lesioned birds suffered a loss of imitation capacity compared to controls of 87% [=1 . 0–0 . 129] ± 9% . Male juvenile birds were obtained from the aviary at the age 65–70 dph and were maintained in sound isolation cages until they began to sing . Upon reaching age the 75–80 dph , pre-surgery songs were recorded , and birds were randomly placed in three experimental groups: deafening , AIV-lesion , and combined deafening and AIV-lesion . Deafening was achieved by bilateral cochlear extirpation , which was carried out by accessing the cochlea through the posterio-lateral cranium . AIV lesions were carried out as described above . Upon recovery from surgery , birds were maintained in sound isolation cages . Undirected song was recorded continuously for 2 weeks . To quantify song degradation , song was recorded every day post-surgery and compared to song recorded prior to surgery . Self-similarity scores were computed using the same algorithm used to evaluate tutor imitation . Single-unit recordings of VTA/SNc-projecting AIV neurons were obtained from seven young adult birds ( 90–120 dph ) during undirected singing . For antidromic identification of AIV neurons , bipolar stimulating electrodes were placed in VTA/SNc , according to the methodology described above . AIV was localized with reference to the boundaries of RA , as electrophysiologically identified by the characteristic tonic activity of RA neurons . Placement of the chronically implanted electrode array was further refined by finding the regions in AIV that exhibited strong antidromic responses to brief unipolar square current pulses ( 0 . 2 ms duration ) applied to the stimulating electrodes in VTA/SNc . Peak currents up to 350 µA were used to search for antidromically activated neurons: threshold currents for antidromic activation of neurons in AIV ranged from 90 µA to 250 µA . Recordings in AIV were made using an array of three recording electrodes ( 200 µm spacing , Microprobe , Gaithersburg , MD , #PI20033 . 0A3 , 3 MΩ ) , mounted in a motorized microdrive ( n = 20 birds implanted ) . Data were acquired and analyzed using custom Matlab software . Disruptive auditory stimuli consisted of 50-ms duration pulses of white noise ( 95 dBSPL ) similar to that previously used to drive song plasticity ( Tumer and Brainard , 2007; Andalman and Fee , 2009 ) , which were presented with random timing during undirected singing .	Most new skills , from playing a sport to learning a language , are acquired through a gradual process of trial and error . While some of this learning is driven by direct external rewards , such as praise , much of it occurs when the individual compares their current performance with their own impression of what a ‘correct’ performance should be . The way that the brain responds to external rewards is relatively well understood , but much less is known about the processes used by the brain to evaluate its own performance . One way to study this process is to examine how songbirds learn their songs . While in the nest , young male birds memorize another bird's song , usually that of their father . They learn to sing by comparing their own vocals with this memorized template , tweaking their song until the two versions match . Now , Mandelblat-Cerf et al . have identified a pathway in the brain that enables the birds to make this comparison and to use any discrepancies to improve their subsequent attempts . Anatomical labeling experiments revealed that a brain structure called the arcopallium has a key role in this process . The ventral part of this structure ( known as AIV ) receives inputs from the auditory cortex—meaning that it has access to the bird’s own song—and then forms connections with a specific group of neurons in the midbrain . These midbrain neurons , which communicate using the chemical transmitter dopamine , project to brain regions that ultimately control the movements involved in singing . This means that the AIV is ideally positioned to be able to evaluate and then adjust the song as required . Consistent with this possibility , young zebra finches were less able to imitate a template song if their AIV was destroyed before they had started practicing . By contrast , destroying the AIV in adult birds who had already learned their song did not impair performance , indicating that the arcopallium circuit supports song learning rather than singing per se . Finally , recordings of neurons in the AIV made during singing revealed that this brain area sends signals about discrepancies between what the young bird tries to sing and what he hears himself sing . In addition to providing further clues as to how the songbirds learn their songs , this work also highlights the fact that dopaminergic neurons in the midbrain—which are best known for being involved in our response to external rewards such as food and drugs—also contribute to learning that is driven internally .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	A role for descending auditory cortical projections in songbird vocal learning
Pollen apertures , the characteristic gaps in pollen wall exine , have emerged as a model for studying the formation of distinct plasma membrane domains . In each species , aperture number , position , and morphology are typically fixed; across species they vary widely . During pollen development , certain plasma membrane domains attract specific proteins and lipids and become protected from exine deposition , developing into apertures . However , how these aperture domains are selected is unknown . Here , we demonstrate that patterns of aperture domains in Arabidopsis are controlled by the members of the ancient ELMOD protein family , which , although important in animals , has not been studied in plants . We show that two members of this family , MACARON ( MCR ) and ELMOD_A , act upstream of the previously discovered aperture proteins and that their expression levels influence the number of aperture domains that form on the surface of developing pollen grains . We also show that a third ELMOD family member , ELMOD_E , can interfere with MCR and ELMOD_A activities , changing aperture morphology and producing new aperture patterns . Our findings reveal key players controlling early steps in aperture domain formation , identify residues important for their function , and open new avenues for investigating how diversity of aperture patterns in nature is achieved . As part of cell morphogenesis , cells often form distinct plasma membrane domains that acquire specific combinations of proteins , lipids , and extracellular materials . Yet how these domains are selected and specified is often unclear . Pollen apertures offer a powerful model for studying this process . Apertures are the characteristic gaps on the pollen surface that receive little to no deposition of the pollen wall exine; during their formation , certain regions of the plasma membrane are selected and specified as aperture domains ( Zhou and Dobritsa , 2019 ) . Pollen apertures create some of the most recognizable patterns on the pollen surface , usually conserved within a species but highly variable across species ( Furness and Rudall , 2004 ) . For instance , in wild-type Arabidopsis pollen , apertures are represented by three long and narrow furrows , equally spaced on the pollen surface and oriented longitudinally ( Figure 1A and A' ) . In other species , aperture positions , number , and morphologies can be different , suggesting that the mechanisms guiding aperture formation are diverse . While the diversity of aperture patterns has captivated scientists for decades ( Furness and Rudall , 2004; Matamoro-Vidal et al . , 2016; Walker , 1974; Wodehouse , 1935 ) , studies of the associated molecular mechanisms have only recently begun ( Dobritsa and Coerper , 2012; Dobritsa et al . , 2018; Lee et al . , 2018; Reeder et al . , 2016; Zhang et al . , 2020 ) . Aperture domains first become visible at the tetrad stage of pollen development , when four sister microspores , the products of meiosis , are held together under the common callose wall and aperture factors , such as INAPERTURATE POLLEN1 ( INP1 ) and D6 PROTEIN KINASE-LIKE3 ( D6PKL3 ) in Arabidopsis and OsINP1 and DEFECTIVE IN APERTURE FORMATION1 ( OsDAF1 ) in rice , accumulate at distinct domains of the microspore plasma membranes ( Dobritsa and Coerper , 2012; Dobritsa et al . , 2018; Lee et al . , 2018; Zhang et al . , 2020 ) . These domains become protected from exine deposition and develop into apertures ( Dobritsa et al . , 2018; Zhang et al . , 2020 ) . Yet how aperture domains are selected and what mechanism guides their patterning remain completely unknown . Recently , we isolated a new Arabidopsis mutant , macaron ( mcr ) , in which pollen , instead of forming three apertures , develops a single ring-shaped aperture , suggesting that the affected gene is involved in specifying positions and number of aperture domains ( Plourde et al . , 2019 ) . Here , we perform a detailed analysis of this mutant and identify the MCR gene . We demonstrate that it belongs to the ancient family of ELMOD proteins , and that together with another member of this protein family in Arabidopsis , ELMOD_A , MCR acts at the beginning of the aperture formation pathway as a positive regulator of aperture domain specification . We provide evidence that aperture domains are highly sensitive to the levels of MCR and ELMOD_A , which can positively or negatively affect their number . We further demonstrate that a third member of this family , ELMOD_E , has an ability to influence the number , positions , and morphology of aperture domains , and we identify specific protein residues critical for this ability . Our study elucidates key molecular factors controlling aperture patterning and functionally characterizes members of the widespread , yet thus far uncharacterized family of the plant ELMOD proteins . In a screen of EMS-mutagenized Arabidopsis plants , we discovered four non-complementing mutants , which , instead of three equidistant pollen apertures , produced a single ring-shaped aperture dividing each pollen grain into two equal parts ( Figure 1B-E' ) . As the mutant phenotype resembled the French meringue dessert , we named these mutations macaron ( alleles mcr-1 through mcr-4 ) . Imaging of mcr microspore tetrads demonstrated that they develop normally and achieve a regular tetrahedral conformation . The ring-shaped aperture domains in mcr microspores , visualized with the help of the reporter INP1-YFP , are positioned so that they pass through the proximal and distal poles of each microspore ( Figure 1G; compare with the INP1-YFP localization in the absence of mcr mutation in Figure 1F ) . Thus , like in wild-type pollen , apertures in mcr are placed longitudinally . However , while aperture positions in each wild-type microspore are coordinated with aperture positions in its three sisters ( Dobritsa et al . , 2018; Reeder et al . , 2016 ) , in mcr , the ring-shaped apertures appear to be placed independently in sister microspores ( Figure 1—figure supplement 1 ) . Occasionally , instead of ring-shaped apertures , mcr pollen displays two unconnected apertures ( Figure 1H and H' ) , suggesting that the ring-shaped aperture is a product of a two-aperture fusion . Thus , mcr mutations reduce the number of apertures , but do not affect their furrow morphology , longitudinal orientation , and equidistant placement . Like all the other previously characterized pollen aperture mutants in Arabidopsis , including inp1 and inp2 , which completely lack apertures ( Dobritsa and Coerper , 2012; Dobritsa et al . , 2011; Lee et al . , 2021 ) , mcr mutants exhibited no obvious fertility defects . We previously showed that aperture number strongly depends on microspore ploidy and is sensitive to cytokinetic defects that disrupt formation of normal tetrahedral tetrads , creating other arrangements of post-meiotic microspores ( Reeder et al . , 2016 ) . While normal haploid ( 1n ) pollen develops three apertures , diploid ( 2n ) pollen produces either four or a mixture of four and six apertures , depending on whether it was generated through tetrads or dyads . In contrast , 2n mcr pollen , produced through either tetrads or dyads , has three equidistant apertures ( Plourde et al . , 2019 ) , suggesting that the increasing effect of higher ploidy on aperture number is counterbalanced by the defect in the MCR function . We have now extended this analysis by assessing the effects of the mcr mutation on aperture formation under additional perturbations of ploidy or post-meiotic microspore arrangement . By creating 1n Mitosis instead of Meiosis ( MiMe ) plants ( d’Erfurth et al . , 2009 ) with the mcr mutation , we generated mcr pollen with normal ploidy ( 1n ) via dyads , and not tetrads . As shown previously ( Reeder et al . , 2016 ) , a majority of the 1n MiMe pollen grains ( ~60% ) develop three normal apertures , with the rest forming mostly six apertures ( Figure 1I , Figure 1—figure supplement 2A–C' ) . Yet , in the pollen of the 1n mcr MiMe plants , the number of apertures was reduced , with ~50–70% of pollen developing the mcr phenotype ( either ring-shaped or two apertures ) and the rest forming three apertures ( Figure 1I , Figure 1—figure supplement 2D–E' ) . We further perturbed microspore formation and ploidy by crossing mcr-1 with a mutant defective in the TETRASPORE ( TES ) gene . In tes mutants , microspore mother cells ( MMCs ) go through meiosis but fail to undergo cytokinesis , producing large pollen grains with four haploid nuclei and a high number ( ~10 or more ) of irregularly placed and fused apertures ( Reeder et al . , 2016; Spielman et al . , 1997 ) . Although in the double mcr tes mutant apertures are often positioned irregularly and fused together , their number was usually lower ( ~4–6 ) than in the single tes mutant ( Figure 1—figure supplement 2F–G' ) . Altogether , these results indicate that mcr mutations have an overall reducing effect on aperture number , manifested across different levels of pollen ploidy and post-meiotic microspore arrangements . In wild-type tetrad-stage microspores , aperture factors INP1 and D6PKL3 localize to the three longitudinal aperture domains of the plasma membrane ( Dobritsa and Coerper , 2012; Dobritsa et al . , 2018; Lee et al . , 2018 ) . Since mcr mutation affects INP1-YFP localization , causing it to migrate to a ring-shaped membrane domain ( Figure 1G ) , we tested whether mcr also affects the localization of D6PKL3 , which likely acts upstream of INP1 . We introgressed the previously characterized transgenic reporter D6PKL3pr:D6PKL3-YFP ( Lee et al . , 2018 ) into the mcr-1 background . In mcr microspores , D6PKL3-YFP re-localized to a single ring-shaped domain ( Figure 2A ) , indicating that MCR acts upstream of both INP1 and D6PKL3 . We also examined the genetic interactions between MCR and other aperture factors , including the recently discovered INP2 ( Lee et al . , 2021 ) , by combining their mutations . d6pkl3 single mutants develop three apertures partially covered by exine ( Lee et al . , 2018 ) . Pollen of the mcr d6pkl3 double mutants developed single ring-shaped apertures that were partially covered by exine , indicating that the two genes have an additive effect on aperture phenotype ( Figure 2B ) . In contrast , pollen grains of mcr inp1 and mcr inp2 completely lacked apertures , phenocopying single inp1 and inp2 mutants ( Dobritsa and Coerper , 2012; Lee et al . , 2021; Figure 2C and D ) . This indicates that INP1 and INP2 are epistatic to MCR , consistent with their roles as factors absolutely essential for aperture formation . We mapped the mcr-1 defect to a 77 kb interval on the second chromosome . One of the 25 genes in this interval , At2g44770 , had a C-to-T mutation converting a highly conserved Pro165 ( see below ) into a Ser ( Figure 3A , Figure 3—figure supplement 1 ) . Sequencing of At2g44770 from the other three mcr alleles also revealed mutations ( Figure 3A , Figure 3—figure supplement 1 ) . mcr-2 had a G-to-A mutation converting Gly129 into an Asp . mcr-3 had a G-to-A mutation affecting the last nucleotide of the fifth intron , disrupting the splice acceptor site and causing a frame shift in the middle of the critical catalytic region ( see below ) . In mcr-4 , no mutations in the coding sequence ( CDS ) of At2g44770 were found; however , there was a G-to-A mutation 310 nt downstream of the stop codon in its 3′ untranslated region ( 3′ UTR ) , suggesting that the 3′ UTR is important for regulation of this gene ( Figure 3A ) . In addition , plants with T-DNA insertions in this gene ( mcr-5 , mcr-6 , and mcr-7 ) all produced pollen with the mcr phenotype ( Figure 3A , Figure 3—figure supplement 2 ) . Yet the T-DNA mutations , likely due to their residence in introns , were hypomorphic , as some pollen with three normal apertures was found in their populations ( 9% in mcr-5 [n = 179] , 13% in mcr-6 [n = 216] , and 22% in mcr-7 [n = 78] ) . This was in contrast to the mcr-1 through mcr-4 mutations , in which the mcr aperture phenotype was fully penetrant . We further verified the identity of MCR as At2g44770 by creating complementation constructs and expressing them in the mcr-1 mutant . The genomic construct MCRpr:gMCR ( driven by the 3 kb DNA fragment upstream of the start codon [referred to as the MCR promoter] and containing introns and the 0 . 8 kb region downstream of the stop codon ) restored three normal apertures in 10/10 T1 transgenic plants ( Figure 3B and B' ) . A similar genomic construct expressing protein fused at the C-terminus with yellow fluorescent protein ( YFP ) also successfully restored apertures ( Figure 3C and C' ) . In contrast , the MCRpr:MCR CDS construct , which contained only the CDS driven by the MCR promoter , did not rescue the mcr phenotype ( 0/6 T1 plants had three apertures restored; Figure 3D and D' ) , indicating that additional regulatory regions are required for expression of this gene , consistent with the notion of the 3′ UTR importance . The MCR promoter and 3′ UTR were then included in all constructs for which we sought MCR-like expression and are herein referred to as the MCR regulatory regions . The protein encoded by At2g44770 contains the Engulfment and Cell Motility ( ELMO ) domain ( InterPro006816 ) ( Figure 3A , Figure 3—figure supplement 1 ) . In animals , proteins with this domain belong to two families: ( 1 ) smaller ELMOD proteins , containing only the ELMO domain , and ( 2 ) larger ELMO proteins , containing , besides the ELMO domain , several other protein domains ( East et al . , 2012 ) . The ELMOD family is believed to be the more ancient of the two , with ELMOD proteins already present in the last common ancestor of all eukaryotes and ELMO proteins appearing later in evolution in the opisthokont clade ( East et al . , 2012 ) . In mammals , ELMOD proteins act as non-canonical GTPase activating proteins ( GAPs ) for regulatory GTPases of the ADP-ribosylation factor ( Arf ) family , a subgroup within the Ras superfamily that includes Arf and Arf-like ( Arl ) proteins ( Bowzard et al . , 2007; Ivanova et al . , 2014; Turn et al . , 2020 ) . Unlike animals , plants only have members of the ELMOD family , and their roles remain essentially uncharacterized . In Arabidopsis , the ELMOD family consists of six members , ELMOD_A through ELMOD_F ( Figure 3E , Figure 3—figure supplement 1 ) , in the nomenclature of East et al . , 2012 . MCR is ELMOD_B . One of the other five proteins , ELMOD_A , shares 86% sequence identity with MCR , and the rest have ~50–55% sequence identity with both MCR and ELMOD_A . Although the ELMOD proteins are broadly expressed in Arabidopsis , young buds at or near the stages when apertures develop express mostly MCR and ELMOD_A ( Figure 3E ) . Given the high similarity between MCR and ELMOD_A , we wondered if ELMOD_A also aids in aperture formation . We disrupted ELMOD_A with CRISPR/Cas9 by introducing a single-nucleotide insertion ( A ) after the codon 64 ( Figure 4A ) . Although this created a shift in the open reading frame and an early stop codon , it did not affect aperture formation ( Figure 4B and B' ) . ( Likewise , a second elmod_a CRISPR allele , in which deletion of the last nucleotide of the codon 64 resulted in a different frame shift and creation of another early stop codon , did not affect the formation of apertures . ) We hypothesized that the lack of phenotype could be due to the ELMOD_A redundancy with MCR . To test this , we crossed the elmod_a mutant ( carrying the CRISPR/Cas9 transgene ) with the mcr-1 mutant . Already in the F1 generation , when all plants were expected to be double heterozygotes , we found several plants producing pollen with the mcr-like aperture phenotype ( Figure 4C and C' ) . Sequencing of the MCR and ELMOD_A genes from these plants showed that , as expected , they were heterozygous for MCR; however , they had homozygous or biallelic mutations in ELMOD_A , indicating that the CRISPR/Cas9 transgene continued targeting the wild-type copy of ELMOD_A in the F1 progeny of the cross . The phenotype of these mcr/+ elmod_a mutants revealed that in the absence of ELMOD_A , MCR displays haploinsufficiency . Notably , when at least one wild-type copy of ELMOD_A is present , MCR is haplosufficient ( Figure 4D and D' ) . Therefore , these paralogs play redundant roles in the formation of aperture domains . Yet , since MCR can specify three normal apertures in the absence of ELMOD_A but not vice versa , its role appears to be more prominent compared to that of ELMOD_A . We also tested how the lack of one wild-type copy of ELMOD_A and both wild-type copies of MCR , as well as the lack of wild-type copies of both genes , would affect aperture formation . In the mcr elmod_a/+ plants , pollen had the mcr phenotype ( Figure 4E and E' ) . However , when the function of both genes was completely disrupted , the resulting pollen produced either one greatly disrupted aperture with an abnormal , circular morphology and partially covered with exine , or formed no apertures ( Figure 4F-G' ) . Thus , the simultaneous loss of function of the two ELMOD family genes has a synergistic effect on aperture formation . To confirm that these defects were caused by mutations in ELMOD_A and not off-site CRISPR targeting events , as well as to identify the ELMOD_A regulatory regions , we created two ELMOD_A genomic constructs driven by the 2 kb region upstream of its start codon – EApr:gELMOD_A ( which also included a 0 . 3 kb ELMOD_A 3′ UTR ) and EApr:gELMOD_A-YFP ( tagged with YFP and lacking the ELMOD_A 3′ UTR ) – and transformed them into the mcr elmod_a double mutant , which no longer carried the CRISPR/Cas9 transgene . Both constructs successfully rescued formation of apertures ( 5/5 and 31/33 T1 plants , respectively , Figure 4H–I' ) , indicating that the selected promoter region is sufficient for ELMOD_A functional expression . In addition , when ELMOD_A was expressed in the mcr single mutant from either its own promoter or from the MCR regulatory regions ( MCRpr:gELMOD_A-YFP-MCR3′UTR ) , it also complemented the loss of MCR ( 12/12 and 14/14 T1 plants; Figure 4J–K' ) . Thus , both ELMOD_A and MCR participate in aperture domain specification . Formation of three apertures in Arabidopsis pollen requires either two intact copies of MCR or at least one copy of each of these two ELMOD family members . According to the publicly available RNA-seq data ( Klepikova et al . , 2016 ) , MCR and ELMOD_A are both expressed in young buds with pollen at or near the tetrad stage of development ( Figure 3E ) . To confirm that in these buds MCR and ELMOD_A are expressed in the developing pollen lineage , we created transcriptional reporter constructs MCRpr:H2B-RFP and EApr:H2B-RFP , expressing the nuclear marker H2B tagged with red fluorescent protein and transformed them into wild-type Arabidopsis . In the resulting transgenic lines , MCR and ELMOD_A promoters were active in the developing pollen lineage ( MMCs , tetrads , and young free microspores ) as well as in somatic anther layers ( Figure 5A ) . To find out if , like the previously discovered aperture factors INP1 and D6PKL3 , MCR and ELMOD_A accumulate at the aperture domains of tetrad-stage microspores , we determined the subcellular localization of the YFP-tagged proteins expressed from the translational reporters MCRpr:gMCR-YFP and EApr:gELMOD_A-YFP , which rescued mutant phenotypes . Consistent with the results from the transcriptional reporters , the YFP signal was present in MMCs , tetrads , and young microspores ( Figure 5B and C ) . This signal was diffusely localized in the cytoplasm and prominently enriched in the nucleoplasm . No specific enrichment near the plasma membrane was observed . Therefore , MCR and ELMOD_A specify positions and number of aperture domains without visibly congregating there . Although ELMOD proteins do not have the typical GAP domain associated with the canonical Arf GAP proteins , they contain a conserved stretch of 26 amino acids , with 13 residues exhibiting a particularly high degree of conservation and forming the consensus sequence WX3G ( F/W ) QX3PXTD ( F/L ) RGXGX3LX2L . In mammalian ELMODs , this region is proposed to mediate their Arf/Arl GAP activity ( East et al . , 2012 ) . The presence of the invariant Arg in this region is of particular importance since the activity of many GAP proteins of the Ras GTPase superfamily , including canonical Arf GAPs , relies on a catalytic Arg ( Scheffzek et al . , 1998 ) . Indeed , in mammalian ELMODs , the Arg in this putative GAP region was shown to be essential for their GAP activity , consistent with its role as the catalytic residue ( East et al . , 2012 ) . Even relatively small changes at this position , such as conversion to Lys , resulted in the complete loss of GAP activity . Although plant ELMODs have only limited similarity to mammalian proteins ( e . g . , the Arabidopsis and human ELMODs have ~20% sequence identity ) , they contain the same conserved region and invariant Arg residue ( Figure 6A ) . To test if this region is essential for function in MCR and ELMOD_A , we created constructs in which the invariant Arg ( R127 ) was substituted with Lys ( MCRpr:gMCRR127K-YFP and EApr:gELMOD_AR127K-YFP ) . These constructs were then expressed , respectively , in the mcr and mcr elmod_a mutants . Unlike the constructs with the wild-type MCR and ELMOD_A , the R127K constructs , although expressed normally , completely failed to restore the expected aperture patterns ( 0/8 T1 plants for MCRR127K; 0/12 T1 plants for ELMOD_AR127K ) , indicating that , like in mammalian ELMODs , the Arg in the putative GAP region is critical for the activity of MCR and ELMOD_A ( Figure 6B and C ) . While working with MCR-YFP and ELMOD_A-YFP transgenic lines , we made a surprising discovery . We noticed that while most of these lines had apertures restored to the expected number ( i . e . , three apertures for mcr MCRpr:gMCR-YFP and a ring-shaped aperture/two apertures for mcr elmod_a EApr:ELMOD_A-YFP ) , in some transgenic T1 lines the number of apertures exceeded the expectations: with up to six apertures forming in mcr MCRpr:gMCR-YFP and up to four apertures in mcr elmod_a EApr:gELMOD_A-YFP ( Figure 7A and B' ) . To test if different aperture numbers could be due to different levels of transgene expression , we examined YFP fluorescence in homozygous lines producing different aperture numbers . For both MCR and ELMOD_A transgenes , the number of apertures positively correlated with the level of YFP signal in the microspore cytoplasm and nucleoplasm ( Figure 7C–E , Figure 7—figure supplement 1A ) . In addition , in some lines , the number of apertures further increased in T2 or T3 generations compared to the numbers in T1 , consistent with the transgene dosage increasing in later generations due to attaining homozygosity . To further test the notion that aperture number depends on the MCR/ELMOD_A gene dosage/levels of expression , we modulated the dosage of MCR , starting with a defined transgene . We crossed a homozygous mcr MCRpr:gMCR-YFP plant from line 7-2 , commonly producing >6 apertures ( Figure 7C ) , with ( 1 ) mcr and ( 2 ) wild type . In the resulting transgenic F1 progeny of the first cross , MCR should be expressed from one source – a single copy of the transgene . In the F1 progeny of the second cross , it should be expressed from two sources – one copy of the transgene plus one of the endogenous gene . In the pollen of these F1 plants , the number of apertures correlated with the number of functional copies of MCR: pollen of gMCR-YFP/- mcr/mcr produced on average 4 . 68 ± 1 . 08 apertures compared to 5 . 85 ± 1 . 52 apertures in gMCR-YFP/- mcr/+ ( Figure 7F , Figure 7—figure supplement 1B ) . We further assessed aperture phenotypes in the progeny of these plants that had a homozygous transgene and either zero or two copies of endogenous MCR . Both genotypes with the homozygous transgene produced many more apertures compared to plants with the hemizygous transgene , but they also differed significantly from each other , with the number of apertures correlating with the presence of endogenous MCR ( 8 . 08 ± 1 . 57 in MCR-YFP/MCR-YFP mcr/mcr vs . 9 . 34 ± 1 . 50 in MCR-YFP/MCR-YFP +/+ ) ( Figure 7F , Figure 7—figure supplement 1B ) . These results indicate that the process of aperture domain specification is highly sensitive to the levels of MCR and ELMOD_A in developing microspores . To examine the evolutionary history of the plant ELMOD family , we retrieved 561 ELMOD sequences belonging to 178 species across the plant kingdom and used them for a detailed phylogenetic analysis . ELMOD proteins are widespread in plants , suggesting that they perform important functions ( Figure 8A ) . Green algae as well as non-vascular land plants ( liverworts , mosses , and hornworts ) typically have a single ELMOD protein , but an ancestor of lycophytes and ferns had a gene duplication ( Figure 8A and B ) . Beginning with gymnosperms , the ELMOD family expanded and diversified , with distinct protein groups clustering with the A/B/C clade , the E clade , and the F clade ( Arabidopsis proteins were used as landmarks in naming the clades ) . In early angiosperms , ELMOD proteins separated into four well-supported clades: A/B , C , E , and F ( Figure 8A and B , Figure 8—figure supplement 1 ) . The split within the aperture factor-containing A/B clade into the separate ELMOD_A and ELMOD_B ( MCR ) lineages happened late – in the common ancestor of the Brassicaceae family ( Figure 8A , Figure 8—figure supplement 1 ) . Yet , in many other angiosperm species , including magnoliids , monocots , basal eudicots , and multiple asterids and rosids , the A/B clade also contains at least two proteins ( Figure 8—figure supplement 1 ) . This shows that independent duplications in this lineage happened multiple times , suggesting the existence of strong evolutionary pressure to maintain more than one gene of the A/B type . The extensive number of the retrieved ELMOD sequences allowed us to evaluate conservation of the residues disrupted in MCR by the mcr-1 and mcr-2 mutations . Pro165 , converted into Ser in mcr-1 ( Figure 3A , Figure 3—figure supplement 1 ) , was present in each of the 553 ELMOD sequences containing this region , suggesting a critical role in protein function . This Pro belongs to the highly conserved WEYPFAVAG motif ( Figure 3—figure supplement 1 ) found in all six Arabidopsis ELMODs , as well as in the majority of ELMODs from other plants , including green algae . The appearance of Asp at position 129 in mcr-2 ( Figure 3A , Figure 3—figure supplement 1 ) affects a site within the putative GAP region , neighboring the critical catalytic Arg127 . This change is also highly unusual from the evolutionary perspective . Except for one likely pseudogene ( see below ) , none of the other 560 ELMOD sequences from across the plant kingdom has an Asp at that site , consistent with the notion that an Asp at this position is not tolerated by natural selection and could be detrimental for all plant ELMOD proteins . Our analysis of residues occupying position 129 in the GAP region across the angiosperm ELMOD proteins led to an interesting discovery . In the 365 analyzed angiosperm sequences , this site is occupied by only three amino acids: Cys , Gly , or Ala . ( Earlier diverged plants have Ala or Gly at this site . ) Strikingly , we found that all proteins with Cys129 cluster with the E clade , whereas nearly all proteins with Gly129 cluster with either the A/B or the F clades , and nearly all proteins with Ala129 cluster with the C clade . ( Only six exceptions were found among the 365 sequences: in five cases , proteins containing Ala129 clustered with the A/B or the F clades , and in one case , a protein with Gly129 clustered with the C clade . ) This suggested the intriguing possibility , tested later , that , in angiosperms , residues at position 129 are important for functional differentiation of the ELMOD proteins . Besides mcr-2 , the only protein with Asp at position 129 is the Arabidopsis ELMOD_D . However , it has several other features that suggest it is likely a pseudogene . At 213 amino acids , ELMOD_D is markedly shorter than the other five Arabidopsis ELMODs ( 265–323 aa-long ) : it misses stretches of 52 aa upstream of the GAP region , 4 aa in the vicinity of the GAP region , and 22 aa at the very C-terminus of the protein ( Figure 3—figure supplement 1 ) . It also has major substitutions unique to this protein within or near its GAP region , which change the conserved Gly119 and Leu138 residues into Arg ( the numbering within the GAP region is based on the MCR and ELMOD_A sequences; Figure 3—figure supplement 1 ) . ELMOD_D clusters with the C clade and is most closely related to the Arabidopsis ELMOD_C , indicating that it is a product of a very recent duplication ( Figure 8—figure supplement 1 ) . While some plants have more than one protein in the C clade , most others , including close relatives of Arabidopsis , have just a single C protein ( Figure 8—figure supplement 1 ) , suggesting that a single C-type activity is sufficient for most species . These findings , combined with the extremely low levels of ELMOD_D expression ( Figure 3E ) , support the hypothesis that this member of the Arabidopsis ELMOD family is likely non-functional . To test if other ELMOD family members , besides MCR and ELMOD_A , might be involved in aperture formation , we examined the phenotypes of their single mutants and double mutants with mcr ( Figure 9—figure supplement 1 ) . All single mutants displayed normal aperture patterns ( Figure 9—figure supplement 1B–F' ) , while the double mutants exhibited mcr phenotypes ( Figure 9—figure supplement 1G–J' ) , indicating that , unlike ELMOD_A , the other four ELMOD genes do not interact synergistically with MCR in aperture formation . We also assessed the ability of these genes to rescue the mcr aperture phenotype by expressing them , tagged with YFP , from the MCR regulatory regions . ELMOD_C showed some limited ability to restore three apertures in mcr ( Figure 9—figure supplement 2A–D' ) , while ELMOD_D and ELMOD_F were unable to do it ( 6/6 and 22/22 T1 plants; Figure 9—figure supplement 2E–F' ) . Based on the YFP signal , ELMOD_F-YFP was expressed at a level comparable with that of the MCR-YFP and ELMOD_A-YFP transgenes ( Figure 9—figure supplement 2G and G' ) . ELMOD_D-YFP , however , was undetectable ( Figure 9—figure supplement 2H and H' ) , consistent with the hypothesis that ELMOD_D is a pseudogene . Most interestingly , the expression of ELMOD_E in mcr led to a neomorphic phenotype: instead of narrow longitudinal furrows , pollen of all seven T1 plants developed multiple short , round apertures ( Figure 9A and A' ) . To see if this effect was limited to the mcr background , we transformed the MCRpr:ELMOD_E-YFP construct into wild-type Col-0 plants . The T1 plants showed a range of aperture phenotypes ( Figure 9B–D' ) , yet multiple round apertures were commonly present , suggesting that ELMOD_E exerts a dominant negative effect when misexpressed in developing microspores . We then tested whether ELMOD_E would have the same effect on aperture patterns when expressed from its own promoter . However , none of the 12 T1 transgenic mcr plants expressing the ELMOD_Epr:ELMOD_E-YFP construct had any changes in the mcr aperture phenotype ( Figure 9E and E' ) . Analysis of the YFP signal showed that this gene is expressed in tetrad-stage microspores at much lower levels from its own promoter than from the MCR promoter ( Figure 9F–H' ) . Thus , while ELMOD_E can influence aperture patterns , it is likely not normally involved in this process in Arabidopsis . To test if differences in the MCR levels could impact the ability of transgenic ELMOD_E to produce round apertures , we crossed a mcr MCRpr:gELMOD_E-YFP line with the above described mcr MCRpr:gMCR-YFP lines 10-3 , 13-5 , and 7-2 that express MCR , respectively , at low , medium , and high levels ( Figure 7C ) . In the F1 progeny of crosses with low and medium expressors , 10-3 and 13-5 , pollen still produced round apertures ( Figure 9I–J' ) . Yet , in the F1 progeny of the cross with the high expressor 7-2 , furrow aperture morphology was restored ( Figure 9K and K' ) , although these plants produced less pollen with a high number of furrows ( >4 ) compared to the original 7-2 plants ( Figure 9L ) . Together , these data support the idea that high level of MCR can counteract the neomorphic activity of ELMOD_E and suggest that MCR and ELMOD_E may compete for the same interactors . The different aperture phenotypes of mcr MCRpr:gMCR-YFP and mcr MCRpr:gELMOD_E-YFP lines gave us an opportunity to test the hypothesis that residues at position 129 are important for functional differentiation of ELMODs from different clades . For E-clade proteins , we also noticed that Cys129 was always found together with Asn121 . These residues are unique to this clade: 100% of the retrieved E-clade sequences ( n = 69 ) have Asn121/Cys129 vs . 0% of sequences from the other clades ( n = 297 ) . Thus , this combination could be important for the E-clade functions . In the other three clades , position 121 is always occupied by Asp . To investigate the importance of sites 121 and 129 for MCR and ELMOD_E functions , we created six constructs in which one or both residues at these positions were replaced with the residues typical of the other clade and expressed them in the mcr mutant . The MCR proteins carrying the E-specific residues at both positions ( MCRD121N/G129C ) or at the position 121 ( MCRD121N ) still retained most of the MCR function , with most T1 plants producing three or more furrow apertures in most of their pollen grains ( 7/9 and 12/12 T1 plants , respectively; Figure 10A and B , Figure 10—figure supplement 1A ) . However , when the E-specific residue was present only at position 129 ( MCRG129C ) , MCR protein became less active , with only 5 out of 11 T1 plants producing three furrow apertures in all or most of their pollen ( Figure 10C ) . In the rest of these T1 plants , the mcr phenotype was not rescued or was rescued poorly , with <30% of pollen grains forming three apertures . Experiments with the ELMOD_E proteins carrying the MCR residues at positions 121 and 129 confirmed the importance of Asn121 and Cys129 for the ELMOD_E neomorphic activity . In the case where both residues were replaced with the MCR residues ( ELMOD_EN121D/C129G ) , ELMOD_E largely lost its ability to create round apertures and instead often restored three furrow-like apertures , thus acting like MCR ( Figure 10D , Figure 10—figure supplement 1B ) . In the cases when only one residue was changed ( ELMOD_EN121D and ELMOD_EC129G ) , the mutant ELMOD_E proteins were still often able to produce multiple round apertures , although three normal furrows or a mixture of furrows and round apertures were also produced , suggesting that the single mutations reduced the ELMOD_E activity , but did not eliminate it entirely ( Figure 10E and F , Figure 10—figure supplement 1B ) . While the number of apertures produced in these experiments varied and some pollen grains had both furrows and round apertures ( Figure 10—figure supplement 1B ) , in general , there was a strong correlation between the number of apertures per pollen and their shape . Analysis of 207 pollen grains from across the ELMOD_EN121D/C129G transgenic lines showed that when pollen had three apertures , they were mostly represented by furrows ( 95% , n = 468 apertures ) , whereas in pollen with four or more apertures , apertures were mostly round ( 84% , n = 239 apertures; Figure 10—figure supplement 1C ) . Similar trends were also observed in ELMOD_EN121D and ELMOD_EC129G transgenic lines ( Figure 10—figure supplement 1C ) . Taken together , these results show that residues at positions 121 and 129 in the GAP region provide important contributions to the specific functions of MCR and ELMOD_E . Yet they are less critical for MCR , in accord with the fact that Asp121 and Gly129 are not unique to the A/B clade . In the case of ELMOD_E , the E-clade-specific combination of Asn121/Cys129 appears to be essential for its distinct activity . When both residues undergo MCR-like changes , ELMOD_E loses its neomorphic activity , instead becoming capable of carrying out the MCR role in aperture formation . How developing pollen grains create beautiful and diverse geometrical aperture patterns has been a long-standing problem in plant biology ( Fischer , 1889; Ressayre et al . , 2002; Wodehouse , 1935 ) . In this study , we uncovered the first set of molecular factors , belonging to the ELMOD protein family , that have a clear ability to regulate the number , positions , and morphology of aperture domains . MCR and its close paralog ELMOD_A act as ( somewhat ) redundant positive regulators of furrow aperture formation in Arabidopsis . Our genetic analysis places MCR and ELMOD_A at the beginning of the aperture formation pathway , upstream of the previously discovered aperture factors D6PKL3 , INP1 , and , likely , INP2 , the recently identified partner of INP1 . Previous studies showed that INP1 and INP2 act as the executors of the aperture formation program , absolutely essential for aperture development but not able on their own to influence the number and positions of aperture domains ( Dobritsa et al . , 2018; Lee et al . , 2021; Li et al . , 2018; Reeder et al . , 2016 ) . D6PKL3 was proposed to act upstream of INP1 , defining the features of aperture domains , yet it also largely lacks the ability to initiate completely new domains ( Lee et al . , 2018; Zhou and Dobritsa , 2019 ) . In mcr microspores , D6PKL3 and INP1 re-localize to the ring-shaped aperture domains ( Figures 1G and 2A ) , indicating that they become attracted to the newly specified aperture domains and the ELMOD proteins act as patterning factors , contributing to symmetry breaking and selection of sites for aperture domains . Our data demonstrate that the aperture domains forming in each microspore are highly sensitive to the ELMOD_A/MCR protein dosage ( Figure 4C–G' , Figure 7 , Figure 7—figure supplement 1 ) . Increased dosage leads to a higher number of apertures , while decreased dosage results in fewer , and the reducing effect of the loss in MCR activity on aperture number is maintained across different levels of ploidy and post-meiotic arrangements of microspores ( Figure 1I , Figure 1—figure supplement 2 ) , which were previously shown to be important factors in aperture patterning ( Reeder et al . , 2016 ) . Thus , modulation of ELMOD protein levels might offer a mechanism for creating different aperture patterns in different species . Interestingly , within the genus Pedicularis , some species display the mcr-like ring-shaped apertures , while others produce three apertures ( Wang et al . , 2009; Wang et al . , 2017 ) . Our findings suggest that such variations in aperture patterns could conceivably be due to variations in ELMOD proteins or their effectors or regulators . Importantly , while great diversity of pollen aperture numbers is found across plant species , within a species , this trait tends to be very robust . For example , in wild-type Arabidopsis , the number of apertures rarely deviates from three ( Reeder et al . , 2016 ) . Our results , therefore , imply that , normally , levels of MCR and ELMOD_A are very tightly controlled , and there must exist mechanisms to ensure this control . The discovery that ELMOD_E can also influence aperture patterns in Arabidopsis and create multiple round apertures instead of three furrows ( Figure 9A–B' ) suggests that the regulation of ELMOD_E might also contribute to the diversity of aperture patterns in nature . In Arabidopsis , ELMOD_E does not seem to be usually involved in aperture formation . Yet , when misexpressed from the MCR regulatory regions , it interferes with MCR and ELMOD_A activity ( Figure 9I–L ) , resulting in the formation of new aperture domains . ELMODs are ancient proteins , predicted to have been present in the last common ancestor of all eukaryotes ( East et al . , 2012 ) . In animals , these proteins act as non-canonical GAPs , regulating activities of both Arf and Arl GTPases ( Bowzard et al . , 2007; Ivanova et al . , 2014; Turn et al . , 2020 ) . Arf GTPases are commonly associated with the recruitment of vesicle coat proteins to different membrane compartments to initiate vesicle budding and trafficking , while the roles of the related Arl proteins are less understood and likely more diverse ( Sztul et al . , 2019 ) . Although the function of ELMOD proteins in plants is unknown , their presence in green algae and other basal plants lineages suggests that they have been playing important roles in plant cells since their inception . Our phylogenetic analysis indicates that this family in plants is monophyletic , and the genes have duplicated and diversified over the course of plant evolution . The angiosperm ELMOD family has four distinct clades ( Figure 8A and B , Figure 8—figure supplement 1 ) . In many species , the A/B clade , containing MCR and ELMOD_A , has two or more proteins due to independent duplications that occurred multiple times in evolution . This suggests that species might be under a selective pressure to keep more than one A/B type protein , implying that the processes in which these proteins are involved ( e . g . , aperture formation ) benefit from genetic redundancy and , thus , are highly important . Further studies will be required to establish the biochemical role of plant ELMOD proteins . Like their animal counterparts , plant ELMODs may be involved in regulation of Arf/Arl activities . Arabidopsis has 19 ARFs and ARLs , which , with few exceptions , mostly remain uncharacterized ( Delgadillo et al . , 2020; Gebbie et al . , 2005; McElver et al . , 2000 ; Singh et al . , 2018; Vernoud et al . , 2003; Xu and Scheres , 2005 ) . The roles attributed to members of this family – for example , in secretion , endocytosis , activation of phosphatidyl inositol kinases , and actin polymerization ( Singh and Jürgens , 2018; Sztul et al . , 2019 ) – are all potentially fitting with the formation of distinct aperture domains . The protein region proposed to be the GAP region in mammalian ELMODs ( East et al . , 2012 ) is conserved in plant proteins , and the invariant Arg residue believed to be catalytic in mammalian ELMODs is also necessary for function in MCR and ELMOD_A ( Figure 6B and C ) . Interestingly , some positions within the conserved GAP region show strict residue specificity in different clades , suggesting that they could be important for functional diversity of these proteins . Consistent with this , we found the combination of Asn121/Cys129 to be key for the ELMOD_E neomorphic aperture-forming activity ( Figure 10 ) . Alternatively , plant ELMODs could have evolved functions different from their animal counterparts and regulate targets other than ARFs/ARLs . Interestingly , the only study done so far on an ELMOD protein in plants ( Hoefle and Hückelhoven , 2014 ) pulled out the barley homolog of ELMOD_C in a yeast two-hybrid screen as an interactor of a ROP GAP , a GAP for a different class of small GTPases , Rho-of-plants ( ROPs ) . Rho GTPases ( including ROPs ) are well-known regulators of cell polarity and domain formation ( Feiguelman et al . , 2018; Yang and Lavagi , 2012 ) , so their involvement in aperture formation cannot be excluded . In summary , we presented critical players in the process of patterning the pollen surface . These players belong to the ELMOD protein family , which , while undoubtedly important , has not yet been characterized in plants . Future studies should focus on identifying the interactors of the ELMOD proteins and on understanding the mechanisms through which they specify positions and shape of aperture domains without noticeably accumulating at these regions . Arabidopsis thaliana genotypes used in this study were either in Columbia ( Col ) or Landsberg erecta ( Ler ) background . Pollen from wild-type Col- and Ler has indistinguishable aperture phenotypes . The following genotypes were also used: mcr-1 , mcr-2 , mcr-3 , mcr-4 , mcr-5 ( CS853233 ) , mcr-6 ( SALK_205528C ) , mcr-7 ( SALK_203827C ) , elmod_c ( SALK_076565 ) , elmod_d ( SALK_031512 ) , elmod_e ( SALK_082496 ) , elmod_f ( SALK_010379 ) , inp1-1 ( Dobritsa and Coerper , 2012 ) , inp2-1 ( Lee et al . , 2021 ) , d6pkl3-2 ( Lee et al . , 2018 ) , inp1-1 DMC1pr:INP1-YFP ( Dobritsa et al . , 2018 ) , d6pkl3-2 D6PKL3pr:D6PKL3-YFP ( Lee et al . , 2018 ) , tes ( SALK _113909 ) , MiMe ( d’Erfurth et al . , 2009 ) , and cenh3-1 GFP-tailswap ( CS66982 ) . mcr-1 through mcr-4 mutants were discovered in a forward genetic screen performed on an EMS-mutagenized Ler population ( Plourde et al . , 2019 ) . mcr-5 through mcr-7 mutants and elmod_c through elmod_f mutants were ordered from the Arabidopsis Biological Resource Center ( ABRC ) . Plants were grown at 20–22°C with the 16 hr light/8 hr dark cycle in growth chambers or in a greenhouse at the Biotechnology support facility at OSU . To generate the 2n mcr tes plants , mcr-1 mutant was crossed with heterozygous tes , double heterozygotes were recovered in F1 by genotyping ( primers listed in Supplementary file 1 ) , and double homozygotes were identified in F2 population . The generation of haploid mcr MiMe plants was similar to the procedure previously described ( Reeder et al . , 2016 ) . In brief , mcr-1 mutant was first crossed with plants that were triple heterozygotes for atrec8-3 , osd1-3 , and atspo11-1-3 ( MiMe heterozygotes ) , then the quadruple heterozygotes were identified among the F1 progeny by genotyping and crossed as males with cenh3-1 GFP-tailswap homozygous plants that were used as haploidy inducers ( Ravi and Chan , 2010 ) . 1n F1 progeny of this cross were identified by their distinctive morphology as described ( Ravi and Chan , 2010; Reeder et al . , 2016 ) , and the triple 1n MiMe and quadruple 1n mcr MiMe mutants were identified by genotyping ( primers listed in Supplementary file 1 ) . Unlike other 1n genotypes generated by this cross , which were sterile , the 1n plants with MiMe mutations were fertile and produced 1n pollen via mitosis-like division and dyad formation . Mapping of the MCR locus mcr-1 mutant with Ler background was crossed with Col-0 , and individual F2 plants were screened under a dissecting microscope for the presence of the distinctive angular mutant phenotype in their dry pollen . In total , 369 plants with mutant phenotype were selected , and their genomic DNA was isolated . To map the MCR locus , we first conducted bulked segregant analysis , followed by the map-based positional cloning ( Lukowitz et al . , 2000 ) . The insertion-deletion ( InDel ) molecular markers were developed based on the combined information from the 1001 Genomes Project database ( 1001 Genomes Consortium , 2016 ) and the Arabidopsis Mapping Platform ( Hou et al . , 2010 ) . The MCR locus was mapped to a 77 kb region between markers 2–18 . 39 Mb ( 18 , 395 , 427 bp ) and 2–18 . 47 Mb ( 18 , 472 , 092 bp ) on chromosome 2 . Molecular markers used for mapping are listed in Supplementary file 1 . Out of the 25 genes located in this interval , we sequenced 11 genes , prioritized based on their predicted expression patterns and gene ontology , and found that one of them , At2g44770 , contained a missense mutation . Sequencing of the other three non-complementing EMS alleles identified in the forward genetic screen ( mcr-2 to mcr-4 ) also revealed presence of mutations in At2g44770 . Two guide RNAs against target sequences at the beginning of the ELMOD_A and ELMOD_E CDS were selected with the help of the CRISPR-PLANT platform ( https://www . genome . arizona . edu/crispr/ Xie et al . , 2014 ) and individually cloned into the BsaI site of the pHEE401E vector ( Wang et al . , 2015 ) as described ( Xing et al . , 2014 ) , using , respectively , two sets of complementary primers: elmod_a sgRNA-F/R and elmod_e sgRNA-F/R ( Supplementary file 1 ) . The resulting constructs were separately transformed into the Agrobacterium tumefaciens strain GV3101 , and then used to transform Arabidopsis Col-0 plants or mcr-1 mutants ( the latter only with the anti-ELMOD_E construct ) using the floral-dip method ( Clough and Bent , 1998 ) . The T1 transformants were selected on ½ strength MS plates supplemented with 1% ( w/v ) sucrose , 0 . 8% ( w/v ) agar , and 50 µg/mL hygromycin , their DNA was extracted , and the regions surrounding the target sequences were sequenced . For ELMOD_A , 5 of 25 T1 plants had homozygous , biallelic , or heterozygous mutations . Sequencing the progeny of these plants demonstrated that all homozygous/biallelic mutants developed frame shifts in the ELMOD_A CDS after the codon 64 ( by acquiring either a 1-nt insertion three nucleotides before PAM or a 1-nt deletion two nucleotides before PAM ) . An elmod_a mutant with a single A insertion , as shown in Figure 4A , and still carrying CRISPR/Cas9 transgene , was crossed with the mcr-1 mutant to obtain the mcr elmod_a double mutant . For ELMOD_E , 1 out of 12 and 1 out of 20 T1 plants had biallelic mutations , respectively , in Col-0 and mcr-1 backgrounds . In T2 generation , homozygous mutants with a frame shift in the CDS were identified: in elmod_eCR , a 13-nt region located four nucleotides before PAM was deleted and replaced with a different 9-nt sequence; in mcr elmod_eCR , a single A was inserted four nucleotides before PAM . These mutants were used to observe the aperture phenotypes . A 3076 bp fragment upstream of the start codon of MCR was used as the MCR promoter for all MCRpr constructs . To generate the MCRpr:gMCR construct , the promoter and the 2868 bp genomic fragment from the MCR start codon to 798 bp downstream of the stop codon were separately amplified from Col-0 genomic DNA and cloned into SacI/NcoI sites in the pGR111 binary vector ( Dobritsa et al . , 2010 ) through In-Fusion cloning ( Takara ) . An AgeI site was introduced in front of the MCR start codon for ease of subsequent cloning . For MCRpr:MCR CDS , the genomic fragment was replaced with the MCR coding sequence , which was amplified from the MCR cDNA construct CD257409 obtained from ABRC . For MCRpr:gMCR-YFP construct , the genomic fragment of MCR was amplified without the stop codon and cloned upstream of YFP into the pGR111 binary vector ( Dobritsa et al . , 2010 ) . Additionally , a 497 bp 3′ UTR region from MCR was then cloned downstream of YFP . Since we achieved phenotypic rescue and observed strong YFP signal with this construct , we used this combination of regulatory elements in all subsequent constructs for which we wanted to achieve the MCR-like expression . The constructs MCRpr:gELMOD_A/C/D/E/F-YFP were created in a similar way . For all EApr constructs , a 2163 bp fragment upstream of the start codon of ELMOD_A was amplified from Col-0 genomic DNA and used as the ELMOD_A promoter . For EApr:gELMOD_A , a 2833 bp fragment , which included a 296 bp region downstream of the stop codon , was subcloned into pGR111 downstream of EApr . A BamHI site was introduced in front of the start codon for ease of subsequent cloning . For EApr:gELMOD_A-YFP , a 2534 bp genomic fragment ( from the ELMOD_A start codon to immediately upstream of the stop codon ) was cloned between the EApr and YFP . For ELMOD_Epr:ELMOD_E-YFP , a 1469 bp fragment upstream of the start codon of ELMOD_E was amplified from Col-0 genomic DNA and used as the ELMOD_E promoter to replace the MCR promoter in MCRpr:gELMOD_E-YFP . To generate the reporter constructs MCRpr:H2B-RFP and EApr:H2B-RFP , the H2B-RFP fusion gene was cloned into the BamHI/SpeI sites downstream of the respective promoters in pGR111 . To create constructs with single and double nucleotide substitutions , PCR-based site-directed mutagenesis was performed with IVA mutagenesis ( García-Nafría et al . , 2016 ) using gMCR-pGEM-T Easy , gELMOD_A-pGEM-T Easy , and gELMOD_E-pGEM-T Easy as templates . The mutated sequences then replaced the respective wild-type sequences in MCRpr:gMCR-YFP-pGR111 , EApr:ELMOD_A-YFP-pGR111 , and MCRpr:gELMOD_E-YFP-pGR111 . All primers used for creating constructs are listed in Supplementary file 1 . All constructs were verified by sequencing and transformed by electroporation into the Agrobacterium strain GV3101 together with the helper plasmid pSoup . Agrobacterium cultures confirmed to contain the constructs of interest were then transformed into mcr or mcr elmod_a by floral dip ( mcr elmod_a was verified to lack the anti-ELMOD_A CRISPR/Cas9 transgene ) . Preparation and imaging of mature pollen grains , MMC , tetrads , and free microspores were performed as previously described ( Reeder et al . , 2016 ) . Imaging was done on a Nikon A1+ confocal microscope with a 100× oil-immersion objective ( NA = 1 . 4 ) , using 1× confocal zoom for anthers , 3× zoom for pollen grains , 5× zoom for MMC and tetrads , and 5× or 8× zoom for free microspores . For imaging mature pollen grains , pollen was placed into an ~10 μL drop of auramine O working solution ( 0 . 001%; diluted in water from the 0 . 1% [w/v] stock prepared in 50 mM Tris-HCl ) , allowed to hydrate for ~5 min , covered with a #1 . 5 cover slip , and sealed with nail polish . Exine was excited with a 488 nm laser and fluorescence was collected at 500–550 nm . To count aperture number , images from the front and back view of pollen grain were taken . If some apertures were present on sides of a pollen grain not directly visible by focusing on the front and on the back , then z-stacks were taken ( step size = 500 nm ) and 3D images were reconstructed using NIS Elements software v . 4 . 20 ( Nikon ) and used for counting aperture number . For imaging cells of the developing pollen lineage , anthers were dissected out of stage 9 flower buds and placed into a small drop of Vectashield antifade solution supplemented with 0 . 02% Calcofluor White and 5 μg/mL membrane stain CellMask Deep Red . Cells in the pollen lineage were released by applying gentle pressure to the coverslip placed over the anthers . To obtain fluorescence signals , the following excitation/emission settings were used: RFP , 561 nm/580–630 nm; YFP , 514 nm/522–555 nm; Calcofluor White , 405 nm/424–475 nm; CellMask Deep Red , 640 nm/663–738 nm . Z-stacks of tetrads were obtained with a step size of 500 nm and 3D reconstructed using NIS Elements v . 4 . 20 ( Nikon ) . To compare the YFP fluorescence intensity in three different lines of mcr MCRpr:gMCR-YFP or mcr elmod_a EApr:gELMOD_A-YFP , tetrads were prepared simultaneously and imaged on the same day under identical acquisition settings on Nikon A1+ confocal microscope . The mean YFP signal intensities in nucleoplasm and cytoplasm of tetrads ( n ≥ 15 ) were separately measured with the help of the region of interest ( ROI ) statistics function in NIS Elements v . 4 . 20 ( Nikon ) . For each tetrad , a single optical section showing both nucleoplasm and cytoplasm was selected and analyzed . ELMOD family members in Arabidopsis have the following accession numbers: ELMOD_A , At3g60260; MCR , At2g44770; ELMOD_C , At1g67400; ELMOD_D , At3g43400; ELMOD_E , At1g03620; ELMOD_F , At3g03610 . The phylogenetic tree of Arabidopsis ELMOD proteins in Figure 3E was built using the neighbor-joining ( NJ ) algorithm of MEGA X ( Kumar et al . , 2018 ) , with bootstrap support calculated for 1000 replicates . Sequences of ELMOD proteins from species across the plant kingdom were retrieved from the Phytozome v . 12 database ( https://phytozome . jgi . doe . gov/pz/portal . html ) and the 1000 Plants ( 1KP ) database ( https://db . cngb . org/onekp/ , last accessed in May 2020; Wickett et al . , 2014 ) . MCR protein sequence was used as a query for an online BLASTP search of these databases with default parameters . The protein sequences with the E-value ≤1e-10 , sequence identity ≥30% , and Bit-Score ≥ 60 were identified as ELMODs and further confirmed by a local BLASTP search using each of the other Arabidopsis ELMODs as a query . In the cases when two or more proteins were potentially translated from the same gene , the one providing the best match with the query was selected . In total , 561 ELMOD protein sequences from 178 representative species belonging to eudicots ( 36 species/195 sequences ) , monocots ( 14/94 ) , magnoliids ( 20/64 ) , basal angiosperms ( 5/13 ) , gymnosperms ( 17/62 ) , ferns ( 17/44 ) , lycophytes ( 20/29 ) , bryophytes ( 37/47; including 18 sequences from 15 liverworts , 6 sequences from 5 hornworts , and 23 sequences from 17 mosses ) , and green algae ( 12/13 ) were retrieved and used for phylogenetic analysis . Multiple sequence alignment was performed using MAFFT v7 . 017 ( Katoh and Standley , 2013; Katoh et al . , 2002 ) with the L-INS-i algorithm and default parameters . Sites with greater than 20% gaps were trimmed by TrimAl ( Capella-Gutiérrez et al . , 2009 ) and manually inspected for overhangs . ModelFinder ( Kalyaanamoorthy et al . , 2017; accessed through IQ-TREE [Nguyen et al . , 2015] ) was run to find the best-fit amino acid substitution model . The alignment in Figure 3—figure supplement 1 was visualized with Espript3 . 0 ( Gouet et al . , 1999 ) . Phylogenetic trees were constructed using IQ-TREE with the maximum likelihood ( ML ) method , SH-aLRT test , and ultrafast bootstrap with 1000 replicates . For the tree on Figure 8A , containing sequences from across the plant kingdom , 267 sequences were used , including all sequences retrieved from green algae , bryophytes , lycophytes , ferns , gymnosperms , and basal angiosperms , as well as 24 sequences from magnoliids , 19 sequences from three monocots , and 16 sequences from three eudicots . For the tree on Figure 8—figure supplement 1 , containing only angiosperm sequences , we used all 366 sequences retrieved for this group . Phylogenetic trees were visualized in iTOL v . 5 ( Letunic and Bork , 2021 ) and can be accessed at http://itol . embl . de/shared/Zhou3117 . RNA-seq data for different tissues/developmental stages of six Arabidopsis ELMOD genes were obtained from the TRAVA database ( http://travadb . org/; Klepikova et al . , 2016 ) . The ‘Raw Norm’ option was chosen for read counts , and default settings were used for all other options . The retrieved RNA-seq data were presented as a bubble heatmap using TBtools ( Chen et al . , 2020 ) . Quantification of aperture numbers and YFP signal was done with NIS Elements v . 4 . 20 software ( Nikon ) . For each line , the aperture number of 160 pollen grains from at least three different plants was counted and the mean YFP fluorescence of at least 15 tetrads from the same plants was measured . Graphs were generated using Microsoft Excel or Origin version 2018 software . Binary comparisons were performed using a two-tailed Student’s t-test in Microsoft Excel; results with the p values below 0 . 05 were judged significantly different . The p values are represented as ( *p<0 . 001 ) , ( p<0 . 01 ) , *p<0 . 05 . All error bars represent standard deviation ( SD ) . For all boxplots , the box defines the first and third quartile , the central line depicts the median , and the small square in the box represents the mean value . Whiskers extend to minimum and maximum values . Outliers are indicated as ∗ . Different shapes show individual samples . Details of statistical analysis , number of quantified entities ( n ) , and measures of dispersion can be found in the corresponding figure legends .	Zooming in on cells reveals patterns on their outer surfaces . These patterns are actually a collection of distinct areas of the cell surface , each containing specific combinations of molecules . The outer layers of pollen grains consist of a cell wall , and a softer cell membrane that sits underneath . As a pollen grain develops , it recruits certain fats and proteins to specific areas of the cell membrane , known as ‘aperture domains’ . The composition of these domains blocks the cell wall from forming over them , leading to gaps in the wall called ‘pollen apertures’ . Pollen apertures can open and close , aiding reproduction and protecting pollen grains from dehydration . The number , location , and shape of pollen apertures vary between different plant species , but are consistent within the same species . In the plant species Arabidopsis thaliana , pollen normally develops three long and narrow , equally spaced apertures , but it remains unclear how pollen grains control the number and location of aperture domains . Zhou et al . found that mutations in two closely related A . thaliana proteins – ELMOD_A and MCR – alter the number and positions of pollen apertures . When A . thaliana plants were genetically modified so that they would produce different levels of ELMOD_A and MCR , Zhou et al . observed that when more of these proteins were present in a pollen grain , more apertures were generated on the pollen surface . This finding suggests that the levels of these proteins must be tightly regulated to control pollen aperture numbers . Further tests revealed that another related protein , called ELMOD_E , also has a role in domain formation . When artificially produced in developing pollen grains , it interfered with the activity of ELMOD_A and MCR , changing pollen aperture shape , number , and location . Zhou et al . identified a group of proteins that help control the formation of domains in the cell membranes of A . thaliana pollen grains . Further research will be required to determine what exactly these proteins do to promote formation of aperture domains and whether similar proteins control domain development in other organisms .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "plant", "biology", "cell", "biology" ]	2021	Members of the ELMOD protein family specify formation of distinct aperture domains on the Arabidopsis pollen surface
Tauopathies feature progressive accumulation of tau amyloids . Pathology may begin when these amplify from a protein template , or seed , whose structure is unknown . We have purified and characterized distinct forms of tau monomer—inert ( Mi ) and seed-competent ( Ms ) . Recombinant Ms triggered intracellular tau aggregation , induced tau fibrillization in vitro , and self-assembled . Ms from Alzheimer’s disease also seeded aggregation and self-assembled in vitro to form seed-competent multimers . We used crosslinking with mass spectrometry to probe structural differences in Mi vs . Ms . Crosslinks informed models of local peptide structure within the repeat domain which suggest relative inaccessibility of residues that drive aggregation ( VQIINK/VQIVYK ) in Mi , and exposure in Ms . Limited proteolysis supported this idea . Although tau monomer has been considered to be natively unstructured , our findings belie this assumption and suggest that initiation of pathological aggregation could begin with conversion of tau monomer from an inert to a seed-competent form . Amyloids are ordered protein assemblies , typically rich in beta sheet , that underlie multiple disorders such as Alzheimer’s disease ( AD ) . Amyloid-forming proteins include tau , synuclein , and expanded polyglutamine proteins such as huntingtin , among many others . It is unknown how or why intracellular proteins such as tau transition from a relatively inert form to one that efficiently self-assembles into ordered structures in vivo . This process begins with the formation of a pathogenic ‘seed , ’ a structure that serves as a template for homotypic fibril growth . This structural transition could be a critical event in the pathogenesis of neurodegeneration . Under defined conditions and relatively high concentrations ( typically micromolar ) , recombinant tau monomer will form amyloid fibrils in vitro . However the basis of spontaneous assembly in cells is unknown . The conversion of a protein from a monomer to a large , ordered multimer could occur by several mechanisms , but the first step probably involves the formation of a seed . This event , and indeed the actual conformation or assembly state of the protein that constitutes the ‘minimal’ seed , has remained obscure . This has led to the idea that a seed is potentially transitory , arising from an equilibrium between two states: one relatively aggregation-resistant , and another that is short-lived . A seed could be a single molecule , or several . Based on extrapolation from kinetic aggregation studies , it has been suggested that a critical seed for tau and polyglutamine peptide amyloid formation is a single molecule ( Chirita et al . , 2005; Bhattacharyya et al . , 2005; Kar et al . , 2011 ) , while an earlier study ( among others [Ramachandran and Udgaonkar , 2013] ) has proposed a tau multimer ( Friedhoff et al . , 1998 ) . Isolation of the seed-competent form of tau could be critical to understanding the initiation of disease and the design of more effective diagnostics and therapeutics . Tau forms amyloids that underlie neurodegeneration in a variety of neuropathological syndromes , collectively termed tauopathies ( Lee et al . , 2001 ) . These include AD and frontotemporal dementias , among many others . Multiple groups , including ours , have now observed that tau will propagate an aggregated state from the outside to the inside of a cell , between cells , across synapses , and within brain networks ( Sanders et al . , 2016 ) . In prior work , we used size exclusion chromatography ( SEC ) to define tau trimers as the minimal unit of spontaneous cellular uptake and intracellular amyloid formation , and proposed this as the smallest particle capable of propagating aggregates between cells ( Mirbaha et al . , 2015 ) . This work involved application of ‘naked’ protein assemblies derived from recombinant protein or human brain onto cultured ‘biosensor’ HEK293 cells or primary neurons that express a tau aggregation reporter ( Frost et al . , 2009a; Holmes et al . , 2014 ) . Biosensor cells and primary neurons alike take up tau aggregates via macropinocytosis ( Holmes et al . , 2013 ) . The aggregates subsequently serve as highly specific templates to trigger intracellular amyloid formation ( Holmes et al . , 2014; Sanders et al . , 2014 ) . We have also determined that preincubation of cationic lipids such as lipofectamine with tau seeds facilitates their direct transduction into a cell , bypassing the physiologic uptake mechanism ( Holmes et al . , 2014; Furman et al . , 2015 ) . Lipofectamine-mediated delivery into biosensor cells allows direct quantitation of seed titer for both tau and α-synuclein ( Holmes et al . , 2013 ) . Tau is intrinsically disordered upon isolation from bacteria or mammalian cells and is relatively inert in terms of spontaneous self-assembly . However under various conditions , including exposure to polyanions such as heparin , tau will form aggregates via nucleated self-assembly ( Goedert et al . , 1996; Pérez et al . , 1996 ) . It is unknown how these experimental conditions relate to the initiation of aggregation in human brain . We have now purified various stable forms of full-length tau monomer from recombinant sources and human brain . One is relatively inert and is stable for long periods . Another is ‘seed-competent , ’ triggers amyloid formation in cells and in vitro , and exhibits intrinsic properties of self-assembly . We have used crosslinking with mass spectrometry ( XL-MS ) to probe the structures of these molecules . Models of discrete regions within the RD predict that differential exposure of hexapeptide motifs previously known to be important for amyloid formation distinguishes the two forms of tau . These models are supported by limited proteolysis studies . The identification of distinct and stable forms of tau monomer , including some that are uniquely seed-competent , bears directly on how we understand the initiation of protein aggregation in the tauopathies . We initially sought to define the tau seeding unit that would trigger intracellular aggregation upon direct delivery to the cell interior . We had previously observed that a tau trimer is the minimal assembly size that triggers endocytosis and intracellular seeding ( Mirbaha et al . , 2015 ) . These experiments depended on spontaneous cell uptake , since no lipofectamine was added to the reactions . A prior study had also indicated the role of disulfide linkages in promoting tau aggregation , potentially by dimer formation ( Friedhoff et al . , 1998 ) . Thus , for our initial studies we engineered and purified full-length ( FL ) tau monomer that lacks any internal cysteines due to alanine substitutions ( C299A and C322A ) , termed tau ( 2A ) . FL tau ( 2A ) cannot self-associate based on disulfide linkages , which helped prevent the formation of cryptic dimers that could have confounded our studies . These substitutions did not affect tau purification , heparin-induced fibrillization , and sonication protocols , which we performed as described previously ( Mirbaha et al . , 2015 ) . We treated fibril preps with sonication , prior to isolation of recombinant FL tau ( 2A ) assemblies of various sizes by size exclusion chromatography ( SEC ) ( Mirbaha et al . , 2015 ) . In parallel , we also studied FL wild type ( WT ) tau . We purified unfibrillized recombinant FL tau ( 2A ) monomer by SEC ( Figure 1A ) , and isolated SEC fractions of sonicated fibrils that contained putative monomer , dimer , trimer and ~10 mer ( Figure 1B ) . To test the seeding activity of the tau preparations , we used a previously described ‘biosensor’ cell reporter line ( Holmes et al . , 2014 ) . These cells stably express 4R tau repeat domain ( RD ) containing the disease-associated P301S mutation . All cells express 4R-RD-Cyan fluorescent protein and 4R-RD-yellow fluorescent protein ( RD-CFP/YFP ) . Exogenously applied seeds induce intracellular aggregation with resultant fluorescence resonance energy transfer ( FRET ) between CFP and YFP that can be measured via flow cytometry ( Holmes et al . , 2014; Furman et al . , 2015 ) . The degree of aggregation is scored using ‘integrated FRET density’ ( IFD ) , which is the product of the percent positive cells and the mean fluorescence intensity of FRET-positive cells , and from this we determine a titer of tau seeding activity ( Holmes et al . , 2014 ) . Lipofectamine directly transduces tau assemblies across the plasma membrane and increases the assay’s sensitivity by approximately 100-fold . Upon incubation with Lipofectamine , we were surprised to observe seeding by monomer and larger assemblies alike , whether FL WT or 2A . ( Figure 1C , D ) . Epifluorescence microscopy confirmed the presence of intracellular inclusions after FL WT tau monomer seeding ( Figure 1D ) . We termed the inert monomer ‘Mi , ’ and the seed-competent monomer ‘Ms . ’ To rule out higher order assemblies of tau within the putative monomer fraction , immediately prior to the seeding assay we passed fractions through a 100 kDa cutoff filter to eliminate anything larger than a monomer . While monomer fraction retained ~80% of seeding activity , only ~20% of dimer seeding activity remained , and ~1–2% of trimer seeding activity remained ( Figure 1E ) . To exclude an artifact related to Lipofectamine transduction into cells , we tested FL ( 2A ) tau preparations in an in vitro seeding assay that induces fibril formation by full-length tau ( 0N4R ) through iterative polymerization and agitation steps ( Morozova et al . , 2013 ) . Mi had no intrinsic seeding activity . However Ms induced amyloid formation , albeit more slowly than trimer or unfractionated fibrils ( Figure 1F ) . This slow aggregation process may reflect inefficient fibril assembly , and a predominance of small nucleated assembly events from the added monomer . We concluded that the Ms fraction contained seeding activity that enabled intracellular aggregation of tau RD-CFP/YFP in cells , or full-length tau in vitro . Finally , we tested whether contamination of very small amounts of seeds could somehow account for the seeding activity in monomer fractions by carrying out dose-response titrations of the various preparations . Ms had an EC50 of ~10 nM ( Figure 1G ) , which was very similar to dimer and trimer ( Figure 1H ) . Thus to account for signal observed in the seeding assay , contamination of an otherwise inert monomer with larger seed-competent assemblies would have to be substantial . We tested for obvious structural differences between Mi and Ms using CD spectroscopy , which revealed none ( Figure 2A ) . We re-tested the assemblies using fluorescence correlation spectroscopy ( FCS ) , which measures particle diffusion through a fixed volume . As we previously observed ( Mirbaha et al . , 2015 ) , we accurately estimated the units of small assemblies ( ≤10 mer ) , but not larger assemblies ( >10 mer ) ( Figure 2B ) . In an additional effort to detect cryptic multimers within the Ms preparation , we used double-label FCS . We engineered a cysteine onto the amino terminus of FL tau ( 2A ) to enable its covalent modification ( Cys-Tau ( 2A ) ) . We then prepared Cys-tau ( 2A ) fibrils , or monomer , and labeled them simultaneously with Alexa488 ( green ) and tetramethylrhodamine ( TMR ) via maleimide chemistry . We carried out sonication and purification by SEC as before , isolating assemblies of various sizes . We evaluated each for cross-correlation between red and green signal , which indicates the presence of at least two tau molecules in a particle . We analyzed >300 events for each assembly . When we evaluated Mi and Ms , 100% of events in each case showed a diffusion time consistent with a tau monomer ( Figure 2C , D ) . Furthermore , we observed no cross-correlation between red and green signal , indicating that neither preparation had detectable multimeric assemblies ( Figure 2C , D , H ) . By contrast , when we evaluated larger species such as dimer , trimer , or ~10 mer , we observed longer diffusion times consistent with the predicted assembly sizes , and significant cross-correlation values ( Figure 2E–H ) , consistent with the presence of multimers . The FCS studies supported the conclusion that Mi and Ms are comprised predominantly of monomer . To rule out cross-contamination of assemblies within the SEC column , we tested its ability to exclude larger seeds from the monomer fraction . We first isolated Ms and larger assemblies from a sonicated fibril preparation ( Figure 3 , Group 1 ) . Removing the fraction that contained Ms ( B5 ) , we then pooled the remaining fractions , and spiked them with Mi . We re-fractionated the material on SEC to isolate the monomer in fraction B5 again ( Figure 3 , Group 2 ) . As previously observed , Ms and other fibril-derived assemblies in Group 1 had seeding activity ( Figure 3 ) . However , in Group 2 , while we observed seeding activity in larger assemblies , the monomer ( which we take to be Mi ) re-isolated from a pool of larger fibril-derived assemblies had no seeding activity ( Figure 3 ) . This confirmed that larger , seed-competent assemblies do not appreciably contaminate the monomer fraction during SEC . Although prior controls had essentially excluded the presence of tau multimers in the sample , we used heat-mediated dissociation of oligomeric assemblies as an additional test for the possibility that Ms in fact represents a uniquely compact multimer that somehow purifies as a monomer . We collected Ms by SEC , and heated the sample to 95°C for 3 hr . We then re-isolated the sample via SEC . We carried out the same procedure with trimer and ~20 mer . In each case , we tested the resultant fractions for seeding activity . In the first instance , after heating we re-isolated Ms purely as monomer that retained virtually all of its seeding activity ( Figure 4A ) . The trimer assembly ( fraction B8 ) broke down to smaller assemblies , predominantly monomer , each of which retained seeding activity ( Figure 4B ) . The ~20 mer ( fraction A5 ) was largely stable following heat treatment , and retained its seeding activity ( Figure 4C ) . These experiments highlighted the lability of small multimers ( i . e . trimer ) , and a surprising persistence of seeding activity in heat-treated monomer . In the preceding experiment Ms retained seeding activity even after 3 hr at 95°C , a condition sufficient to dissociate trimers . These experiments implied that Ms consists of a stable seed-competent structure , resistant to heat denaturation . Consequently , we used more nuanced heat denaturation of seeding activity to probe the relative stabilities of Ms , dimer , trimer , and larger assemblies of FL WT tau . We first isolated tau monomer , dimer , trimer , ~10 mer , and ~20 mer on SEC . We then incubated the various assemblies at a range of temperatures ( 65 , 75 , 85 , 95°C ) and times ( 0 , 3 , 12 , 18 , 24 , 48 , 72 hr ) before measuring seeding activity . Lower temperatures only slightly reduced seeding activity , whereas exposure of Ms , dimer , and trimer to temperatures ≥ 85°C for 18–24 hr eliminated it at roughly the same rate for each ( Figure 4D–G ) . By contrast , the seeding activities of ~10 mer and ~20 mer were relatively heat-resistant ( Figure 4D–G ) . This was consistent with our prior observations that tau seeds derived from cultured cells are resistant to boiling ( Sanders et al . , 2014 ) . To determine a putative energy barrier between Ms and Mi , we evaluated the denaturation data for Ms by integrating the data from the prior experiments ( Figure 4H ) . We compared two models for the transition of Ms to an inert form ( which we assumed to be an unfolding reaction ) : a unimodal unfolding model vs . a multimodal model that assumes intermediate seed-competent states . The unimodal model did not account for the data at early time points , which indicated a lag phase in denaturation , whereas the multimodel model performed better ( Figure 4H ) . The lag phase in denaturation implied an ensemble of seed-competent states that define Ms , each separated by smaller energy barriers . Using the multimodal model , we calculated the barrier to conversion of Ms to an inert form to be ~78 kcal/mol . Aggregation of Mi in vitro is relatively slow , requires high protein concentration ( micromolar ) , and polyanions such as heparin ( Goedert et al . , 1996; Pérez et al . , 1996 ) . Based on the seeding activity of Ms , we predicted that it might more readily self-associate . We incubated FL WT tau Mi and Ms alone , or dimer or trimer at equimolar ratios , keeping total particle concentration constant at 500 nM . We then monitored change in assembly size over 24 hr . Mi , dimer , and trimer showed no evidence of self-association in this timeframe ( Figure 5A , C , D ) . By contrast , when incubated alone , Ms readily formed larger assemblies ( Figure 5B ) . When we incubated Mi with dimer or trimer , we saw no change in the assembly population over 24 hr ( Figure 5E , F ) . By contrast , when we mixed Ms with dimer or trimer we observed a growth of larger assemblies with a concomitant reduction in dimer and trimer peaks ( Figure 5G , H ) . We conclude that Mi , dimer , and trimer do not form larger assemblies at an appreciable rate , while Ms self-assembles and adds on to larger assemblies . The preparation of Ms based on sonication of fibrils raised two important issues . First , it left uncertain whether Mi could be converted to a seed-competent form without previously being incorporated into a fibril . Second , we observed that sonication could create fragments from tau monomer that might potentially act as seeds ( Figure 6—figure supplement 1 ) . Consequently , we used heparin to induce the formation of Ms , thereby avoiding sonication . We exposed FL WT tau to heparin for varying amounts of time before purifying different assembly sizes by SEC and testing for seeding activity . After 15 min of heparin exposure , we detected low but significant amounts of seed-competent monomer , and much fewer larger assemblies ( Figure 6A ) . Crosslinking of purified , heparin-induced Ms revealed no evidence of multimers or an increase in fragments ( Figure 6—figure supplement 1 ) . Recombinant monomer not treated with heparin had no seeding activity at any time point ( Figure 6A ) . At longer heparin treatment times ( 1 hr , 4 hr ) monomer fractions as well as larger assemblies all had strong seeding activity ( Figure 6A ) . Ms derived from heparin exposure was relatively resistant to heat denaturation at 95°C , albeit less so than fibril-derived Ms ( Figure 6B ) . Relative seeding efficiency of the various forms of Ms as well as sonicated or unsonicated fibrils were relatively similar ( Figure 6C ) . We noted also that sonication of Mi and purification by SEC did not produce any seed-competent species , eliminating the possibility that small assemblies of sonication-induced fragments accounted for seeding activity of Ms ( Figure 6C ) . These experiments also indicated that it is not necessary for tau monomer to be part of a fibril or to be exposed to sonication to produce an efficient seed-competent monomer . Heparin , presumably by catalyzing a transition from an inert to a seed-competent form , enables this critical conformational change . To probe the structures of Mi and Ms , we employed cross-linking with mass spectrometry ( XL-MS ) , which uses DSS-mediated crosslinking of proteins ( monomer or larger assembly ) followed by trypsin proteolysis , enrichment of resultant fragments by SEC , and identification by capillary liquid chromatography tandem mass spectrometry . This method creates restraints for structural models of single proteins or protein complexes ( Leitner et al . , 2012; Lasker et al . , 2012; Joachimiak et al . , 2014 ) . We assigned the complex fragment ion spectra to the corresponding peptide sequences using xQuest ( Rinner et al . , 2008 ) . Denaturation of recombinant tau with 8M urea prior to crosslinking produced no intramolecular cross-links ( data not shown ) , indicating that crosslinks observed under native conditions represented local structure . We studied Mi , fibril-derived Ms and heparin-derived Ms using XL-MS . Short reaction times ensured the production of only intra-molecular crosslinks as monitored by SDS-PAGE ( Figure 6—figure supplement 1 ) . XL-MS for each sample was carried out in triplicate , and only considering consensus crosslinks present in each replicate . Mi exhibited crosslink patterns which indicated local and distant intramolecular contacts ( Figure 7A ) . In Ms , we observed a consistent crosslinking of K150 with K254 , K267 , K274 or K280 all located between RD 1 and 2 . These crosslinks tracked exclusively with Ms , both fibril- or heparin-derived ( Figure 7B , C ) . We never observed these crosslinks in Mi . To test the relationship of this crosslink with seed function , we carried out heat denaturation at 95°C for 3 or 24 hr , followed by XL-MS . Heating samples results in a decrease in crosslink frequency ( Figure 7—figure supplement 1 ) . Importantly , however , we observed a parallel persistence of this crosslink pattern with seeding activity ( Figure 7B , C ) . The XL-MS results indicate a distinct structure and seeding activity for Ms that is surprisingly resistant to denaturation at 95°C . Given our experiments with recombinant Mi and Ms , we wished to test whether similar structures exist in vivo . We extracted AD and control brain samples using a dounce homogenizer to avoid liberating significant monomer from fibrils . We immunoprecipitated tau using an antibody that targets the amino-terminus ( HJ8 . 5 ) , and resolved the eluates by SEC , followed by ELISA to determine tau levels ( Figure 8A , B ) . Tau from control brain purified in the monomer fraction ( Figure 8A ) , while tau from AD brain distributed across multiple fractions , corresponding to monomer and larger assemblies ( Figure 8B ) . When we tested each fraction for seeding activity , we observed none in any control brain fraction ( Figure 8C ) . However , all AD fractions contained seeding activity , including monomer ( Figure 8C ) . To exclude the possibility that the brain homogenization protocol liberated Ms from neurofibrillary tangles , we spiked tau KO mouse brain samples with recombinant fibrils in vitro , or fibril-derived Ms . We then used dounce homogenization and immuno-purification as for human brain . We evaluated the seeding activity in total lysate , supernatant following 10 , 000 x g centrifugation , and SEC fractions ( Figure 8D ) . We readily observed monomer seeding activity in tau KO brain spiked with Ms , however we observed none in fractions that had been spiked with fibrils ( Figure 8D ) . The homogenization protocol for human brain was thus unlikely to have liberated Ms from pre-existing tau fibrils . To test for self-association of control-derived Mi vs . AD-derived Ms , we purified these species by SEC , and divided each monomer fraction in two . We snap-froze one fraction and incubated the other overnight at room temperature . Then , we again resolved the assemblies via SEC and tested each fraction for seeding activity . Control monomer was inert , even after incubation at RT ( Figure 8E ) . AD-derived Ms that was purified , frozen , and re-purified by SEC exhibited seeding activity exclusively in the monomer fraction ( Figure 8E ) . By contrast , AD-derived Ms incubated at RT formed seed-competent assemblies of increasing size ( Figure 8E ) . We concluded that , as for other types of Ms , AD-derived Ms exhibited an intrinsic capacity for self-association into seed-competent assemblies . To compare structures of control vs . AD-derived monomer via XL-MS , we isolated tau from brains of 3 AD patients and three age-matched controls . In control-derived monomer , we observed no evidence of the crosslink that marked Ms ( Figure 8G ) . However , in each AD-derived Ms sample we observed a discrete set of crosslinks between aa150 and aa259-290 ( Figure 8H ) . This essential finding did not change , no matter what method of homogenization we used ( Figure 8—figure supplement 1 ) , and implied a common structure that unifies ensembles of seed-competent tau monomer , whether produced in vitro or in vivo . Based on intramolecular FRET and electron paramagnetic resonance spin labeling Mandelkow et al . have previously proposed native tau structure to be in a ‘paperclip’ configuration , with the C-terminus folded over the RD ( Jeganathan et al . , 2006 ) . To understand how core elements of tau control its aggregation , we employed Rosetta to create models of tau structure for Mi and Ms using restraints from the crosslink patterns and length of the DSS crosslinker . The overall energetics and radii of gyration in the models were comparable for Mi and Ms ( Figure 9—figure supplement 1 ) , indicating global structural similarity . We thus focused on the RD , given its high frequency of intramolecular crosslinks , and primary role in aggregation ( Figure 9A ) . We observed differences in the predicted interface structure between R1/R2 and R2/R3 which encode two core VQIINK and VQIVYK motifs critical for tau amyloid formation ( von Bergen et al . , 2000; von Bergen et al . , 2001 ) . The Mi structural model predicted masking of VQIINK and VQIVYK sequences in compact ‘hairpin’ structures ( Figure 9B ) , similar to the structure of microtubule-bound tau previously determined by NMR ( Kadavath et al . , 2015 ) . By contrast , within Ms the model predicted relative exposure of VQIINK and VQIVYK ( Figure 9C ) . We next evaluated XL-MS-guided predictions of patient-derived tau , although lower sample quality and fewer high confidence crosslinks ( possibly due to protein heterogeneity ) limited our accuracy . As for recombinant protein , Mi from control patients also featured VQIINK/VQIVYK sequences in a less accessible configuration ( Figure 9D ) . In AD-derived Ms , long-range contacts from aa150 to R2 influenced the model , and predicted an exposed configuration of VQIINK/VQIVYK ( Figure 9E ) . With important caveats , the models guided by XL-MS imply that the general difference between Mi and Ms derives from relative shielding vs . exposure of VQIINK/VQIVYK sequences . As an orthogonal comparison of the structures of Mi and Ms , we used limited proteolysis with trypsin . Mi or Ms ( heparin-exposed ) that had been passed through a 100kD filter immediately prior were subjected to a fine time course of limited proteolysis ( Figure 10A ) . Each sample was prepared in triplicate with matched protein quantities to facilitate label-free analysis . We then used mass spectrometry to evaluate the production of tau fragments and mapped these to specific cleavage sites ( Figure 10B ) . We identified 60 peptides common across the two conditions ( Figure 10—figure supplement 1 ) . To summarize enrichment of peptides across the two datasets we compared the ratio of averaged kinetic profiles ( Figure 10—figure supplement 1 ) . Differences between the Mi and Ms primarily localized to the RD ( Figure 10—figure supplement 1 ) . In Mi , an R1R2 fragment was enriched ( Figure 10C ) while only the R2 portion of that fragment was enriched in Ms ( Figure 10D ) . We observed similar patterns in R2R3 ( Figure 10F , G ) . By contrast , other domains outside of these regions had similar cleavage kinetics in Mi and Ms ( Figure 10E , H , Figure 10—figure supplement 1 ) . Mapping these cleavage sites onto our structural models revealed that proteolysis in Mi preferentially occurred outside the hairpin that includes VQIINK and VQIVYK amyloid sequences , while cleavage in Ms occurred adjacent to the amyloid sequences ( Figure 10I , J ) . The cleavage patterns were thus consistent with structural models of VQIINK and VQIVYK regions , which predicted relative inaccessibility of hairpin-associated sequences in Mi , and accessibility in Ms . We propose that tau monomer occupies two distinct and stable conformational ensembles . One set of structures ( collectively termed Mi ) is relatively inert , while another has intrinsic ability to self-assemble , and acts as a template , or seed , for fibril growth in vitro and in cells ( collectively termed Ms ) . Multiple controls indicated that our original preparation of fibril-derived Ms is in fact a monomer , uncontaminated by larger assemblies . Tau monomer purified from AD brain also had intrinsic seeding activity , and self-associated to produce larger seed-competent assemblies . A model restrained by the XL-MS data , and consistent with biochemical studies , predicts that VQIVYK and VQIINK sequences assume an open configuration in all types of Ms ( fibril-derived , heparin-induced , and AD-derived ) . By contrast , the model predicts lack of VQIINK/VQIVYK exposure in Mi . Limited proteolysis studies are consistent with this idea , although clearly more detailed biochemical , biophysical , and structural analyses will be needed to test its validity . Taken together , these data establish a new concept for tau: this intrinsically disordered protein has multiple , stable monomeric states , functionally distinguished by the presence or absence of seeding activity . Amyloid proteins form progressively larger assemblies over time , and it has been difficult to define the composition of the minimal seed . Mandelkow and colleagues studied tau aggregation in vitro and concluded that a seed of 8–12 molecules existed in their experimental system ( Friedhoff et al . , 1998 ) . By contrast , Kuret and colleagues posited an ‘intermediate’ of tau that could subsequently initiate self-assembly , and their data , based on extrapolation of tau concentrations needed to enable development of thioflavin fluorescence in vitro , were consistent with a monomeric seed ( Chirita et al . , 2005 ) . Wetzel and colleagues also proposed that a monomer is the basis of a ‘thermodynamic nucleus’ that templates the aggregation of synthetic polyglutamine peptides ( Bhattacharyya et al . , 2005 ) . However , no prior study has previously identified stable forms of tau monomer that seed amyloid formation . The actual cause of tau aggregation in tauopathies is unknown . It has been proposed that dissociation of tau monomer from microtubules , possibly due to phosphorylation , allows high concentration and self-association to form pathogenic assemblies ( Mandelkow and Mandelkow , 2012 ) . In this study , using a single source of recombinant protein , we define distinctly structured seed-competent and inert forms of tau . We have similarly identified seed-competent species in human brain . In reality ‘seed-competent’ and ‘inert’ forms of tau almost certainly represent multiple structural ensembles separated by defined energy and/or kinetic barriers . The barrier for conversion of an inert to a seed-competent form of tau can apparently be overcome by incubation with heparin and/or incorporation into a fibril . In neurons , other factors such as post-translational modifications and heterologous binding events likely play a role . Identification of the factors that trigger conversion from inert to seed-competent forms will thus have obvious implications for understanding disease mechanisms . Isolation of seed-competent monomer from AD brain , with a very mild purification that explicitly excludes sonication or vigorous tissue homogenization , strongly suggests that this form of tau exists in vivo . Furthermore , we observed that both recombinant Ms and AD-derived Ms build multimeric assemblies in vitro far more efficiently than Mi or control-derived monomer . Thus , we hypothesize that a uniquely structured form of tau may be required for efficient assembly growth in cells . This contrasts with the idea that multimeric assemblies uniquely stabilize the conformation of otherwise unstructured proteins as they incorporate into the growing fibril , or that liquid-liquid phase separation with extremely high local concentration underlies tau aggregation ( Wegmann et al . , 2018 ) . Instead , we imagine that the initiation of aggregation in human brain might begin with a stable transition of tau monomer from an inert to a seed-competent form . To fully study this process will require more extensive biochemical purification of tau Ms from the earliest stages of disease . Ms has a remarkably stable structure , as it resists heat denaturation at 95°C for up to 3 hr . This suggests a heretofore unrecognized conformation of tau that , to account for its slow denaturation , likely involves multiple intra-molecular interactions involving short and long range amino acid contacts . XL-MS provides some indication of what these might be , and crosslinks between aa150 and R1/R2 appear to mark a seed-competent conformation . In agreement with the XL-MS results , we observed that heat inactivation of Ms seeding activity occurs with a lag phase , rather than first order time-dependent decay . This implies a complex tertiary structure in which Ms has multiple seed-competent intermediates . Future XL-MS studies performed at different temperatures could reveal these structures . With more advanced methods to interrogate the structure of monomeric tau in patient material , we imagine that ‘seed-competent monomer’ will in fact represent myriad structures , depending on the underlying disease . This could provide an explanation for how a single tau protein might self-assemble into diverse amyloid strains . We note with excitement a recent study of the yeast prion Sup35 from the Tanaka laboratory . Like tau , Sup35 is intrinsically disordered , yet they have observed local structure that influences the conformations of fibrils it can form ( Ohhashi et al . , 2018 ) . Without further studies to identify structures of tau at higher resolution , we cannot know for certain why one form acts as a seed and another does not . However , we gained important insights when we modeled the configurations of R1R2 and R2R3 using Rosetta , with crosslinks as restraints . With obvious caveats , our models predicted that the local environment surrounding two hexapeptide motifs , VQIINK and VQIVYK , which are required for tau to form amyloid structures , may explain the differences between seed-competent and inert forms . In the models of Mi , and control brain-derived tau , these motifs lie buried in hairpin structures . By contrast , in Ms and AD-derived tau , both are exposed . VQIINK and VQIVYK thus might mediate intermolecular interaction in a growing assembly . In support of our structural model , the proteolysis experiments corroborate differences in exposure of the VQIINK and VQIVYK sequences in the R1R2 and R2R3 regions between Mi and Ms . We note with great enthusiasm the recent study of Fitzpatrick et al . ( Fitzpatrick et al . , 2017 ) , which defined critical sequences of tau within the amyloid core that are based on VQIVYK and adjacent amino acids . Indeed , it has been recently observed that heparin binding involves residues spanning 270–290 , and promotes expansion of the remainder of the molecule ( Zhao et al . , 2017 ) . This is consistent with our predictions of relative exposure of VQIINK/VQIVYK . The diversity of exposed core elements ( almost certainly beyond VQIINK/VQIVYK ) could specify the formation of assemblies that give rise to distinct strains , as suggested by work from the Tanaka laboratory ( Ohhashi et al . , 2018 ) . Consistent with this idea , the Fitzpatrick et al . study indicates that in AD-derived tau fibrils the VQIVYK sequence plays a key role in the core amyloid structure ( along with adjacent amino acids ) , but the VQIINK sequence does not ( Fitzpatrick et al . , 2017 ) . We also note that multiple disease-associated mutations in tau affect residues in close proximity to VQIINK/VQIVYK . For example , our models predict that serine or leucine substitutions at P301 ( which cause dominantly inherited tauopathy ) would uniquely destabilize the local structure and promote exposure of the VQIINK/VQIVYK sequences . Future experiments will test these ideas more definitively . We utilized several forms of recombinant tau . Full-length ( FL ) , wild-type ( WT ) tau contains two cysteines that create disulfide bridges and could complicate isolation of monomer . Thus in addition to preparing FL WT tau ( 2N4R ) as previously described ( Frost et al . , 2009b ) , we purified FL tau ( 2N4R ) that contains two cysteine/alanine substitutions ( C291A , C322A ) , termed tau ( 2A ) . We used the 2A and WT forms of tau in our initial studies , before exclusively studying WT . Additionally , for fluorescence correlation spectroscopy ( FCS ) , we engineered a single cysteine at the amino terminus ( Cys-Tau ( 2A ) ) for labeling via maleimide chemistry . These modified proteins have fibrillization and seeding properties similar to FL WT tau . To initiate fibrillization , we incubated 8 µM tau in 10 mM HEPES , 100 mM NaCl , and 8 µM heparin ( 1:1 ratio of FL tau to heparin ) at 37°C for 72 hr without agitation . For cysteine labeling , we incubated 200 µL of 8 µM fibrils ( monomer equivalent ) and monomer with 0 . 025 mg of Alexa Fluor-488 ( AF488 ) C5-maleimide ( Invitrogen ) and 80 µM Tetramethylrhodamine-5-maleimide ( Sigma-Aldrich ) overnight at 4°C with gentle rotation . We quenched excess dye with 10 mM DTT for 1 hr at room temperature . For limited heparin exposure , recombinant tau at 1 µM was incubated with heparin at 1 µM for 15 min , 1 hr and 4 hr at 37°C before purification of monomer via Superdex 200 column . We employ the following terminology: Mi: This refers to ‘inert’ tau monomer , whether recombinant or derived from control brain . Ms: This refers to ‘seed competent’ monomer , whether derived from sonicated fibrils , heparin-treated monomer , or AD brain . We sonicated labeled and non-labeled fibrils using a Q700 Sonicator ( QSonica ) at a power of 100–110 watt ( Amplitude 50 ) at 4°C for 3 hr . Samples were then centrifuged at 10 , 000 x g for 10 min and 1 mL of supernatant was loaded into a Superdex 200 Increase 10/300 GL column ( GE Healthcare ) and eluted in PBS buffer at 4°C . After measuring the protein content of each fraction with a Micro BCA assay ( Thermo Scientific , Waltham MA ) and/or fluorescence using a plate reader ( Tecan M1000 ) , we aliquoted and stored samples at −80°C or immediately used them in biochemical studies and cell seeding assays . Each aliquot was thawed immediately before use . The molecular weight/radius of proteins in each fraction was estimated by running gel filtration standards ( Bio-Rad ) : Thyroglobulin ( bovine ) 670 kDa/8 . 5 nm; γ-globulin ( bovine ) 158 kDa/5 . 29 nm; Ovalbumin ( chicken ) 44 kDa/3 . 05 nm; myoglobin ( horse ) 17 kDa/2 . 04 nm; and vitamin B121 . 35 kDa/0 . 85 nm . In a prior publication ( Mirbaha et al . , 2015 ) , Figure 1E , we demonstrated through use of crosslinking with SDS-PAGE that the SEC protocol used in this work reliably purifies monomer , dimer , and trimer . Monomer , dimer and trimer fractions were passed through a 100 kDa MWCO filter ( Corning ) as instructed by the manufacturer ( centrifuged at 15 , 000 x g for 15 min at 4°C ) . Filtered material was immediately collected and used in seeding assay along with the non-filtered samples of the same fraction at a final concentration of 100 nM , or analyzed by limited proteolysis . Protein concentration was determined before and after filtration by determining absorption at 205 nm . Circular dichroism ( CD ) measurements were performed at 25°C on a Jasco J-815 spectropolarimeter using a 0 . 1 cm optical path length . 200 μL of 2 μM Ms or Mi monomer was dialyzed onto 10 mM NaP and the spectra were measured at 0 . 10 nm intervals , with a band width of 1 . 0 nm , and scan speed of 10 nm/min . The spectrum represents the average of 4 scans in the range of 195 to 250 nm . A total tau ‘sandwich’ ELISA was performed similarly to that described previously ( Acker et al . , 2013 ) . Antibodies were kindly provided by Dr . Peter Davies ( Albert Einstein College of Medicine ) . 96-well round-bottom plates ( Corning ) were coated for 48 hr at 4°C with DA-31 ( aa 150–190 ) diluted in sodium bicarbonate buffer ( 6 µg/mL ) . Plates were rinsed with PBS three times , blocked for 2 hr at room temperature with Starting Block ( Pierce ) , and rinsed with PBS five additional times . SEC fractions were diluted in SuperBlock solution ( Pierce; 20% SuperBlock , diluted in TBS ) , and 50 µL sample was added per well . DA-9 ( aa 102–150 ) was conjugated to HRP using the Lighting-Link HRP Conjugation Kit ( Innova Biosciences ) , diluted 1:50 in SuperBlock solution , and 50 µL was added per well ( 15 µg/mL ) . Sample +detection antibody complexes were incubated overnight at 4°C . Plates were washed with PBS nine times with a 15 s incubation between each wash , and 75 µL 1-Step Ultra TMB Substrate Solution ( Pierce ) was added . Plates were developed for 30 min , and the reaction quenched with 2M sulfuric acid . Absorbance was measured at 450 nm using an Epoch plate reader ( BioTek ) . Each plate contained a standard curve , and all samples were run in triplicate . FCS measurements were conducted on a Confocal/Multiphoton Zeiss LSM780 Inverted microscope ( Carl Zeiss-Evotec , Jena , Germany ) , using a 40X water immersion objective as previously described ( Chattopadhyay et al . , 2002 ) . Fluorescently labeled tau from SEC fractions ( in PBS ) was excited at 488 nm and 561 nm for 30 s , recording 10 times ( Buschmann et al . , 2003 ) . The data analysis was performed with Origin 7 . 0 ( OriginLab , Northampton , MA ) . Stable cell lines were plated at a density of 35 , 000 cells per well in a 96-well plate . After 18 hr , at 60% confluency , cells were transduced with protein seeds . Transduction complexes were made by combining [8 . 75 μL Opti-MEM ( Gibco ) +1 . 25 μL Lipofectamine 2000 ( Invitrogen ) ] with [Opti-MEM + proteopathic seeds] for a total volume of 20 μL per well . Liposome preparations were incubated at room temperature for 20 min before adding to cells . Cells were incubated with transduction complexes for 24 hr . Biosensor cells were confirmed as HEK293T by PowerPlex sequencing . Mycoplasma contamination was ruled out by PCR analysis using VenorGem ( Sigma ) . Cells were harvested with 0 . 05% trypsin and fixed in 2% paraformaldehyde ( Electron Microscopy Services ) for 10 min , then resuspended in flow cytometry buffer . The MACSQuant VYB ( Miltenyi ) was used to perform FRET flow cytometry . To measure CFP and FRET , cells were excited with a 405 nm laser , and fluorescence was captured with 405/50 nm and 525/50 nm filters , respectively . To measure YFP , cells were excited with a 488 nm laser and fluorescence was captured with a 525/50 nm filter . To quantify FRET , we used a gating strategy similar to that previously described ( Holmes et al . , 2014 ) . The integrated FRET density ( IFD ) , defined as the percentage of FRET-positive cells multiplied by the median fluorescence intensity of FRET-positive cells , was used for all analyses . For each experiment , ~20 , 000 cells were analyzed in triplicate . Analysis was performed using FlowJo v10 software ( Treestar ) . Recombinant full length ( 0N4R ) tau monomer was purified as previously described ( Morozova et al . , 2013 ) at 1 mg/mL in BRB80 buffer ( 80 mM PIPES , 1 mM MgCl2 , 1 mM EGTA , pH 6 . 8 with 0 . 3M NaCl ) and boiled at 100°C for 5 min with 25 mM β-mercaptoethanol . The tau protein solution was then rapidly diluted 1:5 and cooled to 20°C in PBS , pH 7 . 4 , to a final concentration of 0 . 2 mg/mL of tau and 5 mM β-mercaptoethanol . This solution was supplemented with Thioflavin T ( ThT ) to a final concentration of 20 µM and filtered through a sterile 0 . 2 µm filter . Reaction sizes of 195 µL were aliquoted from the prepared protein stock and thoroughly mixed with 5 µL of each sample at 100 nM monomer equivalent , or 5 µL of buffer control . For each sample , three different technical replicates were prepared . An opaque 96-well plate was prepared with a 3 mm glass bead added to each well to increase agitation . The recombinant tau solution was added to the plate in 200 µl reaction volumes . The plate was sealed with sealing tape to prevent evaporation and incubated in the plate reader ( SpectraMax M2 ) at 37°C . ThT fluorescence was monitored over time with excitation and emission filters set to 444 nm and 485 nm , respectively . Fluorescence readings were taken every 5 min , with agitation for 5 s before each reading . 0 . 5 g frontal lobe sections from AD patients at late Braak stage ( VI ) and age-matched controls lacking evident tau pathology were gently homogenized at 4°C in 5 mL of TBS buffer containing protease inhibitor cocktails ( Roche ) using a dounce homogenizer . Samples were centrifuged at 21 , 000 x g for 15 min at 4°C to remove cellular debris . Supernatant was partitioned into aliquots , snap frozen and stored at −80°C . Immunopurification was performed with HJ8 . 5 anti-tau antibody ( Yanamandra et al . , 2013 ) at a ratio of 1:50 ( 1 µg mAb per 50 µg of total protein ) , incubating overnight at 4°C while rotating . To each 1 mL of mAb/brain homogenate we added 200 µL of a 50% slurry protein G-agarose beads ( Santa-Cruz ) . We washed the bead with TBS buffer before overnight incubation at 4°C . We then centrifuged the complexes at 1000 x g for 3 min and discarded the supernatant . Beads were washed with Ag/Ab Binding Buffer , pH 8 . 0 ( Thermo Scientific ) three times . Tau bound to the beads was eluted in 100 µL low pH elution buffer ( Thermo Scientific ) , incubated at room temperature for 7 min , followed by neutralization with 10 µL Tris-base pH 8 . 5 . This elution step was repeated once more with 50 µL elution buffer and 5 µL Tris-base pH 8 . 5 for a total of 165 µL . Samples were then centrifuged at 10 , 000 x g for 10 min , and the supernatant loaded onto a Superdex 200 Increase 10/300 GL column ( GE Healthcare ) . SEC fractions were frozen at −80°C after evaluation of protein content by Micro BCA assay ( Thermo Scientific ) . To compare different extraction methods , fresh frozen frontal lobe section from an AD patient brain was suspended in TBS buffer containing protease inhibitor cocktails ( Roche ) at 10% w/vol in four portions . Samples were homogenized using three different devices: a dounce homogenizer , probe sonicator ( Omni International ) , and tissue homogenizer ( Power Gen 125 , Fischer Scientific ) . We also included one more condition of homogenizing with tissue homogenizer followed by probe sonication for 10 min . Samples were centrifuged at 21 , 000 x g for 15 min at 4°C to remove cellular debris . Supernatant was partitioned into aliquots followed by immunopurification . To control for release of tau Ms from fibrils in AD brain , a tau KO Mouse brain was divided into two halves , followed by spiking one half with recombinant fibrils and the other with fibril-derived Ms , both at final concentration of 10 µM monomer equivalent . Each was dounce homogenized , centrifuged , immunoprecipitated with HJ8 . 5 anti-tau antibody , and fractionated by SEC with identical techniques as used for human brain processing . SEC fractions were then used in seeding experiments . We analyzed the IFD from measurements of temperature-dependent seeding using global fits to a proposed unimolecular heat denaturation reaction . This analysis rests on the Arrhenius equation Laidler , 1984:kU=Ae-ERTwhere kU is the unfolding rate constant , E is the activation energy , R is the gas constant , T is the temperature , and A is the pre-exponential factor . For the unimodal model , the data were fit globally to:IFD ( t ) =100e-t/τ Here , t is the heat denaturation time and τ = 1/kU is the unfolding time . A second , multimodal model was deployed to account for discrepancies in the early time points which appeared to suggest the presence of a lag phase in denaturation . In this model , the data were fit globally toIFD ( t ) =100;t≤ltIFD ( t ) =100e− ( t−lt ) /τ;t>ltwhere lt is the lag time given by1/lt=Be-ERTand B is a pre-exponential factor . We used the Akaike information criterion ( AIC ) to evaluate the best model as it quantifies the trade-off between goodness of fit and the complexity of the model ( von Bergen et al . , 2001 ) . For least squares model fitting , AIC can be reduced to:AIC=2p+nln ( RSS/n ) where p is the number of parameters in the model , n is the number of observations , and RSS is the residual sum of squares . The preferred model is the one with the minimum AIC . Here , we find AIC = 123 for the unimodal model and AIC = 105 for the multimodal model , which suggests the multimodal model is a better description of the denaturation data . Mi and Ms tau samples were prepared as described above . In all cases , tau preparations were crosslinked at a total protein concentration of ~0 . 1 mg/mL using 10–20 µg starting material . The crosslinking buffer was 50 mM HEPES-KOH ( pH 7 . 4 ) containing 150 mM NaCl and 1 mM DTT . The crosslinking reaction was initiated by adding disuccinimidyl suberate ( DSS ) stock solution ( 25 mM DSS-d0 and –d12 , Creative Molecules ) in DMF to a final concentration of 1 mM . Samples were incubated at 37°C for 1 min . For the heparin-derived Ms sample , heparin sulfate ( Sigma ) was added to a final concentration of 5 µM , followed by 1 mM DSS and the samples were incubated for 1 min at 37°C . Excess reagent was quenched by addition of ammonium hydrogen carbonate to 50 mM and incubation at 37°C for 30 min , and then flash frozen at −80°C . Absence of higher molecular species was confirmed by SDS-PAGE and coomassie stain . After the quenching step , samples were evaporated to dryness in a vacuum centrifuge and resuspended in 8M urea . Proteins were reduced with 2 . 5 mM TCEP ( 37°C , 30 min ) and alkylated with 5 mM iodoacetamide ( 30 min , room temperature , protected from light ) . The sample solutions were diluted to 1M urea with 50 mM ammonium hydrogen carbonate and trypsin ( Promega ) was added at an enzyme-to-substrate ratio of 1:50 . Proteolysis was carried out at 37°C overnight followed by acidification with formic acid to 2% ( v/v ) . Samples were then purified by solid-phase extraction using Sep-Pak tC18 cartridges ( Waters ) according to standard protocols . Samples were fractionated by size exclusion chromatography ( SEC ) on a Superdex Peptide column as described elsewhere ( Leitner et al . , 2012 ) . Two fractions collected from SEC were evaporated to dryness and reconstituted in water/acetonitrile/formic acid ( 95:5:0 . 1 , v/v/v ) to a final concentration of approximately 0 . 5 µg/µl . 2 µL each were injected for duplicate LC-MS/MS analyses on an Eksigent 1D-NanoLC-Ultra HPLC system coupled to a Thermo Orbitrap Fusion Lumos Tribrid system ( Thermo Scientific ) . Peptides were separated on self-packed New Objective PicoFrit columns ( 11 cm x 0 . 075 mm I . D . ) containing Magic C18 material ( Michrom , 3 µm particle size , 200 Å pore size ) at a flow rate of 300 nL/min using the following gradient . 0–5min = 5 %B , 5–95min = 5–35%B , 95–97 min = 35–95%B and 97–107min = 95 %B , where A = ( water/acetonitrile/formic acid , 97:3:0 . 1 ) and B = ( acetonitrile/water/formic acid , 97:3:0 . 1 ) . The mass spectrometer was operated in data-dependent mode by selecting the five most abundant precursor ions ( m/z 350–1600 , charge state 3 + and above ) from a preview scan and subjecting them to collision-induced dissociation ( normalized collision energy = 35% , 30 ms activation ) . Fragment ions were detected at low resolution in the linear ion trap . Dynamic exclusion was enabled ( repeat count 1 , exclusion duration 30 s ) . Thermo . raw files were converted into the open . mzXML format using msconvert ( proteowizard . sourceforge . net ) and analyzed using an in-house version of xQuest ( Rinner et al . , 2008 ) . Spectral pairs with a precursor mass difference of 12 . 075321 Da were extracted and searched against the respective FASTA databases containing Tau ( TAU_HUMAN P10636-8 ) . xQuest settings were as follows: Maximum number of missed cleavages ( excluding the crosslinking site ) =2 , peptide length = 5–50 aa , fixed modifications = carbamidomethyl Cys ( mass shift = 57 . 021460 Da ) , mass shift of the light crosslinker = 138 . 068080 Da , mass shift of mono-links = 156 . 078644 and 155 . 096428 Da , MS ( Chirita et al . , 2005 ) tolerance = 10 ppm , MS ( Kar et al . , 2011 ) tolerance = 0 . 2 Da for common ions and 0 . 3 Da for crosslink ions , search in ion-tag mode . For brain-derived samples we also included variable modifications including: Methionine oxidation = 15 . 99491 , Ser/Thr/Tyr phosphorylation = 79 . 96633 and lysine ubiquitylation = 114 . 043 with nvariable_mod = 1 . Post-search manual validation and filtering of the recombinant samples was performed using the following criteria: xQuest score >16 , mass error between −4 and +7 ppm , %TIC >10 , and a minimum peptide length of six aa . In addition , at least four assigned fragment ions ( or at least three contiguous fragments ) were required on each of the two peptides in a crosslink . False discovery rates ( FDR ) for the identified crosslinks were estimated using xprophet ( Rinner et al . , 2008 ) . For the recombinant samples , Mi and Ms , the FDR ranged from 6–10% . Post-search manual validation of the brain-derived samples was performed using the following criteria: xQuest score >7 , mass error between −5 and +7 ppm , %TIC >10 , and a minimum peptide length of six aa . In addition , at least four assigned fragment ions ( or at least three contiguous fragments ) were required on each of the two peptides in a crosslink . The FDRs for the brain samples were much higher and ranged between 20–25% . For triplicate datasets ( N = 3 ) corresponding to the Mi and Ms boiling time course we computed consensus crosslink profiles enforcing that at least two of the three datasets contain a crosslink . Crosslink data were visualized using Xvis ( Grimm et al . , 2015 ) . Average contact distance was computed by averaging the sequence separation between crosslink pairs in a given dataset . See Figure 7—source datas 1 and 2 . High confidence crosslink pairs identified above were used to generate an ensemble of possible structures using a Rosetta protocol employing the crosslink pairs as structural restraints . The integration of XL-MS derived restraints have been previously used to refine structural models of large complexes ( Leitner et al . , 2012 ) and simpler heterodimeric complexes ( Kahraman et al . , 2013 ) . Based on distance distributions of crosslink pairs mapped onto crystallographic structures we set a lower bound of 15 Å and an upper bound of 25 Å for lysine Cα pairs in our simulations . Importantly , in our simulations we weighted the constraint pairs as to allow some distances above the upper bound limit . The fragment library was supplanted by using chemical shifts derived from fibrillar tau ssNMR assignments ( bmrb entry 17920 ) using csrosetta ( Lange et al . , 2012 ) . We generated 1000 models for each of the four XL-MS datasets on a high performance cluster ( biohpc . swmed . edu ) . Representative structures were selected according to the low Rosetta score and radius of gyration . All plots were generated with gnuplot . All figures were generated using Pymol . See Figure 9—source data 1 . AbinitioRelax . default . linuxgccrelease -in:file:fasta tau . fasta -file:frag3 tau . frags3 . dat -file:frag9 tau . frags9 . dat -nstruct 1000 -abinitio::increase_cycles 0 . 5 -abinitio::relax -score::weights score13_env_hb -abinitio::rg_reweight 0 . 5 -abinitio::rsd_wt_helix 0 . 5 -abinitio::rsd_wt_loop 0 . 5 -disable_co_filter true -out:file:silent csrosetta . out -constraints:cst_fa_file tau . cst -constraints:cst_file tau . cst -constraints:cst_weight 0 . 1 -constraints:cst_fa_weight 0 . 1 -loopfcst::coord_cst_weight 10 . 0 Group mean values were analyzed by one-way ANOVA with Bonferroni post hoc significant differences test using GraphPad prism five software . Data in text and figures are represented as mean ± SD . Biological Replicates: This refers to separately generated samples , e . g . analysis of samples from different individuals , in which the input represents a distinct biological source , or analysis of fibril preparations formed independently . Technical Replicates: This refers to independently analyzed samples in which each would be expected otherwise to be identical , e . g . multiple wells in which the same cell line was treated with the same sample , or when a single brain is broken up into three fractions for identical analyses . Outliers: We did not exclude outliers in any case . Inclusion/Exclusion of data: In no case did we include or exclude data . Figure 1: Biochemical purification of Mi and Ms and seeding into cells was carried out ~50 times over 3 years . In Figure 1 the data shown is a representative dataset for Mi and Ms on SEC , seeding , microscopy and seeding following filtration . In vitro Ms and Mi aggregation was done once ( using N = 3 independent technical replicates per fraction studied ) . Data from the titration of Mi and Ms seeding in cells is from a single representative experiment ( from at least three independent studies ) , using technical triplicates for each sample . Primary data is available in Figure 1—source data 1 . Figure 2: CD experiments were carried out with three biological replicates . A single representative experiment is shown , and the data for Mi and Ms is an average of 4 technical replicates . The FCS experiment was repeated three times with biological replicates . A single representative scan is shown . Primary data is available in Figure 2—source data 1 . Figure 3: Seeding activity was measured in technical triplicates . Primary data is available in Figure 3 source data 1 . Figure 4: Heat denaturation was done in two biological replicates , for which a representative experiment is shown . Seeding activity was performed in technical triplicates . Primary data is available in Figure 4—source data 1 . Figure 5: Stability of monomers and assemblies in solution was tested in a single experiment , with 32 s analyses . Primary data is available in . Figure 6: This single experiment is representative of at least three distinct biological replicates . Seeding activity was determined in technical triplicates ( N = 3 ) . Primary data is available in Figure 6—source data 1 . Figure 6—figure supplement 1: This experiment was performed in three biological replicates . A representative gel is shown . Figure 7: Each XL-MS dataset was collected in biological triplicate across the different conditions ( Mi: Recombinant; Ms: Fibril-derived; Ms: Heparin-derived ) . The figure indicates consensus crosslinks across triplicates for each condition . Primary data is available in Figure 7—source data 1 , and Figure 7—source data 2 . Figure 7—figure supplement 1: This represents a histogram of consensus pairs across triplicates for each condition depicted in Figure 7 . Primary data is available in Figure 7—source data 1 , and Figure 7—source data 2 . Figure 8: SEC was performed at least three times , and a single representative experiment is shown . The seeding assay was performed in technical triplicate for each fraction . Spiking of tau KO brain with fibrils or Ms was performed once , and the SEC was performed once , with the seeding assay performed in technical triplicate . XL-MS was performed in biological triplicate for Controls vs . AD . Primary data is available in Figure 8—source datas 1 and 2 . Figure 8—figure supplement 1: XL-MS data from different methods of homogenization was performed in four individual experiments on a single biological sample . Primary data is available primary data available in . Figure 9: Rosetta simulations produced models from 1000 independent trajectories across four different conditions: Mi , Ms , Control , AD . The pymol session is available in Figure 9—source data 1 . Figure 9—figure supplement 1: Plot of the energetics of the ensemble of 1000 Rosetta models built for each condition . Primary data is available in Figure 9—source data 2 . Figure 10: Proteolysis experiments were carried out for Mi and Ms in technical triplicate for each time point . Primary data is available in Figure 10—source data 1 . Figure 10—figure supplement 1: Summary data from Figure 10 , comparing peptide abundance between Mi and Ms is represented as technical triplicates for each data point presented . Primary data is available in Figure 10 source data 2 . Limited proteolysis of Mi and Ms using trypsin was carried out in triplicate ( N = 3 ) in 50 mM TEAB at 25°C . The enzyme to tau ratio was adjusted to 1:100 ( wt/wt ) with around 11 ug of Mi/Ms present initially . The total reaction mixture volume was 60 µl . Aliquots were withdrawn from the reaction mixture at 1 , 5 , 15 , 30 , 60 and 120 min by using 10 µL of 10% trifluoroacetic acid ( TFA ) to quench the reaction ( PH <3 ) . The trypsin-digested peptides were then desalted using an Oasis HLB plate ( Waters ) and eluted with 100 µL 80% acetonitrile ( ACN ) containing 0 . 1% TFA . The solvent was evaporated in a SpeedVac concentrator and the dried samples were reconstituted in 20 µl of 2% acetonitrile , 0 . 1% TFA and 2 µl solution was used for by LC/MS/MS analysis , the analysis were performed on an Orbitrap Elite mass spectrometer ( Thermo Electron ) coupled to an Ultimate 3000 RSLC-Nano liquid chromatography systems ( Dionex ) . Samples were injected onto a 75 µm i . d . , 15 cm long EasySpray column ( Thermo Scientific ) , and eluted with a gradient from 1 to 28% buffer B over 60 min . Buffer A contained 2% ( v/v ) ACN and 0 . 1% formic acid in water , and buffer B contained 80% ( v/v ) ACN , 10% ( v/v ) trifluoroethanol , and 0 . 1% formic acid in water . The mass spectrometer operated in positive ion mode with a source voltage of 2 . 8kV and an ion transfer tube temperature of 275°C . MS scans were acquired at 240 , 000 resolution in the Orbitrap and up to 14 MS/MS spectra were obtained in the ion trap for each full spectrum acquired using collision-induced dissociation ( CID ) , with charge one ions rejected . Dynamic exclusion was set for 15 s after an ion was selected for fragmentation . Raw MS data files were searched against the appropriate protein database from Uniprot , and reversed decoy sequences appended ( Elias and Gygi , 2010 ) by using Protein Discovery 2 . 2 ( Thermo Fisher Scientific ) . Fragment and precursor tolerances of 20ppm and 0 . 6 Da were specified , and 12 missed cleavages were allowed . Carbamidomethylation of Cys was set as a fixed modification and oxidation of Met was set as a variable modification . Label-free quantitation of proteins across samples was performed . Average peptide intensity values were computed for all time points for each peptide across triplicates ( N = 3 ) . To estimate differences in kinetic profiles we calculated the median value of each profile and compared the Mi to Ms ratio . See Figure 10—source datas 1 and 2 .	When doctors perform autopsies to look at the brain tissue of people with Alzheimer’s disease , they find toxic buildups of certain proteins – in particular , a protein called tau – in structures called ‘aggregates’ . People with more severe dementia have more tau aggregates in their brain . Aggregates form when individual proteins stick together in repetitive patterns , much like the way a single Lego block might attach to another identical one . Like all proteins , tau is built from a string of amino acids that folds into a specific shape . Normally folded tau proteins do not form aggregates . It was not clear that an individual tau protein had two distinct forms—structures associated with health ( “good” ) or disease ( “bad” ) . Mirbaha et al . have now studied the folding pattern of purified tau proteins with a sophisticated technology called mass spectrometry . This technique can measure changes in tiny amounts of protein . Some of the purified proteins had been extracted from human brains ( from people with and without Alzheimer’s ) . To detect which of the proteins were toxic , Mirbaha et al . also grew simple human cells in a dish that were designed to react specifically to the bad forms of tau . This allowed the good and bad forms of tau to be isolated . Mirbaha et al . discovered that in the good form of tau the parts of the protein that allow it to stick to itself are hidden , folded inside . By contrast , the bad form of tau exposes the parts that allow it to aggregate , enabling the protein to build upon itself to form a large , toxic assembly . The shape-shifting concept established by Mirbaha et al . might apply to other proteins that form toxic aggregates . This could help us to better understand how many other neurodegenerative diseases develop and progress . Recognizing that the shapes that tau forms can be categorized as either ‘good’ or ‘bad’ may also help to develop new treatments for Alzheimer’s disease . Drugs could be designed to stabilize the good form of tau , or to help remove the bad form from the brain . Furthermore , if the shape-shift described by Mirbaha et al . can be measured early enough in patients , it may allow treatments for Alzheimer’s before people have developed any detectable symptoms .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "neuroscience" ]	2018	Inert and seed-competent tau monomers suggest structural origins of aggregation
LINE-1/L1 retrotransposon sequences comprise 17% of the human genome . Among the many classes of mobile genetic elements , L1 is the only autonomous retrotransposon that still drives human genomic plasticity today . Through its co-evolution with the human genome , L1 has intertwined itself with host cell biology . However , a clear understanding of L1’s lifecycle and the processes involved in restricting its insertion and intragenomic spread remains elusive . Here we identify modes of L1 proteins’ entrance into the nucleus , a necessary step for L1 proliferation . Using functional , biochemical , and imaging approaches , we also show a clear cell cycle bias for L1 retrotransposition that peaks during the S phase . Our observations provide a basis for novel interpretations about the nature of nuclear and cytoplasmic L1 ribonucleoproteins ( RNPs ) and the potential role of DNA replication in L1 retrotransposition . Retrotransposons are genetic elements that move within the host genome through a ‘copy and paste’ process utilizing an RNA intermediate . 17% of the human genome is made up of copies of the Long Interspersed Nuclear Element-1 ( LINE-1 or L1 ) retrotransposon ( Lander et al . , 2001 ) . L1 is the only autonomous human retrotransposon able to ‘jump’ through a process called retrotransposition . The full length , active L1 retrotransposons consists of a 5’ untranslated region ( UTR ) containing a bidirectional promoter ( Speek , 2001; Swergold , 1990 ) , two open reading frames ( ORFs ) , ORF1 and ORF2 , separated by a short inter-ORF region and a 3’ UTR with a weak polyadenylation signal ( Doucet et al . , 2015; Dombroski et al . , 1991; Burns and Boeke , 2012; Burns , 2017 ) . A short , primate specific , third ORF of unknown function called ORF0 was also described in the 5’UTR of L1 with an antisense orientation compared to ORF1 and ORF2 ( Denli et al . , 2015 ) . Upon transcription by RNA polymerase II , the 6 kb bicistronic L1 mRNA is poly-adenylated and exported into the cytoplasm . LINE-1 exhibits ‘cis-preference’ ( Wei et al . , 2001 ) , the mechanism of which is unknown . A popular model posits that newly translated ORF1 and/or ORF2 proteins preferentially bind the mRNA molecule that encoded them ( Boeke , 1997 ) . Interestingly , translation of the second ORF , ORF2p , is mediated and regulated by an unconventional process that remains poorly understood ( Alisch et al . , 2006 ) . ORF1p is an RNA binding protein with chaperone activity ( Martin and Bushman , 2001; Khazina et al . , 2011 ) while ORF2p contains domains with endonuclease ( EN ) and reverse transcriptase ( RT ) activity ( Feng et al . , 1996; Mathias et al . , 1991; Weichenrieder et al . , 2004 ) . It is believed that ribonucleoprotein particles ( RNPs ) , consisting of many ORF1p molecules , as few as one or two ORF2ps and one L1 mRNA ( Khazina et al . , 2011; Basame et al . , 2006; Dai et al . , 2014 ) form in the cytoplasm . The RNP is then imported into the nucleus through a still uncharacterized process . L1 RNPs accumulate in cytoplasmic stress granules ( SGs ) and processing bodies ( PBs ) ( Goodier et al . , 2007; Hu et al . , 2015; Doucet et al . , 2010 ) , but the role of these cellular structures in the L1 lifecycle is still controversial . L1 RNPs accumulate mainly in the cytoplasm but nuclear ORF1p was also observed in a certain percentage of cells using cell lines and cancer specimens ( Sokolowski et al . , 2013; Sharma et al . , 2016; Doucet et al . , 2016; Goodier et al . , 2007; Harris et al . , 2010 ) . While several specific antibodies against ORF1p have been raised ( Taylor et al . , 2013; Wylie et al . , 2016; Doucet-O'Hare et al . , 2015 ) , a highly effective antibody against ORF2 protein that would allow definitive observation of protein localization is still lacking . Antibodies against LINE-1 ORF2p have been recently developed ( De Luca et al . , 2016; Sokolowski et al . , 2014 ) but the much lower amount of expressed ORF2p compared to ORF1p makes the study of ORF2p expression and localization difficult . To overcome these difficulties , tagged ORF2ps have been employed ( Taylor et al . , 2013; Doucet et al . , 2010 ) . In the nucleus , L1 endonuclease nicks the DNA at A/T rich consensus target sites ( 5’-TTTT/AA-3’ ) ( Feng et al . , 1996 ) and through a process called TPRT ( Target Primed Reverse Transcription ) ( Cost et al . , 2002; Luan et al . , 1993 ) inserts a DNA copy into the new genomic target locus . During TPRT , ORF2p EN domain nicks the DNA and the newly formed 3’OH end is then used by the RT domain of ORF2p to prime the synthesis of a complementary DNA using the L1 mRNA as template . A second strand of cDNA is then synthetized and joined to adjacent genomic DNA . L1 lifecycle is extensively entwined with host cellular processes . Several proteins and cellar pathways have been shown to restrict or support L1 retrotransposition and life cycle ( Goodier , 2016 ) . RNA metabolism ( Belancio et al . , 2008; Dai et al . , 2012 ) , DNA damage response ( Servant et al . , 2017 ) and autophagy ( Guo et al . , 2014 ) are a few cellular processes shown to affect LINE-1 retrotransposition . Progression through the cell cycle was shown by several groups to promote L1 retrotransposition ( Shi et al . , 2007; Xie et al . , 2013 ) but a molecular understanding of this aspect has been elusive . Because of the importance of the cell cycle in efficient retrotransposition , it has been proposed that , as for some exogenous retroviruses ( Goff , 2007; Suzuki and Craigie , 2007 ) , nuclear breakdown during mitosis could represent an opportunity for entrance of L1 into the nucleus ( Xie et al . , 2013; Shi et al . , 2007 ) . This hypothesis , never directly tested previously , was challenged by studies demonstrating effective retrotransposition in non-dividing and terminally differentiated cells ( Kubo et al . , 2006; Macia et al . , 2017 ) . To shed light on the role of the cell cycle on different aspects of L1 life cycle we explored the nuclear localization of L1 proteins and the retrotransposition efficiency of L1 in different stages of the cell cycle in rapidly dividing cancer cells . Here , we use imaging , genetic and biochemical approaches to show that in these cells , the L1 lifecycle is intimately coordinated with the cell cycle . LINE-1 encoded proteins enter the nucleus during mitosis and retrotransposition appears to occur mainly during S phase . Previous works ( Doucet et al . , 2016; Luo et al . , 2016; Sokolowski et al . , 2013; Goodier et al . , 2007; Goodier et al . , 2004; Sharma et al . , 2016; Rodić et al . , 2014; Taylor et al . , 2013; Branciforte and Martin , 1994 ) have shown varying localization of ORF1 protein in cells growing in culture or in mammalian specimens from various organs and tumors . In particular , the nuclear localization of LINE-1 encoded proteins has been sparsely studied and the mechanisms driving import of L1 retrotransposition intermediates into the nucleus are largely unknown . We therefore set out to characterize the cellular localization of L1 ORF1p and ORF2p in human cells overexpressing recoded ( ORFeus ) or non-recoded L1 ( L1rp ) with a 3xFlag tag on the ORF2p C-terminus ( Taylor et al . , 2013 ) . Immuno-fluorescence staining of HeLa-M2 cells ( Hampf and Gossen , 2007 ) overexpressing ORFeus , showed clear expression of ORF1p and ORF2p ( Figure 1A and D and Figure 1—figure supplements 1 , 2 , Videos 1 and 2 ) . As previously observed ( Taylor et al . , 2013 ) , ORFeus ORF1p was detected in virtually all the cells ( ≅ 97% ) whereas ORF2p , encoded by the same bicistronic L1 mRNA also expressing ORF1p , was detected in just a subset of cells ( ≅10% using a rabbit anti-Flag antibody and ≅20% using a more sensitive mouse anti-Flag antibody ) ( Figure 1B , bar graph and inset ) . This pattern of expression is most likely due to an unknown mechanism controlling ORF2p translation ( Alisch et al . , 2006 ) . Interestingly , when non-recoded L1 was over-expressed , only 44% of cells displayed ORF1p expression . Overexpression of ORFeus-Hs and L1rp with or without the L1 5’ untranslated region ( 5’ UTR ) excluded the possibility that the reduced expression of L1rp compared to ORFeus was due to the presence of the 5’ UTR ( Figure 1—figure supplement 3 ) ( Chen et al . , 2012 ) . Due to the overall lower expression of L1rp , ORF2p was barely observable in cells expressing non-recoded L1 ( Figure 1A and B ) . ORF1p was mainly present in the cytoplasm but some cells ( ≅26% ) also showed clear nuclear staining ( Figure 1A , Figure 1D , white arrowheads , Figure 1—figure supplements 1 , 2 , Videos 1 and 2 ) . Z-stack movies of cells expressing L1 and stained for recoded and non-recoded ORF1p and ORF2p are also presented ( Videos 1 and 2 ) . Confocal images recapitulated our observations and confirmed our conclusions ( Figure 1—figure supplement 2 ) . Interestingly , the cells with nuclear ORF1p fluorescence were usually observed as pairs of cells in close proximity . This observation suggested to us that cells with nuclear ORF1p may have undergone mitotic division immediately prior to fixation and observation . In HeLa cells , this observation was made more evident by the fact that daughter cells displaying nuclear ORF1p were sometimes connected by intercellular bridges often containing filaments of DNA and persisting from incomplete cytokinesis during the previous mitosis ( Figure 1D , lower panels ) ( Steigemann et al . , 2009; Carlton et al . , 2012 ) . To quantify our initial qualitative observation of closer cell proximity for cells displaying nuclear ORF1p , we employed automated image acquisition of HeLa cells expressing recoded L1 and stained for ORF1p ( see ‘quantification and statistical analysis’ section ) . We compared the distance of pairs of cells expressing nuclear ORF1p versus the distance of the same number of cell pairs expressing ORF1p just in the cytoplasm . This ‘proximity analysis’ , described in more details in the methods section , clearly shows that cells with nuclear ORF1p are statistically closer to each other than cells expressing ORF1p exclusively in the cytoplasm ( Figure 1C ) . The cytoplasmic ORF1p localization pattern was often dominated by previously described cytoplasmic foci ( Doucet et al . , 2010; Goodier et al . , 2007 ) . The formation of these foci is particularly enhanced by L1 overexpression . The nature and role of these cytoplasmic structures is still debated ( Goodier et al . , 2007; Martin and Branciforte , 1993; Guo et al . , 2014 ) and in our system they may be induced by non-physiological expression levels of ORF1p . We therefore performed the localization experiments reported here using Tet-inducible constructs , and we induced L1 expression at lower concentrations of doxycycline ( 0 . 1 μg/ml ) compared to the concentrations typically used to induce full induction of retrotransposition or production of RNP intermediates for proteomic studies ( 1 μg/ml ) ( Taylor et al . , 2013 ) . We observed very heterogeneous localization of ORF1p and ORF2p in cells overexpressing L1 upon treatment with 0 . 1 μg/ml doxycycline for 24 hr . ORF1p is observed to be only cytoplasmic ( Figure 1D top panel ) , or both cytoplasmic and nuclear ( Figure 1D , top panel ) but is never exclusively nuclear when immunostaining is performed using 4H1 , 4632 and JH74 antibodies ( Figure 1—figure supplements 1 , 2 ) . ORF2p also had heterogeneous nuclear/cytoplasmic localization with cells displaying ORF2p only in the cytoplasm ( Figure 1A and Figure 1D middle panel ) , only in the nucleus ( Figure 1D , top panel ) or in both the cytoplasm and the nucleus ( Figure 1D , lower panel ) . Nuclear ORF2p had a punctate pattern ( Figure 1D top panel ) and nuclear ORF1p was usually excluded from nucleoli ( Figure 1D , arrowheads ) . Also , a small population of cells ( usually less than 0 . 1% of L1 expressing cells ) showed a strong and clear nuclear signal of ORF2p in the absence of nuclear ORF1p ( Figure 1D , top panel and Figure 1—figure supplement 2 ) consistent with a possible nuclear form of L1 RNPs that lacks ORF1p ( see also accompanying paper by Taylor et al . , 2018 ) . On the contrary , in the cytoplasm , ORF2p always co-localized with ORF1p ( Figure 1A and D ) . We compared several antibodies against ORF1p that confirmed a consistent pattern of nuclear/cytoplasmic localization of ORF1p ( Figure 1—figure supplement 1 ) . The corresponding secondary-only controls and IF staining of cells not expressing L1 are presented in Figure 1—figure supplement 4 . The mouse 4H1 monoclonal antibody ( mAb ) recognizing the N-terminus of ORF1p and the rabbit JH74 antibody against the C-terminus of ORF1p have an identical pattern . The polyclonal 4632 antibody ( pAb ) displayed a high nuclear non-specific signal that renders this antibody less sensitive than 4H1 and JH74 ( Figure 1—figure supplements 1 , 4 ) . The JH73g rabbit Ab was distinct from the other antibodies and recognized mainly nuclear ORF1p ( Figure 1—figure supplement 1A–B and Figure 4—figure supplement 1 ) . This nuclear form of ORF1p is also recognized by 4H1 and JH74 antibodies but is much more easily observed using JH73g because of its higher affinity for nuclear ORF1p and much lower affinity for cytoplasmic ORF1p . Indeed , quantification of nuclear ORF1p upon staining with JH74 and JH73g revealed a comparable percentage of cells displaying nuclear ORF1p using the two antibodies ( ≅14–17% ) . The percentage of cells detected overall as expressing ORF1p using JH73g Ab is much lower than the percentage recognized as ORF1p expressing cells by JH74 Ab ( 37% versus 93% respectively ) because JH73g is able to recognize mainly the nuclear form ( Figure 1—figure supplement 1B ) . This unusual staining pattern suggests that the nuclear form of ORF1p may be highly enriched for a conformation specifically recognized by JH73g . Staining of L1rp expressing cells , on the other hand , reveals a lower threshold of sensitivity for the JH73g antibody that is unable to detect the lower amount of non-recoded ORF1p . Interestingly , ORF1p immunoprecipitated with JH73g Ab is impaired in binding to ORF2p protein , consistent with the possibility that most of the nuclear ORF1p species fail to bind ORF2p ( Figure 1—figure supplement 1D ) . Our immunofluorescence staining of ORF1p and ORF2p suggests that L1 RNPs enter the nucleus during mitosis , when the nuclear membrane breaks down . To better explore this hypothesis , previously put forward by other studies ( Xie et al . , 2013 ) , we exploited well-characterized markers of the cell cycle: geminin , expressed only in S/G2/M phases , and Cdt1 , which specifically marks the G1 phase ( Arias and Walter , 2007 ) . Co-staining of ORF1p with geminin and Cdt1 clearly showed that ORF1p is nuclear in cells expressing Cdt1 ( G1 phase ) , and completely cytoplasmic in cells expressing geminin ( Figure 2A and quantification in B ) . Confocal images of Cdt1 , geminin and ORF1p staining confirmed our conclusions ( Figure 2—figure supplement 1 ) . We also verified these results using the FUCCI system ( Fluorescent Ubiquitin Cell Cycle Indicator ) , that exploits Cdt1 and geminin fragments fused to mAG and mKO2 ( Monomeric Azami-Green and Monomeric Kusabira-Orange 2 fluorescent proteins ) respectively ( Sakaue-Sawano et al . , 2008 ) . Expression of an ORF1p-HaloTag7 in HeLa . S-FUCCI cells treated with the Halo tag ligand JF646 , ( Grimm et al . , 2015 ) shows that ORF1p is nuclear only in cells in G1 phase with orange nuclei resulting from expression of mKO2-tagged Cdt1 fragment ( Figure 2C ) . In this setting , no instances of cells with nuclear ORF1p-Halo and green nucleus ( cells in S/G2/M ) were observed ( total number of cells counted = 3309; cells with nuclear ORF1p and green nuclei = 0; cells with nuclear ORF1p and red nuclei = 80 ) . ORF2p-Halo was expressed in cells scattered throughout the cell cycle that displayed either orange ( Cdt1 expressing ) or green ( geminin expressing ) nuclei ( Figure 2—figure supplement 2 ) . This result suggests that ORF2p expression is not cell cycle regulated . Overall , our analyses strongly suggest that ORF1p , most likely together with ORF2p , enters into the nucleus during mitosis and is retained there when the nuclear membrane reforms after cell division . To gain insight into the nuclear/cytoplasmic distribution of L1 mRNA , we also performed RNA-FISH followed by immunofluorescence staining of ORF1p and ORF2p in cells expressing recoded ( ORFeus ) or non-recoded L1 ( L1rp ) ( Figure 3 ) . As expected , the detected L1 mRNA mostly co-localized with ORF1p staining in both ORFeus and L1rp ( Figure 3A and additional images in Figure 3—figure supplement 1A ) . Interestingly , and in line with our results on ORF1p localization ( Figure 1C and Figure 2 ) , nuclear signal of L1 mRNA was particularly bright in cells exiting mitosis identified by cytoplasmic bridges still connecting the two daughter cells ( Figure 3—figure supplement 1B , Figure 3—figure supplement 2 and Videos 3 and 4 ) . To verify that high nuclear L1 mRNA was detected in post-mitotic cells in G1 phase , we performed RNA-FISH of ORFeus followed by Cdt1 and Geminin co-staining ( Figure 3B–C and Figure 3—figure supplement 2 ) . As shown for ORF1p and in line with our hypothesis , nuclear L1 mRNA co-localized with cells with low Geminin and higher Cdt1 staining . These results confirmed our ORF1p localization results , and suggest that L1 RNPs formed by ORF1p , L1 mRNA and , most likely , ORF2p enter the nucleus during mitosis and remain ‘trapped’ in the nucleus upon nuclear membrane reformation in G1 phase . A direct consequence of these conclusions would be that prolonged expression of L1 in dividing cells should eventually lead to a population with all cells displaying nuclear ORF1p because a longer time of L1 induction will allow all the cells to undergo mitosis while expressing ORF1p . To test this hypothesis , we quantified the percentage of cells displaying nuclear ORF1p after 24 and 48 hr of L1 expression , considering that HeLa cell doubling time is about 24 hr . Automated picture collection ( Arrayscan HCS , Cellomics ) and software based nuclear/cytoplasmic analysis ( HCS studio cell analysis software ) was implemented as for proximity analysis . We set very stringent negative fluorescence thresholds ( limit in Figure 4A ) determined from cells not treated with doxycycline and therefore not expressing L1 , as described in the methods section . These stringent parameters were necessary to avoid interference of the strong ORF1p cytoplasmic signal with the measurement of nuclear ORF1p signal . The analysis of ORF1p nuclear and cytoplasmic distribution surprisingly showed that the percentage of cells with nuclear ORF1p does not increase but actually decreases after 48 hr of L1 induction compared to the 24 hr time point ( Figure 4A ) . The decrease in nuclear ORF1p after 48 hr induction is probably due to the decreased growth rate of a more confluent cell population . The absence of an increase of cells with nuclear ORF1p with increased time of L1 induction , suggests that , after entering the nucleus in M phase , nuclear ORF1p is either degraded and/or exported from the nucleus during or after G1 phase . To better explore potential cytoplasmic/nuclear shuttling of ORF1p and ORF2p we took advantage of a known inhibitor of exportin 1 ( XPO1/CRM1 ) , leptomycin b . We treated HeLa cells expressing LINE-1 with leptomycin for 18 hr . Two different concentrations of leptomycin were used and several antibodies ( Abs ) were utilized to detect ORF1p in immunofluorescence assays ( Figure 4B–E ) . At both leptomycin concentrations , and using any of the Abs recognizing ORF1p we observed an increased number of cells with nuclear ORF1p after leptomycin treatment , suggesting that at least a subset of ORF1p is exported from the nucleus in a CRM1-dependent manner ( Figure 4E ) . As control , a known CRM1 regulated protein ( MEK-1 ) ( Dave et al . , 2014 ) tagged with GFP was used to show nuclear retention upon leptomycin treatment ( Figure 4—figure supplement 2 ) . Our results suggest that ORF1 protein , in a ribonucleoprotein complex with L1 mRNA ( and presumably ORF2p ) , is able to enter the nucleus during mitosis and it accumulates in the nucleus in early G1 phase of the cell cycle . Following early G1 , ORF1p is then exported to the cytoplasm through a CRM1 dependent mechanism . We therefore asked whether L1 retrotransposition occurred in a cell cycle-dependent manner and more specifically during M phase or G1 phase , when we observed ORF1p in the nucleus and when chromatin is accessible to L1 RNPs . To answer this question we performed retrotransposition assays using a previously described ORFeus-GFP-AI reporter ( Taylor et al . , 2013; An et al . , 2011 ) . HeLa cells expressing the retrotransposition reporter were treated for increasing times with nocodazole ( Figure 5A ) , a cell cycle inhibitor that blocks cells in M phase interfering with microtubule assembly ( Ma and Poon , 2017; Rosner et al . , 2013 ) . Treatments were performed for no longer than 21 hr , a time sufficient to allow cells passage through just one cell cycle . Increased time of nocodazole treatment , and therefore longer time in M phase , fails to increase the percentage of M phase green cells ( Figure 5A–B ) , suggesting that L1 retrotransposition does not occur during M phase . Longer times of nocodazole treatment ( 21 hr ) increased cell death , detected by an increase of propidium iodide-positive cells , and a consequent decrease in retrotransposition ( Figure 5B , dotted line ) . Similar experiments were also performed using thymidine and mimosine treatments to interrogate possible biases of L1 retrotransposition toward G1 phase ( Ambrozy , 1971; Lalande , 1990 ) . The effects on cell cycle progression of increased times of 4 mM thymidine and 1 mM mimosine treatments are reported in Figure 5—figure supplement 1 . Treatment with excess thymidine inhibits DNA synthesis blocking cells in late G1 . As with nocodazole treatments , cells treated with thymidine , showed no increase in GFP positive cells compared to untreated cells , suggesting that L1 does not preferentially retrotranspose in late G1 phase ( Figure 5C ) . Mimosine treatments , also did not increase retrotransposition but actually decreased L1 hopping ( Figure 5D ) . Mimosine is a non-protein amino acid that potently inhibits cell cycle . Despite mimosine’s well-established role in blocking cells in late G1/early S phase , the molecular mechanisms affected by mimosine to induce cell cycle arrest are still debated . Mimosine was shown to affect both DNA synthesis initiation and elongation and to induce depletion of deoxynucleotides through chelation of iron and consequent inhibition of ribonucleotide reductases ( RNR ) and serine hydroxymethyltransferase ( SHMT ) . Mimosine was also shown to inhibit viral replication consistent with a general metabolic mechanism ( Nguyen and Tawata , 2016; Kalejta and Hamlin , 1997; Dai et al . , 1994; Park et al . , 2012 ) . Our data showing a decreased retrotransposition in cells treated with mimosine , suggest that mimosine inhibits L1 retrotransposition not only arresting the cell cycle but probably through additional ( metabolic ? ) mechanisms ( Figure 5D ) . We then expanded our analysis , measuring retrotransposition during a single cell cycle in a population of HeLa cells synchronized by nocodazole treatment , subsequent ‘mitotic shake off’ and released into the cell cycle in the absence of nocodazole ( Figure 6A ) . Measurements of the percent of cells that underwent retrotransposition were performed every three hours starting after release from nocodazole synchronization . The cell cycle stage of the cells at each time point was determined by propidium iodide staining ( Figure 6—figure supplement 1 ) . A linear increase of retrotransposition should be observed if retrotransposition is unbiased towards specific cell cycle stages , while a non-linear increase represents a specific stage at which retrotransposition is enhanced . Calculation of the slope of the increase of GFP+ cells should therefore produce a clear peak at the time during which most retrotransposition occurs ( Figure 6D ) . This approach allowed us to identify a peak of retrotransposition in the S phase ( Figure 6E top and bottom left panels ) . Control non-synchronized cells , as expected , showed a linear increase in retrotransposition and no clear peaks were identified ( Figure 6F top , bottom right panels ) . To evaluate whether the cell cycle controls retrotransposition using a method independent of cell synchronization , we developed a fluorescent-AI reporter that introduces a temporal component to canonical retrotransposition reporters . To this end , we utilized the previously characterized monomeric fluorescent timers ( FT ) ( Subach et al . , 2009 ) . These derivatives of mCherry change their fluorescence emission from blue to red over 2 to 3 hr ( fast-FT ) . We introduced an antisense intron within the coding region of the ‘fast-FT’ and inserted this cassette in the 3’UTR of a recoded L1 ( Figure 7A ) . Transfection of the L1-fastFT-AI construct into HeLa cells allowed us to identify cells that underwent retrotransposition within a ~ 3 hr period preceding the analysis , as reported by previous work ( Subach et al . , 2009 ) . Our quantification also supports the previously reported timing of FT maturation ( Subach et al . , 2009 ) with an average conversion time from blue to red of 2 . 35 ± 0 . 52 hr ( Figure 7A and Figure 7—figure supplement 1B ) . Immediately after L1-fastFT-AI retrotransposition , the fast FT is expressed and the cells emit blue fluorescence . Upon translation the blue proteins begin turning red in less than 3 hr . To roughly quantify the time needed for the visualization of a fluorescent protein after induction of transcription , we measured GFP expression upon doxycycline induction . Quantification of cells expressing GFP under control of a Tet CMV-inducible promoter revealed that 50% of the cells expressed visible GFP at 2 . 71 ± 0 . 46 hr and 90% of the cells expressed visible GFP within 8 . 01 ± 0 . 63 hr of doxycycline treatment ( Figure 7A and Figure 7—figure supplement 1A ) . This quantification , even when performed in an over-expression setting , suggests that transcription from a strong promoter , translation and accumulation of a fluorescent protein can be fast enough for the detection of retrotransposition events within approximately 3 hr from the event itself . In cells expressing L1 FT-AI , upon retrotransposition of L1 , the FT gene is transcribed and expressed in <3 hr in at least 50% of the expressing cells . After 2 . 35 ± 0 . 52 hr from translation , the blue proteins maturate into red emitting proteins . The cells are now marked by a blue population of proteins continuously transcribed by the constitutive CMV promoter and a red population of aged proteins matured from the blue form ( Figure 7A ) . Analysis of the cell cycle stage of FACS sorted ‘blue only’ cells ( cells that underwent retrotransposition within about 3 hr of the analysis , also considering the time needed for transcription of the marker ) compared to fluorescence negative cells FACS sorted from the same population ( cells that did not undergo retrotransposition before analysis ) revealed a strong enrichment in S phase cells , a partial enrichment in cells in the G2/M phase and a strong de-enrichment of cells in G1 phase ( Figure 7B and Figure 7—figure supplement 2A–B ) . These results confirmed the strong bias of retrotransposition towards S phase that we measured using nocodazole synchronization ( Figure 6 ) . However , the over-representation of G2/M cells that underwent retrotransposition pushed us to design experiments to better dissect the population of cells comprising the peak of blue-only cells presented in Figure 7B . To better dissect the cell cycle stage of cells that underwent retrotransposition , we implemented a second approach that does not involve cell sorting but simply allows direct analysis of the cell cycle in cells expressing the FT-AI reporter ( Figure 7C and Figure 7—figure supplement 2C–D ) . After 24 hr of doxycycline treatment , cells expressing the L1-fastFT-AI reporter were directly stained with SYTO61 DNA labeling dye and analyzed . The main population of blue negative cells that did not undergo retrotransposition , showed cells distributed throughout the cell cycle ( G1 = 49 . 5% , S = 34 . 2% , G2/M = 15 . 2% ) ( Figure 7C , black line and Figure 7—figure supplement 2C–D ) . Using this analysis , we were able to divide the population of blue positive cells ( blue+ ) in two subpopulations: cells with relatively higher red fluorescence ( Figure 7C , purple profile and Figure 7—figure supplement 2C–D ) , and cells with lower/undetectable red fluorescence ( Figure 7C , blue profile and Fig . Figure 7—figure supplement 2C–D ) . The former group of cells underwent retrotransposition in a time closer to the time of analysis compared to the blue+ cells with higher red fluorescence in which few FT molecules had time to mature into the red form . Consistent with the previous experiments , the blue+ cells with lower red fluorescence ( blue peak ) are mainly in S phase ( G1 = 9 . 38% , S = 78 . 1% , G2/M = 12 . 5% ) . Blue+ cells with higher red fluorescence ( purple peak ) , which had more time to proceed through the cell cycle after retrotransposition and before analysis ( from S to G2/M ) , were mainly in G2/M phase ( G1 = 0% , S = 10 . 9% , G2/M = 89 . 1% ) . This result clearly shows that the wide peak of ‘blue only’ sorted cells that spread across S and G2/M phases ( Figure 7B ) actually consists of two subpopulations/peaks: a population of cells in S phase that underwent retrotransposition a short time before analysis and a second population of cells in G2/M phase that underwent retrotransposition earlier relative to analysis . These observations , together with the data presented in Figures 5 and 6 , collectively indicate that L1 retrotransposition has a strong cell cycle bias and preferentially occurs during the S phase . To gain biochemical insight into the timing of L1 retrotransposition we investigated the timing with which ORF2p was recruited onto chromatin , a necessary step for retrotransposition . We isolated nuclear soluble and chromatin bound proteins from cells synchronized and released into the cell cycle , as in Figure 6C . Immunoblot analysis showed no differences in the amount of histone H3 and ORF1p present on chromatin and , as expected , the analysis revealed chromatin recruitment that peaked in S phase for PCNA ( Strzalka and Ziemienowicz , 2011 ) and Upf1 ( Azzalin and Lingner , 2006 ) . Supporting our previous results , ORF2p was recruited on chromatin in S phase in a similar manner to Upf1 and PCNA ( Figure 8A–B ) . It is worth noting that , despite our observation that ORF1p is less nuclear in late G1 ( Figure 4 ) , we did not observe changes in nuclear and cytoplasmic ORF1p during cell cycle progression ( Figure 8A , right panel ) . This is probably due to the contamination of cytoplasmic stress granules ( highly enriched in ORF1p ) in nuclear fractions as shown by detection of G3BP1 , a marker of stress granules ( Figure 8—figure supplement 1 ) . We previously showed that ORF2p binds PCNA through a PIP domain in the ORF2 protein , sandwiched between the EN and RT domains , and that the PCNA-ORF2p interaction is necessary for retrotransposition in HeLa and HEK293 cells ( Taylor et al . , 2013 ) . The PCNA-ORF2p complex is mainly chromatin bound ( Figure 8—figure supplement 1 ) , supporting the idea that PCNA binds ORF2p during retrotransposition . We therefore followed up on these previous findings investigating the interactome of the ORF2p-PCNA complex specifically . To this end , we engineered a V5-tag at the N-terminus of PCNA in HCT116 cells stably expressing a doxycycline inducible ORFeus . HCT116 cells were chosen because of their near-diploid number of chromosomes compared to HeLa cells , characterized by unstable karyotype . We performed sequential immunoprecipitation of ORF2p followed by V5/PCNA IP ( Figure 9A ) , and we analyzed the interacting partners of the ORF2p-PCNA complex by mass spectrometry . Among the identified ORF2p/PCNA interactors ( 279 from the first experiment and 158 from the second experiment ) we identified several MCM proteins ( MCM3 , MCM5 and MCM6 ) as well as TOP1 ( DNA topoisomerase 1 ) , PARP1 ( Poly [ADP-ribose] polymerase 1 ) and RPA1 ( Replication Protein A1 ) ( Figure 9B ) . These proteins are known to be co-recruited with PCNA on the origins of DNA replication before S phase ( MCM proteins ) and during S phase on the replication fork ( MCM , PCNA , TOP1 , RPA1 and PARP1 proteins ) ( Czubaty et al . , 2005; Remus et al . , 2009; Ying et al . , 2016 ) . Co-immunoprecipitation of Flag/ORF2 or ORF1 proteins from HEK293 cells expressing ORFeus , recapitulated the interaction of ORF2p with MCM6 and PCNA ( Figure 9B ) . As expected , immunoprecipitation of ORF2p pulled down a fraction of ORF1p , but also MCM6 and PCNA proteins . Interestingly , as expected from our previous observations revealing that ORF1p is not necessarily in the complex ( es ) with chromatin bound nuclear L1 RNPs , immunoprecipitation of ORF1p pulled down only a small amount of ORF2p , and also a smaller amount of MCM6 and PCNA proteins . These observations suggest that the nuclear L1 complex contains ORF2p , PCNA and components of the replication fork such as MCM6 , and is depleted of ORF1 proteins . To verify that the ORF2p-PCNA-MCM complex identified here also contained L1 mRNA , component of the L1 RNPs essential for retrotransposition , we performed IP experiments followed by RT-qPCR ( IP-RT-qPCR ) for L1 . As expected , IP of ORF2p pulled down L1mRNA ( Figure 9D , top panel ) as well as direct IP of PCNA also showed interaction of this protein with the L1 mRNA ( Figure 9D , top panel ) compared to control IPs performed using normal mouse IgG antibodies . In line with our hypothesis that the ORF2p-PCNA complex is potentially retrotransposing , sequential IP of ORF2p followed by PCNA IP also pulled down L1 mRNA ( Figure 9D , lower panel ) compared to control IgG IP . Control qPCR of samples not treated with reverse transcriptase ( -RT ) displayed no or extremely low amplification ( data not shown ) . Finally , we performed immunofluorescence staining of ORF2p and PCNA in HeLa cells synchronized in S phase by double thymidine synchronization . A subset of nuclear ORF2p puncta overlapped with PCNA foci , marking potential regions of active DNA replication ( Figure 9D ) . Collectively , our biochemical and proteomic work ( Figure 9 ) support the hypothesis that ORF2p binds PCNA on sites of DNA replication during S-phase ( model in Figure 10 ) , and this fraction is most likely engaging in retrotransposition as demonstrated by our functional assays ( Figure 7 ) . Despite the increasingly appreciated relevance of L1 retrotransposon to normal cellular physiology and disease etiology , many of the steps of L1 retrotransposon lifecycle in human cells are largely unknown . This lack of insight about L1 retrotransposons in human cells is unsurprising considering that many technical challenges hinder studies of this highly repetitive but poorly expressed element , which is effectively repressed by host somatic cells ( Goodier , 2016 ) and therefore overexpression approaches can reveal otherwise hidden pathways . It makes sense that nuclear localization of a L1 RNP particle comprising at least ORF2p , with EN and RT activity , bound to L1 mRNA , is essential for L1 retrotransposition to gain access to its target , genomic DNA . Despite this obvious observation , the nature of nuclear L1 RNPs and the process by which L1 gains entry into the nucleus are unknown . No functional nuclear localization signal ( NLS ) has been identified in the two L1 proteins ORF1p and ORF2p suggesting that their import into the nucleus is either mediated by interacting partners or by cellular processes such as the cell cycle and progression through mitosis during which the nuclear membrane breaks down , allowing the possible entrance of L1 RNPs into the nucleus . The former hypothesis is supported by several studies that show an essential role of cell division on retrotransposition and retrotransposition rate in tissue culture cells ( Xie et al . , 2013; Shi et al . , 2007 ) . On the other hand , other work showed the possibility of L1 retrotransposition in differentiated and non-dividing cells such as human neurons and glioma cells , albeit at substantially lower rates ( Macia et al . , 2017; Kubo et al . , 2006 ) . These seeming incongruities may be explained with a possible major mechanism of entry into the nucleus during mitosis and a less frequent mode of nuclear localization for L1 RNPs that is independent of the cell cycle and specific for some cellular state or cell type . Other possible explanations for the discrepancies between our conclusions and works showing retrotransposition in non-dividing cells , are potential cell cycle artifacts caused by the adenovirus vectors or the low rate of proliferation of the cells used . Through imaging , genetic and biochemical approaches , we show that L1 nuclear import as well as L1 retrotransposition has a strong cell cycle bias ( Figure 10 ) . We also show that ORF1p and L1mRNA , probably in a complex with ORF2p , enters the nucleus during mitosis , accumulating in cells in the G1 phase ( Figure 2 ) . The nuclear localization of L1 RNPs upon transition through mitosis may be due to simple diffusion of L1 RNPs in the nuclear proximity and subsequent sequestration of the particles into the nucleus upon nuclear membrane formation . Another interesting possibility , is that , during mitosis , the L1 RNPs may weakly interact with chromatin , most likely through the positively charged ORF1p . These interactions could increase the chances of L1 RNPs being ‘trapped’ in the nucleus after reformation of the nuclear membrane . This hypothesis is supported by our unpublished observations that GFP-tagged ORF1p strongly interact with chromatin during metaphase and this process seems to increase the amount of ORF1p observed into the nucleus in early G1 . Moreover , as shown in Figure 8 and Figure 8—figure supplement 1 of Figure 8 , we always observe ORF1p in chromatin fractions even in the absence of ORF2p ( unpublished data ) . This observation strongly suggests interaction of ORF1p with chromatin independent of retrotransposition itself . We observe a CRM1 mediated/leptomycin sensitive nuclear export of ORF1p ( Figure 4 ) that keeps the nuclear level of ORF1p low and helps explain the observation that ORF1p is always mostly cytoplasmic ( Figure 1 and Figure 4 ) . Future studies will need to explore the CRM1/ORF1p interaction and the role of this interaction in L1 RNP cellular dynamics and retrotransposition . The decoupling of the CRM1 role on the cell cycle from its importance on L1 cellular localization will be challenging but essential for the understanding of ORF1p nuclear export . We also show that even if L1 enters the nucleus in M phase , retrotransposition does not happen during cell division ( M phase ) but it is during the following S phase in which retrotransposition peaks ( Figures 5–7 ) . The finding that L1 retrotransposition has a strong bias for S phase is , in retrospect , not entirely surprising considering that deoxynucleoside triphosphates ( dNTPs ) , critically necessary for reverse transcription , are at high levels during the S phase and are greatly restricted during the other cell cycle stages ( Hofer et al . , 2012; Stillman , 2013 ) . This layer of metabolic regulation may reflect an ancient adaptation to limit the proliferation of retroelements . dNTP concentration is tightly controlled by ribonucleotide reductase ( RNR ) , the enzyme that converts ribonucleotide diphosphates ( rNDPs ) into dNDPs and by SAMHD1 ( sterile alpha motif and HD-domain containing protein 1 ) that cleaves dNTPs to deoxynucleosides ( Stillman , 2013 ) . Indeed , SAMHD1 expression was found to restrict replication of lentiviruses such as HIV , by restricting availability of dNTPs ( Hrecka et al . , 2011; Goldstone et al . , 2011 ) . It is therefore not surprising that SAMHD1 was also shown to restrict LINE-1 retrotransposition ( Zhao et al . , 2013 ) directly supporting the idea that dNTP concentration can profoundly limit L1 jumping . Interestingly , we also found that mimosine , a compound that blocks the cell cycle in G1/early S and that inhibits RNR , also strongly inhibits L1 retrotransposition ( Figure 5 ) . Our findings suggest that inhibition of L1 retrotransposition mediated by mimosine involves multiple mechanisms other than cell cycle inhibition . It is possible that the depletion of dNTPs by mimosine further mediates inhibition of L1 retrotransposition . In complete accord with these observations is our finding that retrotransposition happens in S phase , during which dNTP concentration peaks , allowing efficient reverse transcription and thus L1 retrotransposition . This may thus be viewed as an adaptation of the retroelement to a host defense . Our functional studies showing S phase bias of L1 retrotransposition are also corroborated by our biochemical observations that show that ORF2p is recruited to chromatin during S phase ( Figure 8 ) and suggest that ORF2p is recruited to a subset of sites of DNA replication with PCNA and MCM proteins ( Figure 9 ) . Mass spectrometry analysis revealed interaction of the previously described PCNA/ORF2p complex ( Taylor et al . , 2013 ) with TOP1 , RPA1 and PARP1 proteins ( Taylor et al . , 2018 ) all of which associate with replication forks . Interestingly , the PARP1 interaction suggests that L1 specifically interacts with stalled replication forks ( Berti et al . , 2013 ) . Our co-localization of PCNA and ORF2p also supports the presence of ORF2p at potential sites of DNA replication , marked by PCNA staining during the S phase . PCNA and ORF2p immunofluorescence revealed that only some PCNA foci of replication overlap with ORF2p nuclear foci , suggesting that , at least in some instances , L1 , possibly engaged in TPRT , may specifically interact with a subset of perhaps stalled replication forks . Our data do not clarify whether the replication fork stall is caused by ORF2p nicking of the DNA or alternatively , whether the retrotransposing L1 complex is recruited specifically to previously stalled replication forks . Conversely , not all ORF2p nuclear foci overlap with PCNA sites , supporting a model in which L1 interaction with the replication fork may represent just one of several modes used by L1 to select a DNA target site and retrotranspose . Our previous work ( Taylor et al . , 2013 ) showed that PCNA does not interact with ORF2p mutated in its endonuclease ( EN- ) or reverse transcriptase ( RT- ) domain . These observations led us to hypothesize that PCNA is recruited by ORF2p after the first steps of TPRT ( nicking of genomic DNA and beginning of L1 mRNA reverse transcription ) , and not vice-versa ( chromatin recruitment of ORF2p by PCNA ) . Together with the observations presented in this manuscript we envision a model in which L1 RNPs , comprising at least ORF2p and L1mRNA , are recruited to replication forks in S phase during DNA replication . A subset L1 RNPs subsequently mediate productive retrotransposition into target loci , perhaps aided by stalling of the replication fork . During TPRT , PCNA , readily available at the site of DNA replication , can be recruited in the latter steps of retrotransposition perhaps to mediate repair of the newly synthesized L1 cDNA/genomic DNA junctions . It is intriguing to postulate that ligases involved in DNA replication such as ligase 1 ( LIG1 ) may also help seal the final nicks in L1 retrotransposition events . An alternative model that could , at least partially , explain our data hypothesizes that a replication fork collides with a nicked DNA formed by retrotransposing L1 . In this latter model , the co-localization of ORF2p with PCNA and MCM proteins would happen after endonuclease cut and initiation of RT by ORF2p . Future work , most likely based on single molecule observation of the retrotransposing L1 RNPs , will be needed to validate this still speculative models . Interestingly , our observations on ORF1p cytoplasmic/nuclear dynamic suggest deeper implications . The fact that ORF1p is exported from the nucleus before S phase , leads to the conclusion that retrotransposition , happening mainly during DNA replication , is mediated by RNPs depleted of ORF1p and constituted only or predominantly by ORF2p and L1 mRNA , a conclusion also supported by data presented by Taylor et al . ( Taylor et al . , 2018 ) . In vitro studies of TPRT show that the first and presumably critical steps in retrotransposition can efficiently occur in vitro in the absence of ORF1p ( Cost et al . , 2002 ) . The hypothesis that ORF1p is dispensable for the actual DNA cutting and reverse transcription steps in vivo , is supported by our observation that nuclear ORF1p can be specifically recognized by one of our antibodies ( JH73g ) . We interpret this observation to mean that this antibody recognizes a specific conformational state of ORF1p unique to the nucleus . Not surprisingly , the nuclear form of ORF1p that is recognized by the JH73g Ab , has impaired binding to ORF2p , suggesting that once inside the nucleus , ORF1p may dissociate from the L1 RNPs destined to carry out the critical endonuclease/reverse transcription steps of retrotransposition during S phase . Moreover , most of the ORF1p does not interact with PCNA and MCM6 that , instead , interact mainly with ORF2p . We also observed ( rare ) instances of cells clearly expressing ORF2p in the nucleus in the absence of detectable ORF1p ( Figure 1 and ( Taylor et al . , 2018 ) ) . Finally , observations of HeLa . S-FUCCI cells expressing Halo tagged ORF1p show that ORF1p is never nuclear in cells in the S/G2/M phase . Halo tagged ORF2p , in contrast , was observed in the nucleus of certain cells in S/G2/M ( Figure 2—figure supplement 2 ) suggesting that , during these cell cycle stages , ORF2p is in the nucleus without ORF1p . Overall , these data suggest that chromatin bound and retrotransposition-competent L1 particles are depleted of ORF1p and mainly consist of ORF2p in complex with L1 mRNA and host factors involved in retrotransposition . Future studies are necessary to better understand the differences of nuclear and cytoplasmic ORF1p and the molecular processes that may mediate ORF1p depletion from L1 RNPs . An attractive possibility is that ORF1p dissociation from L1 RNPs might be associated with delivering an ORF2-RNA RNP to chromatin , although we do not have direct evidence for this . It is tempting to imagine that in the nucleus , the absence of ORF1p trimers , thought to bind L1 mRNA every 50 nucleotides in the cytoplasm ( Khazina et al . , 2011 ) , may promote ORF2p’s unhindered movement during reverse transcription of L1 mRNA in the process of TPRT . This hypothesis will need to be better explored in future studies , by examining ORF2p and L1 mRNA dynamics during the cell cycle . The lack of a sensitive Ab against ORF2p , the fact that most cells expressing ORF1p do not express ORF2p due to an unknown post-transcriptional mechanism controlling ORF2p expression ( Taylor et al . , 2013; Alisch et al . , 2006; Luke et al . , 2013 ) and the difficulties of detecting ORF2p even in the context of overexpression ( Doucet et al . , 2016 ) , continue to technically challenge the study of L1 cellular dynamics . More recent advances in the study of L1 retrotransposon , such as the construction and characterization of ORFeus with its increased expression and function ( An et al . , 2011; Han and Boeke , 2004 ) , the use of smaller and brighter fluorescent tags that allow the exploration of the temporal axis of retrotransposition and the implementation of sensitive biochemical approaches ( Sakaue-Sawano et al . , 2008; Subach et al . , 2009; Grimm et al . , 2015 ) enabled us to discover new and unexpected interactions between L1 and the ‘host’ cell . It is not surprising that retrotransposons , evolved within the human genome for millions of years , have ‘learned’ to leverage important cellular pathways , such as the cell cycle and DNA replication , for their own purpose of spreading and increasing their genomic content ( Boissinot et al . , 2000; Boissinot and Sookdeo , 2016 ) . On the other hand , it is also increasingly clear how cells respond to L1 expansions during evolution , engaging in an ongoing genetic arms-race ( Daugherty and Malik , 2012; Molaro and Malik , 2016 ) . For example , as previously proposed , the nuclear membrane may have represented one of many barrier that retrotransposons had to overcome to maintain effective retrotransposition frequency ( Boeke , 2003; Koonin , 2006 ) . HeLa M2 cells ( a gift from Gerald Schumann , Paul-Ehrlich-Institute; ( Hampf and Gossen , 2007 ) were cultured in DMEM media supplemented with 10% FBS ( Gemini , prod . number 100–106 ) and 1 mM L-glutamine ( ThermoFisher/Life Technologies , prod . number 25030–081 ) ( complete medium ) . Cells were routinely split in fresh medium upon reaching 80–90% confluency . During routine culture of the cells the medium was changed every 2–3 days . 293TLD cells adapted to suspension ( Taylor et al . , 2013 ) were used for transfection with PEI and collected to generate cell grindates as previously described in ( Taylor et al . , 2013 ) . Cell grindates were used for IPs presented in Figure 9C . HCT116 colorectal carcinoma cells were cultured in McCoy’s 5A media ( Life Technologies/Gibco , prod . number 16600–108 ) supplemented with 10% FBS and 1 mM L-glutamine . A subline of these cell line and expressing a V5-PCNA protein was generated using a CRISPR approach . A single Cas9-gRNA vector ( PM192 ) was generated by Golden Gate reaction ( Ran et al . , 2013 ) using vector pX459V2 . 0 ( a gift from Feng Zhang , Addgene , plasmid number 62988 ) and an annealed DNA duplex: JB17486-F2: CACC G GTCTAGCTGGTTTCGGCTTC JB17487-R2: Aaac GAAGCCGAAACCAGCTAGAC C A gBlock DNA purchased from IDT integrated DNA technologies was used as donor DNA . PM-GB3: ccgtgggctggacagcgtggtgacgtcgcaacgcggcgcagggtgagagcgcgcgcttgcggacgcggcggcattaaacggttgcaggcgtagcagagtggtcgttgtctttctagGTCTCAGCCGGTCGTCGCGACGTTCGCCCGCTCGCTCTGAGGCTCGTGAAGCCGAAACCAGCTAGACTTTCCTCCTTCCCGCCTGCCTGTAGCGGCGTTGTTGCCACTCCGCCACCATG GGT AAG CCT ATC CCTAACCCTCTCCTCGGTCTC GATTCTACGGGAGAAGGGCAAGGGCAAGGGCAAGGGCCGGGCCGCGGCTACGCGTATCGATCCTTCGAGGCGCGCCTGGTCCAGGGCTCCATCCTCAAGAAGGTGTTGGAGGCACTCAAGGACCTCATCAACGAGGCCTGCTGGGATATTAGCTCCAGCGGTGTAAACCTGCAGAGCATGGACTCGTCCCACGTCTCTTTGGTGCAGCTCACCCTGCGGTCTG The donor DNA was transfected together with PM192 plasmid using Fugene-HD reagent ( Promega , Madison , WI; prod . number E2311 ) in HCT116 plated in a six well plate . Cells were transfected with 200 ng of donor DNA , 1 . 5 µg PM192 and 6 . 8 μl Fugene-HD in 100 μl Opti-MEM ( Thermo Fisher scientific , prod . number 31985088 ) . 24 hr after transfection , cells were selected in 1 ug/ml puromycin for additional 48 hr . Single clones were picked after serial dilution of the cells in complete media without puromycin . Clones were screened in 96 well plates for expression of V5 by immunofluorescence staining . The positive clones were then validated by genomic DNA PCR using primers flanking the site of CRISPR cut . The amplified band was gel isolated and sequenced by Sanger sequencing . The primers used are the following: JB17488-PCNACseqF CTGCAGATGTACCCCTTGgt JB17489-PCNACseqR GACCAGATCTGACTTTGGACTT The positive clones were then also validated by western blotting analysis using an antibody against PCNA and V5 tag . The HCT116-V5-PCNA cell line was then used to generate stable cell lines expressing the rtTA transactivator . The pTet-ON advance vector ( Clonetech , prod . number 631069 ) was transfected and cells were then selected for several weeks using media supplemented with 250 μg/ml neomycin . Clones were screened using a construct encoding GFP under the control of a tetracycline activated promoter . Finally , HCT116-V5-PCNA-rtTA cell lines stably maintaining episomal pCEP-puro-plasmids expressing ORFeus ( LD401 ) were generated and cultured under puromycin selection ( 1 μg/ml ) to prevent the loss of the L1 plasmids and neomycin ( 250 μg/ml ) to prevent the loss of rtTA transactivator . 120 15 cm plates of HCT116-V5-PCNA-rtTA-LD401 cells were used to immunoprecipitate ORF2p-Flag using Flag-M2 antibodies ( Sigma , prod . number F1804 ) . The immunoprecipitated protein complexes were then used as input to immunoprecipitate V5-PCNA with V5 antibodies ( Invitrogen , prod . number 46–1157 ) . Immunoprecipitation assays were conducted as described below . HeLa . S-Fucci cells were purchased from the Riken BRC cell bank ( RIKEN BioResource Center , Japan , prod . number RCB 2812 ) . Stable HeLa . S-FUCCI cells expressing rtTA were generated transfecting a pTet-ON advance vector ( Clontech , prod . number 631069 ) subcloned to carry a blasticidin resistance cassette instead of a neomycin resistance cassette . Cells were selected for several weeks in 15 μg/ml blasticidin and several clones screened for the expression of firefly luciferase under a control of a doxycycline promoter ( gift from S . K . Logan laboratory ) . The selected stable HeLa . S-FUCCI-rtTA cell lines were cultured in complete DMEM media supplemented with 15 ug/ml blasticidin . LD401 ( ORFeus with 3xFlag ORF2 ) and MT302 ( L1rp with 3xFlag ORF2 and L1 5’UTR ) plasmids were previously described and characterized in ( Taylor et al . , 2013 ) . PM160 ( ORfeus with L1rp 5’UTR ) plasmid was obtained through Gibson reaction of a PCR L1 5’ UTR DNA fragment from MT302 used as template . PM226 ( L1rp with 3xFlag ORF2 and without L1 5’UTR ) was generated by ligation of a synthetic 5’ end of L1rp without UTR to the BsiWI-PmlI cut MT302 vector . EA79 ( untagged ORFeus with GFP-AI cassette ) was constructed subcloning ORFeus under a Tet inducible promoter into a pCEP-4 plasmid ( ThermoFisher scientific , prod . number V04450 ) in which the hygromycin resistance cassette was substituted with a puromycin resistance cassette , as previously described in ( Taylor et al . , 2013 ) ( pCEP-puro plasmid ) . The fluorescent timer FT-AI cassette was built with Gibson assembly using two synthetic DNA fragments ( Quinglan Biotech and Twist Bioscience ) ligated into the BstZ17I 3’UTR of untagged ORFeus ( PM260 ) . Sequences of the fast fluorescent timer mCherry variant are the same of plasmid pTRE-Fast-FT ( a gift from Vladislav Verkhusha , Addgene plasmid number 31913 ) ( Subach et al . , 2009 ) . Constructs expressing Halotag7-ORF1p ( PM285 ) and Halotag7-ORF2p ( PM283 ) were made by Gibson assembly of a PCR DNA fragment encoding the HaloTag7 ( pLH1197-pcDNA3 . 1-PfV-Halo , a gift from Liam Holt ) and pCEP-puro-ORFeus vector . The tag was inserted right before ORF1p and ORF2p stop codon downstream of a G4S linker . The MEK1-GFP expression plasmid was purchased from Addgene ( plasmid #14746 ) . All constructs were verified by Sanger sequencing ( Genewiz ) . HeLa M2 cells were lysed in SKL Triton lysis buffer ( 50 mM Hepes pH7 . 5 , 150 mM NaCl , 1 mM EDTA , 1 mM EGTA , 10% glycerol , 1% Triton X-100 , 25 mM NaF , 10 μM ZnCl2 ) supplemented with protease and phosphatase inhibitors ( Complete-EDTA free , Roche/Sigma prod . number 11873580001; 1 mM PMSF and 1 mM NaVO4 ) . NuPage 4XLDS sample buffer ( ThermoFisher Scientific , prod . number NP0007 ) supplemented with 1 . 43M β-mercaptoethanol was added to the samples to reach a 1X dilution ( 350 mM β-mercaptoethanol final concentration ) before gel electrophoresis performed using 4–12% Bis-Tris gels ( ThermoFisher Scientific , prod . number WG1402BOX ) . Proteins were transferred on Immobilon-FL membrane ( Millipore , prod . number IPFL00010 ) , blocked for 1 hr with blocking buffer ( LiCOR prod . number 927–40000 ) :TBS buffer ( 50 mM Tris Base , 154 mM NaCl ) 1:1 and then incubated with primary antibodies solubilized in LiCOR blocking buffer:TBS-Tween ( 0 . 1% Tween in TBS buffer ) 1:1 . Secondary donkey anti-goat antibodies conjugated to IRDye680 ( anti-rabbit ) or IRDye800 ( anti-mouse ) dyes ( LiCOR prod . number 926–32210 and 926–68071 ) , were used for detection of the specific bands on an Odyssey CLx scanner ( LiCOR ) . 293TLD grindates used in Figure 9C were lysed in extraction buffer ( 20 mM Hepes pH 7 . 4 , 500 mM NaCl , 1% Triton X-100 ) supplemented with protease and phosphatase inhibitors . Immunoprecipitations were performed using 5–10 µl of dynabeads conjugated to primary antibodies ( Dynabeads Antibody Coupling kit , Life Technologies , prod . number 14311D ) incubated with lysates for at least 1 hr nutating at 4°C . After five washes in lysis buffer the immunocomplexes were eluted either in sample buffer , shaken with beads for 10 min at 70°C or using 3xFLAG ( Sigma , prod . number F4799 ) or V5 ( Sigma , prod . number V7754 ) peptides shaken with beads for 1 hr at 4°C . After elution , supernatants were collected and β-mercaptoethanol added to a final concentration of 350 mM . Antibodies against ORF1p used in this study are: We also compared the previously described JH73 antibody ( Taylor et al . , 2013 ) with the JH73g antibody ( Figure 4—figure supplement 1 ) . These two antibodies were derived from the same rabbit immunization using a purified globular ORF1p C-terminus ( Dr . Jeffry Han , unpublished data ) . We confirmed that JH73g Ab is distinct from JH73 Ab displaying clear differences in the light chain migration and base peak chromatogram ( Figure 4—figure supplement 1 ) . Immunoprecipitation of L1 complexes used for the analysis of L1 mRNA were conducted exactly as for the IPs used for mass spectrometry analysis of ORF2p-PCNA complexes . It is worth noting that the buffer used to IP the ORF2p-PCNA complex ( SKL Triton lysis buffer ) is different from the elution buffer ( EB500 ) used in ( Taylor et al . , 2013 ) and optimized for the detection of ORF1p and ORF2p interaction . We could detect a much lower amount of PCNA interacting with ORF2p using EB500 buffer and we were not able to detect L1 mRNA after sequential ORF2p and PCNA IP , using EB500 buffer . HCT116-V5-PCNA expressing ORFeus ( LD401 ) ( about 1 × 106 cells for single IP and about 70 × 106 cells for sequential IP ) were collected by trypsinization and lysed in SKL Triton lysis buffer ( 50 mM Hepes pH7 . 5 , 150 mM NaCl , 1 mM EDTA , 1 mM EGTA , 10% glycerol , 1% Triton X-100 , 25 mM NaF , 10 μM ZnCl2 ) supplemented with 1 mM DTT , 400 µM Ribonucleoside Vanadyl Complex ( VRC ) ( NEB , prod . number S1402 ) , 400U per ml of buffer of RNASEOUT ( Thermo , prod . number 10777019 ) , protein inhibitor and phosphatase inhibitor ( Complete-EDTA free , Roche/Sigma prod . number 11873580001; 1 mM PMSF and 1 mM NaVO4 ) . The cells-lysate mixture was stored at −80°C over-night . The next day the mixture was thawed on ice and centrifuged for 15 min at 16 , 000 rcf at 4°C . ORF2p complexes were immunoprecipitated over-night at 4°C using a mouse FLAG-M2 antibody ( Sigma , pred . number F1804 ) coupled to magnetic beads ( Dynabeads Antibody Coupling kit , Life Technologies , prod . number 14311D ) . Beads were washed three times in Triton buffer and protein complexes were then eluted using 100 µl of 1 mg/ml 3xFLAG peptide ( Sigma , product . Number F4799 ) supplemented with protease , phosphatase and RNase inhibitors for 1 hr at 4°C under shaking . The 3xFLAG eluate was split into half and used for the second immunoprecipition using normal mouse IgG control ( Santa Cruz , prod . number sc2025 ) or V5 ( Invitrogen , prod . number 46–1157 ) coupled beads for 4 hr . Beads were washed three times in Triton buffer , resuspended in 100 µl of Triton buffer containing 30 µg of proteinase K ( Invitrogen , prod . number 25530049 ) . The mixture was incubated at 55°C for 30 min . 1 ml of Trizol ( Life Technologies , prod . number 15596026 ) was directly added to the beads mixture and mRNA was purified using RNA clean-up and concentration columns ( Norgen Biotek , prod . number 23600 ) . cDNA was generated from RNA ( 380 ng for single IP and 245 ng for sequential IP ) using the USB First-Strand cDNA synthesis kit for Real-Time PCR ( Affimetrix , prod . number 75780 ) . Q-PCR was performed using a standard curve of LD401 plasmid . Each Q-PCR reaction contained 2 . 5 µl of Sybr Green mastermix 2X ( Roche , LightCycler 480 SYBR Green I Master , prod . number 04887352001 ) , 25 nl of forward primer ( 100 µM ) , 25 nl of reverse primer ( 100 µM ) , 100 nl of cDNA for the single IP and 500 nl of cDNA for the sequential IP and water to 5 µl ( final volume ) . Q-PCR was performed using a Light Cycler 480 ( Roche ) with standard conditions . The primers used for qPCR amplify a 185 bp amplicon in ORF1 and are reported below: JB13415 ( forward ) : GCTGGATGGAGAACGACTTC JB13416 ( reverse ) : TTCAGCTCCATCAGCTCCTT Samples were reduced and alkylated with DTT ( 1 hr at 57°C ) and iodoacetamide ( 45 min at room temperature ) . The samples were then loaded on a NuPAGE 4–12% Bis-Tris gel ( ThermoFisher Scientific , prod . number WG1402BOX ) and ran for only 10 min at 200V . The gel was then stained with GelCode Blue Staining Reagent ( ThermoFisher , prod . number 24590 ) and the protein bands were excised . The gel plugs were cut into 1 mm3 pieces , washed and destained with 1:1 ( v/v ) ( methanol:100 mM ammonium bicarbonate ) . After at least five solvent exchanges the gel plug was dehydrated by aspirating 100 µl of acetonitrile ( ACN ) and further dried in the SpeedVac . In-gel digestion was performed by adding 250 ng of trypsin ( ThermoFisher , prod Number 90057 ) onto the dried gel plug followed by 300 μl of ammonium bicarbonate 100 mM . Digestion was carried out overnight at room temperature with gentle shaking . The digestion was stopped by adding 300 μl of R2 50 μM Poros beads in 5% formic acid and 0 . 2% trifluoro acetic acid ( TFA ) and agitated for 2 hr at 4°C . Beads were loaded onto equilibrated C18 ziptips and additional aliquots of 0 . 1% TFA were added to the gel pieces and the wash solution also added to the ziptip . The Poros beads were washed with additional three aliquots of 0 . 5% acetic acid and peptides were eluted with 40% acetonitrile in 0 . 5% acetic acid followed by 80% acetonitrile in 0 . 5% acetic acid . The organic solvent was removed using a SpeedVac concentrator and the samples reconstituted in 0 . 5% acetic acid . An aliquot of each sample was loaded onto an Acclaim PepMap100 C18 75 μm x 15 cm column with 3 μm bead size coupled to an EASY-Spray 75 μm x 50 cm PepMap C18 analytical HPLC column with a 2 μm bead size using the auto sampler of an EASY-nLC 1000 HPLC ( Thermo Fisher Scientific ) and solvent A ( 2% acetonitrile , 0 . 5% acetic acid ) . The peptides were eluted into a Thermo Fisher Scientific Orbitrap Fusion Lumos Tribrid Mass Spectrometer increasing from 5% to 35% solvent B ( 80% acetonitrile , 0 . 5% acetic acid ) over 60 m , followed by an increase from 35% to 45% solvent B over 10 m and 45–100% solvent B in 10 m . Full MS spectra were obtained with a resolution of 120 , 000 at 200 m/z , an AGC target of 400 , 000 , with a maximum ion time of 50 ms , and a scan range from 400 to 1500 m/z . The MS/MS spectra were recorded in the ion trap , with an AGC target of 10 , 000 , maximum ion time of 60 ms , one microscan , 2 m/z isolation window , and Normalized Collision Energy ( NCE ) of 32 . All acquired MS2 spectra were searched against a UniProt human database using Sequest within Proteome Discoverer ( Thermo Fisher Scientific ) . The search parameters were as follows: precursor mass tolerance ±10 ppm , fragment mass tolerance ±0 . 4 Da , digestion parameters trypsin allowing two missed cleavages , fixed modification of carbamidomethyl on cysteine , variable modification of oxidation on methionine , and variable modification of deamidation on glutamine and asparagine and a 1% peptide and protein FDR searched against a decoy database . The results were filtered to only include proteins identified by at least two unique peptides . Proteins identified exclusively in the V5-PCNA IP ( 279 in analysis 1 and 158 in analysis 2 ) and not in the parallel IP with the control IgG were considered components of the PCNA-ORF2 complex ( Figure 9 and Figure 9—source data 1 ) . These proteins were queried in the STRING database ( Szklarczyk et al . , 2015 ) to identify known interactions between these candidates . The most enriched GO Biological Process classes were RNA splicing ( GO ID = 0008380; FDR = 1 . 18e-14 ) for the first analysis and viral process ( GO ID = 0016032; FDR = 1 . 05e-16 ) for the second analysis . Mass spectrometry analysis of JH73 and JH73g antibodies was performed as follows: the antibodies were buffer exchanged to 100 mM ammonium bicarbonate using 7K molecular weight cutoff Zeba Spin Desalting columns ( Thermo Scientific ) as per the manufacturer recommended protocol to remove glycerol . Following buffer exchange , samples were denatured with 8M Urea in Tris-HCl solution . Denatured samples were reduced with dithiothreitol ( 2 μl of a 1 M solution ) at 37°C for 1 hr and then alkylated with iodoacetic acid at room temperature in dark for 45 min ( 12 μl of a 1M solution ) . Samples were diluted to final urea concentration of 2M to facilitate enzymatic digestion . Each sample was split into two aliquots and digested using trypsin and pepsin . For trypsin digestion , 100 ng of sequencing grade-modified trypsin was added and digestion proceeded overnight on a shaker at RT . To inactivate the trypsin , samples were acidified using trifluoroacedic acid ( TFA ) to final concentration of 0 . 2% . For pepsin digest , samples were acidified to pH <2 using 1M HCl and digested using 100 ng of pepsin ( Promega ) for 1 hr at 37°C . Pepsin was heat inactivated by incubating the samples at 95°C for 15 min . Peptide extraction was performed by addition of 5 μl of R2 20 μm Poros beads slurry ( Life Technologies Corporation ) to each sample . Samples were incubated with agitation at 4°C for 3 hr . Peptide extraction was performed as described above . Digested peptides were loaded onto the column using the HPLC set up described above . The peptides were gradient eluted directly into a Q Exactive ( Thermo Scientific ) mass spectrometer using a 30 min gradient from 5% to 30% solvent B ( 80% acetonitrile , 0 . 5% acetic acid ) , followed by 10 min from 30% to 40% solvent B and 10 min from 40% to 100% solvent B . The Q Exactive mass spectrometer acquired high resolution full MS spectra with a resolution of 70 , 000 , an AGC target of 1e6 , maximum ion time of 120 ms , and a scan range of 400 to 1500 m/z . Following each full MS twenty data-dependent high resolution HCD MS/MS spectra were acquired using a resolution of 17 , 500 , AGC target of 5e4 , maximum ion time of 120 ms , one microscan , 2 m/z isolation window , fixed first mass of 150 m/z , Normalized Collision Energy ( NCE ) of 27 , and dynamic exclusion of 15 seconds . For immunofluorescence staining , cells were grown on coverslips or chamber-slides ( Nunc , prod . number 154534 ) coated with 10 μg/ml fibronectin ( ThermoFisher , prod . number 33016–015 ) in PBS for 4 hours-over night at 37°C . After plating cells were treated with 0 . 1 μg/ml doxycycline to induce expression of L1 ( ORFeus or L1rp ) . After induction cells were prefixed adding formaldehyde ( Fisher , prod . number F79-500 ) 11% directly to the culture media to a final concentration of 1% . After 10 min at room temperature the media/formaldehyde mixture was discarded and cells were fixed for 10 min at room temperature with formaline 4% in PBS ( Life Technologies , prod . number 10010–049 ) . For PCNA/ORF2p immunofluorescence presented in Figure 9 cells were fixed for 20 m in cold methanol . Cells were then washed twice in PBS supplemented with 10 mM glycine and three times in PBS . Cells were then incubated for at least 1 hr at room temperature in LiCOR blocking buffer ( LiCOR prod . number 927–40000 ) . Upon blocking , cells were incubated over night at 4°C with primary antibodies diluted in LiCOR blocking buffer . The next day cells were washed five times in PBS with 0 . 1% Triton-X 100 and then incubated in secondary antibodies ( Invitrogen , prod . number A11029 , A11031 , A11034 , A11036 , A32733 ) for 1 hr in the dark at room temperature . Cells were then washed five times in PBS with 0 . 1% Triton-X100 and 3 times in PBS and then coverslips or chamber slides mounted using VectorShield mounting media with DAPI ( Vectorlab , prod . number H1200 ) . Pictures were taken using an EVOS-FL Auto cell imaging system ( Invitrogen ) . Pictures of HeLa-S . FUCCI live cells expressing Halotag7 ORF1p and ORF2p were obtained incubating the cells for 15 min with 100 nM JF646 dye ( a gift from Timothee Lionnet ) ( Grimm et al . , 2015 ) in complete FluoroBrite DMEM media ( ThermoFisher , prod . number A1896701 ) . After incubation with the dye , the cells were washed twice in PBS before observation under the microscope . Live cell imaging was performed using an EVOS-FL auto cell imaging system with on stage incubator . Live cell images and Z stack movies were obtained using an Andor Yokogawa CSU-x spinning disk on a Nikon TI Eclipse microscope and were recorded with an scMOS ( Prime95B , Photometrics ) camera with a 100x objective ( pixel size 0 . 11 μM ) . Images were acquired using Nikon Elements software and analyzed using ImageJ/Fiji ( Schindelin et al . , 2012 ) . Cytoplasmic/nuclear fractionation was performed using a protein fractionation kit ( Thermo prod . number 78840 ) . HeLa M2 cells were synchronized in M phase by nocodazole treatment and mitotic shake off . Briefly , cells were treated for 12 hr with nocodazole 60 ng/ml and mitotic cells collected in the supernatant after vigorous tapping of the plate . Cells were then washed three times in complete media and released into the cell cycle in fresh complete DMEM media . HeLa cells were synchronized in G1/S boundary by double thymidine synchronization . 0 . 35 × 106 cells were plated in 10 cm culture plates in complete media . After six hours media was exchanged with complete DMEM supplemented with 2 mM thymidine freshly solubilized . After 18 hr , cells were washed three times in PBS and released into the cell cycle with complete media . After 9 hr , 2 mM thymidine medium was added a second time to the cells for additional 15 hr . Cells were then trypsinized , washed twice in PBS and released into the cell cycle in complete DMEM media . About 1 × 106 cells were collected in a 1 . 5 ml tube and fixed at −20°C with 70% cold ethanol for at least 24 hr . Cells were then pelleted by centrifugation at 420 rcf for 5 m and resuspended in HBSS:Phosphate citrate buffer 1:3 ( phosphate citrate buffer: 24 parts 0 . 2M Na2HPO4 and 1 part of 0 . 1M citric acid ) supplemented with 0 . 1% Triton X-100 . Cells were incubated in this buffer for 30 m at room temperature , subsequently pelleted and resuspended in propidium iodide ( PI ) staining buffer ( 20 μg/ml PI , 0 . 5 mM EDTA , 0 . 5% NP40 , 0 . 2 mg/ml RNase A in PBS ) . Cells were incubated in PI staining buffer for 2 hr at 37°C in the dark . After incubation , PI fluorescence was analyzed on an Accuri C6 flow cytometer ( BD bioscience ) to determine the percentage of cells in each stage of the cell cycle . RNA-FISH against recoded and non-recoded L1 was performed using 48 probes against L1 sequence and labelled with cy5 or cy3 fluorophores . The 20 nucleotides probes were designed using Stellaris probe designer from LGC Biosearch Technologies ( minimum spacing length = 2; genomic mask factor = 2 ) . The nucleotide sequence of the probes against L1rp and ORFeus is listed in Supplementary file 1 . In each well of a 24 well plate , 0 . 03 × 10∧6 HeLa-M2 cells expressing ORFeus or L1rp were plated on coverslips coated with 10 μg/ml fibronectin ( ThermoFisher , prod . number 33016–015 ) . Expression of L1 was induced with 0 . 1 µg/ml doxycycline for 24 hr . After induction cells were prefixed adding formaldehyde ( Fisher , prod . number F79-500 ) 11% directly to the culture media to a final concentration of 1% . After 10 min at room temperature the media/formaldehyde mixture was discarded and cells were fixed for 10 min at room temperature with formaline 4% in PBS ( Life Technologies , prod . number 10010–049 ) . Cells were then permeabilized for 10 min in PBS supplemented with 0 . 5% Triton-X 100 followed by a wash in PBS for additional 10 min . Cells were then incubated in pre-hybridization buffer ( 10% deionized formamide in 2X saline-sodium citrate ( SSC ) buffer ) . L1 mRNA was then stained for 3 hr incubating the coverslips in hybridization buffer ( for 6 50 µl reactions ( total volume = 300 µl ) : 30 µl of deionized formamide , 6 µl of competitors ( 5 mg/ml E . coli tRNA +5 mg/ml salmon sperm ssDNA ) , 1 . 5 µl of 60 ng/µl probe , 150µl of 20% dextran sulfate , 30 µl of 20 mg/ml BSA , 52 . 5 µl of water ) at 37°C in the dark . After hybridization , coverslips were washed once for 20 min at 37°C in the dark and once at room temperature on a slow shaker with 10% formamide in 2X SSC buffer . Coverslips were then washed for 10 min in PBS , stained with DAPI ( 0 . 5 µg/ml in PBS ) and mounted on slides using ProLong Gold mounting media ( ThermoFisher , prod . number P36930 ) . We observed a much higher nuclear background signal using L1rp probes compared to using ORFeus probes in L1 non-expressing cells . This high background may be explained with non-specific binding of the probes to genomic L1 sequences . Immunofluorescence staining of ORF1p and ORF2p/Flag was performed after RNA-FISH as reported above . Retrotransposition assays shown in Figure 5 were performed using HeLa-M2 cells stably expressing plasmid EA79 ( pCEP-puro-ORFeus-GFP-AI ) . The experimental design is reported in Figure 5A . Briefly , HeLa M2 cells transfected with EA79 plasmid were selected for at least 5 days in medium containing 1 μg/ml puromycin to generate stable cell lines that maintain episomal EA79 . Expression of recoded L1 was induced with 1 μg/ml doxycycline . Measurements of GFP positive cells were done collecting cells by shake off ( Figure 5B ) or trypsinizing the cells ( Figure 5C and D ) and resuspending them in FACS buffer ( HBSS buffer supplemented with 1% FBS , 1 mM EDTA and 100 U/ml of Penicillin-Streptomycin ) . To exclude dead cells from the measurement of GFP+ cells , 5 μg/ml propidium iodide ( PI ) was added for at least 5 min to the solution containing cells before measuring the percentage of GFP+ cells with an Accuri C6 flow cytometer ( BD bioscience ) . GFP and PI signals were compensated before analysis . Retrotransposition assays presented in Figure 6C–D were performed using HeLa M2 cells transfected with EA79 plasmid and always selected for 5 days in media containing 1 μg/ml puromycin as specified in the experimental design reported in Figure 6A . Measurements of the cell cycle stage of cells that underwent retrotransposition using a fluorescent timer-AI reporter ( PM260 plasmid ) reported in Figure 7B and C were done following two different approaches: Analysis of the percentage of cells in the different cell cycle stages was determined using FlowJo v10 . 2 software . Quantification of western blot protein bands was performed using Image Studio ver . 3 . 1 software on LiCOR CLx scanned images . Quantification of cellular localization of ORF1p and ORF2p was performed counting more than 1000 cells in three different cell preparations . For each experiment , LED intensity , camera gain and exposure were kept constant and set so that background fluorescence was negligible using a negative control sample ( cells not expressing L1 ) that will have no signal using the chosen setting ( Figure 1—figure supplement 4 ) . For quantification reported in Figure 4A , HeLa cells expressing ORFeus were plated on eight wells chamber , induced with doxycycline , fixed and stained as described in the method section . Images of cells stained with JH74 antibody against ORF1p and Alexa 647 conjugated secondary antibody ( Invitrogen , prod . number A32733 ) , were collected with the Arrayscan VTI and quantified using the Compartmental Analysis Bioapplication . Briefly , 16 fields per treatment were acquired at 20x magnification and 2 × 2 binning ( 1104 × 1104 resolution ) . DAPI positive nuclei were identified using the dynamic isodata thresholding algorithm after minimal background subtraction . DAPI images were used to identify cell nuclei and to delineate the nuclear edges ( x = 0 ) . A ‘circle’ smaller than the nucleus ( x=-4 ) was used to identify cells with nuclear ORF1p and a ‘ring’ outside the nucleus ( 1 < x > 8 ) was used to identify cells expressing cytoplasmic ORF1p . Limits of fluorescence were set so that no cells were considered positive for preparations of cells not treated with doxycycline ( negative control ) . The reported parameters are explained below: Total = total number of DAPI nuclei counted; Nuclear ORF1p=cells with fluorescence signal inside the circle ( x=-4 ) higher than the limit ( higher fluorescence signal of the circle in the negative control ) Cytoplasmic/total ORF1p=cells with fluorescence signal inside the ring ( 1 < x > 8 ) higher than the limit ( higher fluorescence signal of the ring in the negative control ) . Because all cells expressing L1 show expression of ORF1p in the cytoplasm , the number of cells with cytoplasmic ORF1p was considered as the number of total cells expressing ORF1p . The percentage of cells with nuclear ORF1p was calculated as: ( number of cells with nuclear ORF1p ) 100/ ( number of cells with cytoplasmic ORF1p=total number of cells expressing ORF1p ) . All results are reported as mean and the error calculated as standard deviation ( S . D . ) or standard error of the mean ( S . E . M ) . Data are considered to be statistically significant when p<0 . 05 by two-tailed Student’s T test . In Figure 5 , asterisks denote statistical significance as calculated by Student’s T test ( p<0 . 05; p<0 . 01; *p<0 . 001 ) . HeLa M2 cells expressing ORFeus for 24 hr were plated on fibronectin treated chamber-slides , treated with 0 . 1 μg/ml doxycycline , fixed and stained using JH74 rabbit primary antibody against ORF1p , Alexa 647 conjugated anti-rabbit secondary antibody and DAPI as described in the method section . Images were collected with the Arrayscan VTI system and cell positions within each slide obtained using the Compartmental Analysis Bioapplication . To quantify if cells with nuclear ORF1 are significantly closer to each other compared to random cells , we first calculated the shortest distance for each nuclear ORF1 cell to another nuclear ORF1 cell . Next , we randomly and repeatedly ( n = 1000 ) select the same number of cells as there are nuclear ORF1 cells and obtain the distribution of distances that correspond to random localization of non-nuclear ORF1 cells . We used the random distribution to calculate the p-value and a false discovery rate ( FDR ) . Finally , we compared the distribution of distances for nuclear ORF1 cells that are significantly closer to each other ( p-value=0 . 1 ) with the distribution of a random sample that did not express nuclear ORF1 using a Wilcoxon rank sum test . Mass spectrometry data for ORF2p and V5-PCNA sequential IP presented in Figure 9 have been deposited in MassIVE archive under submission number MSV000081124 and in ProteomeXchange archive under submission number PXD006628 .	Only two percent of our genetic material or genome are occupied by genes , while between 60-70 percent are made up of hundreds of thousands of copies of very similar DNA sequences . These repetitive sequences evolved from genetic elements called transposons . Transposons are often referred to as ‘jumping genes’ , as they can randomly move within the genome and thereby create dangerous mutations that may lead to cancer or other genetic diseases . LINE-1 is the only remaining active transposon in humans , and it expands by copying and pasting itself to new locations via a process called 'retrotransposition' . To do so , it is first transcribed into RNA – the molecules that help to make proteins – and then converted back into identical DNA sequences . Previous research has shown that LINE-1 can form complexes with a series of proteins , including the two encoded by LINE-1 RNA itself: ORF1p and ORF2p . The LINE-1 complexes can enter the nucleus of the cell and insert a new copy of LINE-1 into the genome . However , until now it was not known how they do this . To investigate this further , Mita et al . used human cancer cells grown in the lab and tracked LINE-1 during the different stages of the cell cycle . The results showed that LINE-1 enters the nucleus as the cell starts to divide and the membrane of the nucleus breaks down . The LINE-1 complexes are then retained in the nucleus while the membrane of the nucleus reforms . Later , as the cell duplicates its genetic material , LINE-1 starts to copy and paste itself . Mita et al . , together with another group of researchers , also found that during this process , only LINE-1 RNA and ORF2p were found in the nucleus . This shows that the cell cycle dictates both where the LINE-1 complexes gather and when LINE-1 is active . A next step will be to further investigate how the ‘copy and paste’ mechanisms of LINE-1 and the two LINE-1 proteins are regulated during the cell cycle . In future , this may help to identify LINE-1’s role in processes like aging or in diseases such as cancer .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2018	LINE-1 protein localization and functional dynamics during the cell cycle
Resident microbes play important roles in the development of the gastrointestinal tract , but their influence on other digestive organs is less well explored . Using the gnotobiotic zebrafish , we discovered that the normal expansion of the pancreatic β cell population during early larval development requires the intestinal microbiota and that specific bacterial members can restore normal β cell numbers . These bacteria share a gene that encodes a previously undescribed protein , named herein BefA ( β Cell Expansion Factor A ) , which is sufficient to induce β cell proliferation in developing zebrafish larvae . Homologs of BefA are present in several human-associated bacterial species , and we show that they have conserved capacity to stimulate β cell proliferation in larval zebrafish . Our findings highlight a role for the microbiota in early pancreatic β cell development and suggest a possible basis for the association between low diversity childhood fecal microbiota and increased diabetes risk . Host-associated microbes play important roles in the development of animal digestive tracts ( Bates et al . , 2006; Semova et al . , 2012; Sommer and Bäckhed , 2013 ) . Using the gnotobiotic zebrafish model , our group has shown previously that resident microbes promote host processes in the developing intestine such as epithelial differentiation ( Bates et al . , 2006 ) and proliferation ( Cheesman et al . , 2011 ) . The role of microbes in the development of other digestive organs remains underexplored , despite the fact that many diseases in peripheral digestive organs are correlated with microbial dysbiosis ( Chang and Lin , 2016; Gülden et al . , 2015 ) . The ability to manipulate resident microbes in the larval zebrafish ( Milligan-Myhre et al . , 2011 ) , combined with the optical transparency and sophisticated genetic tools of the zebrafish model , make it a powerful platform to investigate this question . Here , we use gnotobiotic zebrafish to demonstrate a role for resident microbes in promoting pancreatic β cell development . The zebrafish has a well-characterized program of β cell development , which is highly conserved with that of mammals ( Kinkel and Prince , 2009 ) . In the zebrafish embryo , initial β cells arise from precursors within the dorsal and ventral pancreatic buds ( Biemar et al . , 2001; Field et al . , 2003; Wang et al . , 2011 ) . The two buds fuse by 52 hr post fertilization ( hpf ) , and give rise to the fully fated pancreas with only a single islet of hormone-secreting endocrine cells , by 3 days post fertilization ( dpf ) ( Biemar et al . , 2001; Field et al . , 2003; Kumar , 2003 ) . Coinciding with the approximate time of larval emergence from the chorion by 3 dpf , these newly fated β cells begin to expand ( Chung et al . , 2010; Dong et al . , 2007; Hesselson et al . , 2009; Kimmel et al . , 2011; Moro et al . , 2009 ) . β cells derived from the dorsal bud become quiescent , while ventral bud derived β cells begin to undergo expansion via mechanisms of both proliferation and neogenesis ( Hesselson et al . , 2009 ) . Between 3 and 6 dpf , the number of β cells within the primary islet will almost double ( Moro et al . , 2009 ) . Intestinal colonization with microbes occurs concurrently with this early larval period of β cell expansion . Following development of the gut tube within the sterile embryo , the intestine of the emergent larva becomes open to the environment at both the mouth and the vent by 3 . 5 dpf , allowing for inoculation by environmental microbes ( Bates et al . , 2006 ) . Within the larval gut , bacteria proliferate rapidly , such that a single species in mono-association can reach the luminal carrying capacity within several hours ( Jemielita et al . , 2014 ) . Human post-natal β cell expansion also occurs concurrently with intestinal tract colonization by commensal microbes . In utero , β cells are produced via differentiation from progenitors ( Georgia et al . , 2006; Stanger et al . , 2007 ) and at birth this newly fated cell population begins to expand by self-proliferation ( Georgia and Bhushan , 2004; Gregg et al . , 2012; Kassem et al . , 2000; Teta et al . , 2007 ) . β cell proliferation rates peak at 2 years of age and then steadily decline ( Gregg et al . , 2012 ) . By 5 years of age , most of the β cell mass has become slow cycling and will not expand significantly again unless stimulated by elevated metabolic demands , such as obesity or pregnancy . At birth , infants are exposed to their mothers’ vaginal , fecal and skin associated microbes , which immediately begin to colonize the neonatal intestine ( Biasucci et al . , 2010; Dominguez-Bello et al . , 2010; Palmer et al . , 2007 ) . By 3 years of age , the composition and complexity of the microbiota typically resembles that of an adult associated community ( Murgas Torrazza and Neu , 2011; Palmer et al . , 2007; Yatsunenko et al . , 2012 ) . However , factors such as diet , birth mode and antibiotic exposure can result in reduced microbial taxonomic diversity during these early years of life ( Mueller et al . , 2015 ) . Notably , factors that reduce microbiota diversity are also associated with increased risk for diabetes mellitus ( Knip et al . , 2005 ) . Loss of β cell function through autoimmunity results in abnormal glucose homeostasis and is the cause of type 1 diabetes ( T1D ) in humans . Recent studies have shown that decreased taxonomic diversity of the intestinal microbiota is correlated with T1D ( Brown et al . , 2011; Giongo et al . , 2011 ) . Indeed , loss of bacterial diversity precedes the onset of T1D in children , and may play a causative role in disease ( Kostic et al . , 2015 ) . To our knowledge , no one has yet investigated a role for the gut microbiota in the development of pancreatic β cells . Communication between the intestine and the pancreas is critical for overall homeostasis . The two organs are therefore connected physically , metabolically , and developmentally in order to carry out their essential functions . We propose that this established and important connection might also mediate the influence of resident microbes on developmental processes in the pancreas . Here we examine the effects of microbial colonization on initial expansion of zebrafish primary islet β cells . We find that β cell mass expansion , up to at least 6 dpf , is promoted by the presence of the microbiota . Using a culture collection of zebrafish intestinal bacteria , we show that certain strains can restore β cell expansion in germ free ( GF ) fish . We report the discovery of a secreted protein , shared among these strains and named herein β cell expansion factor A ( BefA ) that is sufficient to recapitulate this effect . Homologs of the befA gene are present in the genomes of a subset of human intestinal bacteria , and we show that two of the corresponding proteins share BefA’s capacity to induce β cell expansion in zebrafish . To investigate a possible role for the microbiota in pancreas development and specifically in β cell expansion , we quantified total β cells in GF and conventionally reared ( CV ) Tg ( -1 . 0insulin:eGFP ) fish ( diIorio et al . , 2002 ) at 3 , 4 , 5 and 6 dpf ( Figure 1A , Figure 1—source data 1 ) . The number of β cells in CV fish increased steadily from 3 to 6 dpf ( Figure 1A ) . However , the average number of β cells in GF fish remained static over this time ( Figure 1A ) . Furthermore , at 6 dpf , the overall structure of β cells within the primary islet also appeared much less densely packed in GF than in CV fish ( Figure 1B ) . This effect is not likely to be due to changes in initial differentiation of the β cell population since the total number of β cells is not different between GF and CV fish at 3 dpf ( Figure 1A ) , a time at which exposure to bacteria is also limited . 10 . 7554/eLife . 20145 . 003Figure 1 . The microbiota are required for normal expansion of the larval β cell mass . ( A ) Total number of β cells per larva in GF ( white box plots ) and CV ( grey box plots ) fish at 3 , 4 , 5 and 6 dpf . In this , and in all subsequent figures , CV data are shown in grey box plots , and GF data , or statistically similar treatment groups , are shown in white box plots . In all relevant panels and remaining figures , box plot whiskers represent the 95% confidence interval of the data set . Single factor ANOVA indicates that gnotobiology of the fish was significant in determining the number of β cells present ( F7=9 . 01 , p=1 . 45e−8 ) . Labels a , ab and b indicate the results of post hoc means testing ( Tukey ) . The difference between GF and CV cell counts became significant at 6 dpf ( t=−5 . 91 , p<0 . 001 ) . ( B ) Representative 2D slices from confocal scans through the primary islets of 6 dpf CV and GF Tg ( -1 . 0insulin:eGFP ) larvae . Each slice is taken from the approximate center of the islet structure . Insulin promoter expressing β cells are in green and nuclei are blue . Scale bar = 40 μM . ( C ) The average amount of glucose ( pmol ) per larva aged 6 dpf ( * t17=−3 . 65 , p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 00310 . 7554/eLife . 20145 . 004Figure 1—source data 1 . Quantifications and statistical analysis of larval β cells corresponding to Figure 1A . Exact values of N , mean , median , SD , and SEM are reported in the source data file and are highlighted in yellow , magenta , blue , green , and cyan , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 00410 . 7554/eLife . 20145 . 005Figure 1—source data 2 . Quantifications and statistical analysis of 6 dpf larval free glucose levels corresponding to Figure 1C . Exact values of N , mean , median , SD , and SEM are reported in the source data file and are highlighted in yellow , magenta , blue , green , and cyan , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 005 Because insulin from β cells functions to reduce levels of circulating glucose , we tested whether the β cell deficiency in GF larvae at 6 dpf affected the metabolic function of the fish by measuring free glucose levels . The amount of glucose detected in GF fish was significantly higher than in CV fish ( Figure 1C , Figure 1—source data 2 ) . These data suggest that GF fish , with a paucity of β cells , are less efficient at importing and processing glucose from the blood due to lower levels of circulating insulin . This is consistent with previous studies showing free glucose levels in zebrafish larvae to be correlated with β cell numbers ( Andersson et al . , 2012 ) . We developed an experimental timeline , depicted in Figure 2A , to test the capacity of individual zebrafish bacterial isolates to induce β cell expansion . We derived embryos GF at 0 dpf and allowed them to develop in this environment until after hatching . At 4 dpf , when the GF larvae have a patent gut tube , we inoculated them with defined microbes and/or microbial derived products by adding these directly to the embryo media . The fish were incubated with the treatment of interest for 48 hr before analysis of the β cell mass at 6 dpf . 10 . 7554/eLife . 20145 . 006Figure 2 . Specific bacterial members of the zebrafish microbiota are sufficient to rescue normal expansion of the GF β cell mass . ( A ) Experimental timeline for all subsequent zebrafish experiments , unless stated otherwise . Experimental manipulations are denoted by red text . Important zebrafish developmental events are denoted by black text . ( B ) Quantification of β cells in CV , GF and GF larvae treated at 4 dpf with either non-sterile tank water ( XGF ) or mono-associated with a specific bacterial strain . Bacterial mono-associations are labeled by genus . Different Aeromonas sp are labeled with a number ( 1 , 2 or 3 ) . p<0 . 05 , p<0 . 01 , *p<0 . 001: Denotes treatment that is significantly different than GF by Tukey analysis . Additionally , here and in all subsequent figures , significant data sets ( p<0 . 05 when compared to GF ) are also highlighted as green box plots . ( C ) Bacterial isolates of the zebrafish gut and related strains are capable of forming mono-associations with larvae from 4 to 6 dpf . Quantification of the colony forming units ( CFUs ) per gut for each bacterial strain , assayed after 48-hr exposure to GF larvae . Dashed line denotes the limit of detection . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 00610 . 7554/eLife . 20145 . 007Figure 2—source data 1 . Quantifications and statistical analysis of larval β cells corresponding to Figure 2B . Exact values of N , mean , median , SD , and SEM are reported in the source data file and are highlighted in yellow , magenta , blue , green , and cyan , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 007 We found that we could rescue β cell numbers to CV levels by the addition of non-sterile , normal fish tank water to GF larvae at 4 dpf ( Figure 2B , Figure 2—source data 1 ) , suggesting that development of the normal number β cells is dependent upon microbes or microbial-derived products present in the water . We next inoculated 4 dpf GF larvae with a selection of bacterial isolates from the zebrafish intestine ( Stephens et al . , 2015 ) as well as one other related strain ( Bomar et al . , 2013 ) . We prioritized bacterial strains that were capable of forming robust mono-associations with larvae between 4 and 6 dpf , as measured by the number of bacteria found within the gut at 6 dpf ( Figure 2C ) . We found that the mono-associations with three different species of the genus Aeromonas and one species of the genus Shewanella was sufficient to rescue GF β cell numbers to levels observed in CV fish ( Figure 2B , Figure 2—source data 1 ) . Importantly , other isolates such as Vibrio sp . and Delftia sp . were not sufficient to rescue this phenotype ( Figure 2B , Figure 2—source data 1 ) , indicating that only specific members of the microbiota are capable of inducing expansion of the β cell mass . Bacterial interactions with host organisms often involve secreted molecules . To test whether a secreted bacterial factor ( s ) could influence β cell expansion , we harvested cell free supernatant ( CFS ) from overnight cultures of each Aeromonas strain shown to rescue β cell expansion ( Figure 2B ) and added these to GF larvae at 4 dpf . For each of the three strains of Aeromonas tested , the CFS alone was able to restore β cell numbers in GF fish ( Figure 3A , Figure 3—source data 1 ) , indicating that a secreted factor ( or factors ) produced by these bacteria is ( are ) sufficient to induce β cell expansion . As a control , we also treated GF fish with CFS from a Vibrio sp . isolate , which colonized the zebrafish gut ( Figure 2C , ) , but did not induce β cell expansion ( Figure 2B , * ) . We found the number of β cells in fish receiving Vibrio CFS was not significantly different from that of GF fish ( Figure 3A , Figure 3—source data 1 ) . Furthermore , the capacity to induce increased β cell numbers was lost when the Aeromonas 1 ( A . veronii ) CFS sample was treated with proteinase K ( Figure 3A , Figure 3—source data 1 ) , indicating that our secreted factor ( s ) of interest was likely to be a protein . Because of existing genetic reagents available for the A . veronii strain ( Bomar et al . , 2013 ) , and its capacity to modulate traits of gnotobiotic zebrafish and other hosts ( Bates et al . , 2006; Cheesman et al . , 2011; Graf , 1999; Rolig et al . , 2015 ) , we focused on this strain for the remainder of our analysis . 10 . 7554/eLife . 20145 . 008Figure 3 . Aeromonas secretes a factor that rescues normal expansion of the GF β cell mass . ( A ) Total β cell numbers in GF , CV and GF fish treated at 4 dpf with different cell free supernatant ( CFS ) samples . 'Aero . ' refers to bacteria of the genus Aeromonas , with each number ( 1 , 2 , 3 ) denoting a separate species . '+ PK' indicates proteinase K addition to the CFS sample prior to treatment . p<0 . 05 , p<0 . 01 , p<0 . 001: Denotes treatment that is significantly different than GF by Tukey analysis . ( B ) Total β cell numbers in CV , GF and GF fish treated at 4 dpf with separate ammonium sulfate fractions ( % AS ) prepared from the Aeromonas 1ΔT2SS CFS . Note that the 60–80% ammonium sulfate fraction resulted in the greatest increase in β cell numbers . ( C ) Total β cells in GF , CV and GF fish treated with purified protein . 10165 represents purified protein from the M001_10165 locus . ( D ) Representative 2D slices from confocal scans through the primary islets of GF , CV and 10165 protein treated Tg ( -1 . 0insulin:eGFP ) 6 dpf larvae . Insulin promoter expressing β cells are shown in green and nuclei are blue . Scale bar = 40 μM . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 00810 . 7554/eLife . 20145 . 009Figure 3—source data 1 . Quantifications and statistical analysis of larval β cells corresponding to Figures 3A . Exact values of N , mean , median , SD , and SEM are reported in the source data file and are highlighted in yellow , magenta , blue , green , and cyan , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 00910 . 7554/eLife . 20145 . 010Figure 3—source data 2 . Quantifications and statistical analysis of larval β cells corresponding to Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 01010 . 7554/eLife . 20145 . 011Figure 3—source data 3 . Quantifications and statistical analysis of larval β cells corresponding to Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 01110 . 7554/eLife . 20145 . 012Figure 3—figure supplement 1 . 10165 ( BefA ) protein purification . SDS-page gel image showing subsequent steps in the purification of BefA ( black arrowhead ) from E . coli cell lysate; lane 1: ladder , lane 2: cell lysate after IPTG induction , lane 3: supernatant from cell lysate after addition of nickel beads , lanes 4-7: elutions of BefA from beads . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 012 To narrow down the list of candidate proteins secreted by A . veronii , we tested whether the activity was present in the CFS of an A . veroniiΔT2SS mutant strain ( Maltz and Graf , 2011 ) lacking a functional type 2 secretion system ( T2SS ) , one of the major protein secretion pathways of Gram-negative bacteria . Despite the fact that it has a reduced secretome , CFS harvested from the A . veroniiΔT2SS strain was sufficient to rescue GF β cell numbers ( Figure 3A , Figure 3—source data 1 ) . Conveniently , this finding significantly decreased the number of candidate secreted A . veronii proteins with β cell expansion capacity . This result also suggested that our protein ( s ) of interest was secreted through an alternative mechanism . We next used ammonium sulfate precipitation to further separate proteins within the A . veroniiΔT2SS CFS . Each of the fractions was able to increase β cells in GF fish ( Figure 3B , Figure 3—source data 2 ) , suggesting that either A . veroniiΔT2SS produces multiple proteins with this activity , or that the effector was present to some extent within each fraction . Since the 60–80% fraction was able to induce the greatest increase in β cell numbers ( Figure 3B ) , we used mass spectrometry to analyze the content of this fraction , which led to the identification of 163 proteins ( Supplementary file 1 ) . To identify promising candidates from this list , we took advantage of the fact that our zebrafish-associated bacterial isolates , for which we have drafted genome sequences ( Stephens et al . , 2015 ) , differed in their capacity to induce β cells ( Figure 2B ) . Using basic local alignment search tool ( BLAST ) we identified those proteins from our candidate list that were , first , predicted to be encoded by the genomes of the four bacterial strains with β cell expansion capacity , and second , absent from the strains lacking this capacity . Our analysis identified one single candidate gene , denoted by the locus tag , M001_10165 ( 10165 ) , predicted to encode a putative protein of 261 amino acids . Consistent with the candidate protein being found in the CFS , the putative protein contained a predicted N-terminal secretion sequence . To test whether 10165 encoded the secreted protein responsible for inducing β cell expansion , we cloned the gene into an inducible expression vector in E . coli strain BL21 , which contains no 10165 homologues in its genome . We expressed and purified the 10165 protein to homogeneity , as confirmed by SDS-page gel electrophoresis ( Figure 3—figure supplement 1 ) . Purified protein was added to flasks of 4 dpf GF zebrafish larvae . This treatment was sufficient to rescue β cell numbers to CV levels by 6 dpf ( Figure 3C , Figure 3—source data 3 ) . The islets of larvae treated with the purified protein were visibly expanded compared to those of GF animals ( Figure 3D ) . Therefore , we have named this protein β cell expansion factor A ( BefA ) after its observed activity in zebrafish . To determine whether the befA ( 10165 ) locus is necessary for A . veronii to induce an increase in β cell numbers , we generated an A . veroniiΔbefA mutant strain by replacing the coding region of befA with a chloramphenicol resistance gene . To ensure that the loss of the befA gene would not affect the ability of A . veronii to form mono-associations with larvae , we performed growth and colonization assays and saw no deficiency in either the in vitro growth rate ( Figure 4—figure supplement 1A ) or the ability of A . veroniiΔbefA to colonize the GF intestine compared to the wild-type ( WT ) strain ( Figure 4A ) . However , when inoculated in a 1:1 ratio together with A . veroniiWT , the A . veroniiΔbefA strain showed a small yet reproducible fitness disadvantage as measured by colonization level and competition index after 48 hr ( Figure 4—figure supplement 1B , C ) . This result indicates that BefA confers some colonization benefit for A . veronii within the larval gut . 10 . 7554/eLife . 20145 . 013Figure 4 . BefA is required for Aeromonas to induce GF β cell expansion . ( A ) Quantification of the colony forming units ( CFUs ) per gut in GF fish mono-associated ( MA ) with either wild type ( WT ) or mutant ( △befA ) A . veronii strains for 48 hr . Dashed line denotes the limit of detection ( B ) Total β cells in GF fish that have been mono-associated with △befA , treated with CFS from either WT or △befA , treated with purified BefA , or have been inoculated with a combination of these . p<0 . 01 , *p<0 . 001: Denotes treatment that is significantly different than GF by Tukey analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 01310 . 7554/eLife . 20145 . 014Figure 4—source data 1 . Quantifications and statistical analysis of larval β cells corresponding to Figure 4B . Exact values of N , mean , median , SD , and SEM are reported in the source data file and are highlighted in yellow , magenta , blue , green , and cyan , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 01410 . 7554/eLife . 20145 . 015Figure 4—figure supplement 1 . BefA confers a colonization advantage in the larval zebrafish gut . ( A ) Growth rates of A . veroniiWT ( black trace ) and A . veroniiΔbefA ( grey trace ) in vitro . Density measurements ( OD600 ) were taken every half hour for 25 hours on three replicate cultures grown in Lauria broth . ( B ) Resulting CFU’s of A . veroniiWT ( WT ) and A . veroniiΔbefA ( ΔbefA ) within the 6dpf larval gut after inoculation with a 1:1 ratio of each strain at 4 dpf . Dashed line denotes the limit of detection . ( C ) Competitive index ( CI ) calculation for data within panel B . CI value was calculated for each fish ( n=22 ) by dividing the ratio of mutant to WT bacteria within each gut by 6 dpf , divided by the ratio of mutant to WT bacteria used to inoculate the fish at 4 dpf . A one-sample t-test indicates that the mean CI value is significantly less than 1 ( dashed line ) ( t21=−3 . 21 , p<0 . 0001 . ) A CI value of 1 is expected if no competition exists . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 015 GF fish were mono-associated with the A . veroniiΔbefA strain , or treated with its CFS from 4 to 6 dpf . Neither treatment was sufficient to rescue β cell numbers to CV levels ( Figure 4B , Figure 4—source data 1 ) . However , mono-associations of A . veroniiΔbefA could be complemented in trans with either CFS from A . veroniiWT or purified BefA protein , which resulted in the restoration of the β cell population ( Figure 4B , Figure 4—source data 1 ) . Taken together , these data demonstrate that the BefA protein is necessary in an A . veronii mono-association for early β cell expansion and suggests that A . veronii only produces a single effector of host β cell expansion . Proliferation is the primary mode of human neonatal β cell expansion ( Gregg et al . , 2012; Kassem et al . , 2000; Teta et al . , 2007 ) . In 4–6 dpf zebrafish larvae , proliferation also contributes to β cell expansion ( Field et al . , 2003; Hesselson et al . , 2009; Moro et al . , 2009 ) . Therefore , we investigated whether CV larvae had higher levels of β cell proliferation than GF larvae . 4 dpf larvae were treated with the thymadine analog , 5-ethynyl-2’-deoxyuridine ( EdU ) for 48 hr to mark cells that underwent proliferation during this time window . We found that , by 6 dpf , CV larvae had significantly more EdU labeled insulin-expressing cells than GF larvae ( Figure 5A , B , Figure 5—source data 1 ) . Next we asked whether treatment of GF larvae with BefA was sufficient to restore β cell proliferation to CV levels . We found that BefA-treated GF larvae had EdU incorporation similar to CV fish and significantly greater than GF ( Figure 5A , B , Figure 5—source data 1 ) . CFS from our A . veroniiΔbefA strain was not sufficient to increase proliferation rates in GF fish ( Figure 5B , Figure 5—source data 1 ) . Our results show that BefA is sufficient to increase cell proliferation that gives rise to an expanded β cell population during early larval development . Furthermore , BefA seems to be the only product of the A . veronii CFS that is capable of inducing this cell proliferation . 10 . 7554/eLife . 20145 . 016Figure 5 . BefA facilitates β cell mass expansion through proliferation . ( A , D & E ) Representative 2D slices from confocal scans through the primary islets of GF , CV and BefA ( 10165 ) protein treated 6 dpf larvae . Scale bars = 40 μM . ( A ) Insulin promoter expressing β cells are shown in green , all nuclei are blue , and EdU containing nuclei are magenta . Left hand panels are a merge of all three markers . For ease of resolving cells that are double positive for both insulin and EdU , the right hand panels show the location of insulin outlined by white dashed lines . ( B ) Percentage of EdU positive β cells in CV , GF or GF treated with either purified BefA or CFS from A . veroniiΔbefA cultures ( △befA CFS ) . p<0 . 001: Denotes treatment that is significantly different than GF by Tukey analysis . ( C ) Total EdU positive exocrine cells quantified from the approximate central longitudinal plane of the pancreas in each fish . ( D ) Insulin promoter expressing β cells are shown in green , all nuclei are blue , and α cells , stained with anti-glucagon antibody are magenta . ( E ) Somatostatin promoter expressing δ cells are shown in white , all nuclei are blue , and β cells stained with anti-insulin antibody are outlined in green . ( F ) Total α cells in GF , CV and GF fish treated with BefA . ( G ) Total δ cells in GF , CV and GF fish treated with BefA . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 01610 . 7554/eLife . 20145 . 017Figure 5—source data 1 . Quantifications and statistical analysis of proliferation of larval β cells corresponding to Figure 5B . Exact values of N , mean , median , SD , and SEM are reported in the source data file and are highlighted in yellow , magenta , blue , green , and cyan , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 01710 . 7554/eLife . 20145 . 018Figure 5—figure supplement 1 . The microbiota increase β cell neogenesis from the EPD . Quantification of EDP localized insulin expressing cells per animal in 6 dpf CV and GF larvae . Error bars represent the standard deviation . t520=3 . 28 , p=0 . 0011 . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 018 In zebrafish larvae , both the proliferation of existing β cells as well as the proliferation of progenitors contribute to the expansion of β cells that occurs between 4 and 6 dpf ( Dong et al . , 2007; Field et al . , 2003 ) . Because our 48-hr EdU pulse labeled β cells born from both events , our experiment did not distinguish the exact cell population undergoing proliferation in response to BefA . Due to their low rates of proliferation , dividing β cells are difficult to detect without pulse labeling . Neogenesis of β cells from progenitors is also rare , but can be detected as the appearance of insulin positive cells in the extra-pancreatic duct ( EPD ) ( Dong et al . , 2007; Hesselson et al . , 2009 ) . We quantified insulin expressing cells in the EPD in 6 dpf CV and GF larvae . In a survey of over 500 Tg ( -1 . 0insulin:eGFP ) larvae , we found a slight but significant increase in EPD-localized insulin expressing cells in CV versus GF fish ( Figure 5—figure supplement 1 ) , suggesting that the microbiota increases endocrine progenitor proliferation . Whether the microbiota also promote proliferation of mature β cells in the islet and whether BefA promotes the proliferation of one or both of these cell populations remains to be determined . To test whether BefA activity was specific to endocrine tissue , or whether it acts as a nonspecific pro-proliferative stimulant in the pancreas , we analyzed its ability to induce proliferation in exocrine pancreatic tissue by treating Tg ( ptf1a:eGFP ) larvae ( Thisse et al . , 2004 ) with EdU and BefA from 4 to 6 dpf and quantifying proliferative eGFP positive cells . We found no difference in the level of exocrine cell proliferation across GF , CV and BefA treatments ( Figure 5C ) . To test whether β cells were the only endocrine cell type in the islet to be responsive to BefA , we also quantified the total number of glucagon-expressing α ( Figure 5D ) and somatostatin-expressing δ ( Figure 5E ) cells in GF , CV and BefA treated fish . We again found no difference in the total numbers of these cells across treatments ( Figure 5F , G ) . These results suggest that in the pancreas , β cells alone are responsive to the presence of BefA . We wondered if BefA-like proteins are produced by the human microbiota . Phylogenetic analysis of related sequences in bacterial genomes uncovered close homologs ( at least 82% amino acid sequence identity ) in many , but not all , species of the Aeromonas , Vibrio , and Photobacterium genera . We also found an example of a highly related sequence in the human-associated species Enterococcus gallinarum , which was likely acquired through a horizontal gene transfer event ( Figure 6A ) . Widening the search to include more distant homologs identified potentially related genes in three additional human-associated genera: Enterobacter , Escherichia , and Klebsiella ( Figure 6B ) . 10 . 7554/eLife . 20145 . 019Figure 6 . Homologs of BefA encoded in the human microbiome have conserved function in zebrafish . ( A ) Close homologs of BefA across microbial species . Each species is represented by its closest BefA homolog , with a minimum allowed amino acid sequence identity of 50% ( relative to the query sequence ) . Notably , the Enterococcus gallinarum homolog clusters among homologs from the Aeromonas genus , which is evidence of a possible lateral gene transfer event . ( B ) A view of the BefA phylogeny including more distant homologs ( sequence identity >20% ) and grouped by genus . The portion of the tree represented in A is contained in the light gray box . In both panels , red numbers indicate branch support ( values closer to 1 are better supported ) ; branches with support values <0 . 5 have been collapsed . Blue clades indicate genera that were associated with humans in metagenomes produced during the Human Microbiome Project ( HMP ) . Black arrowheads indicate genera tested for functional conservation in panel D . Scale bars indicate amino acid substitutions per amino acid site . ( C ) SDS-page gel: 1 = ladder , 2 = CFS from induction of E . coli BL21 carrying an empty vector , 3 = CFS from induction of E . coli BL21 carrying vector with an Enterococcus gallinarum homolog , estimated size of 29 kDa , lane 4 = CFS from induction of E . coli BL21 carrying vector with Enterobacter aerogenes homolog , estimated size of 21 kDa . White arrows indicate induced proteins . ( D ) Total β cells in CV , GF and GF fish that have been treated with either induced BL21 E . coli supernatant dominated by the homologous BefA protein encoded from Enterococcus gallinarum ( E . gal . homolog ) and Enterobacter aerogenes ( E . aero . homolog ) , or induced supernatant from an empty vector control . p<0 . 05 , p<0 . 01 , **p<0 . 001: Denotes treatment that is significantly different than GF by Tukey analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 01910 . 7554/eLife . 20145 . 020Figure 6—source data 1 . Quantifications and statistical analysis of larval β cells corresponding to Figure 6D . Exact values of N , mean , median , SD , and SEM are reported in the source data file and are highlighted in yellow , magenta , blue , green , and cyan , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 02010 . 7554/eLife . 20145 . 021Figure 6—figure supplement 1 . Amino acid sequence alignment of BefA and functionally conserved homologs . Amino acid sequence alignment by MUSCLE . Egal = Enterococcus gallinarum homolog sequence , Eaero = Enterobacter aerogenes homolog sequence , and BefA = original Aeromonas veronii HM21 BefA sequence . Red box contains predicted SYLF domain . Blue box indicates predicted secretion peptides . DOI: http://dx . doi . org/10 . 7554/eLife . 20145 . 021 We tested whether representative BefA-like proteins from human-associated bacteria had the capacity to induce β cell expansion in our gnotobiotic zebrafish model . We cloned into BL21 E . coli two befA-like genes: the more closely related homologue from Enterococcus gallinarum and a more distantly related homologue from Enterobacter aerogenes . The amino acid sequence alignment of these two homologs against the Aeromonas BefA sequence is shown in Figure 6—figure supplement 1 . Both the Aeromonas and Enterococcus sequences contain a short N-terminal hydrophobic secretion signal , which is not predicted in the more distant Enterobacter sequence . The most conserved region of these proteins is the C-terminal portion , which contains a putative SYLF domain of unknown function . Induction of expression of each gene in E . coli yielded CFS that were dominated by each of the respective homologous proteins , in contrast to the CFS from control E . coli expressing an empty vector ( Figure 6C ) . Upon addition of these supernatants to GF larval zebrafish , we observed rescue of β cell numbers to the CV level with both the Enterococcus gallinarum and Enterobacter aerogenes proteins , but not the empty vector control ( Figure 6D , Figure 6—source data 1 ) . These results indicate that members of human-associated microbiota produce secreted proteins capable of inducing β cell expansion . Using a gnotobiotic zebrafish model , we have discovered a class of proteins produced by resident gut bacteria that have the capacity to increase the expansion of pancreatic β cells during early zebrafish development . BefA and related homologues are predicted to contain a C-terminal SYLF domain , which has been described in proteins from organisms in all kingdoms of life , including humans , but for which little is known functionally beyond a possible role in lipid binding ( Hasegawa et al . , 2011 ) . Genes encoding BefA and related proteins are found in a small subset of all bacteria genera , with a predominance in genera of host-associated bacteria , but befA homologues are not ubiquitously present in any of these genera . Our finding of a role for specific secreted bacterial proteins in β cell development raises the possibility of a new link between the resident microbiota and diseases of β cell paucity , such as diabetes mellitus . Type 1 diabetes ( T1D ) , is caused by both genetic and environmental factors , as indicated by the 50% disease discordance among monozygotic twins ( Akerblom et al . , 2002 ) . One environmental factor associated with T1D is microbiota composition ( Gülden et al . , 2015 ) . Mechanistic models for the role of the microbiota in T1D etiology have focused on the capacity of the microbiota to modulate the development and function of the immune system , and thus influence the propensity of genetically susceptible individuals to develop autoimmunity to β cell antigens ( Gülden et al . , 2015 ) . Multiple aspects of host immune cell development and function known to play a role in T1D are altered by the loss of microbes , including development of lymphoid tissue ( Macpherson and Harris , 2004 ) and T cell differentiation and function ( Alam et al . , 2011; Farkas et al . , 2015; Ivanov et al . , 2008 ) . We hypothesize an additional role for the early microbiota in establishing the β cell population size that would either buffer against , or render individuals susceptible to , β cell depletion by autoimmune destruction . In humans , β cells undergo a period of postnatal expansion , before becoming quiescent around age two ( Gregg et al . , 2012 ) . Differences in β cell growth during this time are thought to account for the wide variation in β cell mass observed in adults ( Wang et al . , 2015b ) . The idea that early life β cell census could influence diabetes risk is supported by studies in both rodents and humans , and has been theorized as an important risk factor for type 2 diabetes ( Kaijser et al . , 2009 ) , a disease which is also influenced strongly by microbiota composition ( Cox and Blaser , 2014 ) . Compromised β cell development in rats results in an insufficient number of cells to adequately control glucose metabolism ( Figliuzzi et al . , 2010 ) . In mice , perinatal β cell proliferation rates can be tuned via the modulation of Gi-GPCR signaling ( Berger et al . , 2015 ) . Changes to early β cell proliferation capacity in these mice correlates directly with adult β cell mass , which subsequently impacts glucose regulation ( Berger et al . , 2015 ) . Furthermore , meta analysis of human data has revealed a correlation between an early age of β cell loss and more rapid onset of T1D ( Klinke , 2008 ) , consistent with the model that failure to generate a reserve of β cells early in development increases disease risk . We hypothesize that neonatal microbiomes with a low abundance of BefA equivalents would result in reduced β cell proliferation , lower β cell mass , and increased diabetes risk . We do not know how many different microbiota-derived molecules can stimulate β cell proliferation , but for the case of befA homologues , we know these to be sparsely distributed in bacterial genomes , such that microbiomes of low taxonomic diversity could lack these genes . The idea that microbiota-derived factors capable of protecting against diabetes are not widely conserved is consistent with human microbiota profiling data ( Morgan et al . , 2013 ) , our own functional assays of bacteria in gnotobiotic zebrafish , and other gnotobiotic rodent experiments . For example , specific bacterial lineages have been shown to attenuate disease in diabetes models , including Segmented Filamentous Bacteria ( SFB ) in the non-obese diabetic ( NOD ) mouse ( Kriegel et al . , 2011; Yurkovetskiy et al . , 2013 ) and Lactobacillus johnsonii in the Biobreeding rat model ( Valladares et al . , 2010 ) . Furthermore , Wen and colleagues have shown that certain microbial assemblages , but not others , confer disease protection in neonatal NOD mice ( Peng et al . , 2014 ) . Additional recent work by Wen and colleagues demonstrates early development as a critical window for microbiota modulation of disease risk in NOD mice ( Hu et al . , 2015 ) . We have shown that BefA acts during early developmental stages in zebrafish , and we hypothesize that β cell expansion during this developmental window is important for disease prevention , and may be a critical period for clinical intervention for infants at risk for T1D development . Further work will be required to determine whether BefA is capable of inducing proliferation of adult β cells in zebrafish or other animals . Why certain bacteria produce BefA is unclear . In the context of the zebrafish intestinal environment , BefA confers a slight colonization advantage to A . veronii , however this is unlikely to be related to its capacity to induce β cell mass , because the colonization requirement is only apparent in the context of co-colonization with wild type A . veronii that induce normal β cell numbers . It is possible that bacterial modulation of host β cell number serves a purpose for the bacteria not measured in our assay . Alternatively , bacteria may produce BefA for a purpose independent of β cell expansion and the host simply uses this bacterial molecule as a cue for its own developmental program . Learning the molecular basis for BefA sensing by the host , and whether it interacts directly or indirectly with β cells , will help shed light on the nature and evolutionary conservation of this interspecies signaling . It will also be important to understand the bacterial function of BefA in order to be able to manipulate its abundance for potential therapeutic purposes . The incidence of autoimmune diseases such as T1D has been increasing markedly in developed nations over the past several decades . One theory to explain this phenomenon is the disappearing microbiota hypothesis , which proposes that over time , as our modern lifestyles have become increasingly sterile , we have lost ancestral microbial symbionts important for specific aspects of our health ( Blaser and Falkow , 2009 ) . Our discovery of a specific class of bacterial proteins that promote β cell expansion in early development is consistent with the hypothesis that loss of specific microbial taxa from gut microbiota could underlie increased diabetes risk . Specifically , we suggest that BefA-like proteins promote the establishment of a robust β cell population that is more resilient to subsequent β cell loss . Because befA is a relatively rare component of the microbiome , we cannot measure it directly from available metagenomic sequence data to test our hypothesis that befA abundance correlates with reduced diabetes risk . The low abundance of befA in metagenomes also highlights the challenge of discovering disease determinants from metagenomic data , and emphasizes the importance of functional screening approaches . The larval zebrafish has served as a valuable high-throughput vertebrate model for the identification of new compounds and pathways that can increase β cell numbers exogenously ( Andersson et al . , 2012; Wang et al . , 2015a ) . We have employed the gnotobiotic zebrafish to explore how microbial cues modulate β cell development . Our discovery of BefA highlights the importance of the microbiota in shaping the development of an extra-intestinal tissue and influencing the overall metabolic state of the host . We postulate that resident bacteria are a rich and underexplored source of functionally conserved molecules that shape early host development in ways that impact disease risk in later life . All zebrafish experiments were performed using protocols approved by the University of Oregon Institutional Care and Use Committee and followed standard protocols . Zebrafish embryos were derived germ-free ( GF ) as previously described ( Bates et al . , 2006 ) . XGF and mono-associated larvae were also generated as previously described ( Bates et al . , 2006 ) , except that all bacterial inoculate were added to GF flasks at 4 dpf at a final concentration of 106 CFUs/mL . In experiments quantifying the colonization levels of bacterial isolates , each strain was added to the embryo media ( EM ) and incubated with the larvae for 48 hr at 27°C . Larvae were sacrificed at 6 dpf , immediately before the gut was removed and homogenized in a small sample of sterile EM . Dilutions of this gut slurry were plated onto tryptic soy agar and allowed to incubate overnight at 30°C . Colonies from each gut were quantified . A minimum of 10 guts per mono-association or di-association were analyzed . To measure β cell function in GF and CV zebrafish larvae , levels of free glucose were measured at 6 dpf using a free glucose assay kit ( BioVision , Milpitas , CA ) as described previously ( Andersson et al . , 2012; Gut et al . , 2013 ) except that only 10 larvae were combined per tube . Three to five biological replicates ( sets of 10 larvae ) were completed for both GF and CV treatments each time the assay was conducted . Data shown here were combined from 3 separate experimental assays or technical replicates . GF fish were inoculated with secreted bacterial products at 4 dpf by adding cell free supernatant ( CFS ) at a final concentration of 500 ng/mL to the water of the sterile flasks . CFS was harvested from a 50 mL overnight culture of the specified bacterial strain . The cultures were centrifuged at 7000 g for 10 min at 4°C . The supernatant was then filtered through a 0 . 22-µm sterile tube top filter ( Corning Inc . , Corning , NY ) ; sterile supernatant was concentrated at 4°C for 1 hr at 3000 g with a centrifugal device that has a 10 kDa weight cut off ( Pall Life Sciences , Port Washington , NY ) . For experiments utilizing proteinase K ( Qiagen , Hilden , Germany ) , the enzyme was added to samples of CFS at a final concentration of 100 μg/mL and allowed to incubate at 55°C for 1 hr before inactivating the enzyme at 90°C for 10 min . Ammonium sulfate fractionation was performed on un-concentrated , sterile CFS from a 50 mL overnight culture by slowly adding 100% ammonium sulfate until solutions of 20% , 40% , 60% and 80% ammonium sulfate were achieved . These solutions were prepared at 4°C . Precipitated proteins were collected from each fraction by centrifugation at 4°C and 14 , 000 g for 15 min . The proteins were resuspended in cold EM and dialyzed for 2–3 hr at 4°C before adding them to 4 dpf GF larvae at a final concentration of 500 ng/mL . The 60–80% ammonium sulfate fraction of the A . veroniiΔT2SS CFS was sent to the Proteomics Lab at Oregon Health and Science University in Portland , OR for protein identification ( partial sequencing ) analysis . The nucleotide sequence for the befA gene from was amplified from A . veronii using the following forward and reverse PCR primers respectively: 5’-GCCCATATGatgaacaagcgtaactggttgctg-3’ and 5’-GGCCTCGAGgcggctcgtttcagtcaagtc-3’ . The nucleotide sequences for both the Enterococcus gallinarum and Enterobacter aerogenes befA gene homologs were obtained from NCBI and subsequently synthesized by GenScript , Piscataway , NJ . Each of these two genes was then cloned separately into the pET-21b plasmid ( Novagen , Darmstadt , Germany ) , which contains an IPTG inducible promoter . A His•Tag was added to the C-terminal of the original BefA protein sequence for subsequent purification . As a control , a second version was also constructed lacking the tag . These vectors were then transformed into BL21 Escheria coli ( RRID:WB_HT115 ( DE3 ) ) , treated with 0 . 5 – 1 . 0 mM IPTG during exponential growth phase ( OD600 = 0 . 4–0 . 6 ) and allowed to grow for 3–4 more hours at 30°C . This resulted in both a CFS and cell lysate dominated by our proteins of interest , as confirmed via SDS-page gel electrophoresis by the presence of dark bands of the expected sizes for each protein . These bands were absent from BL21 cultures carrying an empty pET-21b vector . The CFS from these inductions was added to GF zebrafish at 4 dpf at a final concentration of 500 ng/mL . For purification of BefA , IPTG induced BL21 cells were sonicated at 32 , 000 g in a 50 nM Tris , 150 mM NaCl buffer ( buffer A ) . The supernatant was then added to a solution of nickel beads ( Thermo Scientific HisPur Ni-NTA Resin , Waltham , MA ) to capture the His•tag . The beads were washed several times in a 30 mM imidazol solution in buffer A and subsequently eluted in 300 mM imidazole solution in buffer A . The isolation of pure BefA was confirmed with SDS-page gel electrophoresis by the presence of a single band of about 29 kDa in size . Purified BefA was added to 4 dpf GF fish at a final concentration of 500 ng/mL . To create the A . veroniiΔbefA mutant strain , a vector containing a chloramphenicol resistance cassette was transformed into SM10 E . coli . Conjugation between wild-type Aeromonas veronii HM21 and the vector carrying SM10 E . coli strain was carried out , allowing the chloramphenicol resistance gene to replace the befA locus in A . veronii via allelic exchange . Candidate mutants were selected for loss of the plasmid and maintenance of chloramphenicol resistance . Insertion of the chloramphenicol cassette into the befA locus was verified in these candidates by PCR . Joerg Graf graciously provided us with the A . veroniiΔT2SS strain ( Maltz and Graf , 2011 ) . Tg ( -1 . 0insulin:eGFP ) ( RRID:ZFIN_ZDB-GENO-100513-10 , ZIRC , Eugene , OR ) ( diIorio et al . , 2002 ) zebrafish embryos were used to visualize and quantify the total number of β cells in developing larvae . Tg ( insulin:PhiYFP-2a-nsfB , sst2:mCherry ) ( RRID:ZFIN_ZDB-GENO-120217-6 ) ( Wang et al . , 2015a ) were obtained from Jeff Mumm and were used to visualize and quantify δ cells . All experiments were analyzed at 6 dpf unless otherwise specified . At all time points in all experiments , larvae were fixed with 4% paraformaldehyde supplemented with 0 . 01% TritonX-100 ( Thermo Fisher Scientific , Waltham , MA ) at 4°C overnight , or at room temperature for 2–3 hr , and then washed with PBS . The following antibodies were used to distinguish α and β cells: guinea-pig anti-insulin ( Dako Cat# A0564 , RRID:AB_10013624 , Carpinteria , CA ) , mouse anti-glucagon ( Sigma-Aldrich Cat# G2654 , RRID:AB_259852 ) , St . Louis , MO ) , rabbit anti-GFP ( Molecular Probes Cat# A-11122 , RRID:AB_221569 ) , mouse anti-mCherry ( Abcam Cat# ab125096 , RRID:AB_11133266 , Cambridge , MA ) , Alexa Fluor 488 goat anti-rabbit ( Thermo Fisher Scientific , Waltham , MA ) , anti-mouse Cy3 ( Jackson ImmunoResearch Laboratories Inc . , West Grove , PA ) , Alexa Fluor 488 goat anti-guinea-pig ( Thermo Fisher Scientific , Waltham , MA ) , and TO-PRO-3-Iodide ( 642/661 ) ( Thermo Fisher Scientific , Waltham , MA ) . For experiments quantifying proliferation , EdU was added at 4 dpf directly to the EM at a final concentration of 0 . 1 mg/mL . The Click-iT EdU Imaging Kit ( Thermo Fisher Scientific , Waltham , MA ) was used to process the EdU label in whole fixed zebrafish prior to antibody staining , according to the manufacturer’s protocols . Whole , antibody-stained larvae were mounted for confocal microscopy ( BioRad Radiance 2100 ) with their right side facing up against the cover slip , which was flattened sufficiently to spread out the cells within the islet for optimal quantification of individual cells . For quantification of β cells and other primary islet cells , the entire endocrine portion of the pancreas was scanned using a 60X objective ( Nikon Eclipse E600FN ) , and Fiji ( RRID:SCR_002285 ) ( Schindelin et al . , 2012 ) software was used to analyze each image stack . For quantification of pancreatic exocrine tissue proliferation , Tg ( ptf1a:eGFP ) ( RRID:ZFIN_ZDB-GENO-080111-1 , ZIRC , Eugene , OR ) ( Thisse et al . , 2004 ) zebrafish were scanned through the entire pancreas with a 20X objective ( Nikon Eclipse E600FN ) and Fiji was used to analyze the percentage of proliferative cells in single sections from the center of the organ . Images were prepared for publication using the open source Inkscape software ( RRID:SCR_014479 ) . For experiments quantifying insulin-expressing cells in the region of the EPD , zebrafish were processed as described above , and analyzed on a Leica fluorescent microscope using a 2x objective . We screened for BefA homologs across microbial species using a blastp-based ( Altschul et al . , 1997 ) search of the UniProt Knowledgebase ( UniProt Consortium , 2015 ) ( version 6/2015 ) ; default search parameters were changed to allow ( i ) a maximum E-value of 1 . 0 and ( ii ) an arbitrarily large number of database hits . We classified database hits as 'close homologs' if amino acid sequence identity exceeded 50% ( relative to the query length ) and 'distant homologs' if their percent identity exceeded 20% . For phylogenetic analysis at the species level , each species was represented by the hit of highest percent identity to BefA among isolates of that species ( if any ) ; an analogous procedure was used for genus-level analysis . Aligned portions of database sequences were isolated and multiply aligned with MUSCLE ( RRID:SCR_011812 ) ( Edgar , 2004 ) . Phylogenetic trees were constructed from these multiple sequence alignments using PhyML ( RRID:SCR_014629 ) ( Guindon and Gascuel , 2003 ) and visualized within the Phylogeny . fr webserver ( Dereeper et al . , 2008 ) . Microbial genera were classified as 'human-associated' if they occurred with relative abundance >0 . 01% in at least 5 metagenomes from the Human Microbiome Project ( Huttenhower et al . , 2012 ) as profiled by MetaPhlAn ( RRID:SCR_004915 ) ( Segata et al . , 2012 ) . Secretion signal peptides were predicted from amino acid sequences using SignalP ( Petersen et al . , 2011 ) . Appropriate sample sizes for all experiments were estimated a priori using a power of 80% and a significance level of 0 . 05 . From preliminary experiments we estimated variance and effect . For larval β cell quantification , these parameters suggested using a sample size of 30 in order to detect significant changes between treatment groups . Therefore , each experiment contained about 10–15 biological replicates or individual fish per treatment group , although some larger experiments had fewer biological replicates due to limited material . Entire experiments or technical replicates were repeated multiple times , resulting in pooled data sets of about 20–50 biological replicates . These data are represented in the figures as box and whisker plots , which display the data median ( line within the box ) , first and third quartiles ( top and bottom of the box ) , and 95% confidence interval ( whiskers ) . Any data point falling outside the 95% confidence interval is represented as a solid dot . These pooled data were analyzed through the statistical software RStudio . For experiments comparing just two differentially treated populations , a Student’s t-test with equal variance assumptions was used . For experiments measuring a single variable with multiple treatment groups , a single factor ANOVA with post hoc means testing ( Tukey ) was utilized . A p-value of less than 0 . 05 was required to reject the null hypothesis that no difference existed between groups of data .	For a long time , genes and carefully orchestrated chemical signals from the mother’s body have been known to shape the earliest phases of an embryo’s development . More recently , researchers have started to appreciate that environmental cues – such as signals from nearby bacteria – also shape animal development . The body is home to a teeming community of bacteria and other microbes , called the microbiota . Because the gut houses more microbes than any other part of the body , the role that the microbiota plays in gut development has been investigated . Less is known about how the microbiota affects the development of the other organs involved in digestion , such as the pancreas . In developing fish and mammals , the pancreas grows its population of insulin-producing beta cells during the same period of development in which the microbial population of the gut becomes established . Now , Hill et al . show that certain gut bacteria are necessary for the pancreas to populate itself with a robust number of beta cells during development . Normally , the number of beta cells in zebrafish larvae increases steadily in the first few days after hatching . However , developing zebrafish that were reared in a microbe-free environment maintained the same number of beta cells as they had before hatching . Exposing the microbe-free fish to certain bacteria restored their beta cell populations to normal levels . Further investigation revealed that these bacteria release a protein called BefA that causes the beta cells to multiply . Some bacteria in humans produce proteins that are similar to BefA . Hill et al . performed experiments that showed that these proteins also stimulate beta cell development in microbe-free fish . Future studies are now needed to investigate the mechanism by which the proteins affect beta cell development , and to find out whether they have the same effect in humans and other animals . This will help us to understand whether a lack of gut microbes could contribute to the development of diseases , such as diabetes , that are characterized by insufficient numbers of beta cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Material", "and", "methods" ]	[ "developmental", "biology", "microbiology", "and", "infectious", "disease" ]	2016	A conserved bacterial protein induces pancreatic beta cell expansion during zebrafish development
Glioblastomas ( GBM ) are aggressive and therapy-resistant brain tumours , which contain a subpopulation of tumour-propagating glioblastoma stem-like cells ( GSC ) thought to drive progression and recurrence . Diffuse invasion of the brain parenchyma , including along preexisting blood vessels , is a leading cause of therapeutic resistance , but the mechanisms remain unclear . Here , we show that ephrin-B2 mediates GSC perivascular invasion . Intravital imaging , coupled with mechanistic studies in murine GBM models and patient-derived GSC , revealed that endothelial ephrin-B2 compartmentalises non-tumourigenic cells . In contrast , upregulation of the same ephrin-B2 ligand in GSC enabled perivascular migration through homotypic forward signalling . Surprisingly , ephrin-B2 reverse signalling also promoted tumourigenesis cell-autonomously , by mediating anchorage-independent cytokinesis via RhoA . In human GSC-derived orthotopic xenografts , EFNB2 knock-down blocked tumour initiation and treatment of established tumours with ephrin-B2-blocking antibodies suppressed progression . Thus , our results indicate that targeting ephrin-B2 may be an effective strategy for the simultaneous inhibition of invasion and proliferation in GBM . Glioblastoma ( GBM ) is the most common and aggressive type of primary brain tumour and one of the most lethal types of cancer ( Chen et al . , 2012b ) . Current therapies consist of maximally safe surgical resection followed by radio and chemotherapy . However , these are largely ineffective and GBM invariably recur following treatment , resulting in a median survival of <15 months ( Weathers and Gilbert , 2014 ) . A leading cause of GBM recurrence is the diffuse infiltration of the tumour cells into the surrounding brain parenchyma , which limits efficacy of surgery and radiotherapy ( Cuddapah et al . , 2014; Lamszus et al . , 2003 ) . In addition , GBM contain subpopulations of cells with stem cell properties termed glioblastoma stem-like cells ( GSC ) , which are able to self-renew , differentiate into tumour-bulk cells and reconstitute a phenocopy of the original lesion upon transplantation ( Venere et al . , 2011 ) . It is therefore increasingly recognized that GSC are critical players in tumour initiation and progression . Furthermore , GSC were shown to be intrinsically resistant to chemo- and radiotherapy and more invasive than non-stem tumour cells ( Chen et al . , 2012a; Cheng et al . , 2011; Bao et al . , 2006; Venere et al . , 2011; Sadahiro et al . , 2014 ) . This suggests that GSC might be the primary cells within GBM responsible for infiltration and tumour recurrence following therapy and that GSC-targeting therapies should be beneficial for GBM treatment . GBM invasion occurs along three main routes: the white matter tracts , the interstitial space of the brain and the perivascular space surrounding blood vessels ( Scherer , 1938; Cuddapah et al . , 2014 ) . Invasion along the perivascular space is a favourable migration route because endothelial cells secrete chemoattractants , which actively recruit tumour cells to the vasculature ( Montana and Sontheimer , 2011; Cuddapah et al . , 2014 ) . In addition , the perivascular space is enriched in migration-promoting ECM components and is fluid-filled , thereby opposing little physical resistance to invading tumour cells ( Gritsenko et al . , 2012; Cuddapah et al . , 2014 ) . Importantly , within GBM , GSC are particularly prone to perivascular invasion , likely due to their similarities with normal neural progenitors , which preferentially migrate along blood vessels during development and after injury in the adult ( Cuddapah et al . , 2014; Watkins et al . , 2014 ) . In agreement with this , GSC reside in perivascular niches and the majority of invading cells migrate along the host vasculature in xenograft GSC models of both mouse and human origin ( Farin et al . , 2006; Zagzag et al . , 2008; Baker et al . , 2014; Calabrese et al . , 2007; Winkler et al . , 2009; Watkins et al . , 2014 ) . Besides infiltration , GBM/vascular interactions underlie two additional key tumourigenic mechanisms . First , tumour cell migration along pre-existing normal blood vessels is an important tumour vascularisation mechanism , known as vascular co-option , by which tumours gain access to oxygen and nutrients independent of angiogenesis ( Donnem et al . , 2013 ) . Vascular co-option plays crucial roles during initial tumour growth , seeding of satellite lesions and tumour recurrence following therapy . Second , association with the perivascular niche provides GSC with important self-renewal and survival signals , which support GSC tumour-propagating abilities and therapeutic resistance ( Charles and Holland , 2010 ) . Therefore , interactions with the vasculature , particularly within the GSC compartment , are critical throughout gliomagenesis and were indeed proposed as promising therapeutic targets for GBM treatment , but the mechanisms remain poorly defined ( Cuddapah et al . , 2014; Vehlow and Cordes , 2013 ) . GSC share many properties with normal neural stem cells ( NSC ) , such as stem cell markers expression ( Sox2 , Nestin , CD133 , ALDH1 , etc . ) and multilineage differentiation and , importantly , extensive evidence indicates that NSC themselves can be cells of origin in GBM ( Chen et al . , 2012b; Venere et al . , 2011; Chen et al . , 2012a ) . However , GSC also significantly differ from their normal counterparts , in that they harbour transforming mutations that drive their tumourigenic properties , including deregulated proliferation and increased invasiveness ( Engstrom et al . , 2012 ) . Therefore , the comparison of GSC carrying known mutations to otherwise genetically matched normal NSC , should inform disease mechanisms , enable genotype-phenotype correlation and may identify new therapeutic targets for GBM . In this study , we generated a murine GBM model by sequentially introducing oncogenic lesions relevant to the human disease into normal NSC and compared vascular interactions of the resultant transformed GSC-like cells and immortalised parental cells to interrogate mechanisms of perivascular invasion . Using this system , we identified ephrin-B2 as a critical driver of perivascular invasion . ephrin-B2 is a member of the Eph/ephrin family of receptor tyrosine kinases and their membrane-bound ligands , a fundamental cell communication system with widespread roles in tissue development , maintenance and disease ( Pasquale , 2010 ) . Activation of Eph receptors by ephrin ligands on adjacent cells modulates cell behaviour , including migration , proliferation and stemness , by eliciting forward signalling downstream of Ephs and reverse signalling downstream of ephrins ( Kullander and Klein , 2002 ) . Deregulation of the Eph/ephrin system contributes to the pathogenesis of many types of cancer , including GBM ( Pasquale , 2010 ) . Indeed , EphA2 and EphA3 were shown to drive GSC self-renewal , whereas EphA2 , EphA4 , EphB2 , ephrin-B2 and ephrin-B3 have all been linked to GBM invasion , through incompletely understood mechanisms ( Day et al . , 2014; Tu et al . , 2012; Binda et al . , 2012; Day et al . , 2013; Nakada et al . , 2010; 2011 ) . We found that ephrin-B2 expressed on vascular endothelial cells inhibits the migration of non-tumourigenic cells , resulting in cell confinement . In contrast , transformation to GSC-like cells overrides this tumour-suppressive mechanism to drive perivascular invasion . We show that this is caused by upregulation in GSC-like cells of the same ephrin-B2 ligand , which desensitizes the cells to vascular confinement by constitutively activating Eph forward signalling non-cell-autonomously . Furthermore , we discovered that ephrin-B2 reverse signaling also elicits tumour cell proliferation in the absence of normal anchorage signals by driving Rho-A-dependent cytokinesis in a cell-autonomous manner . Consistent with these important roles , ephrin-B2 overexpression was sufficient to fully transform immortalised NSC to the same extent as oncogenic Ras . In human GBM specimens , high Ephrin-B2 levels were detected in perivascular tumour cells with GSC features at the infiltrative tumour margin , indicative of a role in the GSC compartment in primary tumours . Remarkably , EFNB2 knock-down in primary human GSC isolated from patient material or treatment of established tumours derived from these GSC with anti-ephrin-B2 single chain blocking antibodies strongly suppressed tumourigenesis , by concomitantly inhibiting vascular association and proliferation . Thus , ephrin-B2 may be an attractive therapeutic target for the treatment of GBM . To investigate mechanisms of GSC/vascular interactions in the context of syngeneic , immuno-competent brains , we sequentially introduced mutations commonly found in human GBM ( RTK activation , p53 and RB inactivation ) in primary murine SVZ NSC to generate fully transformed , GSC-like cells and genetically-matched immortalised NSC ( Network , 2008 ) . We used two complementary strategies for this . First , we used a ‘classical’ transformation paradigm previously shown to drive gliomagenesis in vivo , whereby NSC were immortalised with SV40 large-T antigen ( imNSC1 ) and transformed with RasV12 ( herein referred to as GSC1 ) to inactivate Trp53 and Rb , and mimic the increased Ras signalling that results from Nf1 loss , respectively ( Blouw et al . , 2003; Hahn et al . , 1999; Sonoda et al . , 2001; Huszthy et al . , 2012 ) . This approach allowed us to readily test candidate effectors by transforming NSCs isolated from mice carrying the specific mutation , as previously reported ( Blouw et al . , 2003 ) . In the second approach , we induced transformation by defined genetic changes in the same pathways to rule out artifacts of oncogene overexpression . Nf1fl/fl NSCs were immortalised with p53 shRNAs and ectopic CDK4 to inactivate p53 and the p16/RB axis , respectively ( imNSC2 ) , and transformed by Cre-mediated Nf1 deletion ( herein referred to as GSC2 ) . Unlike previously reported for SVZ NSC in vitro ( Wang et al . , 2012 ) , increased Ras signalling did not cause premature glial differentiation of the NSC in vitro in either model ( Figure 1—figure supplement 1a and Tables 1 and 2 and Supplementary files 1 and 2 ) . In contrast , GSC1 and GSC2 retained stem cell properties in vitro as judged by high clonal efficiency in neurosphere culture and differentiation into glial and neuronal lineages upon mitogen withdrawal ( Figure 1—figure supplement 1a–c ) . Furthermore , both cell types ( but not their immortalised controls ) formed colonies in soft agar ( Figure 4c , d ) and gave rise to highly aggressive tumours upon intracranial transplantation in immunocompromised mice ( 5/5 animals , 100% penetrance , median survival 24d for GSC1 and 38 . 5 for GSC2 ) , whereas GSC1 also did so in syngeneic animals ( 5/9 animals , 56% penetrance , median survival 73d ) , indicative of a more aggressive phenotype . Consistent with their stem-like properties , clonal dilution experiments revealed that as little as 100 cells of either line was sufficient to generate aggressive tumours ( Figure 1—figure supplement 1d ) . Importantly , the tumours recapitulated the histopathology and gene expression signatures of human GBM , including presence of necrosis , neovascularization , nestin and Sox2 expression and a strong enrichment in the Verhaak mesenchymal subtype gene signature ( Figure 1a–d and Figure 1—figure supplement 1e ) ( Kleihues , 2000; Verhaak et al . , 2010 ) . The tumours presented diffuse borders with the majority of invading cells migrating along blood vessels and displacing astrocyte endfeet and pericytes to come in direct contact with endothelial cells , as previously reported for both murine and human GSC ( Figure 1b and Figure 1—figure supplement 1f ) ( Zagzag et al . , 2008; Farin et al . , 2006; Nagano et al . , 1993; Watkins et al . , 2014 ) . Thus , GSC1 and 2 resemble mesenchymal glioblastoma stem-like cell lines and are highly similar , interchangeable model systems . 10 . 7554/eLife . 14845 . 003Figure 1 . The vasculature compartmentalises immortalised neural but not glioma stem cells through endothelial ephrin-B2 . ( a ) Representative H&E staining of GSC1 tumours 60 days after intracranial injection into nude mice . Neovascularisation ( i ) , focal necrosis ( ii ) and increased cellular density ( iii ) can be observed . ( b ) Representative fluorescent images of GFP-labeled GSC1 tumours stained for the vascular marker CD31 ( red ) and GFP to identify tumour cells ( green , left ) , the stem cell markers nestin ( green , middle ) and Sox2 ( red , right ) , as indicated . The arrows indicate examples of GFP-labeled GSC1 that have migrated away from the tumour mass and are invading along the vasculature as single cells or in small groups . Scale bars = 50 μm . ( c ) Heatmap showing Verhaak subtype classification of 4 GSC1 and 3 GSC2-derived tumours in vivo . Colour scale with corresponding normalised enrichment scores is shown on the right . All tumours classified as mesenchymal with a nes >2 . 6 . ( d ) Mean GSEA enrichment plot for the Verhaak mesenchymal gene signature in the GSC lines . Both nom p-val and FDR q-val = 0 . ( e ) Intravital 2-photon micrographs of GFP-labeled imNSC1 and GSC1 cells injected into the cortex of wildtype ( Efnb2 WT EC ) and endothelial specific Efnb2 knockout mice ( Efnb2-/- EC ) and imaged 7 days later over 6 hr through an intracranial window at a depth of 200 μm . The vasculature was labeled by tail-vein injection of Tx-red conjugated Dextrans ( 3000 MW ) . Arrowheads indicate representative perivascular migration patterns for each genotype . Scale bar = 50 μm . ( f ) Quantification of the migrated distance of the cells depicted in ( c ) . Each dot represents one cell . n indicates number of animals imaged . One way ANOVA with Tukey post hoc test . ( g ) Left: schematic representation of the experimental set up for in vitro migration assays with endothelial cells . Middle: merged fluorescent and phase contrast still images taken from time-lapse microscopy experiments of GFP-labeled imNSC1 and GSC1 ( green ) migrating towards brain microvascular endothelial cells ( bmvEC , unlabeled cells ) at the indicated time points . Right: quantification of boundary length at 60 hr . Students t-test . Scale bar = 500 μm . ( h ) Still fluorescence and phase contrast merged images of GFP-labeled imNSC2 and GSC2 migrating towards bmvEC ( unlabeled , left ) for 60 hr and quantification ( right ) as in ( e ) . Error bars denote s . e . m . , Students t-test . Scale bar = 500 μm . ( i ) Schematic representation of the experimental set up for in vitro migration assays toward recombinant ephrin-B2-Fc ( left ) , phase contrast images ( middle ) and quantification ( right ) of imNSC1 and GSC1 migration against coated ephrinB2-Fc pre-clustered with fluorescently-labelled anti-Fc antibodies at 60 hr . Error bars denote s . e . m . , Students t-test . Scale bar = 500 μm . Green dots denote boundary of ephrin-B2 coating identified by fluorescence . ( j ) Still images ( left ) and quantification ( right ) of GFP-labeled imNSC ( GFP ) migrating towards bmvEC ( unlabeled ) treated with control siRNA ( Scr ) or siRNA against Efnb2 ( siEfnb2 ) for 60 hr . Scale bar = 500 μm . Error bars denote s . e . m . , Students t-test . For this and later figures dots indicate individual data points and *p<0 . 001; p<0 . 01 and *<0 . 05 . See also Figure 1—figure supplement 1 and Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 00310 . 7554/eLife . 14845 . 004Figure 1—source data 1 . Raw data for all quantification of NSC/GSC migrated distance and boundary length shown in Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 00410 . 7554/eLife . 14845 . 005Figure 1—figure supplement 1 . GSC1/2 resemble glioma stem-like cells . ( a ) RNA-seq datasets by Zhang et al . were used to generate sets of astrocyte- , oligodendrocyte- , OPC- and neuron-specific marker genes ( left heatmap ) and expression of these cell-type-specific gene sets was determined in cultured normal neural stem cells ( NSC ) and GSC lines 1 and 2 ( right heatmap ) ( Zhang et al . , 2014 ) . See also Tables 1 , 2 and Supplementary files 1 , 2 for gene set enrichement scores . ( b ) Quantification of GSC1 , GSC2 and GSC1 Efnb2-/- clonal efficiency expressed percentage of colonies formed over total number of cells . n = 3 ( c ) Representative immunofluorescence images of GSC1 and 2 differentiated for 4 days . Cells were stained for the neuronal marker Tuj ( green ) , the astrocyte marker GFAP ( red , left ) and the oligodendrocyte marker O4 ( red , right ) and counterstained with DAPI ( blue ) . n = 3 Scale bar = 60 μm ( d ) In vivo limiting dilution analysis of GSC1 and 2 tumour-initiation potency . As little as 100 were sufficient to generate an aggressive tumour mass for both GSC1 and GSC2 . ( e ) Representative immunofluorescence images of GSC2 tumours stained for Sox2 ( red ) and Nestin ( green , left ) and GFP at the tumour margin ( green , right ) confirming their GBM nature and invasiveness . Scale bars = 50 μm n = 2 . ( f ) Representative immunofluorescence images of GFP labeled GSC1 tumours stained for vascular marker CD31 ( red ) , the astrocyte marker GFAP ( grey , upper panels ) and the pericyte marker NG2 ( grey , lower panels ) . The arrows indicate sites of direct contact between GSC and endothelial cells resulting from displacement of astrocytic endfeet and pericytes . Schematic representation of the interactions is drawn on the right hand side of the picture for clarity . ( g ) Side and top view of intravital 3D reconstructions ( i and ii ) of GSC1 injected into the cortex of syngeneic mice under a chronic cranial window . Note that the cells span a depth of 25–350 μm under the dura mater ( orange autofluorescence ) . Panel iii depicts orthogonal views taken from the same tumour , which show depth at which GSC/vascular interactions have been imaged for all experiments presented . Panel iv is a post-mortem immunofluorescence analysis of a craniotomy tumour stained for GFP ( green ) and CD31 ( red ) , indicating that tumours cells are all contained within the brain parenchyma . Examples of GSC/vascular interactions analysed by intravital imaging are magnified in the boxed areas of the pictures . ( h ) Intravital 2-photon micrographs of GFP-labeled GSC1 and imNSC1 cells injected into the cortex of Efnb2 WT mice and endothelial-specific Efnb2-/- mice and imaged 7 days later over 6 hr through an intracranial window at a depth of 200 μm . The vasculature was labeled by tail-vein injection of Tx-red conjugated Dextrans ( 3000 MW ) . Arrowheads indicate representative perivascular migration patterns for each genotype . Note that single imNSC and GSC cells display identical behaviours as cells in groups ( Figure 1c ) . Scale bar = 50 μm . ( i ) Phase contrast images ( upper panels ) and quantification ( lower panels ) of imNSC1 and GSC1 migration against coated Fc ( left ) , ephrinA5-Fc ( middle ) and ephrinA1-Fc ( right ) ligands . Error bars denote s . e . m . , Student t-test . n = 3 ( j ) Western analysis of ephrin-B2 levels in bmvEC transfected with control ( scr ) or Efnb2 specific ( siEfnb2 ) siRNA , confirming efficiency of the knock-down . n = 2 . See also Figure 1—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 00510 . 7554/eLife . 14845 . 006Figure 1—figure supplement 1—source data 1 . Raw data for all quantitative analyses shown in Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 00610 . 7554/eLife . 14845 . 007Table 1 . Gene set enrichment analysis FDR q-values for astrocyte , oligodendrocyte , OPC and neuron signatures in GSC lines vs NSC . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 007GAGE analysis q-value ( FDR ) GSCvsNSGSC1vsNSGSC2vsNSCell type term gene numberoligo111497neuron111497astro111486opc11149010 . 7554/eLife . 14845 . 008Table 2 . Gene set enrichment analysis FDR q-values for astrocyte , oligodendrocyte , OPC and neuron signatures in NSC vs GSC lines . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 008GAGE analysis q-value ( FDR ) NSvsGSCNSvsGSC1NSvsGSC2Cell type term gene numberastro4 . 06848E-222 . 84603E-232 . 04565E-19490opc1 . 57235E-167 . 01306E-173 . 24522E-14486neuron1 . 51727E-064 . 55938E-074 . 58202E-05497oligo0 . 0649202650 . 067080470 . 222744258497 Next , as only GSC1 formed tumours in syngeneic mice , we used this model to assess interactions with the normal brain vasculature in real time in vivo using 2-photon microscopy . GFP-labeled GSC1 or imNSC1 were injected at a mimimum depth of 150–200 μm into the cortex of syngeneic recipients under a chronic cranial window and 5–7 days later their migration in relation to Dextran-TxRed-labeled blood vessels was imaged over a period of 6 hr ( Figure 1e , f and Figure 1—figure supplement 1g ) ( Holtmaat et al . , 2009 ) . Although imNSC do not form tumors long-term , the cells were still present , viable and migratory at this time-point , as indicated by morphology and motility outside of the perivascular space . Consistent with previous reports , both cell types readily associated to the vasculature ( Kokovay et al . , 2010; Baker et al . , 2014; Farin et al . , 2006; Nagano et al . , 1993; Watkins et al . , 2014; Winkler et al . , 2009 ) . However , while individual GSC1 migrated out of the tumour mass along blood vessels , resulting in cell scattering , imNSC remained in stationary groups ( Figure 1e , f ) . GSC1 perivascular migration was independent of their position relative to the tumour bulk or cell density , with single cells migrating from the main tumour as efficiently and at similar speed as GSC1 migrating at a distance from the tumour margin ( Figure 1e and Figure 1—figure supplement 1h ) . Similarly , rare sparse imNSC further away from the tumour margin failed to migrate along blood vessels ( Figure 1—figure supplement 1h ) . Importantly , this effect was not due to a general impaiment in imNSC migration , as these cells migrated to a similar extent as GSC in vitro ( Figure 1—figure supplement 1i ) . This suggests that signals from the vasculature compartmentalises non-tumourigenic cells to restrict their migration , whereas transformation overrides these signals to enable perivascular spread . To test this more directly and identify potential effectors , we developed an in vitro cell migration assay that mimics the response of infiltrating GBM cells to initial contact with endothelial cells . We seeded both imNSC/GSC models and primary brain microvascular endothelial cells ( bmvEC ) in separate wells of culture inserts and assessed migration of the two cell types towards each other following insert removal by time-lapse microscopy . Remarkably , this assay closely recapitulated in vivo migratory patterns , in that endothelial cells strongly repelled and compartmentalised imNSC forming a sharp boundary ( Figure 1 and Video 1 ) . Similar effects were observed with normal NSC , indicating that compartmentalisation is not caused by immortalisation , but rather reflects the response of normal cells to the vasculature ( not shown ) . In contrast , GSC1 were refractory to compartmentalisation and migrated over the endothelial monolayer , giving rise to an uneven and longer boundary ( Figure 1g and Video 2 ) . The behavior of GSC2 was identical to GSC1 , confirming that it is a general property of transformed cells ( Figure 1h ) . Eph/ephrin signalling is one of most important mediators of cell-cell contact-dependent boundary formation and we reported that endothelial ephrin-B2 , the most abundant ephrin-B2 in the endothelial cells in vivo and in vitro ( Figure 1—figure supplement 1j ) , induces cell sorting of normal NSC ( Cayuso et al . , 2014; Ottone et al . , 2014; Gale et al . , 2001 ) . We therefore tested the role of ephrin-B2 in endothelial-induced compartmentalisation by assessing imNSC1 migration towards recombinant ephrin-B2-Fc . As shown in Figure 1i and Videos 3 and 4 , ephrin-B2-Fc compartmentalised imNSCs , but not GSC , to the same extent as endothelial cells . This effect was highly specific and was not due to general differences in migration between imNSC and GSC cells because imNSC migrated onto Fc peptides , ephrin-A1 or ephrin-A5-Fc ligands to a similar extent as GSC cells ( Figure 1—figure supplement 1i ) . Conversely , Efnb2 knock-down in endothelial cells disrupted boundary formation against imNSC , indicating that endothelial ephrin-B2 is both necessary and sufficient for compartmentalisation ( Figure 1j and Figure 1—figure supplement 1j ) . 10 . 7554/eLife . 14845 . 009Video 1 . Endothelial cells compartmentalise imNSC . Merged fluorescent and phase contrast movie from time-lapse microscopy experiments of cell-tracker labelled imNSC1 ( green ) migrating towards brain microvascular endothelial cells ( bmvEC , unlabelled cells ) . Images were taken every 10 min for 60 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 00910 . 7554/eLife . 14845 . 010Video 2 . GSC escape endothelial compartmentalisation . Merged fluorescent and phase contrast movie from time-lapse microscopy experiments of cell-tracker labelled GSC1 ( green ) migrating towards brain microvascular endothelial cells ( bmvEC , unlabelled cells ) . A slight decrease in fluorescence was observed due to higher proliferation rate of the cells . Images were taken every 10 min for 60 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01010 . 7554/eLife . 14845 . 011Video 3 . imNSC are strongly compartmentalised by ephrinB2 ligand . Merged fluorescent and phase contrast movie from time-lapse microscopy of imNSC1 migrating towards coated ephrinB2-Fc labelled with fluorescent antibody ( green ) . Images were taken every 10 min for 60 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01110 . 7554/eLife . 14845 . 012Video 4 . GSC become insensitive to ephrinB2 ligand . Merged fluorescent and phase contrast movie from time-lapse microscopy of GSC1 migrating towards coated ephrinB2-Fc labelled with fluorescent antibody ( green ) . Images were taken every 10 min for 60 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 012 To assess the role of vascular ephrin-B2 in vivo , we performed intravital imaging of imNSC implanted in inducible endothelial-specific conditional Efnb2 knock-out mice ( Efnb2 i∆EC ) , following postnatal tamoxifen-mediated recombination ( Ottone et al . , 2014 ) . Strikingly , selective deletion of ephrin-B2 in the endothelium elicited robust perivascular migration of imNSC1 , confirming that vascular ephrin-B2 compartmentalises non-tumourigenic cells in vivo ( Figure 1e , f ) . As we reasoned that changes in Eph/ephrin levels might underlie the ability of GSC to escape ephrin-B2-mediated vascular compartmentalisation , we interrogated expression levels of all Ephs and ephrins in imNSC1/2 and GSC1/2 by qRT-PCR ( Figure 2a and Figure 2—figure supplement 1a ) . Normal NSC controls were included to rule out p53 and Rb-dependent changes unrelated to perivascular migration . We found that p53 and Rb inactivation did not change Eph/ephrin expression substantially . In contrast , elevated Ras signalling , triggered by either RasV12 expression or Nf1 loss , strongly downregulated three Ephs ( Epha4 , Ephb1 and Ephb2 ) and upregulated two ephrins ( Efna5 and Efnb2 ) in both models . However , western analysis indicated that of these genes , only Ephb1 , Ephb2 and Efnb2 changed significantly at the protein level ( Figure 2b ) and were thus further assessed in migration assays against ephrin-B2-Fc . Surprisingly , we found that re-introduction of Ephb1 or Ephb2 by overexpression in GSC1 ( Figure 2—figure supplement 1b ) did not affect migration towards recombinant ephrin-B2-Fc , indicating that changes in the complement of Eph receptors do not underlie unimpeded migration , as in other systems ( Figure 2c ) ( Astin et al . , 2010 ) . Instead , genetic deletion of Efnb2 in GSC1 ( Figure 2—figure supplement 1c ) fully rescued boundary formation in response to ephrin-B2-Fc and endothelial cells in vitro and blocked perivascular migration in vivo ( Figure 2c–e ) . Conversely , Efnb2 overexpression ( Figure 2—figure supplement 1d ) conferred imNSC1 with the ability to migrate over recombinant ephrin-B2 and a cultured endothelium and to escape vascular compartmentalisation in vivo ( Figure 2c–e ) . Together , these results show that ephrin-B2 upregulation downstream of Ras underlies GSC evasion of ephrin-B2-mediated endothelial repulsion and invasion along the vasculature . 10 . 7554/eLife . 14845 . 013Figure 2 . Upregulation of ephrin-B2 in GSCs enables perivascular invasion ( a ) Quantitative RT-PCR analysis of indicated Eph receptors and ephrin ligands . Error bars denote s . e . m . , p values of differences in expression between imNSC1 and GSC1 are shown . Multiple t-test analysis . ( b ) Western analysis of levels of the indicated proteins in normal neural stem cells ( NSC ) , imNSC1 and GSC1 . n = 3 ( c ) Kymographs from time-lapse experiments of the indicated cell types migrating against coated ephrinB2-Fc over 60 hr . Quantification of the kymographs is shown on the right . Error bars denote s . e . m . , one way ANOVA with Tukey post hoc test . ( d ) Overlaid fluorescent and phase contrast images ( left ) and quantifications ( right ) of boundaries formed at 60 hr of GFP-labeled imNSC1/GSC1 ( green ) migrating towards bmvEC ( unlabeled ) as indicated . Error bars denote s . e . m . , one way ANOVA with Tukey post hoc test . Scale bar = 500 μm . ( e ) Representative 2-photon microscopy micrographs of GFP-labeled imNSC1-Efnb2 and GSC1Efnb2-/- imaged as in 1e . Quantification of migrated distances is shown on the right . Dots represent single cells measured across multiple animals . One way ANOVA with Tukey correction . p values are given compared to GSC1 for GSC1Efnb2-/- and imNSC1 for imNSC1-Efnb2 . Scale bar = 50 μm . See also Figure 2—figure supplement 1 and Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01310 . 7554/eLife . 14845 . 014Figure 2—source data 1 . Raw data for qRT-PCR analysis and quantifications of kymographs , boundary assays and migrated distance in vivo of NSC/GSC cells shown in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01410 . 7554/eLife . 14845 . 015Figure 2—figure supplement 1 . GSC2 show similar changes in Eph/ephrin levels as GSC1 . ( a ) Quantitative RT-PCR analysis of indicated Eph receptors and ephrin ligands in normal NSC , imNSC2 and GSC2 cells . Error bars denotes s . e . m . Multiple t-test . ( b–d ) Western analysis of EphB2 , EphB1 and Ephrin-B2 in GSC1 cells and imNSC of indicated genotype . β-tubulin served as loading control . See also Figure 2—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01510 . 7554/eLife . 14845 . 016Figure 2—figure supplement 1—source data 1 . Raw data for all quantitative analyses shown in Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 016 To understand the mechanisms involved in ephrin-B2-driven perivascular invasion , we characterised the early events of imNSC1 and GSC1 migration towards ephrin-B2-Fc in greater detail . We noticed that , although all GSC cells eventually migrated over ephrin-B2-Fc , a proportion of the cells ( ~20% ) are repelled on first contact ( Figure 3a ) . Repelled cells invariably migrated towards ephrin-B2-Fc either as single cells or in small groups ( <2 contacts ) , whereas non-repelled cells migrated in larger groups ( >3 contacts ) ( Figure 3b ) . This was not the case in imNSC cultures , where all cells were equally repelled , regardless of the number of homotypic cell-cell interactions at initial contact . This suggested that homotypic cell-cell interactions among GSC , mediated by intrinsic ephrin-B2 , might underpin evasion of extrinsic ephrin-B2 repulsion . To test this , we repeated the migration assay under conditions that disrupt GSC cell-cell contacts by inhibiting cadherin-based junctions ( Ottone et al . , 2014 ) . Both culture in low Ca2+ media and dominant-negative N-cadherin overexpression completely disrupted cell-cell junctions without affecting cell motility ( Figure 3—figure supplement 1a , b ) and blocked GSC migration over ephrin-B2 ligands , indicating that homotypic cell-cell interactions are specifically required ( Figure 3c ) . 10 . 7554/eLife . 14845 . 017Figure 3 . ephrin-B2 upregulation drives perivascular migration by saturating Eph forward signalling through homotypic cell-cell interactions . ( a ) Quantification of cell behavior upon initial contact with coated ephrin-B2 . n = 3 , error bars denote s . e . m . ( b ) Left: representative images taken from videos of GSC1 cells as they first come in contact with coated ephrin-B2 ( green ) . Right: quantification of the number of homotypic GSC1 cell-cell contacts at time of initial interaction with ephrin-B2 . A minimum of 50 cells were counted in each experiment . Error bar denotes St . D , Student t-test . ( c ) Representative kymographs ( left ) and quantifications ( right ) of GSC1 migrating towards ephrin-B2 that were either cultured in normal media ( ctrl ) or low Ca2+ conditions ( low Ca2+ ) , or transduced with GFP control ( GFP ) or dominant negative N-Cadherin ( DN-NCdh ) adenoviral constructs . Dotted lines demarcate coated ephrin-B2 boundary . n = 3 , error bars denote s . e . m . , Student t-test . ( d ) Kymographs and quantifications of the responses of imNSC1 and imNSC1-∆CEfnb2 to coated ephrin-B2 ligands . n = 3 , error bars denote s . e . m . , Student t-test . ( e ) Western analysis of the levels of activated Eph receptors ( p-Eph ) in the indicated cell types cultured in either normal growth media ( ctrl ) or calcium depleted ( low Ca2+ ) conditions . β-tubulin served as loading control . n = 3 ( f ) Western blots of p-Eph levels in indicated cells cultured either with control proteins ( Fc ) , endothelial cells ( Endo ) or ephrinB2-Fc ( ephrinB2-Fc ) for 18 hr . β-tubulin is used as loading control . n = 3 ( g ) Phase contrast images ( left ) and quantification ( right ) of imNSC1 migration towards coated ephrin-B2-Fc following pre-treatment with either control ( Fc ) or clustered ephrinB2-Fc to activate Eph forward signalling . Experiments were stopped at 48 hr to assure maximal stimulation of the cells throughout the assay . Green dots denote boundary of ephrin-B2 coating identified by fluorescence . n = 3 , error bars denote s . e . m . , Students t-test . Scale bar = 250 μm . See also Figure 3—figure supplement 1 and and Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01710 . 7554/eLife . 14845 . 018Figure 3—source data 1 . Raw data for all quantifications of NSC/GSC migration assays shown in Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01810 . 7554/eLife . 14845 . 019Figure 3—figure supplement 1 . Increased repulsion between ephrin-B2 expressing cells leads to greater migration velocity and distance . ( a ) Representative phase contrast images of GSC1 cultured either in normal media ( Ctl ) or low Ca2+ conditions ( low Ca2+ ) , or transduced with GFP control ( Ad-GFP ) or dominant negative N-Cadherin ( Ad-DN-NCdh ) adenoviral constructs . Note the loss of cell-cell contacts in both treatment conditions . ( b ) Quantification of migration velocity of Ad-GFP and Ad-DN-Ncad GSC1 cells , indicating no change in GSC1 cell motility following disruption of cell-cell junctions . ( c ) Quantification of the velocity of indicated cell types 60 min before ( blue ) and after ( grey ) homotypic collisions in sparse cultures . Error bars denote St . D . , One-way ANOVA with Bonferonni post-hoc test . ( d ) Visual representation of the migration path of individual cells tracked over 20 hr . Migration tracks are plotted from a common origin for clarity ( left ) . Right: quantification of total migrated distance . Error bars denote St . D . , One way ANOVA . ( e ) Representative immunofluorescence images of GFP labelled GSC1-derived tumours stained for pEph ( red ) and GFP ( green , left ) and the vascular marker CD31 ( grey , right ) . Arrows indicate examples of active Eph signaling within GFP-labelled tumour cells . See also Figure 3—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 01910 . 7554/eLife . 14845 . 020Figure 3—figure supplement 1—source data 1 . Raw data for all quantitative analyses shown in Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02010 . 7554/eLife . 14845 . 021Figure 4 . ephrin-B2 is a glioblastoma oncogene controlling anchorage independent proliferation . ( a ) Representative bioluminescent images of nude mice 20 days after intracranial injection of imNSC1 , imNSC1-Efnb2 , GSC1 , GSC1Efnb2-/- . ( b ) Kaplan-Meier survival plots of the mice depicted in ( a ) . Significance is given relative to imNSC1 for imNSC1-Efnb2 and to GSC1 for GSC1-Efnb2-/- . Log Rank Mantel Cox . ( c , d ) Left: Representative micrographs of GSC1 ( c ) and GSC2 ( d ) cells of indicated genotype grown in soft agar for 10d . Right: quantification of number of colonies formed in soft agar in all cultures , expressed as percentage over total number of seeded cells . n = 3 , error bars depict s . e . m . One way ANOVA with Tukey correction . ( e ) Left: representative FACS profiles of cells grown in attachment or methylcellulose for 72 hr , showing DNA content by propidium iodide ( PI ) staining . Right: quantification of cell cycle phases from the FACS profiles . n = 3–5 as indicated by the dots . Error bars depict s . e . m . , p values indicate significance of changes in G2/M phase . ( f ) Representative PI FACS profiles and quantifications of imNSC and GSC1 isolated from brain tissue 7 days after intracranial injection . n = 3 . Error bars depict s . e . m . One way ANOVA with Tukey post hoc test shown for G2/M phase . See also Figure 4— figure supplement 1 and Figure 4— source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02110 . 7554/eLife . 14845 . 022Figure 4—source data 1 . Raw data for Kaplan Meier analysis , number of colonies formed in soft agar and cell-cycle analysis presented in Figure 4DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02210 . 7554/eLife . 14845 . 023Figure 4—figure supplement 1 . Anchorage-independent proliferation is independent of homotypic cell-cell contacts . ( a ) Left: representative phase contrast images of imNSC1-Efnb2 cells transduced with GFP- ( Ad-GFP ) or dominant negative N-Cadherin- adenoviruses ( Ad-DN-Ncad ) and cultured in soft agar for 10 days . Right: quantification of the percentage of cells forming colonies in soft agar . Error bars denote s . e . m . n = 3 . ( b ) Western analysis of ephrin-B2 levels in GSC2 transfected with scrambled ( scr ) or Efnb2-specific ( siEfnb2 ) siRNAs , confirming efficiency of the knock-down . n = 2 ( c ) Quantification of activated caspase3 positive cells in the indicated methylcellulose cultures . Values are expressed as percentage of total number of cells . See also Figure 4—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02310 . 7554/eLife . 14845 . 024Figure 4—figure supplement 1—source data 1 . Raw data for all quantitative analyses shown in Figure 4—figure supplement 1DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 024 We next asked whether homotypic ephrin-B2 signalling drives migration through activation of forward or reverse signalling within the GSC population . Ectopic expression of a C-terminal truncated form of ephrin-B2 ( ΔCEfnb2 ) , which lacks reverse signalling ( Pfaff et al . , 2008 ) , promoted imNSC migration over ephrin-B2-Fc to the same extent as full-length Efnb2 , demonstrating that forward signalling between neighboring cells is responsible ( Figure 3d and Figure 2—figure supplement 1d ) . Indeed , western analysis ( Figure 3e ) indicated that while imNSC controls displayed low basal levels of activated Eph receptors ( p-Eph ) , GSC exhinited constitutively high levels of p-Eph , despite their reduced levels of EphB1 and EphB2 ( Figure 2b ) and no increase in other ephrin-B2 receptors in these cells ( Figure 2a and not shown ) . High basal p-Eph could be mimicked by ephrin-B2 overexpression in imNSC and reduced to basal control levels by culture in low Ca2+ ( Figure 3e ) . In addition , the high phosphorylation levels of Ephs in GSC appeared to be saturating , as , in contrast to imNSC , co-culture with endothelial cells or treatment with recombinant ephrin-B2 failed to further activate Eph receptors in these cells ( Figure 3f ) . Importantly , Eph desensitization was again entirely dependent on endogenous ephrin-B2 because genetic deletion of Efnb2 in GSC fully restored Eph stimulation by exogenous ephrin-B2 . Consistent with the known ability of Ephs to trigger repulsion ( Astin et al . , 2010; Pasquale , 2010 ) , such high basal levels of activated Eph resulted in GSC undergoing much greater scattering and faster migration than imNSC following homotypic collisions in sparse monocultures , in an ephrin-B2-dependent manner ( Figure 3—figure supplement 1c , d ) . Thus , high ephrin-B2 levels in GSC result in constitutive activation of Eph forward signalling and desensitization to exogenous stimulation . To further test whether this desensitization mechanism underlies the ability of GSC to override ephrin-B2-mediated repulsion , we pre-stimulated imNSC with recombinant ephrin-B2-Fc to stronly activate Eph forward signalling prior to their migration out of the insert ( Figure 3f ) and assessed migration towards coated ephrin-B2 in boundary assays . Parallel Fc-treated cultures served as controls . Remarkably , we found that ephrin-B2-Fc ( but not Fc ) pretreatment enabled migration over the ephrin-B2 boundary , confirming that activation of Eph forward signaling on first contact with exogenous ephrin-B2 is the mechanism involved . In agreement with these findings , immunofluorescence analysis of GSC1-derived orthotopic tumours revealed high pEph levels in cells migrating along blood vessels at the tumour edge , indicating that constitutive forward signaling also contributes to perivascular invasion in vivo ( Figure 3—figure supplement 1e ) . We conclude that GSC escape from endothelial compartmentalisation depends on continuous activation of Eph forward signalling elicited by elevated ephrin-B2 through homotypic cell-cell interactions within the tumour cell population . This , in turn , desensitises the receptors to further activation by extrinsic ephrins , thereby overriding the repulsion by endothelial ephrin-B2 and enabling unimpeded perivascular migration . Given these important roles in invasion , we asked whether ephrin-B2 might also affect tumourigenesis by performing tumourigenicity studies of luciferase-tagged imNSC1 , imNSC1-Efnb2 , GSC1 and GSC1Efnb2-/- implanted orthotopically in nude mice . We used immunocompromised mice for these experiments because the incomplete tumour penetrance of GSC in syngeneic animals precludes rigorous assessment of their tumourigenicity in the syngeneic model . Strikingly , quantitative imaging and survival analysis revealed that , while imNSC1 did not form tumours and GSC1 formed tumours at full penetrance , Efnb2 deletion strongly suppressed GSC1 tumour growth and , conversely , Efnb2 overexpression was sufficient to fully transform imNSC ( Figure 4a , b ) . Specifically , imNSC1-Efnb2 tumours resulted in a similar median survival as GSC1 ( 28d ) . In contrast , all imNSC animals survived tumour-free beyond 200 days and 7 out of 9 animals of the GSC1Efnb2-/-cohort developed lesions with much slower kinetics , while the remaining 2 mice remained tumour-free , as confirmed by post-mortem examination . These effects were not due to a general loss of stem-like properties due to ephrin-B2 deletion , as clonal efficiency remained unaffected in GSCEfnb2-/- ( Figure 1—figure supplement 1b ) . Thus , ephrin-B2 can substitute oncogenic Ras for transformation . To dissect the mechanisms involved , we performed soft-agar assays , which assess proliferation in the absence of anchorage signals , a property closely linked to in vivo tumourigenicity ( Freedman and Shin , 1974 ) . Consistent with our results in vivo , the majority of imNSC1 and GSC1 Efnb2-/- remained as single cells in soft agar and only less than 5% of the cells generated small colonies , as previously reported for NSC , again indicating that immortalised cells behave like NSC ( Figure 4c ) ( Gursel et al . , 2011 ) . Instead , imNSC1-Efnb2 and GSC1 formed large colonies at similar high efficiency , indicating that ephrin-B2 drives anchorage-independent proliferation . Intriguingly , and in contrast to its role in perivascular invasion , ephrin-B2 effects on proliferation were dependent on reverse signalling and independent of homotypic cell-cell contact . Indeed , imNSC1-ΔCEfnb2 did not form colonies in suspension and imNSC1-Efnb2 overexpressing DN-Ncadherin formed colonies at similar efficiency as GFP-controls , demonstrating that Eph forward signalling is dispensable ( Figure 4c and Figure 4—figure supplement 1a ) . To rule out the possibility that the transforming ability of ephrin-B2 might be a peculiar feature of the lgT-Ras model , we repeated the soft agar assays with imNSC2 , imNSC2-Efnb2 , imNSC2-ΔCEfnb2 , GSC2 and GSC2 knock-down for Efnb2 and obtained identical results , confirming the generality of these findings and the functional equivalence of the two GSC lines ( Figure 4d and Figure 4—figure supplement 1b ) . We next asked how ephrin-B2 overrides anchorage checkpoints by comparing cell cycle progression of immortalised and transformed cells in suspension . We seeded imNSC1 , imNSC1-Efnb2 , GSC1 and GSC1 Efnb2-/- in adhesion or methylcellulose culture for 72 hr and measured their PI-FACS profiles ( Cremona and Lloyd , 2009 ) . As anticipated , all cells proliferated efficiently in attachment ( Figure 4e ) . By contrast , their behavior in suspension was very distinct . While GSC1 continued to proliferate in methylcellulose with similar kinetics , suspended imNSC did not . This was not due to cell death as judged by activated caspase3+ staining ( Figure 4—figure supplement 1c ) . Instead , imNSC arrested with a 4n DNA content . This indicated that while imNSC progress normally through the G1 and S phases of the cell-cycle , their progression through G2/M is blocked by an anchorage checkpoint , which is overridden by activated Ras . Importantly , Ras-mediated progression through G2/M was again dependent on ephrin-B2 reverse signalling , because Efnb2 deletion in GSC re-instated the G2/M arrest ( without additional apoptosis , Figure 4—figure supplement 1c ) , and the cell-cycle arrest of imNSC could be rescued by overexpression of full-length , but not ΔC , Efnb2 ( Figure 4e ) . To assess the relevance of these findings to in vivo tumourigenesis , 7d after intracranial implantation imNSC1 , imNSC1-Efnb2 , GSC1 and GSC1Efnb2-/- , were recovered from the injected brains and their PI profiles analysed ( Figure 4f ) . Strikingly , the cell-cycle profiles of all cells were indistinguishable from the corresponding methylcellulose cultures , confirming that ephrin-B2 drives gliomagenesis in vivo by promoting proliferation in the absence of normal anchorage signals . A previous study reported that human fibroblasts cultured in suspension arrest at cytokinesis and that oncogenic Ras can bypass this arrest ( Thullberg et al . , 2007 ) . We therefore explored the role of cytokinesis in our system by pulsing all attached and suspended cultures with EdU to distinguish G2/M arrested from cycling cells completing mitosis and staining for phalloidin to detect cortical actin . As shown in Figure 5a and b , we found that all methylcellulose cultures devoid of ephrin-B2 reverse signalling ( imNSC1 , imNSC1-ΔCEfnb2 and GSC1 Efnb2-/- ) contained a much larger proportion of EdU- binucleated cells with decondensed chromatin compared to adherent conditions , indicative of a cytokinesis block ( Thullberg et al . , 2007 ) . In contrast , cells with intact ephrin-B2 reverse signalling ( imNSC1-Efnb2 and GSC1 ) had similar low percentages of EdU+ binucleated cells in both suspension and attachment culture . Thus , ephrin-B2 reverse signalling drives anchorage-independent cytokinesis . 10 . 7554/eLife . 14845 . 025Figure 5 . ephrin-B2 mediates cytokinesis in the absence of anchorage through RhoA . ( a ) Representative fluorescent images of binucleated cells grown in suspension for 24 hr and stained with phalloidin ( green ) to label cortical actin and EdU ( red ) to distinguish cycling from arrested cells . Scale bar = 20 μm . ( b ) Quantification of binucleated cells in indicated attached and suspended cultures . Values represent percentages over total number of cells . n = 4 , error bars denote s . e . m . Two way ANOVA with Tukey post hoc test . ( c ) Western analysis of RhoA pulldown assays showing levels of activated RhoA ( RhoA-GTP ) and total RhoA levels in indicated cells grown in suspension for 24 hr . Bottom graph shows quantifications of activated RhoA levels from the western blots . n = 3 . Error bars indicate s . e . m . ( d ) Quantification of the PI cell cycle profile as measured by FACS . Cells transfected with either constitutively active RhoA ( V14 ) or dominant negative RhoA ( N19 ) were grown in suspension for 24 hr . n = 3 , error bars depict s . e . m . One way ANOVA with Tukey post hoc test was used to calculate p values for differences in G2/M phase of each suspended culture relative to corresponding attached cultures . See also Figure 5—figure supplement 1 and Figure 5—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02510 . 7554/eLife . 14845 . 026Figure 5—source data 1 . Raw data for quantifications of binucleated cells and cell cycle analysis presented in Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02610 . 7554/eLife . 14845 . 027Figure 5—figure supplement 1 . Src and FAK do not mediate anchorage-independent proliferation downstream of ephrin-B2 . ( a ) Western analysis of levels of indicated proteins in imNSC1 , imNSC1-Efnb2 and GSC1 cultured in attachment ( a ) and suspension ( s ) . n = 2 ( b ) Western analysis of RhoA pulldown assays showing levels of activated RhoA ( RhoA-GTP ) and total RhoA levels following transfection of imNSC1 with constitutively active RhoA ( V14 ) and GSC1 with dominant negative RhoA ( N19 ) . Note that as expected V14 activates RhoA and N19 inhibits RhoA , confirming efficacy of the mutant constructs . Bottom graph shows quantifications of activated RhoA levels from the western blots . n = 2 . ( c ) Representative DNA cell-cycle profiles of indicated cells transfected with empty vector control ( ctrl ) , dominant negative RhoA ( N19 ) or constitutively active RhoA construct ( V14 ) and cultured in methylcellulose for 24 hr . n = 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 027 Of the known effectors of ephrin-B2 , Src and RhoA have been previously linked to anchorage independent proliferation , with RhoA also regulating constriction of the contractile ring during cytokinesis ( Daar , 2012; Desgrosellier et al . , 2009; Jordan and Canman , 2012 ) . We therefore assessed Src and RhoA in ephrin-B2-mediated anchorage-independent cytokinesis by measuring levels of their activated forms ( phosphorylated Src , p-Src , and GTP-bound RhoA ) in suspension by western blot and pull-down assays , respectively . p-Src was undetectable in suspended imNSC1 , imNSC-Efnb2 and GSC1 and , similarly , levels of activated FAK , a critical Src substrate , were also reduced ( Figure 5—figure supplement 1a ) , indicating that Src does not play a role in our system . In stark contrast , levels of active RhoA were low in suspended cells that lacked ephrin-B2 reverse signalling ( i . e . imNSC , imNSC1-ΔCEfnb2 and GSC1Efnb-/- ) but greatly increased in cells that express high levels of full-length ephrin-B2 ( i . e . imNSC1-Efnb2 and GSC1 ) , suggesting the involvement of RhoA ( Figure 5c ) . To further test this , we overexpressed constitutively active ( RhoA-V14 , which increases levels of active GTP-bound RhoA , Figure 5—figure supplement 1b ) and dominant negative ( RhoA-N19 , which inhibits activation of endogenous RhoA , Figure 5—figure supplement 1b ) ( Qiu et al . , 1995 ) forms of RhoA in imNSC1/GSC1Efnb2-/- and imNSC1-Efnb2/GSC1 , respectively , and assessed the ability of the cells to proliferate anchorage-independently by FACS . Remarkably , RhoA-V14 could rescue the cell-cycle arrest of imNSC1 and GSC1Efnb2-/- , whereas RhoA-N19 arrested imNSC1-Efnb2 and GSC1 in G2/M ( Figure 5d and Figure 5—figure supplement 1c ) . Together , these results demonstrate that ephrin-B2 drives anchorage-independent cytokinesis of GSCs through RhoA-mediated reverse signalling . We have shown that ephrin-B2 drives two key aspects of tumour formation in our murine models: GSC perivascular invasion and proliferation . We thus sought to determine the relevance of these findings to human GBM ( hGBM ) . To this end , we first examined Ephrin-B2 levels at the infiltrative margin of 10 GBM patient specimens and correlated them with GSC marker expression ( Figure 6a , Figure 6—figure supplement 1a and Table 3 ) . Two of these tumours were the original lesions from which G19 and G26 human GSC lines used below have been isolated . Serial sections were stained with H&E and with antibodies against Ephrin-B2 and ALDH1 , a stem cell marker which labels GSC in perivascular and hypoxic niches ( Rasper et al . , 2010 ) . All tumours presented cytoplasmic and membranous Ephrin-B2 expression in 25 to 90% of the tumour cells , as well as neurons , inflammatory cells and vascular endothelial cells , as reported ( Gale et al . , 2001; Sawamiphak et al . , 2010; Ottone et al . , 2014 ) . In contrast , ALDH1 expression was less abundant , with expression detected predominantly in neoplastic cells surrounding blood vessels , in perinecrotic regions and at the infiltrative tumour margin , as expected from a stem cell marker ( Rasper et al . , 2010 ) . Weaker ALDH1 expression was also detected in reactive astrocytes and endothelial cells . Importantly , many perivascular ALDH1+ neoplastic cells at the infiltrative tumour margin co-expressed ephrin-B2 . Thus , ephrin-B2 is expressed in stem-like cells invading along blood vessels in primary human tumours . 10 . 7554/eLife . 14845 . 028Figure 6 . Ephrin-B2 is expressed in human GSC and drives their invasion and proliferation in vitro . ( a ) Haematoxylin-eosin ( H&E ) , ALDH1 and Ephrin-B2 HRP staining of infiltrative margins of patient tumours . Three different glioblastomas of 10 analysed are shown . Atypical tumour cells expressing both ALDH1 and EphrinB2 ( black arrows ) can be identified around both normal and fibrotic co-opted vessels ( white arrows ) in all cases . Images ii and iii for Case 2 are close ups of the artery in i , showing that ALDH1+/EphrinB2+ cells present irregular and hyperchromatic nuclei characteristic of tumour cells . Box denotes magnified area . Case 751 is the original tumour from which G26 cells have been isolated . ( b ) Correlation plot between EFNB2 and mesenchymal gene expression levels ( Z-score ) in 8 primary human GSC lines . Mesenchymal Z-score was calculated using signature marker genes with a variance greater than 0 . 1 ( high-variance genes ) across GSC lines . Verhaak classification of each line into disease subtype using the same cutoff is also shown . Of note , G2 and G26 display dual signatures as indicated . G26 carries a deletion in the NF1 gene . ( c ) Representative kymographs ( left ) and quantifications ( right ) of human primary GSC lines G26 and G19 , stably expressing scrambled shRNA ( ctrl , ) or shRNA directed against EFNB2 ( shEFNB2 ) . Note that while migration of human GSC is less pronounced than in mouse cells , EFNB2 depletion results in complete inhibition of migration . Error bars denote s . e . m . , n = 3 . Students t-test . Green dots denote boundary of ephrin-B2 coating identified by fluorescence . ( d ) Left: Representative micrographs of SCR or shEFNB2 transduced G26 , G166 and G19 cells cultured in soft agar for 14d . Right: quantification of number of colonies formed in soft agar in all cultures , expressed as percentage over total number of seeded cells . n = 3 , error bars depict s . e . m . One way ANOVA with Tukey correction . ( e ) Left: representative FACS profiles of G166 and G19 cells ctrl or shEFNB2 grown in soft agar for 72 hr , showing DNA content by propidium iodide ( PI ) staining . Right: quantification of cell cycle phases from the FACS profiles . n = 3 as indicated by the dots . Error bars depict s . e . m . , One way ANOVA with Tukey post hoc test shown for G2/M phase . See also Figure 6—figure supplement 1 and Figure 6—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02810 . 7554/eLife . 14845 . 029Figure 6—source data 1 . Raw data for quantifications of kymographs , number of colonies formed in soft agar and cell-cycle analysis of human GSC presented in Figure 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 02910 . 7554/eLife . 14845 . 030Figure 6—figure supplement 1 . Ephrinb2 expression is increased and correlates inversely with patient survival in mesenchymal human GBM . ( a ) Representative MRI scans ( Case 7 ) illustrating methodology for the identification of the infiltrative edge of the tumours . The extent of neoplastic infiltration of the cortical ribbon and white matter is assessed at structural MRI , on the T2-weighted FLAIR sequences ( left scan ) as compared to the T1-weighted post-contrast images ( right scan ) ; the axial T1-weighted post-contrast image shows focal enhancement following administration of the paramagnetic agent gadolinium; the area of contrast enhancement indicating disruption of the blood brain barrier appears as a brighter spot and it is indicated by the asterisk . The whole mount tissue section stained with H&E of the right frontal lobe ( bottom ) has the same orientation as the MRI scans and it allows extensive mapping of the infiltrative edge of the lesion . The asterisk indicates the same regions that enhanced after gadolinium administration . ( b ) Top panel: Sorted variance of Verhaak et al Mesenchymal subtype signature genes across 8 GNS lines . Colored lines indicate low variance genes removed at different cutoff values . Grey dotted line indicates the variance of EFNB2 expression . Bottom panel: Mean z-score expression of high variance mesenchymal marker genes at different cuttoff points . Removing low variance tumor derived mesenchymal marker genes that are non-informative in GNS lines , with a variance below 0 . 05 ( Red line , Top panel ) , improves the correlation of sample mean mesenchymal expression estimates across a range of increasing variance cutoff points . ( c ) Relative mRNA expression levels of EFNB2 in the four GBM subtypes of the TCGA dataset . Levels are calculated relative to the average expression levels of EFNB2 in all tumours . One-way ANOVA with Tukey post-hoc test . ( d ) Kaplan Meier analysis of mesenchymal , proneural , neural and classical glioblastoma subtypes stratified on EFNB2 levels defined as above ( high ) and below ( low ) the median EFNB2 levels for each subtype . Proneural tumours are corrected for IDH1 status . n = 128 for mesenchymal , 116 for proneural , 61 for neural and 126 for classical GBM . ( e ) Western analysis of Ephrin-B2 levels in G19 , G26 and G166 human GSC transduced with control ( SCR ) or Efnb2 specific ( shRNA ) lentiviral shRNA constructs . n = 3 . ( f ) FACS profiles of EphrinB2 expression in G26 SCR and G26 shEfnb2 cells transduced with empty vector control plasmid ( Ctl ) or EFNB2 overexpression constructs ( EFNB2 ) . Quantification of soft agar assays on G26 shEFNB2-Ctrl and G26 shEFNB2-EFNB2 cells . Percentage of cells forming colonies over total number of cells is shown . Dashed line indicates mean % of colonies formed in parental G26 SCR cells . Error bars denote s . e . m . Student t-test . n = 4 . ( h ) Quantification of clonal efficiency of G19 and G26 transduced with either SCR and shEFNB2 cells and cultured at limiting dilution . Note that EFNB2 downregulation does not affect clonogenicity of the cells . Error bars denote s . e . m . Student t-test . n = 3 . See also Figure 6—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 03010 . 7554/eLife . 14845 . 031Figure 6—figure supplement 1—source data 1 . Raw data for all quantitative analyses shown in Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 03110 . 7554/eLife . 14845 . 032Table 3 . Patient information for tumours 1–8 used for Ephrin-B2 IHCDOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 032namegenderagesiteIDH1/2MGMTATRXGradecase 1M69left temporalwt0retainedIVcase2M45left temporalwt0retainedIVcase 3M54right temporo-parietalwt<10%retainedIVcase 4M55right temporalwt0retainedIVcase 5F61right parietalwt25%retainedIVcase 6M57right frontalG395A0lostIV ( secondary ) case 7F73left frontal / crossed midlinewt0retainedIvcase 8F39left temporo-parietalwt0retainedIV We next assessed EFNB2 expression in a panel of 8 well-characterised primary human GSC lines isolated from independent patient tumours , including two of the tumours analysed by IHC above ( Caren et al . , 2015 ) . These lines recapitulate the transcriptional sybtypes of primary GBM and predominantly cluster into proneural and mesenchymal subtypes , as previously reported for GSC ( Figure 6 b and Figure 6—figure supplement 1b ) ( Bhat et al . , 2013; Mao et al . , 2013 ) . Consistent with our mouse models , RNA-sequencing analysis revealed frequent upregulation of EFNB2 in GSCs and a significant correlation between EFNB2 and mesenchymal gene expression levels ( Figure 6b ) . In addition , analysis of 402 GBM from the TGCA dataset classified according to Verhaak et al . and corrected for IDH1 status , indicated that EFNB2 expression levels are highest in mesenchymal and classical GBM subtypes ( Figure 6—figure supplement 1c ) ( Verhaak et al . , 2010 ) . When tumours were divided into two groups on the basis of EFNB2 levels relative to median expression within subtypes , EFNB2 correlated with decreased survival in mesenchymal GBM ( Figure 6—figure supplement 1d ) . Together , these results are indicative of a functional role for ephrin-B2 in GSC tumorigenesis within human GBM , specifically of mesenchymal subtype . To test this more directly , we introduced control scrambled shRNAs or shRNAs to EFNB2 in 3 of the mesenchymal GSC lines described above ( Figure 6—figure supplement 1e ) and assessed effects of Ephrin-B2 depletion on invasion and proliferation in vitro . As shown in Figure 6c–h , we found that , similar to the murine GSC models , in the absence of ephrin-B2 , GSC ( but not SCR shRNA-transduced controls ) lost the ability to migrate over coated ephrin-B2 and failed to proliferate anchorage-independently , resulting in a G2/M cell-cycle arrest . These effects were not caused by an impairment of GSC self-renewal following ephrin-B2 downregulation and were highly specific , as overexpression of an shRNA-resistant EFNB2 construct in G26 shEFNB2 cells fully rescued their anchorage-independent proliferation ( Figure 6—figure supplement 1f–h ) . To test effects of ephrin-B2 on hGSC tumourigenicity in vivo , we used two complementary strategies . First , we xenotransplanted luciferase-tagged SCR or shEFNB2 G26 cells intracranially in nude mice and monitored tumour growth and disease-free survival by bioluminescence imaging and survival analysis ( Figure 7a–c ) . We found a dramatic impairment of tumour growth in the EFNB2 knock-down group relative to control , in that vector-transduced G26 formed GBM in all animals ( as previously reported ( Stricker et al . , 2012 ) whilst EFNB2 knock-down cells gave rise to slower growing tumours in only 2 out of 10 animals , with the other 8 mice remaining tumour-free for >1 year . To investigate the mechanisms responsible and confirm the generality of these findings , we repeated these experiments using both G26 and G19 lines and examined vascular association and proliferation of the GSC by immunofluorescence and FACS analysis at 10 days following implantation , a time point at which knock-down cells could still be detected by bioluminescence ( Figure 7b ) . SCR GSC associated with the vasculature at this early time point ( and at later stages of tumour growth ) , resulting in vascular co-option , as reported ( Figure 7d and Figure 7—figure supplement 1 ) ( Watkins et al . , 2014 ) . In contrast , vascular contact was severely compromised in knock-down cells . Furthermore , similar to GSC1Efnb2-/- murine cells , EFNB2 downregulation resulted in a marked decrease in proliferation , accompanied by an arrest in the G2/M phase of the cell-cycle , indicative of a cytokinesis block ( Figure 7e , f ) . Thus , Ephrin-B2 drives tumour initiation by mediating vascular association and proliferation of human GSC . 10 . 7554/eLife . 14845 . 033Figure 7 . EFNB2 silencing abolishes hGSC tumorigenicity . ( a ) Representative bioluminescent images of nude mice injected at d0 with 105 luciferase-labeled G26 cells , stably expressing scrambled shRNA ( SCR , top ) or shRNA directed against EFNB2 ( shEFNB2 , bottom ) . ( b ) Quantification of luciferase bioluminiscence measured at the indicated time points in both groups . n = 9 for ctrl and 10 for shEFNB2 , error bars denote s . e . m . Two-way ANOVA with Tukey correction . ( c ) Kaplan-Meier survival plots of the mice depicted in ( a ) . n = 5 for SCR , 10 for shEFNB2 . Log Rank Mantel Cox test . ( d ) Representative fluorescence images ( left ) and quantification ( right ) of GSC/vascular interactions in tumors derived from G19 SCR and G19 shEFNB2 cells at 10 days post-implantation . Sections were stained for GFP to identify tumor cells and CD31 ( red ) to label pre-existing blood vessels . Arrows indicate vascular association and co-option in SCR tumours , which is reduced in shEFNB2 tumours . ( e ) Quantification of percentages of Ki67+ cells over total GFP+ cells in G19 SCR and shEFNB2 tumors . n = 4 , error bars denote s . e . m . Students t-test . ( f ) PI FACS plots ( left ) and quantifications ( right ) of cell cycle profiles of G26 SCR and shEFNB2 cells retrieved from brain tissue 10 days after intracranial injection . n = 3 error bar denotes s . e . m . Significance is given for G2/M phase , one way ANOVA with Tukey post hoc test . See also Figure 7—figure supplement 1 and Figure 7—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 03310 . 7554/eLife . 14845 . 034Figure 7—source data 1 . Raw data for quantifications of tumour growth by bioluminescence analysis , survival by Kaplan-Meier analysis , tumour cell intractions with the vasculature , Ki67 labelling and cell-cycle analysis of human GSC-derived tumours presented in Figure 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 03410 . 7554/eLife . 14845 . 035Figure 7—figure supplement 1 . G26 cells associate with blood vessels . Representative immunofluorescence images of tumour xenografts derived from GFP-labelled G26 cells stained for the vascular marker CD31 ( red ) and GFP to identify tumour cells . Tumours were collected at the end of the survival studies . At the infiltrative edge of the tumour invading G26-GFP cells are frequently associated with blood vessels ( arrows ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 035 Second , we asked whether Ephrin-B2 inhibition could suppress tumourigenesis of pre-established tumours . For this , we took advantage of an Ephrin-B2 blocking scFv antibody fragment ( B11 ) we previously developed ( Abengozar et al . , 2012 ) . Migration and methylcellulose assays in vitro confirmed that B11 effectively inhibits Ephrin-B2 signalling in GSC1 ( Figure 8—figure supplement 1a , b ) . Luciferase-tagged G26 cells were implanted intracranially in immunocompromised mice and 13d later , once sizeable , exponentially growing tumours had formed , but prior to the onset of neoangiogenesis , B11 was administered intravenously for a total of 9 consecutive days , as reported ( Abengozar et al . , 2012; Binda et al . , 2012 ) . This treatment regimen enabled us to assess direct effects of B11 on the tumour cells in the absence of confounding anti-angiogenic effects known to result from ephrin-B2 inhibition ( Sawamiphak et al . , 2010; Abengozar et al . , 2012 ) . Efficient delivery of the scFv across the blood/tumour barrier to the tumour cells was confirmed in parallel animals using Alexa-680 labelled B11 ( Figure 8—figure supplement 1c ) . Remarkably , B11 strongly suppressed tumour growth in all animals without any evidence of toxicity , with three of six animals showing complete regression , as judged by quantitative bioluminescence imaging , survival analysis and post-mortem examination of the injected brains ( Figure 8a–c ) . Analysis and quantification of the tumours immediately following treatment also revealed that , in contrast to vehicle-treated control tumours , invading tumour cells failed to associate with and coopt the vasculature ( Figure 8d ) . Furthermore , proliferation was reduced in B11 tumours relative to controls , with B11 samples containing many Ki67+ multinucleated cells , indicative of a cytokinesis defect ( Figure 8e ) . Therefore , B11 suppresses G26 tumourigenicity through inhibition of ephrin-B2-dependent perivascular invasion and anchorage-independent proliferation , independent of angiogenesis . We conclude that Ephrin-B2 plays an important role in the pathogenesis of human GBM and its inhibition might be an effective strategy for curtailing GBM progression and recurrence . 10 . 7554/eLife . 14845 . 036Figure 8 . Treatment with anti-EphrinB2 ScFv blocking antibodies suppresses progression of established GBMs . ( a , b ) Representative images ( a ) and quantification ( b ) of bioluminescence radiance of PBS or anti-Ephrin-B2 scFv antibody ( B11 ) injected mice . n = 4 for PBS ctrl and 6 for B11 treatment groups . ( c ) Kaplan-Meier survival plots of the mice depicted in a and b . n = 4 for PBS-treated , 6 for B11-treated G26 tumours . Log Rank Mantel Cox test . ( d ) Representative immunofluorescence images ( left ) of PBS and B11-treated G26 tumours 5 days after the first B11 injection , stained for GFP ( green ) to identify tumour cells and CD31 ( red ) to label the endogenous vasculature . Quantification of vascular association is shown on the right . Arrows indicate co-opted blood vessels in PBS-treated tumours . Scale bar 50 μm . Error bars denote s . e . m . Student t-test . ( e ) Representative immunofluorescence images ( left ) PBS and B11-treated G26 tumours stained for GFP and the proliferation marker Ki67 . Quantification of the percentage of Ki67+ cells over total number of GFP+ tumour cells is shown on the right . Note the presence of multinucleated cells ( arrows and inset ) in B11-treated tumours . Scale bar=25 μm , n = 3 , error bars denote s . e . m . Students t-test . See also Figure 8—figure supplement 1 and Figure 8—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 03610 . 7554/eLife . 14845 . 037Figure 8—source data 1 . Raw data for quantifications of tumour growth by bioluminescence analysis , survival by Kaplan-Meier analysis , tumour cell intractions with the vasculature and Ki67 labelling of human GSC-derived tumours presented in Figure 8 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 03710 . 7554/eLife . 14845 . 038Figure 8—figure supplement 1 . B11 crosses the blood/tumour barrier and inhibits Ephrin-B2 on tumour cells . ( a ) Left: representative fluorescence and phase contrast merged images of GFP-labelled GSC1 ( Green ) migrating towards bmVECs ( unlabelled ) for 60 hr in the presence of 20 µg/ml Fc control ( Fc ) or anti ephrinB2-scFv ( B11 ) , respectively . Right: quantification of boundary length . n = 4 , error bars denote s . e . m . , Students t-test . ( b ) PI FACS quantifications of GSC1 cells treated with 20 µg Fc control ( fc ) or anti ephrinB2-scFv ( B11 ) for 18 hr . n = 3 , error bars denote s . e . m . , Students t-test . ( c ) Representative fluorescence images of G26 tumours treated with vehicle ( PBS ) or Alexa-680 labelled B11 ( red ) , showing efficient delivery of the ScFv antibody to the tumous cells . ( d ) Model of ephrin-B2 functions in GSC tumorigenesis . See also Figure 8—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 03810 . 7554/eLife . 14845 . 039Figure 8—figure supplement 1—source data 1 . Raw data for all quantitative analyses shown in Figure 8—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14845 . 039 Perivascular invasion is a critical mechanism of GBM growth and infiltration , which greatly contributes to the marked therapeutic resistance of these tumours ( Cuddapah et al . , 2014; Scherer , 1938 ) . Here , using a novel mouse ‘progression series’ that mimics transformation of normal NSC to mesenchymal GSC , we identified ephrin-B2 as a critical driver of GSC perivascular invasion . Interestingly , a role for ephrin-B2 and its phosphorylation in the tumour cells has been previously linked to GBM invasiveness , albeit not in the context of GSC . This suggests that ephrin-B2 might promote invasion both by generally enhancing cell-intrinsic invasive potential through reverse signalling ( Nakada et al . , 2010; Tu et al . , 2012 ) and , also , by specifically enabling perivascular spread through forward signalling as we have shown here . We found that ephrin-B2 expressed on vascular endothelial cells compartmentalises non-transformed cells . A similar role for Eph/ephrin signalling in constraining migration of premalignant cells was reported in colorectal cancer , where ephrin-B ligands in the surrounding normal tissue inhibit invasion of incipient lesions through activation of EphB receptors in the tumour cells ( Cortina et al . , 2007 ) . Our study is also consistent with work by Astin et al . , which reported that ephrin ligands on stromal cells repel normal and non-metastatic prostate cancer cells ( Astin et al . , 2010 ) . Thus , the present work further strengthens the notion that ephrins expressed in normal tissues act as tumour-suppressors during early tumourigenesis by promoting tumour confinement and identifies the vasculature as a critical mediator of these effects . It is tempting to speculate that ephrin-B2-mediated vascular compartmentalisation might be a more general tumour-suppressive mechanism , limiting the spread of premalignant lesions across many cancer types . In contrast , malignant transformation overrides vascular compartmentalisation to enable stereotypical GBM invasion along the perivascular space ( Baker et al . , 2014; Farin et al . , 2006; Watkins et al . , 2014; Winkler et al . , 2009 ) . Mechanistically , we found this to be dependent on the upregulation of ephrin-B2 in the GSC themselves , which elicits constitutive activation of Eph forward signalling in the tumour through homotypic cell-cell interactions . Consequently , this promotes perivascular spread in two ways . First , constitutive Eph activation increases repulsion between the tumour cells within the tumour bulk , thereby promoting GSC motility and scattering of single cells away from the tumour , a hallmark of GBM migration ( Vehlow and Cordes , 2013 ) . Second , it desensitizes the cells to extrinsic ephrin-B2 repulsion , thereby circumventing vascular confinement and permitting cells that have detached from the tumour mass to continue their migration along the vasculature ( Figure 8—figure supplement 1d ) . Thus , invasive GSC augment their tumourigenic potential by ‘hijacking’ the same signalling pathway through which the vasculature suppresses tumourigenesis in normal tissues . Intriguingly , the mechanism of invasion we identified here , differs from what has been reported for colorectal and prostate cancer in the studies mentioned above ( Astin et al . , 2010; Cortina et al . , 2007 ) . Indeed , while in those systems tumour spread depended on the downregulation of Eph receptors in a cell-autonomous manner , in our GBM model non-cell-autonomous constitutive activation of Eph forward signalling is responsible , suggesting that the mechanisms that evade ephrin repulsion during tumour progression might be critically dependent on cancer type . We further report that ephrin-B2 is also a critical effector of the ‘Ras-transformed phenotype’ , capable of both transforming immortalised cells and mediating , at least in part , Ras-dependent transformation . Indeed , lgT-immortalised stem cells overexpressing Efnb2 formed aggressive tumours with nearly identical kinetics to Ras-transformed cells and genetic deletion of Efnb2 in GSCs significantly delayed tumourigenesis . We demonstrate that these effects depend on the surprising and previously unknown ability of ephrin-B2 to drive anchorage-independent proliferation through the activation of cytokinesis in a RhoA-dependent manner . It is of note that , in contrast to perivascular invasion , these ephrin-B2 effects are mediated by reverse signalling and independent of cell-cell contact , indicating that ephrin-B2 drives two key aspects of GBM tumourigenesis through a complex interplay of cell-autonomous and non-cell-autonomous mechanisms ( Figure 8—figure supplement 1d ) . Previous studies in endothelial and smooth muscle cells have reported a similar cell-cell contact- and Eph-independent role for ephrin-B2 reverse signalling in driving acto-myosin contractility and cell spread , respectively ( Bochenek et al . , 2010; Foo et al . , 2006 ) . We currently do not know how RhoA becomes activated downstream of ephrin-B2 . A potential candidate is Dishevelled , which mediates activation of Rho/ROCK by ephrin-B1 ( Tanaka et al . , 2003 ) . It would be interesting to explore the role of Dishevelled in our system in the future . Our results in GBM specimens and primary human GSC lines strongly support a critical role for ephrin-B2 in the human disease through the same mechanisms that we identified in the murine models . Indeed , EFNB2 knock-down prior to implantation abrogated tumour initiation and treatment of pre-existing intracranial tumours with Ephrin-B2 blocking antibodies , under conditions that mimic human therapeutic paradigms , strongly reduced the growth and expansion of pre-formed tumours . Significantly , both approaches curtailed tumour growth by impairing both vascular association and cytokinesis . In patient samples stratified by GBM subtype , EFNB2 levels were highest in mesenchymal and classical tumours and correlated , inversely , with survival specifically in mesenchymal GBM . This is in agreement with our mesenchymal mouse models and analysis of human GSC lines , which revealed a robust correlation between EFNB2 and mesenchymal gene expression levels . This recurrent correlation is particularly noteworthy , as EFNB2 was previously identified as a component of the core mesenchymal gene network ( Carro et al . , 2010 ) . Furthermore , increasing evidence suggests that convertion to the mesenchymal signature is associated to progression , radioresistance and recurrence and is thus a fundamental driver of GBM malignancy ( Ozawa et al . , 2014; Bhat et al . , 2013 ) . GSCs are thought to be largely responsible for tumour infiltration and propagation ( Chen et al . , 2012a; Venere et al . , 2011 ) . As our work is based on murine and human GSC models of GBM , and Ephrin-B2 levels are high in the GSC compartment in primary tumours ( Figure 6a ) , our findings indicate that anti-ephrin-B2-based therapies would target the most critical subset of cells within these lesions . As such , by harnessing the complexity of the Eph/ephrin system , inhibition of ephrin-B2 would in itself be an effective ‘combinatorial therapy’ , capable of suppressing two critical GSC-intrinsic properties , namely perivascular invasion and proliferation , and might thus represent an attractive strategy for blocking GBM progression and recurrence . In addition , as ephrin-B2 levels are robustly increased in GSC compared to normal neural stem cells and other tissues , such therapies should be relatively tumour-specific and non-toxic . All animal work was carried out in accordance with the regulations of the Home Office and the ARRIVE guidelines . Efnb2i∆EC mice and recombination protocols were described previously ( Wang et al . , 2010; Ozawa and James , 2010; Ottone et al . , 2014 ) . Tamoxifen-injected Efnb2fl/fl littermates were used as controls . C57Bl6 and CD-1 nude mice for tumourigenicity analysis were obtained from Charles River . For GSC1 tumorigenic studies and B11 treatments of G26 cells , sample size was calculated by power analysis using PASS software based on pilot studies assessing the tumorigenicity of the parental line . For experiments with shEFNB2 human G26 cells sample size was set based on previous studies ( Pollard et al . , 2009 ) . Craniotomies were performed as previously described ( Holtmaat et al . , 2009 ) . Briefly 6–8 week old Efnb2 i∆EC or Efnb2fl/fl mice were anaesthetized with ketamine-xylayzine intraperitoneal injection ( 0 . 083 mg/g ketamine , 0 . 0078 mg/g xylazine ) . The animals were then injected with 0 . 02 ml 4 mg/ml intramuscular dexamethasone to limit an inflammatory response and subcutanaeous bupivacaine ( 1 mg/kg ) a local anaesthetic . Once the skull was exposed a few drops of lidocaine ( 1% solution ) were applied on its surface . Before covering the burr hole with a glass coverslip , 2x104 cells were injected into the cortex at a depth of 150–300 μm using a picospritzer . Mice were left to recover for 7 days and then imaged using a purpose built microscope equipped with a tunable Coherent Ti:Sapphire laser and PrairieView acquisition software . Mice were anaesthetized with isoflurane and secured to a fixed support under the microscope . The eyes were coated with Lacri-lube ( Allergan ) to prevent dehydration , an underlying heat pad used to maintain body temperature ( 37°C ) . To prevent dehydration isotonic saline solution was administered ( i . p . ) during long imaging sessions . Depth of anaesthesia was closely monitored . To visualize blood vessels a 3000MW dextran-Texas Red conjugate was injected intravenously prior to imaging . A pulsed laser beam with a wavelength of 910 nm was used to ensure that both GFP-tumour cells and Texas Red showed sufficient signal intensity . Each imaging session lasted for no longer than 60 min and mice were imaged up to four times daily with cells imaged every 20 min . After image acquisition individual frames were aligned and the displacement of single cells measured using ImageJ software . 5x104 luciferase expressing cells were injected into 6–8 week old CD-1 nude mice into the right putamen ( 1 mm rostral to bregma , 2 mm lateral and 2 . 5 mm depth ) as previously described ( Ozawa and James , 2010 ) . Tumour cells were loaded into the syringe just prior to injection and the needle kept in place for a further 5 min to ensure minimal reflux of the material along the needle tract . Tumour formation , growth and volume were indirectly calculated by sequential images taken with an IVIS Spectrum in vivo imaging system ( Perkin Elmers ) . Following administration of 120 mg/kg D-luciferin ( Intrace medical ) by intraperitoneal injection , mice were anaesthetized ( 3% isoflurane ) and imaged under continuous exposure to 2% isoflurane . Luminescent measures were performed once a week starting 5 days after cell implantation until day 40 . Bioluminescence was detected by the IVIS camera system , integrated , digitised , and displayed . Pseudocolor scale bars were consistent for all images of dorsal views in order to show relative changes at tumour site over time . Tumours were quantified by calculating total flux ( photons/s/cm2 ) using Living Image software ( Xenogen , Caliper Life Sciences ) . Animals were sacrificed when they showed signs of distress or weight loss . For treatment experiments , PBS or B11 administration was started once the tumours reached a minimum signal intensity of 1x106 photons/s/cm2 . Mice were then randomised into two groups prior to intravenous injection of 5 doses of either anti ephrinB2-scFv B11 ( total dose 20 mg/kg ) or PBS control over a period of 9 days . Survival curves were estimated using the Kaplan-Meier method . Significance was calculated using the log-rank Mantel-Cox test . To determine efficiency of delivery , B11 was labeled with Alexa Fluor 680 using the SAIVI rapid antibody kit ( Invitrogen ) according to the manufacturer’s instructions . Primary mouse microvascular endothelial cells were obtained from Caltag Medsystems and subcultured according to the suppliers recommendations ( cells were primary cells isolated directly from normal mouse brain , no further authentication performed by the authors , mycoplasma negative as tested by Mycoalert kit , Lonza ) Human GSC lines were described previously ( Pollard et al . , 2009; Caren et al . , 2015 ) . Cells were originally isolated from patient tumours and are maintained in serum free cultures on laminin ( Pollard et al . , 2009 , no further authentication performed by the authors , mycoplasma negative as tested by Mycoalert kit , Lonza ) . Co-culture experiments with endothelial cells for assessment of p-Eph levels were performed as previously described ( Ottone et al . , 2014 ) . Briefly , endothelial cells were seeded at confluence on PLL-coated dishes and imNSC or GSC cells seeded on top the following day . The cells were separated by differential trypsinisation . For analysis of clonal efficiency , single cells were sorted into individual wells of a low attachment 96 well plate using a FACS Aria III cell sorter and cultured in neural stem cell media for 7d after which percentage of neurosphere formed was calculated . Transient transfections and viral transductions were performed as previously reported ( Ottone et al . , 2014 ) . The following plasmids were used: FUGW was used to label all imNSC and GSC cells ( Lois et al . , 2002 ) . Constitutively active RhoA-V14 and dominant negative RhoA-N19 constructs were a kind gift of Anne Ridley ( Ridley and Hall , 1992 ) . Plasmid encoding full-length cDNAs of human EphB1 was purchased from OriGene Technologies , Inc ( RC214301 ) and mouse EphB2 was a kind gift of E . Batlle . Primary NSCs were isolated directly from the SVZ of postnatal or adult Cdh5 ( Pac ) -CreERT2 or NF1fl/fl transgenic animals and cultured as described previously ( Wang et al . , 2010; Zhu et al . , 2001; Ottone et al . , 2014 ) . To generate imNSC and GSC lines the following plasmids were used: pBabe-largeT + pLXSN-hRasV12 ( Cremona and Lloyd , 2009 ) , shp53 pLKO . 1 ( Godar et al . , 2008 ) , pCMV-Cdk4 ( van den Heuvel and Harlow , 1993 ) ( addgene 1874 ) . Recombination of the NF1fl/fl allele was achieved using Adeno-Cre viruses at an MOI of 100 , as reported ( Ottone et al . , 2014 ) . All experiments were performed on a minimum of two independent batches of GSC1 and 2 generated as described above from independent primary preparations ( no further authentication performed by the authors , mycoplasma negative as tested by Mycoalert kit , Lonza ) . Cells were used within the first 10–15 passages from infection . Western blots and immunocytochemistry were performed as reported previously ( Ottone et al . , 2014 ) . The RhoA activation assay kit ( Millipore ) was used according to manufacturers instructions . For immunohistochemical analysis of tumours , mice were perfused with 4% PFA and the brain post-fixed in 4% paraformaldehyde for 2 hr , placed in a 30% sucrose solution over night , embedded in OCT and snap frozen . Immunohistochemistry was performed on 30 μm cryostat sections . Tissue sections were stained overnight at 4°C with antibodies diluted in 10% goat serum . Hematoxylin and Eosin ( H&E ) staining was performed on 3 μm paraffin embedded sections . The following commercial primary antibodies were used: β-Tubulin ( Sigma T8328 1:5000 ) , BrdU ( Roche 11170376001 1:400 ) , cleaved caspase-3 ( Cell Signaling #9664 1:500 ) , EphA4 ( abcam ab641 1:1000 ) , EphB1 ( abcam ab129103 1:1000 ) , EphB2 ( abcam ab76885 1:1000 ) , EphB3 ( abcam ab133742 ) , EphB4 ( R&D AF446 ) , ephrinA5 ( R&D AF3743 1:500 ) , ephrinB2 ( R&D AF496 1:250 ) , CD31 ( BD 550274 1:400 ) , GAPDH ( abcam ab9483 1:1000 ) , GFAP ( abcam54554 1:400 ) , GFP ( Invitrogen A-21311 1:400 ) , Nestin ( Millipore MAB353 1:400 ) , pan-pEph ( abcam ab61791 1:500 ) , pFAK ( Cell signalling #3283 1:500 ) , p-Src ( Cell Signaling 2101s 1:1000 ) , O4 ( R&D MAB1326 1:400 ) , Sox2 ( Cell signalling 3728s 1:250 ) , SSEA1 ( BD Pharmingen 560079 1:400 ) , Tuj1 ( Covance MMS-435P 1:500 ) and NG2 ( Chemicon , 1:250 ) . For migration assays , endothelial cells and imNSC/GSCs were plated at a density of 2x104 cells into adjacent compartments of cell-culture silicon inserts ( Ibidi ) separated by a 500 μm gap . Alternatively one well of the insert was coated over night with 4 μg/ml of recombinant ephrin-B2-Fc fusion proteins or Fc controls ( R&D ) clustered at a molar ratio of 1:2 with fluorescently labelled anti-Fc antibody for 1 . 5 hr at room temperature . After removal of the insert the cells were cultured in medium supplemented with 1 % Matrigel ( Invitrogen ) and live cell imaging was performed in a heated and CO2 controlled chamber for 60 hr . For stimulation of imNSCs , recombinant ephrinB2-Fc ligands were preclustered with anti-human Fc antibodies as above and added to imNSCs for 24 hr before removal of the insert at a final concentration of 10 μg/ml . Analysis of ephrinb2-stimulated imNSC migration was terminated at 48 hr to achieve maximal stimulation of forward signaling throughout the experiments . Migration was quantified by tracing the boundary between GFP positive NSCs and non-stained bmvECs in ImageJ . To analyse the number of cell contacts individual frames from the videos of cells making initial contact with the ephrinB2-Fc boundary were analysed . The total number of protrusions per cell that were in contact with other cells at the time of entering the ephrin-B2-coated well were counted . All counting was performed blind . Kymograph analysis was performed using an ImageJ macro . Quantification of the kymographs was performed by measuring pixel intensities 200 μm before and after the ephrinB2-Fc boundary at the last imaged time-point to assess the proportion of cells that migrate over ephrin-B2 upon contact ( expressed as relative cell density ) . To analyse scattering behaviour , cells were seeded sparsely at 10 , 000 cells/12 well and their migration tracked for 20 hr . Collisions between single cells were quantified over 200 min and a minimum of 50 cells were counted per condition per biological repeat . For analysis of anchorage independent proliferation , cells were seeded for up to 72 hr in SVZ culture medium containing 1 . 8% dissolved Methylcellulose ( Sigma ) to form a semisolid hydrogel as previously described47 . To retrieve cells , the suspension was diluted fivefold with DMEM and centrifuged at 500 g for 10 min . The pellet was then washed twice with ice-cold PBS . For analysis of binucleated cells , all cells were incubated with EdU ( Life technology ) for 4 hr and placed on PLL-coated coverslips 15 min before fixation and staining . As the metaphase in mammalian cells typically lasts less than an hour this experimental set up enabled us to distinguish between cycling ( EdU positive ) and cytokinesis arrested ( EdU negative ) cells . To assess colony formation in soft agar 5x103 cells/6 well were seeded into 0 . 5% agar on top of a bottom layer of 1% agar ( Sigma ) . Total number of colonies formed was counted after 10 days . Cell pellets retrieved from methylcellulose were resuspended , fixed in 70% EtOH for 4 hr and stained with propidium iodide ( Sigma ) . A minimum of 10 , 000 cells were counted for each condition using a BD LSR II . The PI profile was then analysed using the cell-cycle module of FlowJo X . Cells from tumours were isolated as follows: The GFP positive tumour tissue was dissected using a fluorescent dissection microscope and digested using papain according to the manufactures recommendation ( Worthington Biochemical Corporation ) . Cells were stained with PI as above prior to FACS analysis . All FACS stainings were performed on at least 3 independent cultures/tumours . GFP-labeled GSC1 and GSC2 tumours were dissected and digested as above . RNA was isolated by using RNAeasy Plus Mini Kit ( Qiagen ) and RNA sequencing libraries were constructed using the NEBNext Ultra RNA Library Prep Kit for Illumina ( NEB ) . RNA-seq data from adult brain of normal mice was obtained from GEO [1] ( accession numbers GSM1055111 , GSM1055112 and GSM1055113 ) . Sequencing reads from tumour samples were aligned with the Tophat splice junction mapper ( Kim et al . , 2013 ) , version 2 . 0 . 11 against GRCm37/ mm9 and transcript annotations . All parameters were set to default except inner distance between mate pairs ( r = 100 ) and library type ( fr-firststrand ) . The normal brain data was also aligned using Tophat with default parameter values except for distance between mature pairs ( r = 200 ) . The DESeq2 Bioconductor package ( version 1 . 4 . 5 ) ( Love et al . , 2014 ) was used to perform differential gene expression analysis on read counts obtained with HTSeq-count ( version 0 . 5 . 3p9 ) ( Anders et al . , 2015 ) , and p-values were adjusted for multiple testing with the Benjamini-Hochberg procedure ( Love et al . , 2014 ) . Genes with adjusted p-value <= 0 . 05 were deemed to be differentially expressed . Sequencing reads for NS and GSC cell lines were aligned to mouse genome build GRCm38/ mm10 with STAR 2 . 5 . 2a ( Dobin et al . , 2013 ) using the two-pass method for novel splice detection ( Engstrom et al . , 2012 ) . Read alignment was guided by gene annotation from Ensembl release 84 ( Cunningham et al . , 2015 ) with optimal splice junction donor/acceptor overlap settings . Transcripts were quantified with HTSeq-count ( Anders et al . , 2015 ) based on feature coordinates from Ensembl 84 . Gene set enrichment analysis was carried out with the GAGE Bioconductor package ( version 2 . 18 . 0 ) ( Luo et al . , 2009 ) . Gene sets used corresponded to either biological process ( BP ) terms from the Gene Ontology , or derived from CNS cell type RNA-seq data described by the Barres group ( http://web . stanford . edu/group/barres_lab/brainseq2/brainseq2 . html ) using mean log2 FPKM values for astrocyte , neuron , OPC and oligodendrocyte classes . The top 500 genes for each class were identified by the ratio of individual class average expression to the maximum non-class average expression values . Log2 fold-change values computed by DESeq2 ( version 1 . 8 . 2 ) were used as the input for GSA analysis . Data visualised in Figure 1—figure supplement 1a is derived from log2 FPKM data for the left panel and variance stabilised read counts using the rlog function in DESeq2 . Subtype analysis of human GSC lines was completed using methods and microarray data described by Caren et al . ( Caren et al . , 2015 ) . Sample log2-transformed expression values for the signature centroid genes were produced by taking the mean expression across sample replicates . Centroid genes that could not be assigned to annotated genes were also omitted from further analysis . To ensure the accuracy of subtype expression estimates only subtype genes with high variance across the GNS dataset were carried forward . Subtype scores per sample were computed from mean Z-score transformed levels of overexpressed centroid genes for each subtype . Samples were then classified as belonging to the subtype associated with the highest mean Z-score or mixed subtype when presenting a similarly high expression of another subtype's mean Z-score . Gene set enrichment analysis ( GSEA ) ( Subramanian et al . , 2005 ) was applied to mouse tumor samples to test for Verhaak et al . subtype enrichment using the pre-ranked GSEA tool . DESeq2 computed T statistics from each tumor to control mouse sample differential expression results were used to rank the gene lists . Pre-defined genesets pertaining to Verhaak et al . ( Verhaak et al . , 2010 ) were obtained from the Molecular Signatures Database . TCGA data analysis was performed as described previously ( Binda et al . , 2012 ) survival and expression analysis was performed on the publicly available TCGA data and relative mRNA expression obtained from the TCGA data portal ( http://cancergenome . nih . gov/dataportal/data/about ) . The analysis was performed using samr R package . A two class unpaired model was used with the following parameters: delta 0 . 01 , number of permutation 100 . Quantitative RT-PCR analysis and siRNA transfections were performed as previously described ( Ottone et al . , 2014 ) . See Supplementary file 3 for primer details . For knockdown of EFNB2 in human G26 cells , lentiviral shRNA constructs were used ( Invitrogen ) . Control cells were infected with control vectors . Both cell types were selected with puromycin and knock-down efficiency validated by qPCR and immunoblotting prior to tumourigenic studies . EphrinB2 expression pattern was investigated in brain tissue samples of eight patients operated of supratentorial hemispheric glioblastoma at Imperial College Healthcare Trust between February and June 2015 . None of patients had radiotherapy or chemotherapy prior to surgery or any relevant comorbidities . Patients consent was given for all samples . The project received ethical approval from the Imperial College Healthcare Tissue Bank committee on behalf of MREC ( ICHTB HTA licence: 12275 , REC Wales approval: 12/WA/0196 ) . All 8 tumours were wild type for IDH1 and 2 mutations and ATRX status with the exception of case 6 , which was a secondary glioblastoma bearing the common G395A amino acid substitution in IDH1 and showing loss of ATRX expression . Two additional GBM samples were the original lesions from which G19 and G26 cells have been isolated ( Pollard et al . , 2009 ) . Preoperative structural MRI was available in all patients; the T2-weighted FLAIR sequences were reviewed to assess the extent of invasion into normal tissue . All patients underwent maximally safe surgical debulking and the isolated tissues were extensively sampled in order to represent the bulk of the tumour along with its infiltrative component in the surrounding grey matter . Specifically , areas representative of the invasive front of the tumour were chosen from each case . The tissues were fixed in 4% buffered formalin for 24 hr and then processed using standard method to produce paraffin sections . Sequential three-micron sections cut from each selected block and were used for immunoperoxidase immunohistochemistry . Immunostainings were performed following a standard protocol . Briefly , the sections were dewaxed in xylene and rehyadrated in decreasing alcohols to distilled water . Antigen retrieval was performed incubating the sections for 20 min in steam-heated sodium citrate buffer ( 10 mM Sodium Citrate , 0 . 05% Tween 20 , pH 6 ) at 90°C or 1 mM EDTA at pH 8 . Following antigen unmasking , endogenous peroxidase was quenched with 1% hydrogen peroxide in PBS at -20°C for 15 min . After rinsing in PBS , the sections were incubated overnight at room temperature with the following primary antibodies: anti-EphrinB2 ( clone EFR-163M hybridoma supernatant 1:4 , CNIO ) ( Abengozar et al . , 2012 ) , anti-ALDH1 ( monoclonal , BD Bioscience , dilution 1:500 ) , anti-ALDH1 ( mouse , clone 44 , BD biosciences ) . The ephrin-B2 antibody gave robust membrane and cytoplasmic staining and labelled the endothelium , confirming specificity . Nuclear labelling was however detected in some tumour cases , a feature not present in control samples . Only membranous and cytoplasmic staining was scored as positive in the analysis of the tumours . The SuperSentitive IHC detection system from BioGenex ( Fremon , CA , USA ) was used to visualise antibody binding following the manufacturer’s instructions and counterstained with Mayer’s Haemalum . Statistical analysis was performed using GraphPad Prism statistical analysis software . The precise tests are stated in the figure legends . Shapiro-Wilk test was used to confirm normal distribution of the data . All experiments for which quantifications were performed were carried out a minimum of three times as indicated in the figure legends .	Glioblastoma is the most common and deadly type of brain cancer . On average , patients with glioblastoma only survive 15 months even with aggressive treatment . One of the main reasons that therapy fails is the strong tendency of the tumor cells in this form of cancer to spread into the normal brain tissue . This makes it impossible for surgeons to completely remove the tumor , which means that the disease will almost always recur . Within the brain , invading glioblastoma tumor cells spread along pre-existing structures , like blood vessels . When the tumors use blood vessels as a highway to the rest of the brain it is called “perivascular invasion” . Scientists do not know exactly how glioblastoma cells move along the blood vessels . Learning more about this type of tumor cell movement could help scientists develop treatments to stop the tumor cells from spreading . Now , Krusche et al . show that the glioblastoma tumor cells highjack the system that normally helps keep brain cells in place . Experiments with mouse and human tumor cells grown in the laboratory and injected in mice to produce glioblastoma tumors showed that a family of proteins called ephrins determines whether perivascular invasion occurs . Ephrins are found on the surface of both tumor cells and blood vessels . Normally , the blood vessels use these proteins to block the spread of normal brain cells . However , tumor cells override this normal anti-tumor mechanism and become able to spread along the blood vessels . Specifically , Krusche et al . showed that an increase in the levels of ephrin-B2 in tumor cells allows them to move along the blood vessels . Ephrin-B2 was also found to drive the multiplication of the tumor cells independently of the protein’s interactions with the blood vessels . Using antibodies to block ephrin-B2 in tumors greatly reduced tumor size and extended survival in mice with glioblastoma tumors . The experiments suggest the blocking ephrin-B2 might be a therapy that both stops the glioblastoma cells from spreading and prevents the tumor cells from multiplying .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "cancer", "biology" ]	2016	EphrinB2 drives perivascular invasion and proliferation of glioblastoma stem-like cells
Practically all studies of gene expression in humans to date have been performed in a relatively small number of adult tissues . Gene regulation is highly dynamic and context-dependent . In order to better understand the connection between gene regulation and complex phenotypes , including disease , we need to be able to study gene expression in more cell types , tissues , and states that are relevant to human phenotypes . In particular , we need to characterize gene expression in early development cell types , as mutations that affect developmental processes may be of particular relevance to complex traits . To address this challenge , we propose to use embryoid bodies ( EBs ) , which are organoids that contain a multitude of cell types in dynamic states . EBs provide a system in which one can study dynamic regulatory processes at an unprecedentedly high resolution . To explore the utility of EBs , we systematically explored cellular and gene expression heterogeneity in EBs from multiple individuals . We characterized the various cell types that arise from EBs , the extent to which they recapitulate gene expression in vivo , and the relative contribution of technical and biological factors to variability in gene expression , cell composition , and differentiation efficiency . Our results highlight the utility of EBs as a new model system for mapping dynamic inter-individual regulatory differences in a large variety of cell types . Genome-wide association studies ( GWAS ) have identified thousands of genetic variants associated with human traits and diseases , many of which are located in noncoding regions of the genome and are putatively regulatory in function ( Albert and Kruglyak , 2015 ) . To understand regulatory and functional effects of trait-associated variants , it is necessary to perform molecular assays in the relevant cell types at the relevant stages of development , and potentially to also model different environmental exposures ( Umans et al . , 2020 ) . However , most efforts to identify genetic variants that regulate gene expression ( expression quantitative trait loci , or eQTLs ) have relied on adult tissue samples collected at a single time point . While such efforts have mapped millions of static , steady-state eQTLs across dozens of tissues and cell types ( GTEx Consortium , 2020 ) , most disease-associated variants were not found to also be classified as eQTLs ( Yao et al . , 2020; Aguet et al . , 2017 ) . It is possible that dynamic and variable regulatory genetic effects , including those that are specific to a given cell type , time point , or environment , may underlie the mechanisms for many unexplained phenotypic associations . For example , recent efforts to characterize gene regulatory dynamics in human induced pluripotent stem cells ( iPSCs ) and their derived cell types have identified dynamic eQTLs that are associated with disease risk , supporting the intuitive notion that changes in gene regulation during development may play a role in shaping human adult phenotypes , including disease ( Strober et al . , 2019; Cuomo et al . , 2020 ) . Still , iPSCs are limited in their potential for identifying dynamic regulatory effects . The number of cell types that can be obtained from iPSCs using directed differentiation protocols is quite modest , and time-course experiments , although useful for studying gene regulation at discrete points along a continuum , are inefficient , expensive , and laborious to perform . With these challenges in mind , we wanted to develop and characterize a new in vitro model for studying gene regulation – a model capable of measuring gene expression continuously along the developmental trajectories towards multiple cell types , and to be able to do so in multiple individuals . To do this , we used iPSCs to form embryoid bodies ( EBs ) , which are three-dimensional aggregates of spontaneously and asynchronously differentiating cells . EB formation has been used to verify stem cell pluripotency for decades; yet , until recently , the complexity of EB cellular composition has precluded their use in genomic studies . With single-cell RNA-sequencing ( scRNA-seq ) , it is now possible to characterize the numerous spatially and developmentally distinct cell types within EBs , including transient cell types that would otherwise be inaccessible . Indeed , recent scRNA-seq studies of human EB differentiation have revealed the diversity of cell types composing these structures and the transcriptional dynamics governing early fate decisions ( Han et al . , 2018; Guo et al . , 2019 ) . To date , however , the only studies that have sequenced EB cells have relied on a small sample of cells from a single individual , leaving a gap in our understanding of technical , biological , and inter-individual variation present in this system ( Han et al . , 2018; Guo et al . , 2019 ) . Understanding the sources of variation that affect cell composition and scRNA-seq data from EBs is crucial for evaluating the utility of EBs as a novel system for population-level studies of gene regulation . To this end , we used a batch-controlled study design to generate and sequence EBs from multiple individuals using multiple replicates . This allowed us to measure the degree of technical and biological variability in cell identity and gene expression levels associated with repeated independent EB differentiations . We evaluated the consistency in cell type composition across replicates and individuals , characterized the structure of variation in gene expression across the entire data set , and finally , captured patterns of dynamic gene expression along distinct developmental trajectories . Our results indicate that scRNA-seq of differentiating EBs has the potential to be a powerful model system for the study of inter-individual variation in gene regulation across an array of functionally and temporally diverse cell types . To characterize sources of variation in gene expression in human EBs , we initially differentiated EBs from three human iPSC lines ( 18511 , 18858 , and 19160 ) in three replicates ( see Materials and methods ) . We performed the experiment in three batches , where each batch includes one replicate from each of the three individuals . EBs differentiate quickly , with cell types representing endoderm , mesoderm , and ectoderm present after 8 days ( Han et al . , 2018 ) . In this study , we maintained EBs for 3 weeks after formation , allowing cells to continue to differentiate and mature . After 21 days , we collected scRNA-seq data , targeting equal numbers of cells from each individual in each replicate . After filtering and quality control ( Materials and methods ) , we retained high-quality data from a sample of 42 , 488 cells ( an average of 4721 cells per individual/replicate ) . For these cells , we obtained a median of 16 , 712 UMI counts per cell , which allowed us to measure the expression of a median of 4274 genes per cell ( Figure 1—figure supplement 1 ) . We integrated data from all cells using Harmony , which anchors the data sets by cell type ( Korsunsky et al . , 2019 ) ⁠ . After these initial collections , we found that one individual , 18858 , had lower differentiation efficiency than the other two lines ( Figure 1G , see Differentiation efficiency across individuals ) . Too assess the robustness of differentiation efficiency and cell type composition among a larger sample of individuals from our YRI iPSC panel , we differentiated EBs in one replicate from each of five additional randomly chosen lines ( 18856 , 18912 , 19140 , 19159 , and 19210 ) . After filtering and quality control , we retained an average of 5243 cells per individual in this new set of data , a median of 5983 UMI counts per cell , and a median of 2775 genes per cell ( Figure 1—figure supplement 4 ) . Throughout the text , when we use data from the five lines that were subsequently collected using only a single replicate ( 18856 , 18912 , 19140 , 19150 , and 19210 ) , we refer to them as the ‘additional lines’ . The initial replicated data collected from individuals 18511 , 18858 , and 19160 are used in all analyses throughout this study , while the additional lines are used only to demonstrate consistency of cell type composition across individuals . To validate our expectation that EBs should contain cells from each germ layer , we first characterized the expression of early developmental marker genes . We found cells expressing markers for endoderm ( SOX17 , FOXA2 ) , mesoderm ( HAND1 ) , and ectoderm ( PAX6 ) , in addition to cells still expressing pluripotency markers ( POU5F1 , MYC , NANOG ) . We visualized the data with uniform manifold approximation and projection ( UMAP ) and observed that cells expressing each of these germ layer markers occupied distinct groups in UMAP space ( Figure 1A–D; Becht et al . , 2018 ) . Moreover , we found that every replicate in our experiment , regardless of the individual , includes cells from all three germ layers ( Figure 1E and G ) . We next sought to further explore the heterogeneous cell types present in these EBs . In studies of scRNA-seq data from tissues and samples with well-characterized cell type composition , clustering is often applied to demarcate populations of pure cell types within heterogeneous samples . In these studies , clustering resolution , which determines the number of clusters identified by the algorithm , is typically chosen to recapitulate the expected number of known cell types . The identified clusters can be annotated based on the expression of known marker genes . In our case , however , we had no a priori knowledge of the exact number or types of cells that would result from the spontaneous differentiation of the EBs . Hence , we used three complementary approaches to annotate cells , capturing various perspectives on what might define a cell type in this data set . First , we identified cell types by clustering cells and annotating the cell types based on the genes that are highly expressed in each cluster . Second , we annotated cell types by considering the correlation of gene expression in our data with a reference data set of known primary cell types . For our third approach , we used a different perspective , and applied topic modeling to consider a less discrete definition of cell type . For the first approach , we used a standard clustering analysis , the Louvain algorithm in Seurat , to identify groups of cells with similar transcriptomes ( Blondel et al . , 2008 ) . To avoid making assumptions about the true number of cell types present , we repeated this analysis across different clustering resolutions ( resolution 0 . 1 , 0 . 5 , 0 . 8 , and 1 ) . As expected , the number of clusters we identified varied greatly depending on the resolution ( Figure 1E–F , Figure 1—figure supplement 2 ) . We performed each subsequent analysis using clusters defined at multiple resolutions , to ensure that our qualitative conclusions are robust with respect to the number of clusters identified . For each clustering resolution , we calculated pseudobulk gene expression levels using cells from the same cluster , individual , and replicate . To identify marker genes expressed in each cluster , we used Limma and voom to perform differential expression analysis ( Materials and methods ) using the pseudobulk estimates . For example , considering the gene expression data of the seven clusters identified at resolution 0 . 1 , we found that the most significantly upregulated genes in each cluster included known marker genes for pluripotent cells ( cluster 0 ) , early ectoderm ( cluster 1 ) , mesoderm ( cluster 2 ) , neural crest ( cluster 3 ) , endoderm ( cluster 4 ) , neurons ( cluster 5 ) , and endothelial cells ( cluster 6 ) ( Figure 1E , Table 1 ) . Using this approach provides a confident set of broad cell type categories present in these data . At higher resolutions , DE analysis between clusters enabled annotation of some more specific cell types; for example , at clustering resolution 1 , cluster 19 is characterized by higher expression of hepatocyte marker genes FGB , TTR , and AFP . More generally , however , confident cell type classification of Seurat clusters at higher resolution based on DE alone proved difficult ( Supplementary files 4-7 ) . To pursue the second approach , we annotated cells by comparing our gene expression data to available reference sets of scRNA-seq data from primary fetal tissues , human embryonic stem cells ( hESCs ) , and hESC-derived EBs ( Cao et al . , 2020; Han et al . , 2020 ) . To do this , we first integrated our data set with the reference data sets and visualized cells with UMAP ( Figure 2A , Figure 2—figure supplement 1 , Materials and Methods ) . We observed that reference hESCs cluster closely with pluripotent EB cells . We also observed that the hESC-derived EBs and our iPSC-derived EBs tend to occupy the same areas in UMAP space , implying high overall similarity in cell type composition despite differing protocols for EB differentiation ( and despite the fact that the experiments were performed in different labs ) . EB cells also show overlap with many primary fetal cell types ( Figure 2B–C , Figure 2—figure supplement 2 ) . For example , EB cells annotated as neural crest based on our gene expression analysis , overlap with primary fetal cell types derived from neural crest , such as Schwann cells and enteric nervous system ( ENS ) glia ( Figure 2B ) . EB cells annotated as neurons based on our gene expression analysis overlap with fetal neuronal subtypes , including inhibitory neurons , excitatory neurons , granule neurons , ENS neurons , and others . EB cells also show overlap with populations of cells that are rare in the fetal data set , including AFP_ALBpositive cells ( hepatic cells ) , thymic epithelial cells , and lens fibre cells ( Figure 2C ) . Encouraged by these observations , we expanded the annotation of our EB cells ( which up to this point were based on the expression of known marker genes ) by using the known annotations of the reference primary fetal cell type data set . Specifically , we transferred cell annotations to EB cells based on the nearest reference cells in harmony-corrected PCA space ( Figure 2D ) . Using this approach , we found EB cells representing 66 of the 77 primary cell types present in the reference fetal data set ( Supplementary file 1 ) . The most common annotation was hESC ( 80% of EB cells ) ; this can be partially attributed to the high proportion of pluripotent cells in our EB data set , but also to the fact that the reference fetal data set does not include many early developmental cell types . Indeed , many cells annotated as hESC here are likely to represent immature , differentiating cells which are no longer pluripotent but whose transcriptional profiles more closely match hESCs than the more highly differentiated fetal cell types present in the reference data set . In this sense , EB data sets may capture transient developmental cell types that are difficult or impossible to study even in fetal primary samples . Outside the hESCs , many fetal cell types are only represented by small populations of EB cells . For example , only one EB cell is annotated as a thymocyte , and only one cell is annotated as a myeloid cell . These observations indicate that , in the future , we can benefit from a deeper sampling of EB single cells in order to properly explore their true cell type composition . Overall , annotation based on the reference set revealed the presence of dozens of diverse cell types in EBs . To assess the differentiation efficiency of each individual in each replicate , we calculated the proportion of cells assigned to each cluster as resolution 0 . 1 ( Figure 1G ) . While EBs from two of the individuals in our study differentiated efficiently across all replicates , we observed that 89% of cells from individual 18858 were assigned to cluster 0 , the cluster annotated as pluripotent cells based on differential expression of marker genes ( Table 1 , Supplementary file 3 ) . The EBs from this line do differentiate , producing high quality cells assigned to clusters representing each germ layer ( Figure 1G , Figure 1—figure supplement 3 ) , but these EB have overall markedly lower differentiation efficiency than EBs from individuals 18511 and 19160 . To determine whether individual 18858 is an outlier , and more generally estimate how often EB differentiation is less efficient , we differentiated an additional five human iPSC lines from individuals 18856 , 18912 , 19140 , 19159 , and 19210 ( Materials and methods ) . We were reassured to find that EBs from additional lines differentiated efficiently , with cell type composition similar to 18511 and 19160 ( Figure 1H ) . These results suggest that poor differentiation efficiency is expected to be rare among the YRI iPSC panel . We further explored the robustness of cell type composition by integrating the additional lines with the fetal and hESC reference data sets using the same methodology as was used for the original lines ( Figure 3 , Figure 3—figure supplement 1 , Figure 3—figure supplement 2 ) . We find that annotations assigned to cells from these additional lines represented 66/77 fetal cell types; this set of annotations included several cell types that were missing in the original three lines , but also excludes several that were seen in the original three lines ( Supplementary file 2 ) . Again , we observed that most EB cells from the additional lines are annotated as hESC ( 82% of cells ) , although many no longer express pluripotency markers and do express markers of various germ layers as we previously observed . Together , these results support our conclusions that EBs contain many diverse cell types , many of which likely capture earlier stages of development than are captured in fetal data . Both of the approaches we described above ( clustering , and comparison to a reference data set ) assume that ‘cell types’ are discrete categories . Accordingly , each cell has a single true identity , and cell type categories are assumed to be homogeneous and static . This definition of a cell type is intuitive and makes it practical to consider results from single-cell analysis in the context of the wealth of knowledge previously gained from bulk assays . However , partitioning cells into discrete groups is unlikely to capture the full degree of heterogeneity in gene expression of single cells . A particular cluster or cell ‘type’ may collapse multiple cell states , obscuring functionally distinct subgroups such as cells in different stages of the cell cycle . This problem becomes more apparent in data sets that include differentiating cells , which are expected to show varying degrees of similarity to a terminal cell type . In an alternate paradigm , cell type can be viewed as continuous , with the expression profile of each cell representing grades of membership to multiple categories ( Dey et al . , 2017 ) . One method used to capture cell identity in this paradigm is topic modeling , which learns major patterns in gene expression within the data set , or topics , and models each cell as a combination of these topics . We applied topic modeling using fastTopics at a range of topic resolutions , identifying 6 , 10 , 15 , 25 , and 30 topics in our data . Some topics correspond closely to Seurat clusters , loaded on cells of a given cluster but not on others . For example , in the k = 6 topic analysis , topic 1 is loaded exclusively on cells assigned to Seurat cluster 4 ( cluster resolution 0 . 1 ) which we previously annotated as endoderm ( Figures 4A–D–1E , Table 1 ) . Compared to other topics , topic 1 shows an increase in expression of FN1 and AFP , which are known markers of hepatocytes ( Figure 4E , Table 2 ) . Seurat clustering at higher resolution ( resolution 1 ) results in further categorical division of this large endoderm group of cells into definitive endoderm and hepatocytes ( Figure 1F ) . Topic modeling revealed that these cells actually exhibit variable grades of membership in topic 1 ( in k = 6 topic model ) ; this gradient captures a temporal continuum of differentiation . Certain topics are shared across cells assigned to different Seurat clusters ( Figure 4A , Figure 4—figure supplement 3 ) . For example , topic 6 from the k = 6 topic analysis is loaded across all Seurat clusters; compared to all other topics , topic 6 shows increased expression of many ribosomal genes , housekeeping genes ( GAPDH ) , and genes coding for proteins involved in cellular metabolism ( LDHA ) ( Figure 4—figure supplements 1–3 ) . This indicates that topic six captures patterns of gene expression associated with cellular processes and functions that are not specific to a particular cell type . This again highlights an advantage of topic modeling , enabling us to explore variation in the representation of gene expression profiles associated with processes shared across many cell types , simultaneously with identifying cell-type-specific patterns . Once we functionally annotated EB cells using the three approaches discussed above , we sought to understand the consistency in cell type composition across individuals and between replicates . Here , ‘replicate’ corresponds to a batch of EB differentiations in which each cell line was differentiated , dissociated , and collected in tandem; ‘replicate’ therefore captures technical variation related to differentiation batch , dissociation batch , and single-cell collection batch . We began by calculating the proportion of cells that were assigned to each Seurat cluster at resolution 0 . 1 for each replicate . We then performed hierarchical clustering of the samples based on the proportion of cells in each Seurat cluster ( Figure 5—figure supplement 1 ) . Using this approach , replicate-individual samples cluster first by individual , indicating that cell type composition is distinct between individuals and is consistent between replicates of each individual . We repeated this analysis at a range of cluster resolutions and determined that this finding is robust with respect to the number of clusters ( Figure 5—figure supplement 1 ) . We also repeated this analysis using topic loadings as a measure of cell type composition . We calculated the loading of each topic on each individual-replicate group and performed hierarchical clustering ( Figure 5—figure supplement 2 ) . Again , we found that at varying values of k , samples generally cluster by individual , but using the higher resolution topic-based approach , we also observed substantial variation between replicates ( Figure 5—figure supplement 2 ) . Individual 18858 always clusters away from the other two lines , due to the consistent and distinct distribution of cell types caused by low differentiation efficiency . We further characterized determinants of variation in our system by considering factors that contribute to variation in gene expression levels . Hierarchical clustering of pseudobulk expression estimates of cells from the same Seurat cluster ( res . 0 . 1 ) , replicate , and individual shows that , as might be expected , samples tend to cluster first by cell type ( Seurat cluster ) , then by individual and replicate ( Figure 5A ) . We performed variance partitioning using pseudobulk expression levels to estimate the relative contribution of cell type , individual , and replicate to overall patterns of gene expression variation ( Figure 5B; Hoffman and Schadt , 2016 ) . We found that replicate and individual explained approximately equal proportions of the variance ( each explains a median value of ~5% of variance ) . Cell type identity explained the largest proportion of variation at all clustering resolutions tested ( variance explained median value ~60% at clustering resolution 0 . 1 ) , although this figure is likely exaggerated since cell type identity is determined by clustering , which will inherently maximize variation between cell types . Depending on clustering resolution , a median value of approximately 20–30% of variance is explained by residuals , which can be attributed to noise or other technical variation not specifically modeled ( Figure 5—figure supplement 3 ) . We then partitioned the variance in gene expression at single cell resolution ( instead of using pseudobulk estimates ) and found that replicate explains more variation on average than individuals , with cell type identity continuing to explain more variance than either ( Figure 5C ) . At single-cell resolution , residuals explain a median value of 93% of the variation , which is expected due to the high degree of variance ( both biological and technical ) in gene expression profiles among individual cells . To determine whether biological and technical factors contributed differently to variation between cell types , we also partitioned the variance due to replicate and individual in each Seurat cluster separately ( Figure 5—figure supplements 4–5 ) . The results are not uniform across clusters . At clustering resolution 0 . 1 , individual contributes more to variation , on average , in clusters 0 , 1 , 4 , and 5 , while replicate contributes more to variation in clusters 2 , 3 , and 6 . Notably , clusters 2 , 3 , and 6 include only a few cells from individual 18858 ( Supplementary file 3 ) . Studies that incorporate a larger number of cells will increase representation of rare cell types , which will increase power to study patterns of gene regulation . In every cluster , variation due to replicate dominates the variation of certain genes but not others . This complex structure indicates that , unlike most other eQTL studies – where adding individuals is always preferable to adding technical replicates – future studies of EBs need to implement study designs with multiple replicates to appropriately account for this variation . Because individual variation contributes to overall patterns of variation in gene expression , EBs have the potential to be a powerful model to understand inter-individual variation in gene regulation across cell types and to map dynamic eQTLs . We performed a power analysis to better understand the relationship between power , sample size , and the total number of individual cells analyzed , or the experiment size ( Figure 6 ) . Assuming a simple linear regression to map eQTLs and a conservative Bonferroni correction for multiple testing ( FWER = 0 . 05 , Materials and methods ) , we estimated that an experiment consisting of 58 individuals with 3000 cells collected per individual , collected across three replicates ( experiment size of 174 , 000 cells total ) , would provide 93% power to detect eQTLs with a standardized effect size of 0 . 6 . These assumptions represent an experimentally tractable study design , and a conservative estimate of standard and dynamic eQTL effect sizes , suggesting this could be a powerful system for QTL studies across diverse human cell types . Arguably , the most attractive property of single-cell data from the EB system is the ability to study dynamic gene regulatory patterns throughout differentiation . In order to explore dynamic patterns of gene expression , we inferred developmental trajectories using PAGA ( Wolf et al . , 2019 ) . The PAGA graph shows edges that represent likely connections between cell clusters ( clustering resolution 1 ) and we were able to trace developmental trajectories through these paths ( Figure 7A–C , Figure 7—figure supplement 1A-C ) . Since the EBs still include undifferentiated pluripotent cells , we were able to define rooted trajectories to each germ layer beginning at the known starting point . Trajectories toward endoderm and mesoderm proceed through cluster 22 , which expresses primitive streak marker MIXL1 , showing recapitulation of developmental trajectories defined during gastrulation ( Figure 1F , Figure 7—figure supplement 1 , Figure 7—figure supplement 2 ) . Hepatocytes ( cluster 19 ) , an endoderm-derived cell type , branch off of the endoderm cluster ( cluster 10 ) ( Figure 7B , Figure 7—figure supplement 2 ) . Endothelial cells ( cluster 24 ) , which are derived from mesoderm , branch off from the mesoderm cluster ( cluster 4 ) ( Figure 7C , Figure 7—figure supplement 2 ) , and neurons ( clusters 12 , 15 ) , an ectoderm-derived cell type , branch off from the early ectoderm clusters ( clusters 2 , 3 , 8 , 9 , 13 , 14 , 17 , 20 , 21 , 26 , and 27 ) ( Figures 7A and 1F , Table 1 , Figure 7—figure supplement 2 ) . We then assigned pseudo-time values to each cell using the diffusion pseudo-time method with pluripotent cells ( cluster 1 ) defined as the root ( Haghverdi et al . , 2016; Figure 7—figure supplement 1D , Figure 1F ) . We manually traced high confidence trajectories through the data representing the progression from pluripotent cells to hepatocytes ( clusters 0 , 1 , 5 , 6 , 7 , 10 , 11 , 16 , 18 , 19 , 25 , and 22 ) ( Figure 7B ) , pluripotent cells to endothelial cells ( clusters 0 , 1 , 4 , 5 , 6 , 7 , 11 , 16 , 18 , 22 , 24 , and 25 ) ( Figure 7C ) , and pluripotent cells to neurons ( clusters 0 , 1 , 2 , 3 , 5 , 6 , 7 , 8 , 9 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 20 , 21 , 26 , and 27 ) ( Figure 7A ) . For groups of clusters with a higher degree of connectivity ( e . g . clusters expressing pluripotent markers and clusters expressing early ectoderm markers ) , all clusters within the region with high connectivity were included in the trajectory to avoid choosing an arbitrary path through these clusters . Next , we applied split-GPM , an unsupervised probabilistic model , to infer dynamic patterns of gene expression within a particular developmental trajectory , while simultaneously performing clustering of genes and samples . Split-GPM is built for use with time course , bulk RNA-seq data; therefore , we calculated pseudobulk expression values for individual-replicate groups within decile bins of pseudo-time . We were able to identify gene modules with distinct dynamic patterns of expression along the trajectories to neurons , hepatocytes , and endothelial cells ( Figure 7—figure supplements 3–5 ) . Gene set enrichment analysis of these modules shows expected dynamic patterns ( Figure 7—figure supplements 3–5 ) . For example , we found that gene modules that increase expression through pseudo-time along the differentiation trajectory to hepatocytes , which are the predominant cell type of the liver and are responsible for the production of bile , are enriched for the hallmark bile acid metabolism and fatty acid metabolism gene sets . In the trajectory leading to endothelial cells , which are derived from mesoderm , we found that a gene module with high expression at intermediate pseudo-time values is enriched for hallmark genes expressed during the epithelial-mesenchymal transition , which is essential for mesoderm formation ( Evseenko et al . , 2010 ) . In all three trajectories , gene modules characterized with higher expression at low pseudo-time values show enrichment for gene sets related to the cell cycle; this is expected because pluripotent cells at the lowest pseudo-time values tend to grow and divide faster than more differentiated and more mature cell types , which often exit the cell cycle ( Buttitta and Edgar , 2007 ) . To determine the consistency in dynamic patterns of gene expression between replicates and individuals , we ran split-GPM ten times on cells from the neuron , hepatocyte , and endothelial cell lineages and observed how often each pair of individual-replicate samples clustered together ( Figure 7D–F ) . In the neuronal and endothelial lineages , all three replicates of 18511 always clustered together and often cluster with replicates of 19160 , indicating that these two lines share similar expression dynamics in these trajectories ( Figure 7D and F ) . Replicates of 18858 often clustered together and rarely clustered with the other individuals , suggesting that not only did 18858 have poor differentiation efficiency , but cells that did differentiate show a distinct pattern of expression dynamics . In the hepatocyte lineage , we observed stronger replicate-specific differences ( Figure 7E ) . Replicates of individual 19160 still tended to cluster together and to cluster with replicates 1 and 2 of 18511 . Replicate 3 of 18511 rarely clustered with the other replicates of that individual , indicating that there were replicate-specific effects on dynamic gene expression . This work represents a thorough exploration of heterogeneity in single cell data obtained from human EBs towards the goal of establishing this system as a tool to enable studies of variation in human gene regulation across a range of spatially and temporally diverse cell types . We used iPSC-derived EBs because this in vitro model system circumvents the logistical challenges and ethical barriers associated with studies of primary human developmental tissues . This system has key advantages over studies of primary tissues; for example , we are able to control cellular environment in vitro and intentionally design experiments with respect to biological factors including age , sex , and ancestry . Further , we can generate EBs comprised of the same set of diverse cell types from large samples of individuals , enabling high-powered comparisons of cell-type-specific gene expression . In subsequent studies , we plan to leverage EBs to identify QTLs and dynamic QTLs across diverse terminal and differentiating cell types . This , of course , raises an ostensibly critical question: to what extent do the cell types derived from EBs faithfully model immature , developing cells in vivo ? There is no doubt that the in vitro EB differentiated cells are not a perfect model of primary cell types . The question is whether EB cells are sufficiently representative of primary cell types to be informative . To address this question , we performed several analyses , which suggest that the EB model can be useful . Specifically , we found that EB cells express known cell-type-specific marker genes , including markers of known developmental stages . EB cells also cluster with more than 60 diverse primary cell types from a reference panel of fetal tissues and hESCs , including rare fetal cell types . Lastly , we identified gene modules with dynamic expression patterns that match broad expectations of developmental biology . Together , these results provide evidence that EBs are a suitable model of both terminal and developmental cell types . Moreover , EBs may be a useful model for understanding the genetic underpinnings of human traits and diseases regardless of the degree to which they faithfully model human development . EB-derived cells represent a wealth of previously unstudied cell states and dynamic processes . Hypothetically , QTLs identified in these cell types still represent biologically meaningful differences in genetic control of gene regulation , whether they manifest in human development or in adult tissues upon a particular environmental exposure . To provide an anecdotal example of this reasoning , we considered previously collected data from an in vitro differentiation experiment . We took a closer look at the 28 middle-dynamic eQTLs Strober et al . identified during the differentiation of iPSCs to cardiomyocytes ( Strober et al . , 2019 ) . Middle-dynamic eQTLs have their strongest genetic effects at intermediate stages of the differentiation time course , and most of them ( 25/28 ) were identified exclusively at these intermediate stages of differentiation . Accordingly , these eQTLs are active in early in vitro differentiating cells whose fidelity to primary developing cell types has not been ascertained . These 28 dynamic eQTLs were entirely novel and had not been identified as cis eQTLs in any tissue in the GTEx data set . Strober et al . reported that one of these middle-dynamic eQTLs was also found to overlap a GWAS variant associated with body mass index and red blood cell count . This finding highlights that dynamic eQTLs acting in early cell types in in vitro differentiations may affect long-term disease risk in adults . To further explore the utility of dynamic eQTLs identified in in vitro differentiations , we used GTEx data to ask whether the middle-dynamic eQTLs are associated with inter-individual variation in trans gene expression or cell composition , either of which could indicate lasting downstream effects in adult tissue from transient dynamic cis eQTLs . Trans eQTL associations are more tissue-specific than cis eQTLs , but trans eQTLs are much harder to identify because of their small effect sizes and the requirement to test the association of every locus with every gene . Here , we identified a middle dynamic eQTL SNP ( rs6700162 ) from Strober et al . that , in GTEx data , is associated with fibroblast cell type proportions in HLV ( heart left ventricle; p < 0 . 0009 ) and with cardiac muscle cell proportions in HLV ( p < 0 . 003 ) . This SNP was also found to have a trans eQTL p-value of 1 × 10–5 in coronary artery . Without the prior knowledge provided by dynamic eQTL data from the in vitro differentiated cardiomyocytes , it would have been impossible to identify these associations using adult primary tissues because the burden of multiple testing within the entire GTEx data set is considered is prohibitively large . This example implies that developing EB cells could be used to understand how transient effects on gene expression are propagated into functional , long-lasting consequences downstream . In summary , human EBs have the potential to be a powerful system for the identification of dynamic eQTLs . In this pilot study , we performed foundational analyses to better understand how to appropriately conceptualize heterogeneity in this kind of data and how to best design large-scale studies of EBs . We explored cell type composition of EBs in three paradigms; first , with discrete cell types identified with a traditional clustering algorithm , then with more continuous cell “types” identified with topic modeling , and finally exploring dynamic changes in gene expression along trajectories using pseudotime . Cell types defined by discrete clustering are often easier to interpret because they can be contextualized with marker genes and reference data sets defined with bulk sequencing . We conclude , however , that topic modeling is more appropriate for highly heterogeneous single cell data sets like this one . We also explored sources of variation in cell type composition and gene expression . We found that individual variation primarily contributes to patterns in cell type composition based on both discrete clustering and topic modeling . However , variation between replicates is non-negligible , indicating that future studies should focus on inter-individual variation in cell type composition . We also found that technical variation between replicates contributes to variation in gene expression . Future efforts to map regulatory QTLs in EBs should implement study designs with multiple replicates to appropriately correct for batch effects . Overall , this pilot study has laid the groundwork to transform EBs into a powerful model system for the understanding of human gene regulation . We used iPSC lines from eight unrelated individuals from the Yoruba HapMap population to form EBs . The iPSC lines were reprogrammed from lymphoblastoid cell lines ( LCLs ) and were characterized and validated previously ( Banovich et al . , 2018 ) . The original LCL lines were genotyped by the HapMap project and identity of the stocks used in this study is confirmed by scRNA-seq data collected for this study ( Belmont et al . , 2003 ) . All cell lines used in this study tested negative for mycoplasm . Lines 18511 , 18858 , 18912 , 19140 , and 19159 are from female individuals . Lines 19160 , 18856 , and 19210 are from male individuals . Preprocessing and analysis of lines 18511 , 18858 , and 19160 are described throughout the Materials and methods section . Preprocessing and analysis of lines 18856 , 18912 , 19140 , 19159 , and 19210 is restricted to the Methods section titled ‘Assessment of cell type composition and differentiation efficiency in five additional lines’ . We maintained feeder-free iPSC cultures on Matrigel Growth Factor Reduced Matrix ( CB-40230 , Thermo Fisher Scientific ) with StemFlex Medium ( A3349401 , Thermo Fisher Scientific ) and Penicillin/Streptomycin ( 30 , 002 Cl , Corning ) . We grew cells in an incubator at 37 °C , 5% CO2 , and atmospheric O2 . Every 3–5 days thereafter , we passaged cells to a new dish using a dissociation reagent ( 0 . 5 mM EDTA , 300 mM NaCl in PBS ) and seeded cells with ROCK inhibitor Y-27632 ( ab120129 , Abcam ) . We formed EBs using a modified version of the STEMCELL Aggrewell400 protocol . Briefly , we coated wells of an Aggrewell 400 24-well plate ( 34415 , STEMCELL ) with anti-adherence rinsing solution ( 07010 , STEMCELL ) . We dissociated iPSCs and seeded them into the Aggrewell400 24-well plate at a density of 1 , 000 cells per microwell ( 1 . 2 × 106 cells per well ) in Aggrewell EB Formation Medium ( 05893 , STEMCELL ) . After 24 hr , we replaced half of the spent media with fresh Aggrewell EB Formation Medium . Forty-eight hr after seeding the Aggrewell plate , we harvested EBs and moved them to an ultra-low attachment six-well plate ( CLS3471-24EA , Sigma ) in E6 media ( A1516401 , ThermoFisher Scientific ) . We maintained EBs in culture for an additional 19 days , replacing media with fresh E6 every 48 hr . We performed three replicates of EB formation on different days; each replicate included all three lines . We collected and dissociated EBs 21 days after formation . We dissociated EBs by washing them with phosphate-buffered saline ( PBS ) ( Corning 21–040-CV ) , treating them with AccuMax ( STEMCELL 7921 ) and incubating them at 37 °C in for 15–35 min . After 10 min in Accumax , we pipetted EBs up and down with a clipped p1000 pipette tip for 30 s . We repeated pipetting every 5 min until EBs were completely dissociated . We then stopped dissociation by adding E6 media and straining cells through a 40 µm strainer ( Fisherbrand 22-363-547 ) . We resuspended cells in PBS and counted them with a TC20 Automated Cell Counter ( 450102 , BioRad ) . Before scRNA-seq , we mixed together an equal number of cells from each line . We collected scRNA-seq data using the 10x Genomics V3 . 0 kit . Single-cell collections for this experiment were mixed with cells from a larger experiment in all three replicates . From the first replicate of EB differentiations , we mixed EB cells YRI individuals 18511 , 18858 , and 19160 with EB cells from an additional three humans and chimpanzees ( nine individuals total ) . Even numbers of cells from all nine individuals were collected across nine lanes of a 10x chip , targeting 10 , 000 cells per lane . In this replicate , reagents from three different 10x kits were used . From replicates 2 and 3 of EB differentiation , EBs were only generated from the same three YRI individuals ( 18511 , 18858 , and 19160 ) and the three chimpanzees ( six individuals total ) . In each replicate , we mixed even numbers of cells of each individual and collected cells in four lanes of a 10x chip , targeting 10 , 000 cells per lane , and samples were processed using reagents from a single 10x kit . Libraries were sequenced using paired-end 100 base pair sequencing on the HiSeq 4000 in the University of Chicago Functional Genomics Core . For libraries from replicate 1 , we mixed equal proportions of each of the six EB libraries and sequenced the pooled samples on one lane of the HiSeq 4000 . Preliminary analyses showed that two of these lines were low quality . We remade one of the low-quality libraries and discarded the other . We then mixed equal proportions of the remade library with the remaining three libraries from replicate one and sequenced the pooled samples on one lane of the HiSeq 4000 . Preliminary analyses indicated that three out of four of these libraries were below optimal quality , but would produce usable data . We then pooled together samples from the final eight libraries from replicate 1 , mixing equal parts of each of the five high-quality libraries with half the amount of the other three , and deep-sequenced this pool on eight lanes of the HiSeq 4000 . For replicate two libraries , we mixed equal parts of all four libraries and sequenced them on one lane . After confirming that each library was high-quality , we deep-sequenced the same pool on six additional lanes of the HiSeq 4000 . For replicate three libraries , we mixed equal parts of all four libraries and sequenced them on one lane . After confirming that each library was high-quality , we deep-sequenced the same pool on four additional lanes of the HiSeq 4000 . In all cases , the number of lanes for deep sequencing was calculated to reach 50% saturation . We used STARsolo to align samples to both the human genome ( GRCh38 ) ( Dobin et al . , 2013 ) and the chimpanzee genome ( January 2018; Clint_PTRv2/panTro6 ) . We used gene annotations from ensembl98 and transmapV5 , respectively . In order to separate human cells from chimpanzee cells , we identified discordant reads – those that mapped with different scores in each species . We identified a cell as human if ( 1 ) at least five discordant reads that had a higher human mapping score and ( 2 ) at least 80% of discordant reads had a higher human mapping score . The remainder of analyses in this work were restricted to these human cells . We demultiplexed individual samples and identified doublets using demuxlet ( Kang et al . , 2018 ) . For this demultiplexing with demuxlet , we used previously collected and imputed genotype data for these three Yoruba individuals from the HapMap and 1000 Genomes Project ( Auton et al . , 2015; Belmont et al . , 2003 ) . We ran EmptyDrops to identify barcodes tagging empty droplets and kept only barcodes with a high probability of tagging a cell-containing droplet ( i . e . we kept cells with an EmptyDrops FDR < 0 . 0001 ) ( Lun et al . , 2019 ) . We removed cells labeled as doublets or ambiguous by demuxlet , keeping only barcodes labeled as singlets . We also filtered the data to include only high-quality cells expressing between 3% and 20% mitochondrial reads and expressing more than 1500 genes . We normalized data from each 10x lane using SCTransform in Seurat ( Butler et al . , 2018; Hafemeister and Satija , 2019 ) . In total , we obtained 42 , 488 high-quality cells . We then merged data from each of the 10x lanes from all replicates , scaled the data , and ran principal components analysis ( PCA ) using 5000 variable features . We then integrated data with Harmony to correct the PCA embeddings for batch effects and individual effects ( Korsunsky et al . , 2019 ) . To cluster the data , we applied Seurat’s FindNeighbors using 100 dimensions from the Harmony-corrected reduced dimensions , followed by FindClusters at resolutions 0 . 1 , 0 . 5 , 0 . 8 , and 1 . We performed differential expression analysis using the limma R package ( Ritchie et al . , 2015 ) . First , we filtered genes to include only those expressed in at least 20% of cells in at least one cluster at a given clustering resolution . We then calculated pseudobulk expression values for each individual-replicate-cluster grouping ( i . e . cells from the same individual , same replicate , and same cluster assignment ) . Accordingly , we define pseudobulk expression values as the sum of normalized counts in each group . Next we TMM-normalized pseudobulk expression values and used voom to calculate a weighted gene expression value to account for the mean-variance relationship ( Law et al . , 2014 ) . We then fit the following linear model:Y=0+βclusterx+βreplicatex+βindividual*x We used contrasts to first test for differential expression of each cluster compared to all other clusters and then to test for differential expression between pairs of similar clusters to find distinguishing markers . We annotated cell type identity of each cluster based on significant differential expression of the known marker genes . To evaluate the cell type composition resulting from EB differentiation of YRI iPSC lines more generally , we differentiated five additional randomly chosen lines ( 18856 , 18912 , 19140 , 19159 , and 19210 ) from the YRI iPSC panel . We differentiated and dissociated iPSCs in parallel using the same protocols described above . After dissociation , we mixed cells from each individual in equal proportions and collected scRNA-seq data using the 10x genomics V3 . 1 kit , targeting collection of 10 , 000 cells per lane and 10 , 000 cells per individual . Notably , we mixed cells from these five lines with cells from two additional lines ( each with distinct genotypes ) from a separate experiment during 10x collections . Libraries were sequenced using paired-end 100 bp sequencing on the NovaSeq 6000 at the University of Chicago Genomics Core . We aligned samples to the human genome ( GRCh38 ) using CellRanger ( Zheng et al . , 2017 ) . We then assigned cells to individuals . We used demuxlet to identify doublets with previously collected and imputed genotype data for the five additional YRI individuals; this data originated from the HapMap and 1000 Genomes Projects ( Kang et al . , 2018; Auton et al . , 2015; Belmont et al . , 2003 ) . Finally , we removed cells assigned to individuals that were not a part of this experiment . We filtered out doublets , cells with greater than 15% mitochondrial reads or fewer than 3% mitochondrial reads , and cells with fewer than 1000 unique genes expressed . We then normalized data using SCTransform ( Butler et al . , 2018; Hafemeister and Satija , 2019 ) , identified clusters using the Louvain algorithm in Seurat ( at Resolution 0 . 15 ) , and visualized expression of canonical marker genes and the most significant marker genes of clusters identified in differential expression analysis ( Figure 1—figure supplement 4 ) . Based on marker gene expression , clusters 0 and 2 represent early ectoderm , cluster one represents neural crest cells , cluster three represents pluripotent cells , clusters 4 and 6 represent neurons , cluster 5 represents mesoderm , cluster seven represents endoderm , and cluster 8 represents endothelial cells . We calculated the proportion of cells from each individual that were assigned to each of these cell type categories . We observed that each of these five additional cell lines exhibits high differentiation efficiency , comparable to that of iPSC lines 18511 and 19160 . Additional lines were also integrated with reference data to annotate cell types as described below . We next compared cells to reference data sets of primary fetal cell types , Day 20 hESC-derived EBs , and hESCs ( Cao et al . , 2020; Han et al . , 2020 ) . To integrate our cells with the reference sets , we first subset each reference set to include only protein coding genes . Because the Cao et al . reference set is so large , containing over 4 million cells , we subset cells from this reference set to include a maximum of 500 cells per cell type . We then normalized each reference set using SCTransform ( Butler et al . , 2018; Hafemeister and Satija , 2019 ) . We then merged the data sets using Seurat , re-ran SCTransform regressing out data set specific effects of sequencing depth , scaled the data , and ran PCA . We then ran Harmony to correct PCA embeddings for the effects of each data set to complete the integration ( Korsunsky et al . , 2019 ) ⁠ . We then transferred cell type annotations from cell types present in the fetal reference and hESC to EB cells . For each EB cell , we found the five nearest reference cells in Harmony-corrected PCA space based on Euclidean distance; if at least 3/5 of the nearest reference cells shared an annotation , that annotation was transferred to the EB cell . If fewer than three of the nearest reference cells shared an annotation , the EB cell was annotated as ‘uncertain’ . To assess the quality of our reference integration strategy , we asked whether ( 1 ) datasets are being over-corrected and ( 2 ) EB cells annotated using reference cell types express expected marker genes . We first subsetted EB cells to broad cell type categories identified using clustering ( at resolution 0 . 1 ) and differential expression analysis: Pluripotent ( cluster 0 ) , Early Ectoderm ( cluster 1 ) , Endoderm ( cluster 4 ) , Meso-derm ( clusters 2 , 6 ) , Neural Crest ( cluster 3 ) , and Neurons ( cluster 5 ) . Using each subset of cells , we repeated the reference integration pipeline by merging the EB cells with three reference data sets ( fetal cells , hESCs , and an external set of Day 20 EBs ) , normalizing using SCTransform , running PCA using 5 , 000 variable features , in-tegrating the data using Harmony , and transferring labels based on the five nearest reference cells ( see Materials and methods ) ( Butler et al . , 2018; Hafemeister and Satija , 2019; Korsunsky et al . , 2019 ) . We found that 79% of EB cells are assigned to the same cell type in the full integration and subset integration . Of EB cells that are anno-tated differently in the full and subset integrations , 82% were labeled as ‘hESC’ or ‘uncertain’ in either the full or subset integration . This suggests that differences in these annotations are often be due to slight changes in the positioning of cells between the hESC reference cells and fetal reference cells; this is expected when pluripotent cells are not included in subsets of EB cells . And , importantly , cells are not often annotated as a different fetal cell type . Together , these results suggest that our integration approach is robust to subsetting input cell types and is likely not over-integrating the test and reference data sets . Next , we asked whether annotated EB cells differentially express expected marker genes . We limited this analy-sis to annotations with at least 10 total EB cells from at least two individuals in two replicates . We then calculat-ed pseudobulk expression for cells of the same annotation , individual , and replicate , and filtered genes to in-clude only those with at least 10 counts in at least one sample and at least 15 total counts across all samples . We then TMM-normalized pseudobulk expression values , used voom to calculate a weighted gene expression value , and tested for differential expression between annotations using limma . Of the annotations tested , the most significantly differentially expressed genes often included known cell type markers . For example , cells annotated as cardiomyocytes showed significant upregulation of of MYL7 , MYL4 , and TNNT2 ( Figure 2—figure supplement 3 ) . Cells annotated as hepatoblasts showed significant upregulation of AFP , FGB , and ACSS2 . Cells annotated as mesothelial cells showed significant upregulation of NID2 and collagen genes ( COL6A3 , COL1A1 , COL3A1 , COL6A1 ) . These results provide further support that our reference integration approach yields meaningful annotation of EB cells . We also performed topic modeling using FastTopics to learn major patterns in gene expression within the data set , or topics , and model each cell as a combination of these topics . For this analysis , we used raw counts and filtered the data to include genes expressed in at least 10 cells . We then pre-fit a Poisson non-negative matrix factorization by running 1000 EM updates without extrapolation to identify a good initialization at values of k equal to 6 , 10 , 15 , 25 , and 30 . We used this initialization to fit a non-negative matrix factorization using 500 updates of the scd algorithm with extrapolation to identify 6 , 10 , 15 , 25 , and 30 topics . We then used FastTopics’ diff_count_analysis function to identify genes differentially expressed between topics . We used these differentially expressed genes to interpret the cellular functions and identities captured by each topic . In some cases , differentially expressed genes included known marker genes ( Table 2 ) . To understand how similar cell type composition is between replicates and individuals , we first calculated the proportion of cells from each individual in each replicate assigned to each Seurat cluster at resolution 0 . 1 . Then , using the ComplexHeatmap R package and performing hierarchical clustering based on the complete linkage method , we visualized the clustering of these replicate-individual groups ( Gu et al . , 2016 ) . We repeated this analysis using Seurat clusters at resolution 0 . 5 , 0 . 8 , and 1 to show that the overall patterns of hierarchical clustering are robust to cluster resolution . We performed an analogous analysis using topic loadings instead of cluster proportions . Here , we determined the loading of each topic on cells from the same individual and replicate , then used the same hierarchical clustering with ComplexHeatmap to visualize patterns of similarity between cells of each individual and replicate ( Gu et al . , 2016 ) . We also performed hierarchical clustering on gene expression of individual cells . To do so , we took the pseudobulk expression for each individual-replicate-cluster group and filtered to genes expressed in at least 20% of cells in at least one cluster . We then calculated the log10 counts per million expression of each gene . We then generated a heatmap using the ComplexHeatmap R package , again performing hierarchical clustering based on the complete linkage method ( Gu et al . , 2016 ) . Using the same pseudobulk data and precision weights computed by voom from differential expression analysis , we used the VariancePartition R package to quantify the variation attributable to cluster , replicate , and individual ( Hoffman and Schadt , 2016 ) . We fit a random effect model and modeled cluster , replicate , and individual as random effects . We performed this analysis across all tested Seurat clustering resolutions ( 0 . 1 , 0 . 5 , 0 . 8 , 1 ) . We performed this analysis using both pseudobulk samples of cells from the same cluster , replicate , and individual and at single-cell resolution with each cell as a sample . We also partitioned the variance in each Seurat cluster separately using a random effect model with terms for replicate and individual . For this analysis , we used pseudobulk samples of cells from each individual and replicate . To ascertain the power to detect eQTLs and dynamic eQTLs across a range of sample sizes , standardized effect sizes , and experiment sizes we used a power function as derived in Sarkar et al . , 2019:Pow ( β , α , n , σ ) =Φ ( Φ−1α2+βσn ) where β denotes the true additive significance level , ɑ denotes the significance level , n denotes sample size , and σ represents the phenotype standard deviation . This approach assumes a simple linear regression for eQTL mapping and a conservative Bonferroni correction ( FWER = 0 . 05 , assuming 10 , 000 genes tested , α = 5e-6 ) . To estimate the standard deviation for a given experiment size , we downsampled cells from this experiment , sampling evenly between individuals and replicates to range of experiment sizes from 2700 cells to 21 , 600 cells . For each experiment size , we took 10 random samples of cells and calculated pseudobulk expression of cells from the same cluster ( defined at resolution 1 ) , individual , and replicate . We filtered to include genes expressed in at least 20% of cells in at least one cluster ( at resolution 1 ) in the full set of EB cells . For each sample we then partitioned the median variance attributable to residuals using the variancePartition package . We then took the median of the median variance from the 10 samples at each experiment size and fit an exponential decay model to quantify the relationship between experiment size and residual variance . We used square root of this variance to determine the standard deviation for a given experiment size in our power calculations . We inferred trajectories using PAGA in Scanpy using Seurat clusters at all tested resolutions ( Wolf et al . , 2019 ) . We assigned pseudo-time using diffusion pseudo-time with the pluripotent cells assigned as the root ( Haghverdi et al . , 2016 ) . We then manually traced known developmental trajectories supported by the PAGA graph . At clustering resolution 1 , we traced the trajectory from pluripotent cells to hepatocytes ( clusters 0 , 1 , 5 , 6 , 7 , 10 , 11 , 16 , 18 , 19 , 25 , and 22 ) , pluripotent cells to endothelial cells ( clusters 0 , 1 , 4 , 5 , 6 , 7 , 11 , 16 , 18 , 22 , 24 , and 25 ) , and pluripotent cells to neurons ( clusters 0 , 1 , 2 , 3 , 5 , 6 , 7 , 8 , 9 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 20 , 21 , 26 , and 27 ) ( Figure 7A–C ) . We then isolated cells from each of these three trajectories and used Split-GPM to simultaneously cluster samples and identify dynamic gene modules . For this analysis , we divided data into decile pseudo-time bins and calculated pseudobulk gene expression for cells of the same individual , replicate , and pseudo-time bin . We identified 20 dynamic gene modules in each trajectory and interpreted them using gene set enrichment . To understand the variation in dynamic gene expression between individuals and replicates , we re-ran split-GPM ten times and observed how often cells from each individual and replicate were assigned to the same sample cluster .	One major goal of human genetics is to understand how changes in the way genes are regulated affect human traits , including disease susceptibility . To date , most studies of gene regulation have been performed in adult tissues , such as liver or kidney tissue , that were collected at a single time point . Yet , gene regulation is highly dynamic and context-dependent , meaning that it is important to gather data from a greater variety of cell types at different stages of their development . Additionally , observing which genes switch on and off in response to external treatments can shed light on how genetic variation can drive errors in gene regulation and cause diseases . Stem cells can produce more cells like themselves or differentiate – acquire the characteristics – of many cell types . These cells have been used in the laboratory to research gene regulation . Unfortunately , these studies often fail to capture the complex spatial and temporal dynamics of stem cell differentiation; in particular , these studies are unable to observe gene regulation in the transient cell types that appear early in embryonic development . To overcome these limitations , scientists developed systems such as embryoid bodies: three-dimensional aggregates of stem cells that , when grown under certain conditions , spontaneously develop into a variety of cell types . Rhodes , Barr et al . wanted to assess the utility of embryoid bodies as a model to study how genes are dynamically regulated in different cell types , by different individuals who have distinct genetic makeups . To do this , they grew embryoid bodies made from human stem cells from different individuals to examine which genes switched on and off as the stem cells that formed the embryoid bodies differentiated into different types of cells . The results showed that it was possible to grow embryoid bodies derived from genetically distinct individuals that consistently produce diverse cell types , similar to those found during human fetal development . Rhodes , Barr et al . ’s findings suggest that embryoid bodies are a useful model to study gene regulation across individuals with different genetic backgrounds . This could accelerate research into how genetics are associated with disease by capturing gene regulatory dynamics at an unprecedentedly high spatial and temporal resolution . Additionally , embryoid bodies could be used to explore how exposure to different environmental factors during early development affect disease-related outcomes in adulthood in different individuals .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "genetics", "and", "genomics" ]	2022	Human embryoid bodies as a novel system for genomic studies of functionally diverse cell types
Flows generated by ensembles of flagella are crucial to development , motility and sensing , but the mechanisms behind this striking coordination remain unclear . We present novel experiments in which two micropipette-held somatic cells of Volvox carteri , with distinct intrinsic beating frequencies , are studied by high-speed imaging as a function of their separation and orientation . Analysis of time series shows that the interflagellar coupling , constrained by lack of connections between cells to be hydrodynamical , exhibits a spatial dependence consistent with theory . At close spacings it produces robust synchrony for thousands of beats , while at increasing separations synchrony is degraded by stochastic processes . Manipulation of the relative flagellar orientation reveals in-phase and antiphase states , consistent with dynamical theories . Flagellar tracking with exquisite precision reveals waveform changes that result from hydrodynamic coupling . This study proves unequivocally that flagella coupled solely through a fluid can achieve robust synchrony despite differences in their intrinsic properties . Despite the elegance and apparent simplicity of the eukaryotic flagellum and its shorter ciliary version , the collective motions exhibited by groups of these organelles and the resultant fluid flows are far from trivial . For example , the unicellular biflagellate alga Chlamydomonas reinhardtii executes diffusive ‘run-and-turn’ locomotion ( Goldstein et al . , 2009; Polin et al . , 2009 ) through stochastic switching between synchronized and unsynchronized swimming gaits—a process which could enhance searching efficiency and assist in the avoidance of predators ( Stocker and Durham , 2009 ) . Ensembles of cilia and flagella exhibit stunning temporal coordination , generating flows that transport mucus and expel pathogens ( Button et al . , 2012 ) , establish the left-right asymmetry in developing mammalian embryos ( Nonaka et al . , 2002 ) , and transport ova in human fallopian tubes ( Lyons et al . , 2006 ) . The origin of flagellar synchronization has been the subject of intense theoretical investigation for many decades . One of the earliest experimental results was Rothschild's qualitative observation ( Rothschild , 1949 ) that the flagella of bull spermatozoa tend to synchronize when they swim close to one another , coupled only through the fluid surrounding them . Much more recent observations of self-organised vortex arrays of swimming sea urchin spermatazoa near surfaces ( Riedel et al . , 2005 ) provide further evidence for synchrony mediated purely by hydrodynamic coupling . Motivated by Rothschild's observation , Taylor ( Taylor , 1951 ) developed a mathematical model in which two laterally infinite , inextensible sheets with prescribed sinusoidal travelling waves of transverse deformation interact with each other through a viscous fluid . He found that the rate of viscous dissipation is minimised when the two sheets are in phase . While minimisation of dissipation often holds in real physical systems , it is not in general a fundamental principle from which to deduce dynamical processes . Rather , an explanation for synchronization should capture the forces and torques associated with the underlying molecular motors that drive flagella , their elasticity , as well as the viscosity of the surrounding fluid . Since Taylor's work a myriad of increasingly complex models of flagellar synchronization have been proposed . Hydrodynamically coupled filaments or chains with various internal driving forces exhibit a general tendency towards synchrony ( Machin , 1963; Gueron et al . , 1997; Guirao and Joanny , 2007; Yang et al . , 2008; Elgeti and Gompper , 2013 ) . At the same time , minimal models of coupled oscillators in viscous fluids ( Vilfan and Jülicher , 2006; Niedermayer et al . , 2008; Uchida and Golestanian , 2011 , 2012; Brumley et al . , 2012 ) offer great insight into the emergence of metachronal coordination . Such models have been investigated experimentally with light driven microrotors ( Di Leonardo et al . , 2012 ) , rotating paddles ( Qian et al . , 2009 ) and colloids in optical tweezers ( Kotar et al . , 2010 ) , and have also given rise to interpretations of the synchrony and coupling interactions between pairs of flagella of the model alga Chlamydomonas ( Goldstein et al . , 2009 ) . Although experimentally-derived coupling strengths between micropipette-held Chlamydomonas flagella are consistent with predictions based on direct hydrodynamic coupling ( Goldstein et al . , 2011 ) , it has been proposed ( Friedrich and Jülicher , 2012; Geyer et al . , 2013 ) instead that this coupling is too weak to overcome noise , and that residual motion of elastically-clamped cells could play a role in synchronization . The recent observation ( Leptos et al . , 2013 ) of antiphase synchronization in a non-phototactic mutant of Chlamydomonas points as well to the possible role of internal mechanical coupling between flagella . Clearly , examining the synchronization between flagella on a single cell it is difficult to establish with certainty the origins of the coupling mechanism due to the likely presence of biochemical and elastic couplings of as yet unquantified strength between flagella . In order to disentangle the hydrodynamic from the intracellular contributions to flagellar synchronization we conducted a series of experiments in which two physically separated flagellated cells , which exhibit distinct intrinsic beating frequencies in isolation , are coupled solely and directly through the surrounding fluid . These experiments can be viewed as natural generalisations of earlier work in which vibrating microneedles ( Okuno and Hiramoto , 1976 ) or micropipettes ( Eshel and Gibbons , 1989 ) are used to modulate and entrain the beating of a single sperm flagellum . Owing to the natural distribution of beating frequencies of the flagella of its surface somatic cells , the colonial alga Volvox carteri is ideally suited to this purpose . Each somatic cell possesses two flagella which beat in perfect synchrony , facilitating their treatment as a single entity , henceforth referred to as the flagellum . Somatic cells were isolated from adult Volvox colonies and held with micropipettes at a controllable separation d ( Figure 1A , B ) . The spatial and orientational degrees of freedom associated with this configuration enabled comprehensive analysis over a wide range of hydrodynamic coupling strengths . We found that closely-separated pairs of cells can exhibit robust phase-locking for thousands of beats at a time , despite a discrepancy in their intrinsic frequencies of as much as 10% . Both in-phase and antiphase configurations were observed , depending on the alignment of the directions of flagellar propulsion . Furthermore , with increasing interflagellar spacing we observed for each flagellum a marked change in the beating waveform , a key finding that lends support to models of synchronization that rely on waveform compliance to achieve phase-locking . 10 . 7554/eLife . 02750 . 003Figure 1 . Synchronized pairs of beating flagella . ( A ) Experimental apparatus and ( B ) cell configuration . ( C ) Extracted phase difference Δ= ( ϕ1−ϕ2 ) /2π at four different interflagellar spacings , as indicated . These separations correspond to scaled spacing L = d/l of 0 . 85 , 1 . 22 , 1 . 69 , and 2 . 27 . ( D ) fluctuations during phase-locked periods around the average phase lag , Δ0 , and ( E ) the fluctuations’ probability distribution functions ( PDFs ) , each cast in terms of the rescaled separation-specific variable ( Δ − Δ0 ) /√L . Solid lines represent Gaussian fits . Further details of the phase extraction procedure can be found in Figure 1—figure supplement 1 . Samples of the four processed videos corresponding to the cells in Figure 1C are shown in Video 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 00310 . 7554/eLife . 02750 . 004Figure 1—figure supplement 1 . Phase extraction . ( A–C ) Snapshots from three different processed videos showing the same cells at different interflagellar spacings ( d = 14 . 2 , 20 . 4 , 28 . 2 μm ) . ( D–F ) The signal extracted from the interrogation regions is used to reconstruct the flagellar phases . For each row , the frame on the left corresponds to the time indicated by the dashed line . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 004 We begin by characterising the flow generated by a single beating flagellum . Despite the fact that a flagellum is a spatially-extended object with considerable internal dynamics , it has become clear in recent years that the flow fields generated by its beating may be described , on suitable length scales , in terms of geometrically simpler force distributions . In the simplest case , often used in models of synchronization ( Vilfan and Jülicher , 2006; Niedermayer et al . , 2008; Uchida and Golestanian , 2011 ) , that would be just a single sphere tracing out a closed orbit in space under the action of internal driving forces . The time-averaged flow around the flagellum would then be approximated by the flow from a point force at a suitable average location . A point force F exerted on a viscous fluid at a location x0 produces a velocity field , known as a Stokeslet , of the form ( Blake and Chwang , 1973 ) ui=Fj/8πμr ( δij+rirj/r2 ) , where the vector r = x − x0 , r=\|r\| and δij is the Kronecker delta . A recent study ( Drescher et al . , 2010 ) of freely swimming Chlamydomonas cells has shown that the time-averaged flow field is consistent in its magnitude and topology with a three-Stokeslet model ( one for each flagellum and one for the cell body ) . In our experiments , all flow fields were obtained using particle image velocimetry ( PIV ) ( Raffel et al . , 2007 ) . Figure 2A shows the instantaneous flow field at four different times near a single cell , and it is clear that the magnitude ( colour ) and direction ( vector field ) of the flow vary during the cycle , as expected from the distinct power and recovery strokes . Examining the time-averaged velocity field ( Figure 2B , obtained by averaging data from four cells ) , we see that for distances larger than 20 μm from the flagellar tip , both upstream ( red ) and downstream ( blue ) components of the flow obey a Stokeslet decay ( u ∼ 1/r ) ( Figure 2C ) . This trend is maintained over a range consistent with the distances sampled for our two-cell experiments ( below ) . 10 . 7554/eLife . 02750 . 005Figure 2 . Measured flagellar flow field . ( A ) Time-dependent flow field for an individual cell measured using particle image velocimetry . Results are shown for the first half of the beating cycle . ( B ) Time-averaged flow field 〈u〉t= ( 1/τ ) ∫0τ\|u ( x , t ) \|dt ( averaged across 4 cells with τ ∼ 1000 beats for each ) . The velocity magnitude ( colour ) and streamlines ( white ) are shown . ( C ) Velocity magnitude upstream ( red ) and downstream ( blue ) of the origin ( black dot in B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 005 Let us now examine more closely the time-dependent flows . Figure 3A and Figure 3—figure supplement 1 show a fit of the instantaneous flow fields of each frame to a Stokeslet form , using the position x0 and the magnitude and direction F as fitting parameters . The results of this procedure are illustrated in Figure 3A as the average trajectory x¯0 ( t ) ( the closed white curve ) executed by the Stokeslet over approximately 103 beats . Figure 3B shows x¯0 ( t ) ( solid red line ) together with a scatter plot of x0 ( t ) from individual frames ( red dots ) . The black arrows along the average cycle illustrate the position , orientation and relative magnitude of the Stokeslet at evenly-spaced times along the average cycle . Importantly , the orientation of the point force does not coincide with its direction of motion , a feature to be expected given the anisotropic drag on the flagellum . Figure 3C shows the magnitude of the fitted Stokeslet over all beats . The amplitude of this force exhibits very strong periodic variations , and is approximated by F ( t ) /8πμ=A0 ( 1+A1sin ( 2πt/T ) ) with A0 ≃ 1076 μm2/s and A1 ≃ 0 . 56 . 10 . 7554/eLife . 02750 . 006Figure 3 . Force amplitude of flagellum . ( A ) Fitted instantaneous velocity field at various stages during the first half of one representative flagellar beat . ( B ) The fitted Stokeslet is shown at evenly-spaced times throughout the average flagellar beat cycle . The red dots indicate the Stokeslet position extracted from every frame . ( C ) Amplitude of the fitted point force as a function of time throughout the flagellar beat period T . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 00610 . 7554/eLife . 02750 . 007Figure 3—figure supplement 1 . Time-dependent flow fields . Instantaneous fluid velocity corresponding to various stages during one representative flagellar beat . Also shown are the fitted flow fields for each frame , corresponding to the application of a point force on the fluid . Very good qualitative agreement can be seen . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 007 This determination of the magnitude of the effective Stokeslet describing the flow field around a cell can be compared with an estimate based on the observed motion of the flagellum itself . Figure 4A shows snapshots of a typical flagellum captured over a full beat cycle , superimposed at 2 ms intervals , together with measured instantaneous velocities along the filament . With resistive force theory ( RFT ) , the results of the tracking procedure are used to derive estimates for the forces produced by the flagellum . First proposed by Gray and Hancock ( 1955 ) , RFT considers the anisotropic drag experienced by a long rod-like flagellum moving through a viscous fluid , and assumes that each unit segment of the flagellum experiences a local drag that is proportional to its local instantaneous velocity . The force density f along the flagellum is approximated by ( 1 ) f=C⊥u⊥+C‖u‖ , which is readily computable from experimental data , where the constants of proportionality , C⊥ and C‖ , are the normal and tangential resistance coefficients respectively . We chose C⊥ and C‖ according to the classic model of Lighthill ( 1975 ) , with C⊥=4πμ/ ( ln ( 0 . 18λ/a ) +0 . 5 ) and C‖=2πμ/ ( ln ( 0 . 18λ/a ) −0 . 5 ) , with an aspect ratio λ/a = 80 . 10 . 7554/eLife . 02750 . 008Figure 4 . Resistive force theory analysis . ( A ) Instantaneous velocity distribution along the flagellum during one complete beat cycle ( indexed by frame number , imaged at 1000 fps ) . ( B ) Components of integrated force density produced by a flagellum executing characteristic power and recovery strokes , as a function of arclength along the flagellum measured from the basal to the distal end . ( C ) Integrated vector forces F ( t ) shown localised at centre-of-mass coordinates x ( t ) ( red: per frame , black: averaged over O ( 103 ) frames ) , evolve cyclically around an average trajectory . The average value is \|F\|/8πμ ∼ 1910 μm2/s . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 008 The total instantaneous force F ( t ) produced by the flagellum is given by ∫0lf ( s , t ) ds , where l is the total length of the flagellum and s its arclength parameterisation . In Figure 4B we plot the normal ( blue ) and tangential ( red ) components of f , for characteristic power and recovery stroke waveforms ( solid and dotted lines respectively ) . To construct a limit cycle representation of the cyclic force variation , we define an effective centre-of-mass for the flagellum , x ( t ) =∑iN ( \|fi\|xi ( t ) ) /∑iN ( \|fi\| ) , averaging over all N discretised force vectors fi applied at points xi along the flagellum . Figure 4C depicts the trajectories of integrated force F in this coordinate representation ( red arrows ) . An average limit cycle representation ( black arrows ) is obtained from measurements taken from ∼100 beats: resultant force directions are seen to vary continuously along the cycle . The RFT result overestimates the force production during the recovery stroke , where the assumption of locality breaks down . It is encouraging to see that the amplitude of this force is similar to the value calculated earlier using Stokeslet fitting , though it should be noted that these results correspond to two different cells . To investigate the effect of hydrodynamic coupling on pairs of flagella , we captured pairs of cells and aligned them so that their flagellar beating planes coincided ( Figure 1A ) . Videos of hydrodynamically interacting flagella were first processed by subtracting a 30 frame running average . Median filtering was undertaken using 3 × 3 pixels2 regions . At each cell–cell separation d , we recorded flagellar dynamics over ∼100 s , and extracted flagellar phases ϕ1 , 2 from Poincaré sectioning of the dynamics ( Goldstein et al . , 2009; Polin et al . , 2009 ) by monitoring the signal in respective interrogation regions ( Figure 1B , Figure 1—figure supplement 1 ) , so that the respective flagella passed through precisely once per beat . Recording the passage times between beats allowed reconstruction of the flagellar phase ϕ1 , 2 . The time-dependent interflagellar phase difference Δ ( t ) = ( ϕ1−ϕ2 ) /2π was used to characterise the synchronization properties of the two cells . The measured phase difference Δ ( t ) is shown in Figure 1C for one pair of cells at four different spacings ( see Video 1 ) . We measured beat frequencies ω1 and ω2 for the two flagella in isolation , and define δω=ω1−ω2 to be their intrinsic frequency difference . Calling L = d/l the cell–cell separation normalised by the average flagellar length l of each pair , Figure 1C shows that for L≳2 hydrodynamic coupling is negligible and Δ ( t ) drifts approximately linearly with time depending on δω/ω ( 8 . 2% here ) . For intermediate values of L , the flagella exhibit short periods of synchrony interrupted by brief phase slips . However , when the same cells are brought closer to each other , they phase-lock for the entire duration of the experiment . This conclusively demonstrates that robust and extended flagellar synchronization can arise in physically separated cells purely through the action of hydrodynamics . For different pairs of cells ( n = 11 ) , a similar behaviour is observed . 10 . 7554/eLife . 02750 . 009Video 1 . A pair of hydrodynamically coupled flagella , observed at various cell–cell spacings . Original videos were recorded at 1000 fps , with processed representative segments ( 1000 frames each ) replayed here at 25 fps . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 009 Next we examine in detail the experimental time series Δ ( t ) . Consider first the synchronous periods within the full time series of Figure 1C . Fluctuations about the phase-locked states Δ0 ( Figure 1D ) are Gaussian with a variance proportional to L , as seen by rescaling as ( Δ−Δ0 ) /L1/2 ( Figure 1E ) . Gaussian fluctuations suggest a description of the dynamics of Δ ( t ) based on a Langevin equation with an effective potential V ( Δ ) having a quadratic minimum at Δ0 . We then write ( 2 ) Δ˙=−ν′ ( Δ ) +ξ ( t ) , where ν ( Δ ) =−δνΔ+U ( Δ ) . The quantity δν is the intrinsic frequency difference for the two phase oscillators , U an effective potential which has period one in Δ , and ξ ( t ) is a Gaussian white noise term satisfying 〈ξ ( t ) 〉=0 and 〈ξ ( t ) ξ ( t′ ) 〉=2Teffδ ( t−t′ ) , where Teff is an effective ‘temperature’ . To leading order U=−ϵcos ( 2πΔ ) , where ε is the interflagellar coupling strength . The observed dependence on L of the distribution of Δ fluctuations is a natural consequence of Equation 2 if ϵ∝1/L . We test this scaling below . Intraflagellar biochemical noise leads to stochastic transitions between adjacent minima of the tilted washboard potential ν ( Δ ) ( Goldstein et al . , 2009; Polin et al . , 2009 ) . For each video , the autocorrelation of Δ is used to extract the model parameters ( ϵ , δν , Teff ) as described previously ( Goldstein et al . , 2009; Polin et al . , 2009 ) . Cells aligned so that their power strokes point in the same direction ( as in many ciliates ) exhibit in-phase ( IP ) synchrony ( Δ0 ≃ 0 ) , indicating a coupling strength ϵ>0 . Rotation of pipette P1 ( Figure 1B ) by 180∘ so that the power strokes are opposed ( as in the Chlamydomonas breaststroke ) changes the sign of the coupling strength and gives rise to antiphase ( AP ) synchronization ( Δ0 ≃ 1/2 ) , in agreement with theory ( Leptos et al . , 2013 ) . Figure 5A depicts the nondimensionalised coupling strength κ=ϵ/ω¯ for all experiments , where ω¯ is the average beat frequency across all experiments for a given pair of cells . The dependence on the interflagellar spacing \|κ\|∝L−1 is consistent with the intrinsic flagellar flow field presented in Figure 2 . For both the in-phase and antiphase configurations , we fit \|κ\|=k×L−1 finding kIP = 0 . 016 and kAP = 0 . 014 respectively . At a given L , IP pairs exhibit on average a marginally stronger coupling than AP ones , possibly due to the fact that flagella in IP are on average closer together than in AP . The average values of the other model parameters are 〈Teff/ω¯〉=0 . 005±0 . 003 and 〈δν/ω¯〉=0 . 058±0 . 033 , with 〈ω¯〉=33 . 0 Hz . As a cross-check , we can estimate directly the effective internal noise from the distribution of beating periods of separated cells , and find 〈Teff/ω¯〉=0 . 002 , consistent with the value above . 10 . 7554/eLife . 02750 . 010Figure 5 . Coupling strength . ( A ) Dimensionless interflagellar coupling strength κ=ϵ/ω¯ as a function of the scaled spacing L = d/l ( log–log scale ) . The dotted lines represent fits of the form \|κ\|=k×L−1 with k = 0 . 016 ( in-phase ) and k = 0 . 014 ( antiphase ) . ( B ) Measured beat frequency ω/ω¯cell of each flagellum , nondimensionalised by the average value for that cell across several videos . ( C ) Measured frequency difference δω/δωfar as a function of spacing L . The curves represent the predictions based on the average extracted model parameters in the absence ( orange ) and presence of noise ( green ) . Symbols represent different pairs of cells , with the in-phase ( blue ) and antiphase ( red ) configurations shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 010 The average measured flagellar frequency ω for the two cells in each experiment is shown in Figure 5B , nondimensionalised by the average value for each cell ω¯cell across different spacings . Figure 5C illustrates the measured frequency difference as a function of L . The data exhibit an apparent bifurcation near L = 1 , beyond which phase drifting occurs over time . Integration of Equation 2 in the absence of noise yields a predicted value for the observed frequency difference in terms of the model parameters: δω/δωfar=1− ( 2πϵ/δν ) 2 , for ϵ ( L ) <δν/2π and δω/δωfar=0 otherwise . The orange curve in Figure 5C illustrates this prediction , calculated using the average extracted model parameters . In the presence of noise this sharp bifurcation becomes rounded and shifted ( Risken , 1989 ) , as shown in green in Figure 5C . It is evident that noise plays an important role in determining the observed location of the bifurcation point . Although coupling is established purely through hydrodynamic interactions , the process of synchronization hinges on the ability of the flagella to respond differentially to varying external flows . For sufficiently strong coupling , different cells can adopt a common phase-locked frequency through perturbing one another from their intrinsic limit cycles . Indeed , models of coupled flagella involving hydrodynamically coupled semiflexible filaments ( Gueron et al . , 1997; Guirao and Joanny , 2007; Elgeti and Gompper , 2013; Yang et al . , 2008 ) show a tendency towards metachronal coordination , though the precise role that flexibility plays in facilitating synchrony is unknown . Minimal models in which spheres are driven along flexible trajectories ( Brumley et al . , 2012; Niedermayer et al . , 2008 ) reveal that deformation-induced changes in the phase speed can facilitate synchrony . However , functional variations in the intrinsic flagellar driving forces could lead to synchrony even for fixed beating patterns ( Uchida and Golestanian , 2011 , 2012 ) . Through dynamic tracking ( Wan et al . , 2014 ) , we followed the evolution of the flagellar waveforms for several thousand consecutive beats . One example is shown in Figure 6A where the extracted waveform is shown at various stages through the beating cycle , overlaid onto logarithmically-scaled residence time plots . The same pair of flagella is compared at close and far cell–cell separations ( 7 . 3 μm and 72 . 2 μm respectively ) . In order to characterise flagellar waveform changes as the cells are brought closer together , we define three angles xa , xb , xc ( radians ) with respect to the cell body axis ( Figure 6B ) . Figure 6C shows the temporal evolution of these angles for the right flagellum , corresponding to the close ( red ) , intermediate ( green ) and wide ( blue ) separations . In particular , the most significant difference is observed in the xc component ( distal part of the flagellum ) . Similar results are found for the other cell , indicating that the interaction is mutual . Figure 6—figure supplement 1 and Figure 6—figure supplement 2 illustrate the robustness of these results for multiple cells and different configurations . Taken together , the results in Figure 6 demonstrate that accompanying the robust hydrodynamic phase-locking is a change in the flagellar waveform . For the first time , we have shown by systematically varying the cell–cell spacing that each flagellum can directly alter the beating profile of its neighbour simply through hydrodynamic interactions . 10 . 7554/eLife . 02750 . 011Figure 6 . Waveform characteristics . ( A ) Logarithmically-scaled residence time plots of the entire flagella . The displayed waveforms correspond to 1 ms time intervals over several successive flagellar beats . ( B ) Angles xa , xb , xc ( in radians ) measured and ( C ) their characteristic 3D trajectories . Results are shown for the right flagellum , corresponding to three different interflagellar spacings . As the spacing d is increased , the flagellar waveform exhibits a systematic change . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 01110 . 7554/eLife . 02750 . 012Figure 6—figure supplement 1 . Flagellar filaments are tracked for cells in the ( A ) antiphase state , as well as ( B ) the situation in which one of the cells does not possess a flagellum ( dummy cell ) . For each configuration , the waveform of the left cell is analysed at three different cell–cell separations . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 01210 . 7554/eLife . 02750 . 013Figure 6—figure supplement 2 . Additional waveform data collected for 5 different cells in various geometric configurations . The colours RGB correspond to increasing inflagellar spacing respectively , with distances ( in micrometres ) given by ( A ) {7 . 3 , 10 . 9 , 72 . 2} , ( B ) {6 . 5 , 11 . 0 , 27 . 1} , ( C ) {9 . 2 , 14 . 2 , 34 . 7} , ( D ) {10 . 3 , 31 . 0 , 84 . 0} , ( E ) {12 . 1 , 16 . 2 , 67 . 3} . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 013 The experimental study presented in this article reveals unambiguously the importance of hydrodynamics in achieving flagellar synchronization . Physical separation of the cells precludes any form of chemical or direct mechanical coupling , leaving hydrodynamic interactions as the only mechanism through which synchronization can occur . The process of phase-locking is extremely robust , with cells sufficiently close to one another exhibiting uninterrupted synchrony for thousands of consecutive beats . Accompanying this synchrony is a characteristic shift in the flagellar waveform . The extracted interflagellar coupling strength is consistent with hydrodynamic predictions and the measured flow fields generated by individual flagella . Additional experiments were undertaken using a uniflagellar mutant of the unicellular alga Chlamydomonas . Although its flagellum is shorter and its waveform is different to that of Volvox , we also observed hydrodynamic phase-locking in these experiments . Owing to the ubiquity and uniformity in the structure and function of flagella in various eukaryotic species , the results of the present study are expected to generalise to other systems , and may be of significant value for a wide range of theoretical models . Volvox carteri f . nagariensis ( strain EVE ) were grown axenically in Standard Volvox Medium ( SVM ) ( Kirk and Kirk , 1983 ) with sterile air bubbling , in a growth chamber ( Binder , Germany ) set to a cycle of 16 hr light ( 100 μEm−2s−1 , Fluora , OSRAM ) at 28∘C and 8 hr dark at 26∘C . Individual biflagellate cells were extracted from Volvox colonies using a cell homogeniser , isolated by centrifugation with Percoll ( Fisher , UK ) , and inserted into a 25 × 25 × 5 mm glass observation chamber filled with fresh SVM . Cells were captured using micropipettes and oriented so that their flagellar beating planes coincided with the focal plane of a Nikon TE2000-U inverted microscope . Motorised micromanipulators ( Patchstar , Scientifica , UK ) and custom-made stages facilitated accurate rotation and translation of the cells . The flow field characterisation and pairwise synchronization analyses were imaged using a 40× Plan Fluor objective lens ( NA 0 . 6 ) . A higher magnification 63× Zeiss W Plan-Apochromat objective lens ( NA 1 . 0 ) was used to conduct separate experiments for the waveform analysis . For each experiment , we recorded videos with a high-speed video camera ( Fastcam SA3 , Photron , USA ) at 1000 fps under bright field illumination . The stochastic Adler equation was used to model the dynamics of Δ ( t ) as described in Goldstein et al . ( 2011 ) . Figure 7A , B show the amplitude C0 of the autocorrelation function of Δ and the values of the average synchronous period τsync . Fluctuations of the phase difference Δ about the synchronized states are well described by Gaussian distributions , with variances C0 proportional to the interflagellar spacing L . The coupling strength ε exhibits excellent agreement with the hydrodynamic predictions . Figure 7C , D show the dependence of the effective temperature Teff/ω¯ and intrinsic frequency difference δν/ω¯ as a function of L = d/l for every pair of flagella measured . In order to study the dynamics of hydrodynamically coupled flagella , the two cells were held using orthogonally-positioned glass pipettes . This geometry allowed us to investigate both in-phase and antiphase configurations for the same pair of cells , through the simple rotation of one pipette . At the same time , however , this meant that the two cells were held from different directions with respect to their flagella , and that one of the two pipettes was oriented along the direction of the flagellar power stroke , which is the main flow direction . This can cause two problems . Firstly , the flow field of a cell held by the side could be significantly different from that presented in Figure 2 . Secondly , the holding pipettes could distort the scaling of the flagellar flow with cell–cell separation from the ∼1/r scaling presented in Figure 2C . We investigated these problems with the series of experiments shown in Figure 8 . One cell was held at its posterior pole by a pipette ( Figure 8A ) and the flow field measured . A second micropipette was then moved progressively closer , eventually to the point of contact with the cell ( Figure 8D ) . It is clear that the second pipette affects the flow , but mostly in the region between the two pipettes . 10 . 7554/eLife . 02750 . 016Figure 8 . Effect of nearby pipette . The time-averaged flow field associated with one captured cell is measured as a second pipette slowly approaches . This demonstrates that the precise angle from which the cell is held by the micropipette has very little effect on the resultant flow field . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 016 Let us consider the region upstream of the cell ( above the cell in Figure 8 ) . For a cell held from the side , this is the region where the other cell will be . Here the flow is only minimally affected , with an average relative change between Figure 8D , A below 8% . A large contribution is represented simply by a ∼7% decrease in flow speed . Taking this decrease into account , the average relative change is about 5% . As a result , these experiments allow us to consider the flow generated by a cell held from the side as identical to that generated by a cell held from the back , at least in the region of interest to our experiments . By comparing the flows for different positions of the second pipette , we can also quantify its effect on the flow field that would be experienced by the second cell . For each configuration of pipettes , this can be estimated as the relative difference between the unperturbed and the perturbed flows in the region where the flagella of the second cell would be , here considered to be a 10 × 10 μm2 region 20 μm to the left of the tip of the incoming pipette . The difference ranges from ∼5% to ∼10% and ∼13% for Figure 8B–D respectively ( in the last case we choose a position approximately 10 μm below and 20 μm to the left of the pipette tip ) . These represent the typical error contributions from neglecting , as we have done in the text , the influence of the pipettes on the flows generated by the cells . We used the Stokeslet approximation to the flow field of an isolated cell in Figure 3 , to test the effect of force modulation on synchronization within the class of minimal models which abstract the beating flagellum as a sphere driven along a closed orbit ( Niedermayer et al . , 2008; Uchida and Golestanian , 2011 , 2012 ) . We simulated two spheres of radius a = 0 . 75 μm in an unbounded fluid of viscosity μ = 10−3 Pa·s , driven along coplanar circular orbits of radius r0 = 8 μm by a force F ( ϕ ) /8πμ=A0 ( 1+A1sin ( νϕ+ϕ0 ) ) tangential to the orbit , with A0 = 1076 μm2/s and A1 = 0 . 56 . Notice that this corresponds to assuming that the point forces in Figure 3B are tangential to the cycle . The value a = 0 . 75 μm was chosen to ensure that the orbital frequency matched the mean value observed experimentally . The orbits were separated by d = 20 μm and had a radial stiffness with spring constant λ . The limit λ→∞ corresponds to rigid prescribed trajectories ( holonomic constraint ) . For each value of λ∈{1 pN/μm , 5 pN/μm , ∞} we ran five sets of simulations , corresponding to ν∈{0 , 1 , 2} and ϕ0∈{0 , π/2} . Choosing ν = 2 is equivalent to modulating the driving force with the experimental amplitude but at a frequency double the experimental one . Although this is not what we observed , it is still interesting to consider , since in this configuration it is the frequency that contributes most to synchronization through force modulation . As a consequence of the phase-dependent driving force , the geometric phase ϕi of an individual isolated oscillator does not evolve at a constant rate in time . We thus chose to rescale the phase Φ=Φ ( ϕ ) so that in the absence of hydrodynamic interactions , Φ˙=2π/T=constant . Both the geometric phase difference δ=ϕ1−ϕ2 ( thin curves ) and its rescaled value δrescaled=Φ1−Φ2 ( thick curves ) are shown for each simulation in Figure 9 . These results show clearly that within the boundary of the model we are considering , the two oscillators synchronize through a coupling between hydrodynamic stresses and orbit compliance ( Niedermayer et al . , 2008 ) with no noticeable effect from force modulation . 10 . 7554/eLife . 02750 . 017Figure 9 . Effect of force modulation . Evolution of the phase difference δ=ϕ1−ϕ2 among two identical model oscillators , each composed of a sphere driven around a circular trajectory by a tangential driving force . The trajectories each possess a radial stiffness λ . Smaller values of λ yield rapid convergence towards synchrony ( δ = 0 ) , in a manner essentially independent of the functional form of the driving force . Parameters used are given by a = 0 . 75 μm , r0 = 8 μm , d = 20 μm , A0 = 1076 μm2/s and A1 = 0 . 56 . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 017 Repeating the simulations with a stiffness derived from the flagellar bending rigidity as in the main text , λ = 0 . 05 pN/μm , radius a = 0 . 1 μm , and reducing the force amplitude to A0 = 143 μm2/s to keep the revolution frequency at the experimental value , yields the results in Figure 10 . Again , the synchronization is achieved only through interaction between hydrodynamic stresses and orbit compliance . 10 . 7554/eLife . 02750 . 018Figure 10 . Effect of force modulation . Re-run of the simulations in Figure 9 with properties inspired by real flagella . DOI: http://dx . doi . org/10 . 7554/eLife . 02750 . 018	Sperm cells , as well as many bacteria and algae , propel themselves using whip-like appendages called flagella . Similar , shorter structures called cilia are also found on the surface of many cells , where they perform roles such as moving liquids over the cell . Each cilium or flagellum beats at its own characteristic rhythm , but there are many situations where cilia or flagella must synchronize their beating with other nearby cells . For example , an egg cell is swept along the Fallopian tube by the coordinated beating of the cilia lining the tube . Bull sperm cells are also known to synchronize the beating of their flagella when swimming close to each other . It has been suggested that the movement of the fluid surrounding the beating flagella could be the source of this synchronization . Experiments have produced results that match up with mathematical models describing this fluid movement . However , these experiments have often been designed in ways that didn’t fully exclude other possible sources of synchronization , such as chemical signalling , or—for flagella located on the same cell—a physical connection between the flagella . To overcome this shortcoming , Brumley et al . used high-speed imaging to watch the flagella of cells of Volvox carteri—a species of green alga—that were separated so that they could only communicate through the movement of the fluid around them . The flagella were still able to synchronize their beating , even when the two flagella naturally beat at substantially different rates . The distance between the flagella affects how well the beating synchronizes . When close together , the flagella can lock into the same rhythm for thousands of beats . However , as they move further apart , random biochemical fluctuations within the cells reduce the extent to which the flagella can synchronize . The flagella can also synchronize so that they move in the same direction at the same time , or in opposite directions , depending on how they are oriented relative to each other . Moreover , the results confirm that the fluid flow produced by a beating flagellum is sufficient to synchronize the beating of other nearby flagella .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "physics", "of", "living", "systems" ]	2014	Flagellar synchronization through direct hydrodynamic interactions
Cells use phagocytosis and macropinocytosis to internalise bulk material , which in phagotrophic organisms supplies the nutrients necessary for growth . Wildtype Dictyostelium amoebae feed on bacteria , but for decades laboratory work has relied on axenic mutants that can also grow on liquid media . We used forward genetics to identify the causative gene underlying this phenotype . This gene encodes the RasGAP Neurofibromin ( NF1 ) . Loss of NF1 enables axenic growth by increasing fluid uptake . Mutants form outsized macropinosomes which are promoted by greater Ras and PI3K activity at sites of endocytosis . Relatedly , NF1 mutants can ingest larger-than-normal particles using phagocytosis . An NF1 reporter is recruited to nascent macropinosomes , suggesting that NF1 limits their size by locally inhibiting Ras signalling . Our results link NF1 with macropinocytosis and phagocytosis for the first time , and we propose that NF1 evolved in early phagotrophs to spatially modulate Ras activity , thereby constraining and shaping their feeding structures . Phagotrophic cells feed by performing large-scale endocytosis . A wide range of unicellular eukaryotes grow in this way , suggesting that it is extremely old in evolutionary terms ( Stanier , 1970; Cavalier-Smith , 2002; Yutin et al . , 2009 ) . Typically phagocytosis is used by these organisms to engulf solid particles ( Metchnikoff , 1892 ) , and nutrients are then extracted from them by lysosomal degradation ( De Duve and Wattiaux , 1966 ) . Animal cells and amoebae ingest solid material using F-actin driven projections of their plasma membrane , forming pseudopodia and ultimately cup- or crown-shaped ruffles that enclose adhered particles . These cells can also internalise bulk fluid without the guidance of a particle using a closely related process , macropinocytosis ( Swanson , 2008 ) . Phagocytosis and macropinocytosis are controlled using a large set of cytoskeletal and membrane-associated regulators , notably a variety of small G proteins ( Bar-Sagi and Feramisco , 1986; Ridley et al . , 1992; Peters et al . , 1995; Cox et al . , 1997; Martínez-Martín et al . , 2011 ) . Oncogenes such as Src and phosphatidylinositide 3′-kinase ( PI3K ) have also been linked with regulation of these processes ( Araki et al . , 1996; Veithen et al . , 1996; Buczynski et al . , 1997; Amyere et al . , 2000 ) . In amoebae , growth and endocytosis have obvious connections since phagocytosed material supplies essentially all their nutrients; in contrast vertebrates are specialised to digest food extracellularly in the gut , and so links are less apparent . However , large-scale endocytosis is extremely important in immune cells ( Metchnikoff , 1892; Norbury et al . , 1995; Sallusto et al . , 1995 ) , while tumour cells , released from the normal constraints on growth and proliferation , can display pronounced macropinocytotic or phagocytotic uptake ( Lewis , 1937; Montcourrier et al . , 1994 ) , and can feed by ingesting extracellular protein ( Commisso et al . , 2013 ) . While there are clear similarities between large-scale endocytosis in animal cells and amoebae , neither the regulatory architecture nor evolutionary contexts are adequately understood . Phagotrophic microorganisms can be difficult to study in the laboratory because of their requirement for other organisms as food . This can be overcome if cells can be cultured axenically ( ‘a-xenic’ indicating the absence of organisms of another species ) : in some cases , such as the social amoeba Dictyostelium , which feeds primarily on bacteria in the wild ( Vuillemin , 1903 ) , strains were gradually adapted to growth in complex liquid broth , and ultimately in chemically defined media ( Sussman and Sussman , 1967; Watts and Ashworth , 1970; Franke and Kessin , 1977; Watts , 1977 ) . This process involved the selection of mutants that display increased rates of macropinocytosis ( Watts and Ashworth , 1970; Loomis , 1971; Hacker et al . , 1997 ) . Two important mutations , axeA and axeB , were identified by linkage analysis as being necessary for robust axenic growth ( Williams et al . , 1974a , 1974b ) , but only the latter is strictly required ( Clarke and Kayman , 1987 ) . Although these axenic mutant strains have been very widely used for over 40 years , the genetic basis of their growth has remained mysterious , since the mutations could not be precisely mapped . We used a forward genetic approach to identify mutations that promote axenic growth in Dictyostelium discoideum using whole genome sequencing . We found that the Dictyostelium orthologue of the Ras GTPase activating protein ( RasGAP ) Neurofibromin ( NF1 ) , a tumour suppressor that is mutated in the genetic disorder Neurofibromatosis type 1 ( Xu et al . , 1990 ) , is a key regulator of both macropinocytosis and phagocytosis . To generate fresh axenic strains for sequencing , we cultured wildtype D . discoideum cells in HL5 growth medium after washing them free of food bacteria . This medium supports the growth of axenic strains such as Ax2 and AX4 , but wildtype cells arrest their growth and ultimately die . In order to minimize the number of irrelevant background mutations we avoided mutagenesis and found that spontaneous mutants that are able to grow and proliferate arise frequently among these growth-arrested populations . We selected several independent mutants and sequenced the genomes of three after clonal isolation , along with that of the parental DdB strain , which was chosen because it was also parent to the established axenic laboratory strains ( Bloomfield et al . , 2008 ) . At first , other than two large duplications that do not correlate with axenicity ( Figure 1—figure supplement 1 ) , we could only identify one mutation affecting coding sequence in any of these strains relative to their parent , a seven basepair deletion in strain HM559 ( Table 1 ) . We noted that the reference genome sequence ( Eichinger et al . , 2005 ) , derived from the axenic mutant strain AX4 , also differs from its parent DdB in the same gene model ( annotated as DDB_G0279251 ) . Further analysis demonstrated that AX4 has lost almost nine kilobases of this region on chromosome 3 , resulting in the deletion of most of the coding sequence of a large gene encoding a homologue of the Ras GTPase-activating protein ( RasGAP ) Neurofibromin ( NF1 ) , as well as part of the upstream gene ( Figure 1A ) , with a short segment of extraneous sequence inserted . The 7 bp deletion mutation in HM559 lies within the C-terminal region of this NF1 homologue , and we found that another established axenic mutant , Ax2 , has exactly the same deletion-insertion mutation as AX4 ( Figure 1—figure supplement 2; Table 1 ) . 10 . 7554/eLife . 04940 . 003Table 1 . Mutations in the axeB gene in Dictyostelium discoideum axenic mutantsDOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 003StrainMutationEffect on Dd NF1 proteinPosition in human NF1 proteinAx2c . -1954_6926delinsCM000150 . 2:1390060_1390808Deletion to amino acid 2309to 2358AX4c . -1954_6926delinsCM000150 . 2:1390060_1390808Deletion to amino acid 2309to 2358HM557c . 226_230delDeletion , frameshift66HM587c . 1015A > TNonsense315HM591c . 3033_3040delDeletion , frameshift∼1060 ( in insertion relative to Human ) NP73c . 3508delDeletion , frameshift1228HM589c . 4113G > TK > N1423HM590c . 4227_4459delDeletion , frameshift1461–1533HM558c . 6393_6413invDPVVSAIL > EELQKPND2182–2189HM586c . 6833_7077delDeletion , frameshift2325–2481HM559c . 7137_7143delDeletion , frameshift2525–2529Strains are described fully in Table 2 . Description of changes to the coding sequence of the axeB gene follow the recommendations of the Human Genome Variation Society ( den Dunnen and Antonarakis , 2000 ) ; the effect on the protein sequence is indicated , using the IUPAC one-letter code for amino-acid substitutions . All changes except one are predicted to inactivate the protein either through the introduction of premature stop codons or the substitution of a conserved residue known to be important for function in the human version of the protein . Approximate corresponding locations in the amino-acid sequence of the human orthologue are also indicated . 10 . 7554/eLife . 04940 . 004Figure 1 . Discovery of the D . discoideum axeB locus . ( A ) The region of chromosome 3 spanning the genes DDB_G0279751 and DDB_G0279753 in AX4 genome ( top line ) contains a conversion mutation in which almost 9 kilobases of sequence ( lower line ) were lost and replaced by sequence ( pale blue ) resembling a short region of chromosome 1 . The deleted segment contains most of the D . discoideum orthologue of NF1 , axeB ( brown ) . ( B ) NF1 knockout cells can grow in the standard axenic medium , HL5 . Amoebae of strains Ax2 , DdB ( WT ) , and HM1591 ( axeB , an engineered NF1 knockout strain in the DdB background; in this and subsequent figures , ‘axeB’ refers to this strain ) , were incubated in tissue culture plates in HL5 medium , and growth measured at indicated timepoints using a crystal-violet binding assay . See also Figure 1—figure supplements 1 , 2 . The AX4 reference genome is at dictyBase ( http://dictybase . org ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 00410 . 7554/eLife . 04940 . 005Figure 1—figure supplement 1 . Two new axenic mutant strains possess overlapping duplications on the same chromosome . The samtools ‘depth’ command was used to calculate the depth of coverage at each position along the chromosomes . A rolling median was obtained ( window size 999 ) to remove outliers , then each chromosome examined by sampling every 1000th position and plotting them sequentially using R ( www . r-project . org ) ; the ‘index’ in the plots refers to these 1000 nucleotide divisions . Contiguous segments with approximately double the average depth reflect likely duplication events . Only two such segments could be identified , on overlapping regions at one end of chromosome four in strains HM557 and HM558: shown here are the plots for this chromosome in all four strains resequenced . These duplications are large , spanning hundreds of kilobases and many genes , and it is possible that they contribute to these strains' growth phenotypes; this hypothesis remains to be tested . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 00510 . 7554/eLife . 04940 . 006Figure 1—figure supplement 2 . Two established axenic mutants possess identical complex mutations affecting the axeB gene . Genomic DNA isolated from the Kuspa laboratory stock AX4 and the Kay laboratory stock Ax2 was amplified using primers spanning the deletion-insertion mutation identified in the AX4 reference sequence , and sequence using a primer within the upstream gene . A black vertical line shows the 5′ boundary of the mutation; the boundaries and inserted sequence are identical . The mutation's effects on the parental DdB sequence are annotated in the sequence file deposited in the ENA database as HF565448 . Resequencing of these strains' genomes , to be described elsewhere , confirmed this result . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 006 Reanalysis of our sequencing data aligned against an amended reference containing this deleted region revealed that both of the other two new mutants also possess mutations in this gene: HM557 has a short frameshifting deletion , while HM558 has undergone an inversion leading to a substitution of eight consecutive amino acids in the predicted protein ( Table 1 ) . To examine how frequently this gene is mutated in axenic mutants , we amplified and sequenced it from six further strains: five more new mutants selected from the same parental DdB strain , and one from the V12 genetic background ( strains used in this study are listed in Table 2 ) . All possess mutations in the NF1 homologue ( Table 1 ) : four have frameshifting deletions , one a nonsense mutation , and one has a substitution of a conserved lysine to asparagine . 10 . 7554/eLife . 04940 . 007Table 2 . Strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 007StrainParentGenotypeReferenceAx2DdBaxeA2 axeB2 axeC2 ( Watts and Ashworth , 1970 ) AX4DdBaxeA1 axeB1 axeC1 ( Knecht et al . , 1986 ) DdBNC4Wildtype ( Bloomfield et al . , 2008 , as ‘DdB ( Wel ) ’ ) NP73V12axeB3 ( Williams , 1976 ) HM557DdBaxeB ( GB1 ) This studyHM558DdBaxeB ( GB2 ) This studyHM559DdBaxeB ( GB3 ) This studyHM586DdBaxeB ( GB4 ) This studyHM587DdBaxeB ( GB5 ) This studyHM589DdBaxeB ( GB6 ) This studyHM590DdBaxeB ( GB7 ) This studyHM591DdBaxeB ( GB8 ) This studyHM1591DdBaxeB ( GB9 ) neoRThis studyHM1709DdBnfaA ( GB1 ) hygRThis studyHM1710HM1591nfaA ( GB1 ) axeB ( GB9 ) neoR hygRThis studyThe generally accepted genotype of Ax2 and AX4 is given , although the true number of mutations contributing substantially to their fast axenic growth phenotype remains unknown . AX4 derives from another axenic mutant , AX3 ( or A3 ) , which was isolated from wildtype cells independently from Ax2 ( Loomis , 1971 ) . Extant AX3 and AX4 strains share a large inverted duplication on chromosome 2 ( Eichinger et al . , 2005 ) that is not present in Ax2 . However , the mutation in axeB in Ax2 and AX4 is identical , suggesting that the extant lines of these strains , along with AX3 , had a common ancestor that was axenic . It might not be possible to determine the reason for this discrepancy with the literature; one possibility is that very early in these strains' contemporaneous history one line was contaminated with the other and the slower-growing of the two then lost . In formally numbering alleles we have retrospectively assigned allele number ‘3’ to the axeB mutation in the historic strain NP73 , but follow recent recommendations ( http://dictybase . org/Dicty_Info/nomenclature_guidelines . html ) for new strains , and use the same number for gene disruptions using the same knockout construct . The ubiquity of mutations in the NF1 gene in the axenic strains tested suggested that they must underlie the phenotype we selected for , and the gene's location on chromosome 3 accords with the mapping of the classically defined axeB gene ( Williams et al . , 1974a , 1974b ) . To test whether inactivation of NF1 promotes axenic growth , we engineered a deletion at the locus in a wildtype strain , DdB , and found that the resulting mutant is able to grow axenically in HL5 medium ( Figure 1B ) . However , it grows more slowly than the established Ax2 strain ( Figure 1B ) , and does not grow well in suspension in this medium ( see below ) , confirming earlier findings that additional mutations are necessary to potentiate the basal axenic phenotype ( Williams et al . , 1974a , 1974b ) . Together , the identification of mutations in the original axenic strains on chromosome 3 and demonstration that inactivation of the affected gene results in a phenotype closely resembling axeB single mutants derived parasexually ( Clarke and Kayman , 1987 ) gives adequate reason to believe that we have identified the original causative mutation . We therefore formally retain the name axeB for the locus , while naming the encoded protein NF1 . The Dictyostelium NF1 gene encodes a protein with the same domain organisation as the human version , with CRAL/TRIO and PH-like domains at the C-terminal side of the catalytic RasGAP domain ( Figure 2A ) . It is also of a similar size , with homology extending across most of the two proteins' lengths ( Figure 2B ) . The D . discoideum NF1 orthologue is about as similar to the human protein as are those from the basal metazoa and choanoflagellates ( Figure 2C ) . NF1 is an ancient protein , conserved considerably beyond the metazoan and fungal lineages in which it has been studied to date , with homologues in a variety of unicellular eukaryotes including the excavates Naegleria and Trichomonas as well as other amoebae ( Figure 2C , D and Figure 2—figure supplement 1; Carlton et al . , 2007; Fritz-Laylin et al . , 2010; Clarke et al . , 2013 ) . RasGAPs are more broadly distributed than NF1 , being present in further excavates as well as in certain ciliates , oomycetes , and the foraminiferan Reticulomyxa ( Figure 2—figure supplement 2 and Figure 2—source data 1; van Dam et al . , 2011; Glöckner et al . , 2014 ) . The dictyostelids , Entamoeba , Thecamonas , and Naegleria all possess separate smaller homologues with a similar domain organisation to NF1 but lacking homology outside of the central region; we term these proteins ‘MNF’ ( for ‘miniature neurofibromin’ ) . The D . discoideum NfaA protein ( Zhang et al . , 2008 ) falls into this class ( Figure 2A , D and Figure 2—figure supplement 1 ) , and is discussed further below . 10 . 7554/eLife . 04940 . 008Figure 2 . NF1 is broadly conserved in a range of amoeboid species as well as animals and fungi . ( A ) NF1 and related proteins have a characteristic domain organisation . The RasGAP domain and adjacent CRAL/TRIO and PH-like domains can be used to identify NF1-like proteins , although the PH-like domain is divergent . Approximate locations of mutations identified in axenic mutants are indicated with arrows; these are described precisely in Table 1 . ( B ) The D . discoideum ( Dd ) NF1 sequence shows homology to the Homo sapiens protein along its entire length: the sequence of the Hs protein was split into segments with a sliding window of 200 amino acids , and these globally aligned to the Dd , Takifugu rubripes , and Drosophila melanogaster NF1 orthologues , and the Saccharomyces cerevisiae Ira1p sequence . Dashed lines mark the outermost windows containing parts of the central domains . ( C ) NF1 protein sequences from Takifugu rubripes , Drosophila melanogaster , Trichinella spiralis , Trichoplax adhaerens , Salpingoeca rosetta , Capsaspora owczarzaki , Mortierella verticillata , Saccharomyces cerevisiae ( Ira1p ) , Dd , and Naegleria gruberi ( EFC40840 . 1 ) were globally aligned with the Homo sapiens NF1 sequence . The bars display the percentage similarity and identity of the protein to the human sequence . ( D ) Phylogram of NF1 and MNF homologues; the Dictyostelium AxeB protein is an NF1 homologue , while homologues of NfaA form the MNF class of RasGAP , defined here . The presence of NF1 and MNF in Naegleria and Thecamonas as well as amoebozoans indicates that MNF was ancestral and then lost in a common ancestor of the Holozoa and Holomycota after the divergence of apusozoans . The scale shows substitutions/site . See Figure 2—figure supplement 1 for a version with all species labelled , and also Figure 2—figure supplements 2 and Figure 2—source data 1 for illustration of the wider pattern of conservation of RasGAPs . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 00810 . 7554/eLife . 04940 . 009Figure 2—source data 1 . Examples of RasGAPs and NF1 orthologues in different lineages . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 00910 . 7554/eLife . 04940 . 010Figure 2—figure supplement 1 . Phylogram of NF1 and MNF homologues . This represents the same tree as Figure 2D , displayed rectilinearly instead of radially . Selected NF1 homologues from Metazoa and Fungi are included; outside of these taxa all identified homologues are included . The code used , aligned sequences and tree files have been deposited in FigShare ( with DOIs 10 . 6084/m9 . figshare . 1057805–808 ) . For NfaA-related proteins we suggest the name MNF for ‘miniature Neurofibromin’ to avoid confusion with the unrelated Naegleria Nfa1 protein ( Shin et al . , 2001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01010 . 7554/eLife . 04940 . 011Figure 2—figure supplement 2 . The presence of NF1 homologues and other RasGAPs in the three main eukaryotic supergroups . While the corticates evidently ancestrally possessed RasGAPs ( and Ras signalling ) , no NF1 homologues are detectable in the genomes of any presently available in the public databases . A previously discussed putative homologue in Stramenopiles has a START domain next to its RasGAP domain , not the unrelated CRAL/TRIO domain found in NF1 ( van Dam et al . , 2011 ) . The other two supergroups , podiates and excavates , both possess NF1 ( and MNF ) homologues; if the root of the eukaryotic tree lies between the podiates and either Naegleria or Trichomonas this implies that NF1 was present in the LECA . Examples of RasGAPs and NF1 orthologues in the lineages shown here are given in Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 011 The lysine to asparagine substitution occurring in one of our new axenic mutants ( see above ) has also been found in human cancer and Neurofibromatosis type 1 patients ( Li et al . , 1992 ) , and affects a lysine residue ( numbered 1423 in the human polypeptide ) that is located on the surface of the GAP domain where it contacts Ras , and that is essential for GAP activity ( Poullet et al . , 1994 ) . This strikingly underscores the homology inferred from sequence analysis , and furthermore suggests that overactivation of Ras subfamily small G proteins causes the axenic mutant phenotype . Established axenic mutants have very high rates of macropinocytosis ( Hacker et al . , 1997 ) . To examine fluid uptake in our NF1 knockout mutant , we incubated amoebae with fluorescent dextran and compared them with both the wildtype DdB strain and the established axenic mutant Ax2 . When they are harvested directly from bacterial growth plates , NF1 mutants ingest fluid at about the same rate as an established axenic strain , Ax2 , and more than four times more rapidly than wildtype cells ( Figure 3A ) . After prolonged incubation in HL5 medium without bacteria , Ax2 cells and NF1 mutants increase their uptake further while wildtype cells decrease it , such that after 24 hr mutants take in fluid at a rate more than twenty times higher than the wildtype ( Figure 3A ) . Fluid uptake is linear at the earliest times measured with no evidence for rapid recycling of fluid in any of these strains ( Figure 3—figure supplement 1 ) , in agreement with earlier studies of axenic mutants ( Aubry et al . , 1993; Padh et al . , 1993 ) . Membrane uptake measured using the accumulation of FM1-43 dye was not increased in NF1 mutants ( Figure 3—figure supplement 2 ) , consistent with an earlier comparison of Ax2 with wildtype cells ( Aguado-Velasco and Bretscher , 1999 ) . Since this assay predominantly measures uptake into small vesicles or tubules with a high surface to volume ratio , we conclude that clathrin-dependent and -independent micropinocytotic processes in these cells ( Neuhaus et al . , 2002; Hirst et al . , 2014 ) , are unaffected by NF1 loss . 10 . 7554/eLife . 04940 . 012Figure 3 . NF1 mutants grow axenically in HL5 medium and have increased fluid uptake . ( A ) NF1 knock-out mutants accumulate fluid more quickly than wildtypes . Fluid uptake was measured by shaking cells , either fresh from bacterial growth plates or after 24 hr incubation in axenic medium , with fluorescent dextran in buffer for 1 hr . ( B ) NF1 mutants accumulate fluorescent dextran in large endosomes , and exhibit a flattened phenotype compared to wildtypes . Cells were harvested from bacterial growth plates and incubated in Loflo medium plus TRITC-dextran for 30 min then imaged by confocal microscopy; cells' cytoplasm appears dark since no dextran penetrates it while endosomes are bright as their contents become concentrated . NF1 mutants tend to assume a flattened morphology; since only a single confocal section is shown this will tend to exaggerate the apparent number of endosomes per cell and so these images should not be relied on for comparison of cumulative fluid uptake . ( C ) NF1 knock-out mutants form macropinosomes more frequently than wildtypes , as assessed by confocal imaging . 15 cells of each strain were tracked in total in three independent experiments . Scale = 5 μm . Data points are the means of three independent experiments plus and minus the standard error . See also Figure 3—figure supplements 1–3 . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01210 . 7554/eLife . 04940 . 013Figure 3—figure supplement 1 . No evidence for fast recycling of ingested fluid . Although wildtype cells appear to have a lower rate of macropinocytosis as assessed by confocal microscopy , it was formally possible that their lower rate of fluid accumulation could be explained by a faster rate of fluid release , for instance by recycling of fluid from endosomes before fusion with lysosomes . To test this , we incubated cells in axenic medium overnight before performing a timecourse of FITC dextran uptake . Fast recycling would be revealed by a high initial rate of uptake followed by a more moderate rate as a portion of ingested dextran is expelled . We found no evidence for such recycling , although a small amount might occur . Means plus and minus standard errors are given for three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01310 . 7554/eLife . 04940 . 014Figure 3—figure supplement 2 . In contrast to fluid uptake , membrane uptake is not increased in NF1 mutants . DdB ( WT ) and the NF1 null mutant HM1591 ( axeB ) were harvested from bacterial growth plates , washed , and resuspended in KK2 buffer and shaken at room temperature for 15 min before being added a stirred fluorimeter cuvette containing FM1-43 dye . When cells are added , the fluorescence of the sample increases rapidly as dye enters the plasma membrane; as membrane is internalised , compensating fresh membrane is exposed to the surface enabling more dye to bind and fluoresce . While fluid uptake is several-fold higher in mutants , we found no evidence of increased membrane uptake rates , indicating that most membrane is taken up as small vesicles or narrow tubules with large surface-area to volume ratios . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01410 . 7554/eLife . 04940 . 015Figure 3—figure supplement 3 . Intracellular degradation of proteins occurs normally in NF1 mutants . To image degradation of internalized protein , cells of strains Ax2 , DdB , and HM1591 were incubated in Loflo medium plus 50 µg/ml DQ Green BSA , either ( A ) for 60 min for cells taken directly from bacterial growth , or ( B ) for 15 min for cells incubated for 24 hr in loflo medium before DQ Green BSA was added; cells were imaged for green fluorescence of degraded peptides using the same laser power and gain in each case . Scale = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 015 Confocal microscopy reveals striking differences between the NF1 knockout mutant and wildtype cells . When bathed in fluorescent dextran , NF1 mutants accumulate large dextran-filled endosomes more prominently than wildtypes ( Figure 3B ) , consistent with the fluorimetry of cell populations . NF1 mutants attached to glass coverslips also tend to flatten periodically , a phenotype often observed in established axenic strains such as Ax2 but not in wildtype cells ( Figure 3B ) . In four-dimensional timelapse imaging , when freshly harvested from bacteria mutants and wildtype cells can be observed to enclose large macropinosomes after projecting cup- or crown-shaped ruffles , with the mutant performing macropinocytosis in this way four times as frequently as wildtype cells ( Figure 3C ) . The NF1 mutation therefore accounts for the increased macropinocytotic fluid uptake of axenic strains . Despite taking in a similar amount of fluid , the NF1 knockout mutant grows more slowly than Ax2 ( see above ) . We examined whether the mutant processes ingested medium effectively by incubating cells with BODIPY-labelled bovine serum albumin ( DQ-BSA ) , which becomes fluorescent only after lysosomal degradation releases fluorophores that previously quenched each other . Ax2 and the NF1 null strain rapidly and comparably degrade protein after internalization by macropinocytosis and maturation of endosomes , ( Figure 3—figure supplement 3 ) . Wildtype DdB cells effectively degrade DQ-BSA when taken freshly from bacterial growth but not after overnight incubation in axenic conditions ( Figure 3—figure supplement 3 ) , again suggesting that they shut down endocytic feeding as part of a starvation response . Given the known function of NF1 in regulating Ras , and the conserved mutation affecting the RasGAP domain in one of our mutants , we examined the involvement of Ras signalling in macropinocytosis , and the specificity of individual RasGAPs in controlling it . First , we deleted the closely related MNF RasGAP , nfaA , from wildtype cells , and found that this does not confer the ability to grow axenically ( Figure 4—figure supplement 1 ) , suggesting that NF1 has specific functions not shared by other GAPs . This corroborates earlier findings that NfaA has a distinct function regulating pseudopodium formation during chemotaxis ( Zhang et al . , 2008 ) . We then asked whether inactivation of NF1 results in a global increase in Ras activity . Pulldowns of Ras-GTP using the Raf1 Ras-binding domain ( RBD ) from growing cells indicated no increase in Ras activity in mutants compared to wildtype cells ( Figure 4—figure supplement 2 ) . Similarly , confocal microscopy of cells expressing the GFP-RBD reporter revealed no difference between mutant and wildtypes in overall Ras activity estimated by determining the proportion the cell periphery labelled with the RBD ( Figure 4—figure supplement 3 ) . This is not surprising , since NF1 is only one of twelve putative RasGAPs encoded in the D . discoideum genome ( not including IQ-GAPs , which generally act as small G protein effectors and scaffolds and do not stimulate Ras GTPase activity [Shannon , 2012] ) . Having ruled out a global increase in Ras activity resulting from NF1 inactivation , we examined activity at sites of macropinocytosis . In wildtypes , as well as weak localisation at the leading edge of the cell , the GFP-RBD reporter is recruited intensely to small ruffles as they become concave and close into macropinosomes ( Figure 4A ) . In NF1 mutants , these specifically macropinocytotic ruffles tend to be larger ( Figure 4A ) , 50% of them being greater than 2 µm across upon closure , compared to less than 10% in wildtype cells ( Figure 4B ) . 10 . 7554/eLife . 04940 . 016Figure 4 . NF1 localises to membrane ruffles , its loss potentiates Ras signalling at macropinosomes , and its over-expression represses macropincytosis . ( A ) Ras activity , as reported by GFP-tagged Raf1 Ras-binding domain ( GFP-RBD ) , is exhibited at sites of macropinocytosis ( pointer ) in wildtype DdB cells as well as at the leading edge ( arrow ) as the cells move; the distribution of the reporter is qualitatively similar in NF1 knock-out amoebae , but ruffling is more extensive than in wildtypes . ( B ) The Ras-marked membrane ruffles tend to be larger in knock-out mutants prior to closure into pinosomes . Mutant or wildtype GFP-RBD reporter strains were harvested from bacterial growth plates and Ras-marked ruffles were measured across their longest visible axis just after they closed; data are from 60 events for each strain in total from three independent experiments . ( C ) Introduction of N-terminally GFP-tagged Dictyostelium NF1 proteins into axeB mutants reduces axenic growth in the case of the wildtype sequence ( NF1-WT ) but not when two consecutive arginine residues in the protein's ‘arginine finger’ are mutated to alanine and serine ( NF1-AS′ ) , nor when only the central region of the protein encompassing the RasGAP , CRAL-TRIO , and PH-like domains ( NF1ΔNΔC ) is expressed , when compared to a GFP control . Data are means plus and minus standard error for three independent experiments using the crystal violet assay to assess growth after 7 days incubation in tissue culture plates . ( D ) The active NF1-RR construct almost completely abolishes macropinosome formation when expressed in NF1 mutants , while the inactive NF1-AS form does not inhibit macropinocytosis . Bacterially grown cells were monitored by confocal microscopy as in Figure 3C; rates for nine cells of each line from three independent experiments are shown . ( E ) The NF1-AS mutant protein is recruited to membrane ruffles and sites of macropinocytosis ( examples indicated by pointers ) , whereas the wildtype version ( NF1-RR ) has an even cytoplasmic distribution , as does the truncated NF1ΔNΔC protein . The scale bars represent 5 μm . See also Figure 4—figure supplements 1–5 . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01610 . 7554/eLife . 04940 . 017Figure 4—figure supplement 1 . The axenic growth phenotype is specific to loss of the NF1 RasGAP protein . All axenic mutants we have examined so far possess mutations in axeB , the gene encoding NF1 . To test whether other RasGAPs might also have related functions ( but , for instance , a lower rate of spontaneous mutation ) , we also deleted the related RasGAP gene nfaA in both the wildtype and axeB null background . The nfaA single mutant ( HM1709 ) does not grow axenically , as assessed by the crystal violet binding assay , while the axeB nfaA double mutant ( HM1710 ) has slightly potentiated the axeB phenotype , suggesting that in the absence of NF1 , the NfaA protein can substitute for it to some extent . Data are the means plus and minus the standard error for three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01710 . 7554/eLife . 04940 . 018Figure 4—figure supplement 2 . NF1 mutants do not have an increase in overall Ras activity as assayed using RBD pulldowns . Using GST-Raf1-RBD beads to pull down GTP-bound Ras proteins and an anti-Ras antibody to compare samples by immunoblotting , no increase in Ras activity could be found in vegetative axenic mutants Ax2 and HM1591 ( axeB ) compared to wildtype DdB cells ( WT ) ; a single representative experiment is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01810 . 7554/eLife . 04940 . 019Figure 4—figure supplement 3 . NF1 mutants do not have an increase in overall Ras activity as assessed by confocal microscopy . Using GFP-Raf1-RBD reporter constructs , no increase in plasma-membrane associated active Ras was observed in the axeB null: this was quantified from tilescans of cells from three independent experiments . In all cases , to ensure that the cells were in comparable state , they were used within 30 min of harvesting from bacterial growth . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 01910 . 7554/eLife . 04940 . 020Figure 4—figure supplement 4 . Localisation of GFP-Ras fusion proteins . DdB cells ( WT ) or HM1591 ( axeB ) cells expressing GFP or GFP-tagged RasG ( ‘G’ ) , RasGG12T ( ‘G12’ ) , RasS ( ‘S’ ) , RasSG12V ( ‘S12’ ) , RasSQ61R ( ‘S61’ ) , or RasBG15T ( ‘B15’ ) were imaged by confocal microscopy to confirm proper localisation . All show some degree of enrichment on the plasma membrane except the dominant negative S17N mutants , which are very weakly fluorescent and do not show membrane localisation , presumably because they are deleterious and only cells with restricted expression grow through the selection procedure . Scale = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 02010 . 7554/eLife . 04940 . 021Figure 4—figure supplement 5 . Growth phenotypes of Ras expression lines . GFP-tagged D . discoideum Ras proteins were expressed in either or both DdB and HM1591 to test their effects on growth in HL5 medium in 24-well tissue culture plates; growth was assessed by the crystal violet assay after 7 days' incubation ( B = RasB , G = RasG , S = RasS ) : no active Ras constructs are able to stimulate axenic growth of wildtype DdB cells; expression of constitutively active RasG ( G12 ) or RasS ( S12 ) , or expression of wildtype RasS ( S ) is actually deleterious towards axenic growth . Data are means plus and minus standard error for three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 021 To examine the role of NF1 in regulating macropinocytosis more directly , we expressed full-length constructs of the D . discoideum orthologue tagged with GFP . A construct using the native polypeptide sequence reduced axenic growth of the NF1 null mutant considerably ( Figure 4C ) , almost completely suppressing macropinosome formation as assessed by confocal microscopy ( Figure 4D ) and displayed an evenly cytosolic distribution ( Figure 4E ) . In contrast , a version in which two consecutive arginine residues were mutated in the ‘arginine finger’ motif required for GAP activity ( Ahmadian et al . , 1997 ) ( NF1-AS ) did not retard growth ( Figure 4C ) nor affect macropinosome formation ( Figure 4D ) , and localised transiently to membrane ruffles and macropinosomes ( Figure 4E ) . This further supports a role for NF1 GAP activity in limiting macropinocytosis . A full length construct expressing the human NF1 protein also adversely affected growth of mutant cells , but a version with the critical arginine residue mutated to alanine reduced growth by a similar amount , suggesting non-specific effects ( unpublished data ) . Since the spontaneous inversion mutation we identified ( Table 1 ) affects the C-terminal region of the protein , we also tested whether the core RasGAP-Sec14-PH region of the D . discoideum NF1 protein could rescue the NF1 mutant; like the NF-AS version this did not suppress axenic growth ( Figure 4C ) , and displayed an evenly cytosolic distribution ( Figure 4E ) , suggesting that other regions of the protein are required for its correct localisation . The inactive NF1-AS form of the protein is presumably recruited by strong Ras activity at macropinocytotic crowns that it is then unable to attenuate . This localisation implies that the NF1 protein directly regulates signalling at the plasma membrane during ruffling , and together with the more extensive Ras signalling during macropinocytosis , suggests that this RasGAP regulates a pool of Ras responsible for this feeding process . Two Dictyostelium Ras proteins have been linked to endocytic functions ( Chubb et al . , 2000; Hoeller et al . , 2013 ) ; to examine their involvement in NF1-controlled events we expressed GFP-tagged versions of each in NF1 mutant and wildtype cells . All tested GFP-tagged Ras constructs localised to the plasma membrane , except for dominant negative ( S17N mutant ) RasG and RasS , expression of which was apparently poorly tolerated in these strains ( Figure 4—figure supplement 4 ) . None of the Dictyostelium Ras expression constructs phenocopied the loss of NF1; constitutively active RasG was deleterious to growth , as was expression of wildtype or constitutively active RasS ( Figure 4—figure supplement 5 ) , suggesting that improper activation of these isoforms interferes with endocytosis or other Ras-influenced processes leading to detrimental effects on cell growth . Active Ras at the plasma membrane recruits class 1 phosphoinositide 3′-kinases ( PI3Ks; Rodriguez-Viciana et al . , 1997 ) , allowing the spatially restricted formation of phosphatidylinositol trisphosphate ( PIP3 ) and other inositol phospholipids that occurs during macropinocytosis ( Araki et al . , 1996; Buczynski et al . , 1997; Hoeller et al . , 2013 ) . Confirming that the macropinosomes observed in NF1 mutants are mechanistically similar to those previously documented , we find that Ras activity at membrane ruffles in NF1 mutants is accompanied by recruitment of PH-domain reporters that bind the plasmanyl inositides produced by Dictyostelium class 1 PI3Ks ( Figure 5A; Clark et al . , 2014 ) , as well as by actin polymerisation ( Figure 5B ) . PH domains are also prominently recruited during macropinocytosis in wildtype cells; regions of recruitment tend to be larger in mutants reflecting the increased Ras signalling that results from the absence of NF1 ( Figure 5C ) . This pattern of Ras activity and PIP3 formation is invariably observed in every instance of macropinocytosis in Dictyostelium . The contributions of other Ras effectors remain unclear; for example no increase in ERK activity is observed in NF1 mutants compared to wildtype cells ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 04940 . 022Figure 5 . Downstream signalling: connections between Ras and PI3K activity during macropinocytosis . ( A ) Ras activity ( mCherry-Raf1-RBD reporter , magenta ) is accompanied by phosphoinositide 3-kinase activity ( PH-CRAC-GFP reporter , green ) on macropinosomes in axeB mutants; note the green endosome where PI3K products remain but Ras signalling has terminated . ( B ) Actin polymerisation ( labelled with mRFP-LifeAct , magenta ) occurs around the structures marked by the PH-domain reporter ( green ) . ( C ) PH domains ( GFP-PH10 ) are also recruited to macropinosomes in vegetative wildtype DdB cells; the kinetics of recruitment and retention are similar in axeB cells . The scale bars represent 5 μm . See also Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 02210 . 7554/eLife . 04940 . 023Figure 5—figure supplement 1 . ERK phosphorylation is not increased in NF1 mutants . WT ( DdB ) and axeB knockout ( HM1591 ) cells were harvested from bacterial growth plates then washed and incubated in HL5 medium and shaken for the indicated times before ERK activity was assessed using an antibody raised against a phosphorylated TEY motif . A band of the expected size of D . discoideum ErkB reproducibly increased in intensity over time in both strains , but more intensely in WT cells than NF1 mutants , perhaps reflecting starvation-induced development . A band of the approximate expected size of ErkA varied in its pattern of intensity in different experiments , but showed no tendency to be more intense in mutants . A single representative experiment is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 023 The observations described above indicate that wildtype cells perform qualitatively similar macropinocytosis to NF1 mutants , but on a smaller scale . This results in a markedly different outcome when the cells are incubated in HL5 medium: mutants can grow but the wildtype cannot . One possible explanation is that nutrient-uptake below a certain threshold leads to a growth arrest . To test this idea , we asked whether wildtype cells can maintain growth in an enriched axenic medium , as suggested by earlier work ( Sussman and Sussman , 1967 ) . Wildtype cells incubated in stationary cultures in HL5 supplemented with foetal bovine serum ( or bovine serum albumin , data not shown ) were able to grow , albeit still much more slowly than NF1 mutants cultured in the same medium ( Figure 6A , B ) . The morphology of wildtype cells was not appreciably altered after several days of axenic growth , while NF1 mutants remained consistently more flattened and extensively ruffled than wildtype cells in the same conditions ( Figure 6C ) . Wildtype cells were also found to degrade DQ-BSA efficiently after axenic growth in the presence of serum ( Figure 6—figure supplement 1 ) . Serum addition also stimulated the growth of NF1 mutants in shaking suspension ( Figure 6—figure supplement 2 ) . These findings suggest that the additional nutrients in the richer broth allow these cells to avoid the starvation-triggered growth arrest that can occur in axenic media . 10 . 7554/eLife . 04940 . 024Figure 6 . Wildtype amoebae can grow axenically in medium supplemented with bovine serum . ( A ) Wildtype ( DdB ) and NF1 mutant ( HM1591 ) cells were incubated in HL5 medium supplemented with vitamins and microelements without further additions or with 10% or 20% foetal bovine serum ( FBS and filter-sterilised HL5 mixed in 1:9 or 1:4 ratios ) in 24-well tissue culture dishes at a starting density of 5 × 104 cells per well . After 7 days growth was measured using the crystal violet assay . FBS stimulated growth of both wildtype and NF1 mutant cells , with mutants having a growth advantage in all axenic conditions . ( B ) Time courses of growth in the presence and absence of 10% FBS in the same conditions as above except that the HL5 medium was dissolved in 10% FBS or in water , then filter-sterilised . Data are means plus and minus standard errors of three ( A ) or four ( B ) independent experiments . ( C ) Wildtype amoebae retain their normal vegetative morphology after growth in serum-supplemented HL5 medium and NF1 mutants are still distinguished by a more flattened appearance . Cells were grown in HL5 plus 10% FBS for 4 days before being washed and placed into Loflo plus 10% FBS in presence of TRITC-dextran . After 30 min , the cells were imaged by confocal microscopy . Scale = 5 μm . See also Figure 6—figure supplements 1 , 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 02410 . 7554/eLife . 04940 . 025Figure 6—figure supplement 1 . Wildtype cells degrade extracellular protein effectively after growth in rich axenic media . DdB cells were grown in HL5 medium plus 10% foetal bovine serum for 3 days before being washed , resuspended in Loflo medium plus 50 µg/ml DQ Green BSA and 2 mg/ml TRITC dextran . Images were taken after 20 min . Scale = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 02510 . 7554/eLife . 04940 . 026Figure 6—figure supplement 2 . NF1 mutants are able to grow in suspension in rich axenic media . The established axenic mutant strain Ax2 was selected for high rates of growth in HL5 medium in shaken suspension; deleting axeB in the wildtype background only partially recapitulated this pheontype , the cells only being able to grow well in HL5 when attached to a substratum , indicating that at least one other mutation is required to account for the full axenic phenotype . However the single mutant is able to grow in shaken suspension when HL5 is supplemented with 10% foetal bovie serum . Cells were inoculated at a starting density of 5 × 105 cells per ml in 50 ml of medium in 250 ml Erlenmeyer flasks at 22°C and counted at the stated intervals using a haemocytometer . The means and standard errors are given for three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 026 Finally , since macropinocytosis and phagocytosis are closely related processes we compared phagocytosis in NF1 mutants and wildtype cells . Mutant and wildtype strains grow well on bacteria ( Figure 7—figure supplements 1 , 2 ) and take up bacterium-sized polystyrene microspheres ( 1 μm and 1 . 8 μm diameter ) at very similar rates ( Figure 7A ) , although the standard strain Ax2 is marginally but consistently less effective at internalising smaller beads than the other strains ( Figure 7B ) . Against expectation , we found that wildtype cells cannot efficiently ingest yeast or beads greater than 3 μm in diameter ( Figure 7A , C ) , whereas NF1 mutant cells can ingest beads larger than 4 μm in diameter ( Figure 7A ) or yeast cells very readily ( Figure 7C ) . In line with earlier findings in Ax2 cells ( Clarke et al . , 2010 ) , RBD and PH domain reporters localised to phagosomes as they formed , essentially identically to their behaviour during macropinocytosis ( Figure 7—figure supplement 3 ) . We conclude that , as well as controlling macropinocytosis , NF1 limits the size of nascent phagosomes , supporting the idea that these large-scale endocytic processes share regulatory as well as structural features . The striking improvement in phagocytosis of larger cells after NF1 deletion also suggests that variation in or loss of this gene can have important ecological and evolutionary consequences by enabling predators to target additional prey species ( Porter , 2011 ) . 10 . 7554/eLife . 04940 . 027Figure 7 . NF1 mutants can phagocytose larger particles than wildtypes . ( A ) Axenic mutants ingest small bacterium-sized beads at a similar rate as wildtypes , but wildtype cells are dramatically less efficient at ingesting beads greater than 2 μm in diameter . Cells were harvested from bacterial growth plates , washed , then shaken with fluorescent microspheres of the indicated diameter , then after 1 hr scored for the presence of internalised beads . ( B ) The Ax2 mutant accumulated small 1 . 0 μm beads more slowly than the wildtype DdB or the axeB deletion mutant . ( C ) Axenic mutants can ingest fluorescently labelled budding yeast cells much more easily than wildtype cells . All data are mean ± standard error for three independent experiments . See also Figure 7—figure supplements 1–3 . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 02710 . 7554/eLife . 04940 . 028Figure 7—figure supplement 1 . NF1 mutants grow and develop when grown on bacterial lawns . DdB ( WT ) and HM1591 ( axeB null ) spores were plated clonally on SM agar plates in association with Klebsiella pneumoniae . After 5 days , plaques of amoebae growing outwards on the bacterial lawn were photographed; aggregates and fruiting bodies are visible where the bacteria have been cleared causing the amoebae to enter their asexual developmental cycle ( scale = 5 mm ) . Fruiting bodies in the mutant tend to be smaller than wildtype , but otherwise do not show obvious defects . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 02810 . 7554/eLife . 04940 . 029Figure 7—figure supplement 2 . NF1 mutants grow normally when shaken in suspensions of dead bacteria . To quantify cell growth without complicating factors such as cell motility and susceptibility to harmful bacterial metabolites , cells were grown on heat-killed Escherichia coli strain B/r in shaking suspension . The established axenic strain Ax2 grows consistently more slowly than DdB and HM1591 , with a doubling time on average approximately 5% greater . Data are means plus and minus standard error for three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 02910 . 7554/eLife . 04940 . 030Figure 7—figure supplement 3 . Phagocytosis is accompanied by Ras and PI3K activity in the same way as in macropinocytosis . NF1 mutants ( HM1591 ) were transformed with an expression construct containing both mCherry-Raf1-RBD and PH ( CRAC ) -GFP and imaged in the presence of Klebsiella cells; the initial engulfment occurred out of the plane of acquisition , but Ras and PI3K activity remained visible as the nascent phagosome moved into view . Scale = 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 030 We set out to explain the genetic basis of the axenic growth phenotype of standard laboratory D . discoideum strains , which has remained mysterious for decades despite the widespread use of these cells . In freshly selected mutants , we discovered coding sequence mutations only in the Dictyostelium orthologue of the tumour suppressor NF1 . Importantly , all axenic mutants , across two distinct genetic backgrounds , bear mutations in this gene . While it is possible that mutations in other genes will result in similar phenotypes , it is clear that NF1 mutations must be the most frequent by far that cause axenic growth . The further mutations enabling faster growth in the established axenic strains remain to be identified , and their precise effect is still unclear . Our identification of the axeB gene as NF1 will provide a route towards creating new axenic strains from wild isolates , thus giving strains with minimal background mutations . Vegetative wildtype cells perform macropinocytosis in a qualitatively similar way as axenic mutants but to a lesser extent , and accordingly they can grow axenically when the standard medium is supplemented with bovine serum . NF1 mutants retain a large growth advantage in the more complex medium , and so will still be selected during prolonged culture . Nevertheless , this protocol should be of use for short-term axenic culture of wildtype strains; standard defined medium supplemented with bovine serum albumin also enables slow growth of wildtype strains ( unpublished data ) suggesting that fully synthetic defined media should be attainable . This further supports the idea that the important effect of serum is the provision of bulk nutrients , preventing the nitrogen starvation that initiates Dictyostelium development ( Marin , 1976 ) . One important focus of Dictyostelium research is chemotaxis , and key roles for Ras and PIP3 in steering migrating cells have been proposed ( Kay et al . , 2008; Artemenko et al . , 2014 ) . Although axenic cells chemotax very well to the best-studied chemoattractant , cyclic-AMP , they are much less efficient than wildtype cells in chemotaxis to folic acid , due to interference from large , PIP3-rich , macropinosomes ( Veltman et al . , 2014 ) . It is now apparent that the reason for this poor chemotaxis is likely to be the inactivation of NF1 in axenic cells , leading to Ras and PI3K hyperactivity . Macropinosomes and intense patches of Ras activity and PIP3 also appear in cells chemotaxing to cyclic-AMP ( Parent et al . , 1998 ) and it will be important to disentangle their contribution to chemotaxis by comparative studies of axenic and wildtype cells . Several lines of evidence emphasize the importance of Ras in the feeding process used by Dictyostelium cells . Previous studies in Dictyostelium showed defects in macropinocytosis and phagocytosis after disruption of Ras genes ( Chubb et al . , 2000; Hoeller et al . , 2013 ) , and Ras activates PI3K during macropinocytosis ( Hoeller et al . , 2013 ) . Further , Ras activity reporters localise strongly to membrane ruffles during macropinocytosis as well as to nascent phagosomes ( Sasaki et al . , 2007; Clarke et al . , 2010 ) , and we found that the size of these sites of Ras activity is increased in NF1 mutants . The well-studied NF1 orthologues in mammals and yeast are specific to Ras subfamily small G proteins ( Tanaka et al . , 1990; Zhang et al . , 1991 ) , and one of mutations we found in Dictyostelium corresponds exactly to one that abolishes the GAP activity of the human orthologue by altering the Ras-binding interface of the protein ( Poullet et al . , 1994 ) . These data , along with the effect of mutating the ‘arginine finger’ critical for RasGAP activity , suggest that Dictyostelium NF1 is most likely Ras-specific , but this remains to be demonstrated biochemically . The specificity of Dictyostelium NF1 towards different Ras isoforms remains unclear . Since the rasG and rasS null mutants have defective endocytic feeding but are still viable ( Chubb et al . , 2000; Hoeller et al . , 2013 ) it is likely that multiple Ras isoforms are involved , possibly including other less well-characterised genes . The biochemical functions of NF1 beyond its GAP activity remain poorly understood . A module comprising a CRAL/TRIO and a PH-like domain is conserved and likely mediates an interaction with lipids within cells ( D'angelo et al . , 2006 ) , but has not been tied to any function of consequence . Regions of conserved sequence to either side of the relatively well-characterised core of the protein are even more mysterious , but may be important for its function in Dictyostelium , perhaps by mediating its dynamic localisation . We find that GFP-tagged full length Dictyostelium NF1 localises transiently to macropinocytotic ruffles . This translocation has only been visualised in a mutant form of the protein with the ‘arginine finger’ motif mutated , because over-expression of the active protein almost completely inhibits macropinocytosis . It will be important to identify structural determinants of this localisation as a route towards a better understanding of the cell-biological function of NF1 . Both macropinosomes and phagosomes are significantly larger in NF1 mutants than in wildtypes , suggesting that NF1 stimulates Ras GTPase activity as endocytic ruffles form and spread , thereby limiting their size ( Figure 8 ) . The control of NF1 function is not well understood , but our results suggest that it might be locally inactivated during macropinocytosis and phagocytosis . In an intriguing parallel , growth factor treatment of mammalian cells leads to rapid degradation of NF1 by the proteasome ( Cichowski et al . , 2003 ) and also triggers membrane ruffling and macropinocytosis with similar kinetics ( Brunk et al . , 1976; Mellström et al . , 1983 ) . However , we could find no obvious fluid-uptake phenotype in NF1 null mouse embryonic fibroblasts , suggesting that NF1 inactivation is not sufficient for stimulation of macropinocytosis in these cells ( unpublished data ) . Given the known involvement of Ras signalling in promoting ruffling and macropinocytosis the possibility remains that NF1 function is conserved in metazoa , but in a context in which Ras activity is more heavily regulated , with additional layers of control not present in amoebae ( Casci et al . , 1999; Johnson et al . , 2005 ) . 10 . 7554/eLife . 04940 . 031Figure 8 . Schematic model of NF1 function in Dictyostelium . ( A ) While wildtype NF1+ amoebae ingest bacteria most readily , NF1− cells are also able to ingest larger particles such as yeast cells , and accumulate more fluid in macropinosomes . ( B ) The large concave membrane ruffles formed during phagocytosis and macropinocytosis both are marked by intense Ras signalling ( green ) ; NF1 localises dynamically to these regions , stimulating the GTPase activity of Ras proteins there , inactivating them and thereby limiting the expansion and spread of the ‘activated’ membrane domain . DOI: http://dx . doi . org/10 . 7554/eLife . 04940 . 031 The set of species possessing NF1 homologues indicates the extreme antiquity of this protein: NF1 is present in the excavates Naegleria and Trichomonas as well as the amoebae , animals , and fungi . Recent placements of the root of the eukaryotes within the excavate group ( He et al . , 2014; Derelle et al . , 2015 ) suggest that NF1 evolved very early in the history of eukaryotes , and may have been present in the last eukaryotic common ancestor ( LECA ) , along with the related protein MNF . The presence of these RasGAPs in organisms that display strikingly similar forms of circular ruffling suggests the hypothesis that NF1 and MNF evolved together to control the movements of the exploratory cell projections used in very early eukaryotes during feeding ( Boschek et al . , 1981; John et al . , 1984; Hacker et al . , 1997; Cavalier-Smith , 2013 ) . Macropinocytosis and phagocytosis occur in a well-defined series of stages that are shared between amoebae and vertebrate cells ( Swanson , 2008 ) . The actin cytoskeleton is used to project membrane ruffles outwards to enclose either a bound particle or extracellular fluid , inositol phospholipids accumulate and are then dephosphorylated in a well-defined sequence ( Dormann et al . , 2004; Egami et al . , 2014 ) . The role of Ras is less well understood , but has previously been implicated in macropinocytosis and phagocytosis in Dictyostelium ( Chubb et al . , 2000; Hoeller et al . , 2013 ) , and in circular ruffle formation , macropinocytosis , and trogocytosis in vertebrate cells ( Bar-Sagi and Feramisco , 1986; Martínez-Martín et al . , 2011; Welliver and Swanson , 2012 ) . Our unbiased forward genetic analysis suggests that NF1 has a fundamental role in governing the feeding processes used by amoebae . The mechanistic parallels between large-scale endocytosis in metazoa and amoebae , as well as their shared history , raise the possibility that this function is conserved . The role of Schwann cells , the cell-type of origin of neurofibromas , as non-professional phagocytes during the repair of nerve damage ( Stoll et al . , 1989 ) is striking in this regard . Although the consequences of NF1 mutation are well understood in humans , its cell-biological function is still not well understood ( Stephen et al . , 2014 ) . Our results , for the first time linking NF1 to macropinocytosis and phagocytosis , may provide an important clue . D . discoideum strains ( listed in Table 2 ) were cultivated in association with Klebsiella pneumoniae on SM agar plates at 22°C , harvested and prepared for experiments by removing the bacteria by differential centrifugation in KK2 buffer ( 16 . 5 mM KH2PO4 , 3 . 9 mM K2HPO4 , 2 mM MgSO4 ) . For axenic growth , cells were grown in autoclaved HL5 medium ( Formedium , Hunstanton , UK ) on tissue-culture treated plastic dishes , or shaken at 180 rpm in 250 ml flasks , in both cases at 22°C . To prepare serum-supplemented medium foetal bovine serum ( Hyclone/GE Healthcare , South Logan , Utah ) was either added directly to filter sterilised HL5 medium supplemented with vitamins and microelements ( HL5VME , Formedium ) , or first diluted to 10% vol/vol in Milli-Q water then used to dissolve powdered HL5VME before 0 . 2 μm filter sterilisation . Variable amounts of precipitated material are visible in filtered serum-supplemented HL5VME during incubation with cells at 22°C; these solids may help to stimulate growth ( Watts , 1977 ) . To measure growth rates on bacteria , cells were shaken in a suspension of heat-killed Escherichia coli B/r ( at an OD600 of 10 ) in KK2 . Cells were transformed according to the method of ( Pang et al . , 1999 ) , except that selection was carried out growing cells on bacteria in order to avoid selection for axenic growth: for the axeB gene disruption cells were plated in 2 . 5 ml KK2 buffer containing a suspension of heat-killed K . pneumoniae at an OD600 of 5 and 50 μg/ml tetracycline , 100 μg/ml dihydrostreptomycin , and 40 μg/ml G418 in 6-well tissue culture dishes . Once wells became confluent , cells were passaged again in the same conditions to kill non-transformed cells before being cloned on SM agar plates and screened for the disruption by polymerase chain reaction and sequencing . Later transformations , to introduce expression constructs and to disrupt nfaA , used an improved selection protocol as follows: Klebsiella pneumoniae was grown overnight in a standing bottle of SM/5 broth ( Formedium SM broth diluted fivefold in Milli-Q water ) to stationary phase , then diluted fourfold in fresh SM/5 broth containing 100 μg/ml dihydrostreptomycin and 60 μg/ml hygromycin , giving a final OD600 of approximately 0 . 1–0 . 2 . A total of 2 . 5 ml of bacterial suspension was added per well of a 6-well cluster dish , or 10 ml per 100 mm dish . The bacteria and antibiotics were replenished at least every 2 days until amoebae grew to their maximum density ( typically 3–4 days after the initiation of selection ) . Putative nfaA disruptants were again cloned on SM agar plates before screening as above . DdB cells were grown on mass culture SM agar plates before being washed free of bacteria and resuspended in HL5 medium . Selection was carried out without mutagenesis by incubation in HL5 medium either under shaking suspension from a starting density of 106 cells per ml ( strains HM557 , HM558 , and HM559 ) or in 100 mm tissue-culture treated plastic dishes at a starting density of 107 cells per dish ( subsequent mutants ) , at 22°C . Mutants accumulated over the course of 3–5 weeks , and were cloned on SM agar and axenic growth was retested before genome resequencing or targeted sequencing of the axeB locus . Genomic DNA was extracted from cells starved overnight that were resuspended in lysis buffer ( 20 mM Tris-HCl , 5 mM MgCl2 , 0 . 32 M sucrose , 0 . 02% sodium azide , 1% Triton X-100 , pH 7 . 4 ) at 4°C , vortexed and incubated at 4°C for 15 min . Nuclei were pelleted at 3000×g for 10 min , resuspended in lysis buffer and pelleted again , before freezing the pellets on dry ice . Proteinase K ( 100 μl of a 20 mg/ml stock in water ) was added , followed immediately by 10 ml digestion buffer ( 10 mM Tris-HCl , 5 mM EDTA , 0 . 7% SDS , pH 7 . 5 ) , and the pellet resuspended by gentle pipetting . The lysate was incubated for 1 hr at 60°C and the DNA finally phenol-chloroform extracted using Phase Lock Gel tubes ( 5 Prime , Hilden , Germany ) . Single end Illumina sequencing libraries were constructed according to the manufacturer's instructions . Sequencing was carried out on an Illumina GAII instrument , producing reads of 36 and 45 basepairs across different runs , to a depth of approximately 17–20× after removing of potential PCR and optical duplicates . Reads were aligned against the dictyBase AX4 assembly using Stampy ( Lunter and Goodson , 2010 ) , and duplicates removed and variants called using samtools and bcftools ( Li et al . , 2009 ) . Candidate variants in each strain were pre-filtered ( depth greater than 3 , mapping quality greater than 20 , SNP quality greater than 20 , ‘heterozygous’ calls excluded ) to remove misalignments , then putative variants common to all four resequenced strains , representing real differences between DdB and the reference sequence , were excluded . To display homology across the length of the NF1 protein , the Homo sapiens isoform 2 polypeptide sequence ( NP_000258 . 1 ) was split into segments using a sliding window of 200 residues . These were then aligned using the EMBOSS ‘water’ local alignment software ( Rice et al . , 2000 ) to the D . discoideum , Drosophila melanogaster ( AAB58975 . 1 ) , and Takifugu rubripes ( AAD15839 . 1 ) NF1 orthologues and the Saccharomyces cerevisiae Ira1p protein ( NP_009698 . 1 ) . Alignment scores were normalised to the Hs–Hs comparison such that the self-comparison gives a value of one , and plotted sequentially along the sequence length for each comparison . To display global similarity and identity percentages , the same H . sapiens polypeptide sequence was aligned to the T . rugripes , D . melanogaster , Trichinella spiralis ( XP_003376664 . 1 ) , Trichoplax adhaerens ( XP_002115170 . 1 ) , Salpingoeca rosetta ( EGD75509 . 1 ) , Capsaspora owczarzaki ( EFW43762 . 1 ) , and Batrachochytrium dendrobatidis ( EGF81694 . 1 ) NF1 proteins , S . cerevisiae Ira1p and Ira2p ( NP_014560 . 1 ) , and D . discoideum NF1 and NfaA ( XP_645456 . 1 ) . The EMBOSS ‘needle’ software was used to generate global alignments in order that indels count against overall homology . The axeB disruption plasmid was constructed by inserting the V18-tn5 cassette into the BamHI and HindIII sites of pBluescript2 KS+ , then amplifying 5′ and 3′ flanking regions of the axeB gene from strain DdB using the primers KOKpnI ( 5′-GGTACCAAATGTATACTTGTATATGATG-3′ ) with KOHindIII ( 5′-AAGCTTGAGCTCTTCACCACCATTAAGT-3′ ) , and KOBamHI ( 5′-GGATCCATTGGGTAGTTATCGATC-3′ ) , with KOEagI ( 5′-CGGCCGTGCACAGTCTTTAGAAAATTTTG-3′ ) , and inserting them either side of it . The nfaA disruption plasmid was also based on pBluescript2 KS+ using , between BamHI and XhoI sites , the hygromycin resistance gene driven by the act14 promoter; and the following primer pairs were used to amplify flanking segments: nfa2kpn ( 5′-GGTACCTAATGGTGTAACTCAAGTTTTCG-3′ ) with nfa2xho ( 5′-CTCGAGTGGTAATGTTTTATTTGCTGTTG-3′ ) , and nfa2bam ( 5′-GGATCCAGATATTCATTGTACATCCATCAG-3′ ) with nfa2spe ( 5′-ACTAGTATACTTATAAGAAACCTTCTTCAG-3′ ) . Expression constructs used the act14 or coaA promoter to drive the resistance gene and the act15 promoter to drive the gene fragment of interest , using the vector pDM1005 as backbone , a derivative of previously described extrachromosomal vectors ( Veltman et al . , 2009 ) . The GFP-Raf1-RBD included amino acids 1–134 of the H . sapiens Raf1 polypeptide , with a short linker ( TTSRT ) between GFP and its N-terminal methionine . PH-CRAC-GFP contains residues 1–126 of the D . discoideum CRAC protein . GFP-PH10 contains amino acids 1–103 of D . discoideum PkgE ( Ruchira et al . , 2004 ) ; this fragment is longer than the previously published version , but gives a similar localisation . The GFP-Ras constructs included an RS ( GGS ) 4RS linker between the C-terminus of GFP and N-terminus of each Ras . During the construction of the full length NF1 expression construct , in the same vector backbone including an N-terminal GFP , silent mutations were incorporated at positions 409 and 813 of the D . discoideum axeB cDNA sequence to introduce a XhoI and a XmaI site , respectively . The consecutive arginine codons in the ‘arginine finger’ region were mutated by introducing changes in overlapping PCR primers during cloning . The NF1ΔNΔC construct uses the same vector backbone and tag , and includes amino acids 1189 to 1779 , with the wildtype arginine finger motif . Dual-colour experiments used plasmids containing both protein constructs as previously described ( Veltman et al . , 2009 ) . Since axeB knock-out mutants are unable to grow well under shaking suspension , growth was measured by accumulation of crystal-violet staining . Cells were pregrown on bacteria , washed , and plated at a density of 105 cells per well of 24-well tissue culture plates in 1 ml of HL5 medium; for the experiments to test the effect of serum on cell growth the initial density was 5 × 104 cells per well . At each timepoint the medium was removed , the cells washed once in KK2 buffer , then incubated for 20 min in 0 . 5 ml of 0 . 1% crystal violet ( in 10% ethanol ) . Each well was then carefully washed three times with water , then incubated for a further 20 min in 10% acetic acid . After brief agitation , the absorbance of each well was measured at 590 nm . To measure growth on bacteria , cells were grown overnight in heat-killed E . coli B/r ( OD600 of 10 in KK2 buffer ) to logarithmic phase then diluted to a density of 5 × 105 per ml of fresh bacterial suspension; cells were counted using a haemocytometer every 2 hr for 8 hr . Bacterially grown amoebae were assayed either directly or after adaptation in HL5 medium for 24 hr shaken at 180 rpm from a starting density of 2 × 105 cells per ml . Cells were resuspended at 1 × 107 per ml in KK2C ( KK2 plus 0 . 1 mM CaCl2 ) and the assay initiated by adding FITC dextran ( average MW 70 , 000 ) to 2 mg/ml final with 8 × 105 cells being removed ( in duplicate ) for each data point and mixed with 0 . 75 ml of ice-cold wash buffer ( KK2C plus 0 . 5 mg/ml BSA ) . The cells were pelleted by centrifugation ( 20 , 000×g ) for 12 s , the supernatant removed and the cells resuspended in 1 . 5 ml ice-cold wash buffer . The cells were pelleted and washed once more before 1 ml of lysis buffer ( 0 . 1 M Tris-Cl pH 8 . 6 , 0 . 2% Triton X-100 ) was added . The fluorescent intensity was measured by excitement at 490 nm and emission at 520 nm ( PerkinElemer LS50B Luminescence Spectrometer ) . Bacterially grown cells were shaken at 180 rpm for 15 min . Then 0 . 1 ml was added to a stirred fluorimeter cuvette containing 0 . 9 ml 11 μM FM1-43 ( Life Technologies , Paisley , UK ) in KK2C and data collected every 1 . 2 s at an excitation of 470 nm and emission of 570 nm for approximately 5 min using a PerkinElmer LS50B fluorimeter . TRITC labelled yeast was made as described by ( Rivero and Maniak , 2006 ) and the assay itself was based on that described in the same paper . Bacterially grown amoebae were resuspended at 2 × 106 cells/ml in KK2C and the assay initiated by the addition of TRITC labelled yeast cells to approximately 1 × 107 per ml final . For each data point , 2 × 105 cells were removed and the uningested fluorescent yeast quenched by the addition of 0 . 1 ml of trypan blue solution ( 20 mM sodium citrate , 150 mM NaCl , 2 mg/ml trypan blue ) . The cell suspension was shaken for 3 min at 2000 rpm ( Eppendorf MixMate ) and then pelleted by centrifugation ( 4000×g ) for 2 . 5 min . The cell pellet was resuspended in 1 . 5 ml of KK2C and pelleted as before . Finally , the cell pellet was resuspended in 1 ml of KK2C and the fluorescent intensity measured by excitement at 544 nm and emission at 574 nm ( PerkinElemer LS50B Luminescence Spectrometer ) . For bead uptake experiments Fluoresbrite Bright Blue carboxylate microspheres ( Polysciences , Eppelheim , Germany ) were used . Bacterially grown cells were resuspended at 107 per ml in KK2C containing 0 . 2% ( wt/vol ) BSA ( KK2CB ) ( to reduce non-specific binding ) and 20 beads added per amoebae . Uptake was stopped by adding 0 . 5 ml of the cell suspension to an equal volume of ice-cold KK2CB containing 10 mM NaN3 ( KK2CBA ) . The cells were pelleted at 300×g for 2 min . The pellet was then washed twice more in 1 ml of ice-cold KK2CBA and finally resuspended in 1 ml of ice-cold KK2CBA . 100 μl of this cell suspension was added to 200 μl of KK2CBA containing 40 μg/ml TRITC dextran ( as a cell counterstain ) in a chamber of a 8-well LabTek chambered coverslip and image stacks taken of several fields of cells for analysis . This procedure removes most non-phagocytosed beads up to 3 μm . For 3 . 1 , and 4 . 4 μm beads the cells can be directly scored by phase contrast microscopy without counterstain . Cells were imaged either directly from growth plates in SM/5 medium or after incubation overnight in Loflo medium , as indicated . To image lysosomal degradation of endocytosed protein , cells were incubated in Loflo medium ( Formedium ) plus 20 μg/ml DQ green BSA ( Life Technologies ) . Images were acquired using a Zeiss 780 LSM microscope , with laser power and gain set identically for all strains and the brightness and contrast of images adjusted later identically . For DIC images , brightness and contrast was adjusted for visual clarity using ImageJ . To measure the frequency of macropinosome internalization , cells were harvested from mass-inoculation SM agar growth plates , washed three times in Loflo medium , then 1 × 105 cells plated per chamber of a Lab-Tek II 8-well chambered coverglass ( Thermo , Waltham , MA ) and allowed to settle for 10–15 min . Within 30 min of removal from bacteria , 0 . 4 mg/ml FITC- and 2 mg/ml TRITC-dextran were added , and movies recorded taking 5 Z-sections ( 1 μm apart ) every 5 s . Pinosomes were counted if they appeared adjacent to ruffled cell projections and cups , and if they retained FITC fluorescence ( FITC is rapidly bleached as endosomes that are acidified ) . To enable estimation of the rate of uptake , cells were tracked and included in the analysis only if they remained within the field for at least 5 min; these cells were tracked as long as they remained in the field to at most 10 min . To measure the extent of Ras signalling at the membrane of random growing cells , strains expressing the GFP-Raf1-RBD reporter were harvested during exponential growth in tissue culture dishes containing K . aerogenes in SM/5 broth , washed and plated in fresh SM/5 in Lab-Tek chambers as above . Tile scans were acquired of 25 fields , and cells were outlined manually using a custom built MatLab script ( Source code 1 ) . Normals of 3 pixels long were drawn at equidistant points along the perimeter spaced 2 pixels apart and the highest intensity value along this normal was determined . The patch threshold was set as all membrane values that were more than 3 standard deviations above the mean intensity of the cytosol . Over 100 cells from two independent experiments were analysed for each strain . To measure the extent of Ras signalling during pinocytosis , the same Ras-activity reporter strains were plated in Lab-Tek chambers in SM/5 as above , and movies of a single confocal section through the cells recorded with a 2 . 5 s interval . The maximum extent of GFP-RBD fluorescence across enclosed ruffles was measured in the first frame after they closed using ImageJ . Images of tagged Ras and NF1 proteins , Raf1-RBD , PH-CRAC , and LifeAct reporters were adjusted for brightness and contrast across the whole image of each channel , and cropped , for clarity . Cells were grown in association with K . aerogenes on mass-inoculation SM agar plates , washed three times in KK2 buffer , and 2 × 107 cells resuspended in 10 ml KK2 and shaken at 180 rpm at 22°C for 30 min . The cells were then pelleted at 4°C , and lysed in 1 ml lysis buffer ( 0 . 5% Triton X-100 , 150 mM NaCl , 40 mM Tris , 20 mM MgCl2 , 10% glycerol , 1 mM DTT , 1 tablet per 50 ml Complete EDTA-free protease inhibitors ( Roche Lifescience , Burgess Hill , UK ) , pH 7 . 4 ) . The lysate was cleared by centrifugation at 13 , 000×g for 10 min at 4°C , then the supernatant added to 33 μl GST-Raf1-RBD conjugated to agarose beads ( Millipore , Watford , UK ) suspended in lysis buffer , with BSA added to a final concentration of 1 mg/ml . The mixture was then shaken for 30 min at 4°C , before washing twice with lysis buffer by centrifugation at 2000×g for 1 min . The bound Ras was released by boiling the washed beads in LDS sample buffer ( Life Technologies ) for 5 min , before immunoblotting and detection with mouse monoclonal anti-pan-Ras antibody ( clone RAS 10 , Millipore ) and HRP-conjugated goat anti-mouse secondary antibody using standard techniques . Cells of each strain were harvested from mass inoculation cultures grown in association with K . pneumoniae on SM agar , washed , and resuspended in autoclaved HL5 medium at a density of 2 × 106 cells per ml in a total of 10 ml in 50 ml Erlenmeyer flasks . The flasks were then shaken at 180 rpm and 22°C and aliquots taken at the indicated timepoints . These samples were lysed in LDS sample buffer ( Life Technologies ) in the presence of protease and phosphatase inhibitors ( Roche , as above; Sodium pyrophosphate , sodium orthovanadate , and ß-glycerophosphate ) , and separated on NuPAGE polyacrylamide gels and blotted onto PVDF according to standard protocols . Blots were blocked in 5% bovine serum albumin in TBS-Tween and activated ERK kinases detected using an antibody raised against phospho-ERK ( anti-phospho-p44/p42 MAPK rabbit antibody from Cell Signalling Technology , cat #9101 ) , and HRP-conjugated goat anti-mouse secondary antibody . Sequence data have been deposited in European Nucleotide Archive under the accessions HF565448 ( axeB genomic sequence ) and ERP002043 ( whole-genome resequencing reads ) .	Dictyostelium amoebae are microbes that feed on bacteria living in the soil . They are unusual in that the amoebae can survive and grow in a single-celled form , but when food is scarce , many individual cells can gather together to form a simple multicellular organism . To feed on bacteria , the amoebae use a process called phagocytosis , which starts with the membrane that surrounds the cell growing outwards to completely surround the bacteria . This leads to the bacteria entering the amoeba within a membrane compartment called a vesicle , where they are broken down into small molecules by enzymes . The cells can also take up fluids and dissolved molecules using a similar process called macropinocytosis . With its short and relatively simple lifestyle , Dictyostelium is often used in research to study phagocytosis , cell movement and other processes that are also found in larger organisms . For example , some immune cells in animals use phagocytosis to capture and destroy invading microbes . Most studies using Dictyostelium as a model have used amoebae with genetic mutations that allow them to be grown in liquid cultures in the laboratory without needing to feed on bacteria . The mutations allow the ‘mutant’ amoebae to take up more liquid and dissolved nutrients by macropinocytosis , but it is not known where in the genome these mutations are . Here , Bloomfield et al . used genome sequencing to reveal that these mutations alter a gene that encodes a protein called Neurofibromin . The experiments show that the loss of Neurofibromin increases the amount of fluid taken up by the amoebae through macropinocytosis , and also enables the amoebae to take up larger-than-normal particles during phagocytosis . The experiments suggest that Neurofibromin controls both phagocytosis and macropinocytosis by inhibiting the activity of another protein called Ras . Neurofibromin is found in animals and many other organisms so Bloomfield et al . propose that it is an ancient protein that evolved in early single-celled organisms to control the size and shape of their feeding structures . In humans , mutations in the gene that encodes the Neurofibromin protein can lead to the development of a severe disorder—called Neurofibromatosis type 1—in which tumours form in the nervous system . Given that tumour cells can use phagocytosis and macropinocytosis to gain nutrients as they grow , understanding how this protein works in the Dictyostelium amoebae may help to inform future efforts to develop treatments for this human disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology" ]	2015	Neurofibromin controls macropinocytosis and phagocytosis in Dictyostelium
Insect cuticular hydrocarbons ( CHCs ) prevent desiccation and serve as chemical signals that mediate social interactions . Drosophila melanogaster CHCs have been studied extensively , but the genetic basis for individual variation in CHC composition is largely unknown . We quantified variation in CHC profiles in the D . melanogaster Genetic Reference Panel ( DGRP ) and identified novel CHCs . We used principal component ( PC ) analysis to extract PCs that explain the majority of CHC variation and identified polymorphisms in or near 305 and 173 genes in females and males , respectively , associated with variation in these PCs . In addition , 17 DGRP lines contain the functional Desat2 allele characteristic of African and Caribbean D . melanogaster females ( more 5 , 9-C27:2 and less 7 , 11-C27:2 , female sex pheromone isomers ) . Disruption of expression of 24 candidate genes affected CHC composition in at least one sex . These genes are associated with fatty acid metabolism and represent mechanistic targets for individual variation in CHC composition . Insects comprise the most species-rich class in the animal kingdom . They evolved about 480 million years ago and their fecundity and rapid evolutionary adaptations have made them successful in populating almost every ecological niche on our planet . The evolution of mechanisms for desiccation resistance was critical for insects to colonize dry land . To prevent desiccation insects evolved the ability to produce and accumulate species-specific blends of fatty acid-derived apolar lipids on the epicuticle; the most prominent of these are cuticular hydrocarbons ( CHCs ) ( Jallon et al . , 1997 ) . CHCs are produced continuously by specialized cells called oenocytes , and are transported through the hemolymph and then to the cuticular surface through specialized pore canals ( Romer , 1991; Schal et al . , 1998; Blomquist and Bagnères , 2010 ) . The primary role of CHCs is desiccation resistance ( Gibbs , 1998; 2002 ) , but they have been co-opted to serve as chemical signals and cues mediating intra- and inter-specific social interactions ( Venard and Jallon , 1980; Jallon , 1984; Ferveur , 2005 ) . These interactions include species and nest-mate recognition , assessment of reproductive status , and mate choice , and CHCs play a prominent role in camouflage and mimicry that mediate inter-specific parasitic relationships ( Stanley-Samuelson and Nelson , 1993; Blomquist and Bagnères , 2010 ) . Adaptive evolution acts through selection on phenotypic variation within a population . Thus , understanding the genetic basis for individual variation in CHC composition will shed light on evolutionary mechanisms of assortative mating ( Noor et al . , 1996; Ishii et al . , 2001; Rundle et al . , 2005; Gleason et al . , 2005 ) and evolution of social organization ( Howard and Blomquist , 2005; Richard and Hunt , 2013 ) . The extensive genetic resources available for Drosophila melanogaster make it a valuable model for studying the genetic basis of CHC production and natural variation in CHC composition . Mature D . melanogaster have sexually dimorphic CHCs ranging from chain lengths of 21 to 31 carbons ( C21–C31 ) ( Antony and Jallon , 1982; Jallon and David , 1987 ) . Males produce predominantly shorter-chain CHCs ( < C26 ) and they use two of the CHCs , 7-C23:1 and 7-C25:1 , as sex pheromones . Females produce predominantly longer-chain dienes , among which 7 , 11-C27:2 and 7 , 11-C29:2 act as the primary female sex pheromone components ( Antony and Jallon , 1982; Cobb and Jallon , 1990; Arienti et al . , 2010 ) . CHCs are produced from the fatty acid biosynthetic pathway , which begins with an acetyl-CoA . Acetyl-CoA carboxylase ( ACC ) then catalyzes the synthesis of malonyl-CoA , and the multifunctional protein fatty acid synthase ( FASN ) successively incorporates malonyl-CoA units onto the acetyl-CoA , elongating the chain by two carbons each time and forming long chain fatty acids ( LCFA ) . RNAi-knockdown of ACC in the oenocytes completely eliminates CHCs in both male and female D . melanogaster ( Wicker-Thomas et al . , 2015 ) . Products of insect FASN , including Drosophila FASN , can be 14 , 16 or 18 carbon fatty acids . A thioesterase that is part of the multienzyme FASN removes the elongated chain as a free fatty acid , and fatty acid elongation and desaturation use the CoA derivative and take place in the microsomal fraction ( endoplasmic reticulum ) . Members of a family of tissue-specific elongases ( ELOVL ) catalyze the incorporation of malonyl-CoA units to form very long chain fatty acids ( VLCFA ) . After condensation of the malonyl-CoA with the fatty acyl-CoA the next three steps in each elongation cycle include reduction of a carbonyl to an alcohol ( by a 3-keto-acyl-CoA-reductase; KAR ) , dehydration ( by a 3-hydroxy-acyl-CoA-dehydratase; HADC ) , and reduction of the carbon-carbon double bond by a trans-enoyl-CoA-reductase ( TER ) . The VLCA as the CoA derivative is readuced to an aldehyde by a fatty acyl-CoA reductase ( FAR ) . The one-carbon chain-shortening conversion of aldehydes to hydrocarbons is catalyzed by a cytochrome P450 enzyme , Cyp4G1 ( Qiu et al . , 2012 ) . Desaturation reactions to introduce double bonds , leading to unsaturated fatty acids , appear to occur on the 16 or 18 carbon fatty acyl-CoAs . A comprehensive analysis of CHCs in both sexes segregating in a panel of recombinant inbred lines derived from a natural population identified 25 quantitative trait loci ( QTL ) in females and 15 in males contributing to variation in CHCs ( Foley et al . , 2007 ) , but this study did not have the power to resolve QTLs to individual genes . QTL mapping analyses have also identified genomic regions called small monoene quantities ( smoq ) and seven pentacosene ( sept ) , respectively , associated with variation in the proportions of 7-C23:1 and 7-C25:1 in males ( Ferveur and Jallon , 1996 ) . In a mutagenesis study , another genomic region , nerd , drastically reduced 7-C23:1 production in males and altered courtship behavior ( Ferveur and Jallon , 1993 ) . However , none of these male loci have been resolved to specific genes . Several additional genes affecting CHC biosynthesis have been described in D . melanogaster: an acetyl-CoA carboxylase ( ACC ) ( Wicker-Thomas et al . , 2015 ) ; two fatty acid synthases , FASN2 ( CG3524 ) which is involved in the synthesis of precursors of 2-methylalkanes ( Chung et al . 2014 ) , and FASN1 ( CG3523 ) which is expressed in the fat body but nonetheless contributes to the CHC pool ( Wicker-Thomas et al . , 2015 ) ; an elongase ( EloF , Wicker-Thomas et al . , 1997; Chertemps et al . , 2007 ) ; KAR ( CG1444 ) , TER ( CG10849 ) and HADC ( CG6746 ) whose knockdown eliminates or reduces CHC synthesis ( Wicker-Thomas et al . , 2015 ) ; a cytochrome P450 ( Cyp4G1 [CG3972] , Qiu et al . , 2012 ) ; and three desaturases , Desat1 , Desat2 and DesatF ( Labeur et al . , 2002; Marcillac et al . , 2005; Chertemps et al . , 2006; Fang et al . , 2009 ) . Here , we used the sequenced , inbred lines of the D . melanogaster Genetic Reference Panel ( DGRP ) ( Mackay et al . , 2012; Huang et al . , 2014 ) to perform genome wide association ( GWA ) analyses for nearly all detectable CHCs in both sexes in a scenario where all common genetic variants are genotyped and local linkage disequilibrium ( LD ) is sufficiently low to identify candidate genes and causal polymorphisms . We found considerable heritable genetic variation in a majority of male and female CHCs , distilled the axes of genetic variation into several principal components ( PCs ) , and performed GWA analyses on each PC . We identified 24 candidate genes plausibly associated with CHC biosynthesis and for all of them disruption of their expression altered CHC profiles in males , females , or both sexes . Surprisingly , we also found that the DGRP lines are segregating for the ancestral and deletion alleles in Desat2 , previously associated with CHC profiles thought to be unique for African flies . Finally , our results provide a new perspective and highlight the complexity of the biosynthetic and catabolic pathways that regulate the dynamics of CHC composition and provide the stage for adaptive evolution . In agreement with other recent studies of complex traits , our results demonstrate that the genetic architecture underlying potentially adaptive traits can consist of many , even hundreds , of polymorphic loci with small effects affecting different aspects of the phenotype . In this article , linear alkanes are referred to by the abbreviation n-Cx , where ‘‘x’’ is the total carbon number . For example , tricosane is designated n-C23 . Methyl-branched alkanes are referred to with a y-Me-Cx prefix , where 'y' is the carbon onto which the methyl group is bound . For example , 9-Me-C23 is a 23 carbon chain with a methyl on the 9th carbon . For monoenes ( z-Cx:1 ) and dienes ( z , z-Cx:2 ) the number of double bonds is indicated after the colon and the double bond position/s are 'z' or 'z , z" ) . For example , 7-C23:1 has one double bond between the 7th and 8th carbons . We identified and quantified 71 female CHCs and 42 male CHCs in 169 and 157 DGRP lines , respectively ( Table 1 , Figure 1 , Supplementary file 1 ) . Sixteen of these CHCs have not been described previously in D . melanogaster . Eight of the new compounds were methyl-branched CHCs , seven were dienes , and one was a monoene . Nine of these compounds were detected only in females . 10 . 7554/eLife . 09861 . 003Table 1 . Cuticular lipids identified by GC-MS in DGRP males and females . NI = not identified; nd = not detected; bold typeface = not previously identified in D . melanogaster . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 003#Cuticular componentRetention index#Cuticular componentRetention index♀♂♀♂1n-C212100210033NI2516nd2x-C22:1 ( quantified only in ♂ ) 2179217934NI2521nd3x-C22:1nd218435 13-Me-C25 11-Me-C25 2533ndccis-vaccenyl acetatend218936 5-Me-C25 2550nd4n-C222200220037 8 , 12-C26:2 2555nd57 , 11-C23:22259nd387 , 11-C26:22560nd62-Me-C2222632263392-Me-C25256225627NInd2267406 , 10-C26:22566nd89-C23:12273227341 9-C26:1 ( only in ♀ ) 3-Me-C25 2572257297-C23:122802283427-C26:125772577106-C23:122852286436-C26:1 + impurity ( i ) 25812581115-C23:12291229144n-C262600260012x-C23:12294nd459 , 13-C27:22652nd13n-C2323002300467 , 11-C27:2 ( only in ♀ ) 2-Me-C262664266314 11-Me-C23 9-Me-C23 23362336475 , 9-C27:2 ( only in ♀ ) 9-C27:126752675157 , 11-C24:22355nd487-C27:12682268216 x , y-C24:2 23632363495-C27:12693nd173-Me-C23 9-C24:1 ( only in ♀ ) 2373237350n-C2727002700187-C24:1 ( quantified only in ♂ ) 2377237751 8 , 12-C28:2 2756nd196-C24:123802380522-Me-C27 7 , 11-C28:2 ( only in ♀ ) 2761276120NI ( Variable in DGRP lines but not detected in GC-MS samples ) ndnd53 6 , 10-C28:2 3-Me-C27 9-C28:12768nd215-C24:12386238654NI2772277222n-C242400240055n-C2828002800239 , 13-C25:22451nd569 , 13-C29:22852nd247 , 11-C25:22460nd572-Me-C28 7 , 11-C29:2 ( only in ♀ ) 28642862252-Me-C2424632463589-C29:1 5 , 9-C29:1 ( only in ♀ ) 2875287526 6 , 10-C25:2 24682468597-C29:128822882275 , 9-C25:2 ( only in ♀ ) 9-C25:12474247460n-C2929002900288-C25:12478nd61 8 , 12-C30:2 ( only in ♀ ) 2-Me-C29 29612961297-C25:124822483622-Me-C303060306030 6-C25:1 2485nd637 , 11-C31:23065nd315-C25:12492249264n-C313100310032n-C2525002500ISn-C323200320010 . 7554/eLife . 09861 . 004Figure 1 . Representative male and female chromatograms from the DGRP . Male cuticular lipids of DGRP_38 are shown on the top ( blue ) and female CHCs of DGRP_786 are mirrored below ( red ) . All peaks for both sexes were assigned a unique number based on its corresponding compound determined by GC-MS; thus compounds shared between the sexes carry the same number . See Table 1 for the list of compound names . Compounds not previously described in D . melanogaster are shown in bold typeface . Some CHC isomers were not resolved by conventional GC , so a few chromatogram peaks contain more than one CHC . pA = picoAmperes , c = cis-vaccenyl acetate , * = contaminants from CHC extraction , IS = internal standard ( n-C32 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 004 We assessed the extent to which the CHCs were genetically variable in the DGRP . All but three female CHCs and one male CHC ( female peaks 26 , 39 , and 52 and male peak 59 ) showed significant among-line variation in a univariate ANOVA ( Supplementary file 2 ) . Broad sense heritabilities in females ranged from 0 . 98 for 7 , 11-C25:2 ( peak 24 ) to 0 . 22 for 6-C25:1 ( peak 30 ) . Broad sense heritabilities for males ranged from 0 . 97 for 7-C25:1 ( peak 29 ) to 0 for 9-C29:1 ( peak 58; Supplementary file 2 ) . The Desat locus was the first desaturase gene sequence described in insects ( Wicker-Thomas et al . , 1997 ) and has been implicated in the biosynthesis of pheromones . It consists of two desaturase genes , Desat1 and Desat2 . Desat1 is expressed in both sexes and encodes a Δ-9 desaturase that catalyzes the synthesis of palmitoleic acid , an ω-7 fatty acid and precursor to 7-monoene and the first double bond of 7 , 11-dienes ( Labeur et al . , 2002; Marcillac et al . , 2005 ) . Desat2 has a female-specific effect on CHC production and has been associated with adaptive divergence of African and Cosmopolitan races of D . melanogaster ( Greenberg et al . , 2003 , but see Coyne and Elwyn , 2006 ) . Desat2 encodes a functional desaturase in African females , but is inactive in Cosmopolitan females due to a 16-bp deletion in the promoter region ( Coyne et al . , 1999; Dallerac et al . , 2000 , Takahashi et al . , 2001 ) . Females with an intact Desat2 gene produce altered CHC profiles which are high in the pheromonal CHC isomers , 5 , 9-C27:2 and 5 , 9-C29:2 , and low in the 7 , 11-isomers . African D . melanogaster females also exhibit a strong behavioral bias against non-African males ( Fang et al . , 2002; Takahashi and Ting , 2004; Grillet et al . , 2012 ) . Surprisingly , females of 15 DGRP lines expressed the African phenotype; i . e . , they had high levels of 5 , 9-C27:2 and low levels of the primary female sex pheromone , 7 , 11-C27:2 ( Figure 2A ) . Since Desat2 ( cytological position 87B10 ) is located within the common African inversion In3R ( K ) ( computed breakpoints: 86F1-86F11;96F11-96F14 ( St . Pierre et al . , 2014 ) we expected the Desat2 allele status ( ancestral African or a Cosmopolitan 16-bp deletion in the promoter ) , CHC phenotype and inversion status to perfectly co-segregate in the DGRP lines . However , this was not the case . A total of 17 DGRP lines with female CHC phenotypes contained the ancestral , functional Desat2 allele ( Table 2 , Figure 2B ) . There was significant variation in the 7 , 11- and 5 , 9-C27:2 peaks among these lines ( 5 , 9-C27:2 & 9-C27:1 F = 35 . 17 , p<0 . 0001; 7 , 11-C27:2 & 2-Me-C26 F = 16 . 09 , p<0 . 0001; Supplementary file 3 ) . We obtained female CHC data for two African lines , Z53 and Z30 , for comparison . Females had significantly less 7 , 11-C27:2 according to deletion status and correspondingly more 5 , 9-C27:2 ( Figure 2C ) . Females from DGRP lines that were either homozygous or heterozygous for the ancestral Desat2 sequence had intermediate amounts of the 7 , 11- and 5 , 9-C27:2 compared to the DGRP lines homozygous for the deletion or the African lines , respectively . 10 . 7554/eLife . 09861 . 005Figure 2 . DGRP lines segregate for the female African CHC phenotype , Desat2 allele , and In3R ( K ) inversion status . ( A ) Overlaid chromatograms of African D . melanogaster CHCs ( Z30 and Z53 ) , a DGRP line with an African-like CHC phenotype ( DGRP_235 ) , and a Cosmopolitan DGRP line ( DGRP_714 ) . ( B ) DGRP lines with at least one ancestral Desat2 allele exhibit natural variation in the percentage of each CHC peak for the isomeric sex pheromones 7 , 11-C27:2 ( 2-Me-C26 co-elutes with 7 , 11-C27:2 ) ( gray ) and 5 , 9-C27:2 ( 9-C27:1 co-elutes with 5 , 9-C27:2 ) ( red ) . ( C ) Box-plots of the proportion of each sex pheromone peak for DGRP and African lines according to Desat2 allele and In3R ( K ) genotypes . DGRP_105 and DGRP_551 , which have more Cosmopolitan-like phenotypes despite having the functional ancestral Desat2 allele , are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 00510 . 7554/eLife . 09861 . 006Table 2 . Phenotypes and In ( 3R ) K genotypes for females from DGRP lines with functional Desat2 alleles . Red text indicates "mismatched" Desat2 genotype ( '+' = ancestral; '-' = 16-bp deletion ) and inversion status ( ‘INV’ = In ( 3R ) K; ‘ST’ = Standard karyotype ) . Blue background indicates "mismatched" Desat2 genotype and phenotype . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 006DGRP lineDesat 2 genotype% 7 , 11-C27:2% 5 , 9-C27:2RatioIn ( 3R ) K statusDGRP_31 + / -23 . 718 . 71 . 27INV / STDGRP_38 + / -19 . 424 . 90 . 78INV / STDGRP_48 + / -19 . 125 . 00 . 76INV / STDGRP_100 + / +22 . 332 . 30 . 69INV / INVDGRP_136 + / -14 . 733 . 70 . 44INV / STDGRP_309 + / -19 . 117 . 81 . 07INV / STDGRP_440 + / -20 . 014 . 01 . 43INV / STDGRP_559 + / -23 . 510 . 52 . 24INV / STDGRP_646 + / +17 . 324 . 60 . 70INV / INVDGRP_802 + / -18 . 523 . 40 . 79INV / STDGRP_732 + / -19 . 317 . 31 . 12INV / STDGRP_367 + / +12 . 636 . 80 . 34ST/ STDGRP_776 + / +16 . 78 . 981 . 86ST / STDGRP_509 + / +17 . 827 . 50 . 65ST / STDGRP_235 + / +10 . 322 . 20 . 46ST / STDGRP_551 + / -26 . 64 . 296 . 20ST / STDGRP_105 + / +21 . 45 . 104 . 20INV / INV The co-segregation of the functional allele and In3R ( K ) was not perfect: five of the lines with the ancestral Desat2 sequence were homozygous for the standard karyotype ( Figure 2C , Table 2 ) . Furthermore , two lines with the ancestral and presumably functional Desat2 , DGRP_105 and DGRP_551 , did not exhibit the African CHC phenotype . In total , six of the 17 lines with the ancestral sequence had mismatched inversion and CHC status in females . It is possible that the lines with homozygous standard karyotypes are actually segregating for the inversion at low frequency and thus only the homozygous standard flies were sampled for karyotyping . Alternatively , the Desat2 deletion may have occurred prior to the inversion event . We next checked Desat2 for potentially damaging genetic variants ( Mackay et al . , 2012; Huang et al . , 2014 ) . We identified five Desat2 alleles unique to DGRP_105 . Two variants are synonymous coding ( 3R_8262545_SNP and 3R_8263020_SNP ) , one is a deletion causing a frameshift ( 3R_8263023_DEL ) , and two are nonsynonymous coding ( 3R_8263031_MNP and 3R_8263048_SNP ) . The frameshift and nonsynonymous variants are potentially damaging and could explain why this line produced the Cosmopolitan phenotype despite having the functional Desat2 allele . We did not find such evidence for DGRP_551 , suggesting this line may contain unknown genetic variants that inhibit the production of 5 , 9-C27:2 . Most of the CHCs belong to homologous series in which the chain length increases by two carbons; thus , these compounds may be genetically correlated due to a shared biosynthetic pathway and the data may be confounded by multi-colinearity ( Martin and Drijfhout , 2009 ) . We visualized the correlations between CHCs using modulated modularity clustering ( MMC ) ( Stone and Ayroles , 2009 ) . The MMC algorithm clusters highly correlated variables based on the Spearman's rank correlation coefficients ( ρ ) . As expected , some CHCs were highly correlated within each sex ( Figures 3 and 4; Supplementary file 4 ) . 10 . 7554/eLife . 09861 . 007Figure 3 . MMC modules of DGRP female CHCs based on Spearman's rank correlation coefficients ( ρ ) . Correlations are color-coded from +1 ( dark red ) to -1 ( dark blue ) . Correlated CHCs are clustered into groups ( modules ) . Modules ( outlined in black ) are arranged along the diagonal according to the average strengths of the correlations within each cluster; the most strongly correlated modules are on the top left and the weakly correlated modules are on the bottom right . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 00710 . 7554/eLife . 09861 . 008Figure 4 . MMC modules of DGRP male CHCs based on Spearman's rank correlation coefficients ( ρ ) . Correlations are color-coded +1 ( dark red ) to -1 ( dark blue ) . Correlated CHCs are clustered into groups ( modules ) . Modules ( outlined in black ) are arranged along the diagonal according to the average strengths of the correlations within the groups; the most strongly and weakly correlated are on the top left and bottom right , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 008 In the first two female modules , seven dienes had strong positive correlations with each other . There was also one peak ( 47 – 5 , 9-C27:2 and 9-C27:1 ) that strongly negatively correlated with those dienes and the shorter-chain ( ≤ C25 ) alkanes of module 3 ( Figure 3 , Supplementary file 4 ) . Similarly , the module 3 shorter-chain alkanes had strong positive correlations with each other , some dienes of modules 1 and 2 , and monoenes and dienes in module 5 . These four modules ( 1 , 2 , 3 , and 5 ) all had weak to moderate negative correlations with module 7 , which consisted of strongly intercorrelated longer-chain ( ≥ C25 ) alkanes and dienes . We found similar trends in the male CHCs ( Figure 4; Supplementary file 4 ) . Module 1 consisted of longer-chain alkanes that negatively correlated with the shorter-chain alkanes of module 2 . This could also be seen between module 2 and the longer-chain monoenes in module 7 . These negative correlations were a consistent trend between other long- and short-chain CHCs and exemplify the tradeoff between short- and long-chain compounds , as the latter are produced through the elongation of fatty acids that serve as precursors for the former . The correlations among CHCs are plausible given the biology of CHC production . We computed PCs of genetically variable CHCs within each sex to reduce the dimensionality of the data to orthogonal PCs . The first seven and five PCs accounted for 98 . 00% and 98 . 12% of the total variation among the DGRP lines for female and male CHCs , respectively ( Table 3 , Figure 5 , Supplementary file 5 ) . We hypothesized that the GWA results would provide insights into the genetic architecture underlying the MMC trends . 10 . 7554/eLife . 09861 . 009Figure 5 . Principal component biplots for PC1 and PC2 of DGRP CHCs . ( A ) Female and ( C ) male PC1 and PC2 . ( B ) Female and ( D ) male PC1 and PC2 eigenvectors . The percent of variance explained by each PC is indicated on the x- and y-axes . In ( A ) and ( C ) DGRP lines are color-coded ( Supplementary file 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 00910 . 7554/eLife . 09861 . 010Table 3 . Percent of CHC variation in the DGRP explained by PCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 010SexNumberEigenvaluePercentCumulative percentFemale 10 . 006141 . 1641 . 1620 . 004329 . 4770 . 6330 . 002114 . 5085 . 1340 . 00096 . 2291 . 3550 . 00053 . 0794 . 4260 . 00032 . 3096 . 7270 . 00021 . 2998 . 01Male 10 . 017075 . 5275 . 5220 . 003314 . 5790 . 1030 . 00104 . 5994 . 6940 . 00052 . 0496 . 7350 . 00031 . 3998 . 12 We performed GWA analyses using the PCs of natural variation in individual chromatographic peaks to identify novel components of the CHC metabolic pathways in D . melanogaster . None of the female or male PCs were affected by the presence of the endosymbiont Wolbachia pipientis in some of the DGRP lines ( Supplementary file 6 ) . Female PC1 and PC2 and male PC1 were affected by the In ( 2L ) t inversion; female PC1 , PC2 and PC3 , and male PC5 were affected by the In ( 3R ) K inversion; and the In ( 3R ) P inversion only affected male PCs ( PC1 , PC3 , PC5 ) ( Supplementary file 6 ) . We corrected for these effects prior to conducting the GWA analyses . Although these inversions contain genetic variants affecting CHC production , they cannot be resolved by GWA analysis due to elevated linkage disequilibrium within the inversions ( Mackay et al . , 2012; Huang et al . , 2014 ) and are thus excluded from consideration . We identified genetic variants in or near 305 ( 173 ) genes nominally ( P≤10–5 ) associated with female ( male ) PCs ( Supplementary file 7 ) . Although all of the top variants did not reach individual Bonferroni-corrected significance levels , most quantile-quantile plots indicated no systematic inflation of the test statistic and a clear departure from random expectation below P<10–5 , justifying our choice of this reporting threshold and suggesting that the top associations were enriched for true positives ( Figure 6 ) . 10 . 7554/eLife . 09861 . 011Figure 6 . QQ-plots of CHC PCA GWA P-values . ( A–G ) Female PCs . ( H–L ) Male PCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 011 Several of the top variants associated with each PC are in or near candidate genes with plausible roles in CHC metabolism ( Supplementary file 7 ) . In females , these include a single nucleotide polymorphism ( SNP ) in the fourth intron of Lipase2 ( Lip2 ) associated with variation in PC1; a SNP 641 bp upstream of the cytochrome P450 gene , Cyp49a1 , associated with PC2; SNPs in and near Desiccate ( Desi ) , a gene previously shown to contribute to desiccation resistance in D . melanogaster , associated with PC4 and PC6; variants in and near two peroxidase genes , Immune-regulated catalase ( Irc ) and Peroxidase ( Pxd ) , associated with PC5; and variants in three fatty acid metabolism genes ( CG14688 , CG9458 , CG8814 ) associated with variation in PC6 . Several notable candidate genes were also implicated by variants associated with PC7; CG9801 contained the 3 most significant SNPs . One of the top associated SNPs causes a missense mutation in Cyp6w1 . Cyp4s3 and a predicted dihydrolipoamide branched chain acyltransferase ( St . Pierre et al . , 2014 ) , CG5599 , were down- and up-stream , respectively , of associated variants . We did not find any associated variants within or near Desat1 or Desat2 . In the case of Desat2 the ancestral allele was not tested for association , because there were only six DGRP lines homozygous for the insertion , so its minor allele frequency ( MAF = 6/169 = 0 . 035 ) did not reach the MAF cutoff ( 0 . 05 ) for evaluation . Further , any effect of this variant would have been minimized by correcting for the effect of the In ( 3R ) K inversion . In males , nearly all of the variants associated with variation in PC1 were in or near CG13091 , a putative fatty acyl-CoA reductase , of which one , a nonsynonymous coding variant ( 2L_8521314_SNP ) , was the most significant variant in this study ( P = 2 . 19E-11 ) and passes the Bonferroni-correction for multiple tests ( Supplementary file 7 ) . Many of the variants in CG13091 were in perfect or near-perfect linkage disequilibrium , and thus it is not possible to discern which was/were the causal variant ( s ) . Variants in CG13091 were also associated with variation in PC4 and PC5 . Genes tagged by variants associated with PC2 include approximated ( app ) , a palmitoyltransferase; PHGPx , a peroxidase; and CG16979 , a thiolester hydrolase . Top variants in the PC3 GWA analysis included two genes predicted to be fatty acid elongases ( CG18609 , CG30008 ) and an NADH dehydrogenase ( CG8680 ) . The most significant variant in the PC5 GWA analysis was a nonsynonymous SNP ( 3R_8220563_SNP , P = 1 . 97E-09 ) in CG10097 , which also encodes a fatty acyl-CoA reductase . Interestingly , there are two cytochrome P450 genes located upstream of CG10097 , one functional ( Cyp9f2 ) and one pseudogenized ( Cyp9f3Psi ) . We selected 24 candidate genes with plausible contributions to various stages in the biosynthesis or turnover of CHCs and tested the effects of disruption of expression of these genes on CHC composition . All but one of these genes had publicly available transgenic UAS-RNAi lines ( Dietzl et al . , 2007 ) . We used an oenocyte-specific GAL4 driver line , PromE ( 800 ) -GAL4 ( Billeter et al . , 2009 ) , to restrict reduction in candidate gene expression to the CHC-producing cells . For the gene for which no RNAi line was available , CG10097 , we used a PiggyBac insertion line to study the effects of this mutation on CHC composition ( Supplementary file 8 ) ( Thibault et al . , 2004 ) . The candidate genes implicated by our GWA analyses that we selected for further functional assessment are annotated to encode a palmitoyl transferase ( app ) , fatty acyl reductases ( CG13091 and CG10097 ) , a thiol hydrolase ( CG16979 ) , thioredoxin peroxidases ( PHGPx and Prx6005 ) , fatty acid elongases ( CG30008 , CG18609 , and CG9458 ) , cytochrome P450s ( Cyp49a1 , Cyp9f2 , and Cyp4s3 ) , peroxidases ( Irc , Pxd and Pxn ) , an NADH dehydrogenase ( CG8680 ) , and a dihydrolipoamide branched chain acyltransferase ( CG5599 ) , which is involved in the metabolism of branched chain amino acids leucine , isoleucine and valine , which serve as precursors for methyl-branched alkanes ( Blomquist and Bagnères , 2010 ) . Disruption of expression of any of these genes by targeted RNA interference resulted in altered CHC compositions , often with sexually dimorphic effects . The alterations in CHC composition as a consequence of gene disruption were often complex and unexpected ( Figures 7 and 8 , Figure 7—figure supplements 1–24 ) . 10 . 7554/eLife . 09861 . 012Figure 7 . Summary of RNAi and mutant experiments for female CHCs . UAS-RNAi target gene and the CG10097e00276 mutant are indicated on the horizontal axis . CHC names and numbers are listed on the y-axis . Data are color coded to represent P-values ( P ≤ 0 . 05 ) from t-tests for the mean differences of the experimental and the control lines . Black = no significant change; blue = significant decrease; green = significant increase; gray = not applicable ( peaks 46 and 57 split into two peaks for the CG10097 mutant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01210 . 7554/eLife . 09861 . 013Figure 7—figure supplement 1 . Functional validation PCA and total CHCs for RNAi-app . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , *: P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01310 . 7554/eLife . 09861 . 014Figure 7—figure supplement 2 . Functional validation PCA and total CHCs for RNAi-CG5599 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01410 . 7554/eLife . 09861 . 015Figure 7—figure supplement 3 . Functional validation PCA and total CHCs for RNAi-CG7724 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01510 . 7554/eLife . 09861 . 016Figure 7—figure supplement 4 . Functional validation PCA and total CHCs for RNAi-CG8680 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01610 . 7554/eLife . 09861 . 017Figure 7—figure supplement 5 . Functional validation PCA and total CHCs for RNAi-CG8814 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01710 . 7554/eLife . 09861 . 018Figure 7—figure supplement 6 . Functional validation PCA and total CHCs for RNAi-CG9458 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01810 . 7554/eLife . 09861 . 019Figure 7—figure supplement 7 . Functional validation PCA and total CHCs for RNAi-CG9801 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 01910 . 7554/eLife . 09861 . 020Figure 7—figure supplement 8 . Functional validation PCA and total CHCs for mutant CG10097 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02010 . 7554/eLife . 09861 . 021Figure 7—figure supplement 9 . Functional validation PCA and total CHCs for RNAi-CG13091 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02110 . 7554/eLife . 09861 . 022Figure 7—figure supplement 10 . Functional validation PCA and total CHCs for RNAi-CG14688 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02210 . 7554/eLife . 09861 . 023Figure 7—figure supplement 11 . Functional validation PCA and total CHCs for RNAi-CG16979 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02310 . 7554/eLife . 09861 . 024Figure 7—figure supplement 12 . Functional validation PCA and total CHCs for RNAi-CG18609 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02410 . 7554/eLife . 09861 . 025Figure 7—figure supplement 13 . Functional validation PCA and total CHCs for RNAi-CG30008 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02510 . 7554/eLife . 09861 . 026Figure 7—figure supplement 14 . Functional validation PCA and total CHCs for RNAi-Cyp4s3 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02610 . 7554/eLife . 09861 . 027Figure 7—figure supplement 15 . Functional validation PCA and total CHCs for RNAi-Cyp9f2 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02710 . 7554/eLife . 09861 . 028Figure 7—figure supplement 16 . Functional validation PCA and total CHCs for RNAi-Cyp49a1 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02810 . 7554/eLife . 09861 . 029Figure 7—figure supplement 17 . Functional validation PCA and total CHCs for RNAi-Desi . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 02910 . 7554/eLife . 09861 . 030Figure 7—figure supplement 18 . Functional validation PCA and total CHCs for RNAi-Irc . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 03010 . 7554/eLife . 09861 . 031Figure 7—figure supplement 19 . Functional validation PCA and total CHCs for RNAi-Lip2 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 03110 . 7554/eLife . 09861 . 032Figure 7—figure supplement 20 . Functional validation PCA and total CHCs for RNAi-Nrt . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 03210 . 7554/eLife . 09861 . 033Figure 7—figure supplement 21 . Functional validation PCA and total CHCs for RNAi-PHGPx . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 03310 . 7554/eLife . 09861 . 034Figure 7—figure supplement 22 . Functional validation PCA and total CHCs for RNAi-Prx6005 . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 03410 . 7554/eLife . 09861 . 035Figure 7—figure supplement 23 . Functional validation PCA and total CHCs for RNAi-Pxd . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 03510 . 7554/eLife . 09861 . 036Figure 7—figure supplement 24 . Functional validation PCA and total CHCs for RNAi-pxn . ( A ) PCA biplots for females and males , ○ = female , ○ = male , and ● = control samples . ( B ) PC1 and PC2 eigenvectors . ( C ) Box-plots of female and male total amount of CHCs ( µg/fly ) . P-values are reported for the Satterthwaite test , : P < 0 . 05 , : P < 0 . 01 , : P < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 03610 . 7554/eLife . 09861 . 037Figure 8 . Summary of RNAi and mutant experiments for male CHCs . UAS-RNAi target gene and the CG10097 e00276 mutant are indicated on the horizontal axis . CHC names and numbers are listed on the y-axis . Data are color coded to represent P-values ( P ≤ 0 . 05 ) from t-tests for the mean differences of the experimental and the control lines . Black = no significant change; blue = significant decrease; green = significant increase; gray = not applicable ( peaks 46 and 57 split into two peaks for the CG10097 mutant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 037 In insects , the evidence to date suggests that the elongated fatty acyl-CoA is reduced to an aldehyde prior to oxidative decarbonylation , the latter step being catalyzed by CYP4G1 in Drosophila ( Qiu et al . , 2012 ) . While it is possible that the elongated fatty acyl-CoA is reduced from acyl-CoA to aldehyde to alcohol and then reoxidized to aldehyde before oxidative decarbonylation , to date there is no evidence for this . CG13091 and CG10097 encode fatty acyl-CoA reductases , which are both expressed at high levels in males and at low levels in females . FARs reduce fatty acyl-CoA to an alcohol ( Howard and Blomquist 2005 ) . Inhibition of the production of these FARs promoted the production of longer chain CHCs . Males in which CG13091 expression was targeted with RNAi had higher relative amounts of longer-chain CHCs in general and reduced shorter-chain monoenes and methyl-branched CHCs . In particular , 7-C25:1 and 7-C27:1 and 2-MeC26 , and 2-MeC28 increased substantially ( Figures 8 and 9 ) . The increase in 7-C25:1 is of particular interest , as this compound acts as a male sex pheromone that mediates female mate choice . Females showed a similar trend but to a lesser extent; they also had some increased longer-chain dienes and , like the males , increased 7-C27:1 , 7 , 11-C27:2 & 2-MeC26 , and 2-Me-C28 & 7 , 11-C29:2 ( Figures 7 and 9 ) . Both sexes of the CG10097 mutant were similar , with higher relative amounts of longer-chain CHCs . They also had lower amounts of shorter-chain monoenes and methyl-branched CHCs . In the CG10097 mutant and control females , we were able to separate the 7 , 11-C27:2 and 2-Me-C26 peaks as well as the 2-Me-C28 and 7 , 11-C29:2 peaks ( Figure 9 ) , enabling us to infer decreased 7 , 11-C27:2 and 7 , 11-C29:2 and increased 2-Me-C28 relative to the control . Our results suggest that the FARs encoded by CG13091 and CG10097 may be specifically associated with alkane and monoene synthesis . 10 . 7554/eLife . 09861 . 038Figure 9 . Example chromatograms of oenocyte-specific RNAi knockdowns and mutants – CG13091 and CG10097 . ( A ) and ( B ) PromE ( 800 ) -GAL4 x UAS-CG13091 . ( C ) and ( D ) Exelixis mutant CG10097 e00276 . pA = picoAmperes , IS = internal standard , ↑ CHCs significantly increased or ↓ decreased according to the individual t-tests . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 038 The complexity of effects on CHC composition through RNAi-targeting is illustrated by the diverse effects on CHC composition of the three different fatty acid elongases encoded by CG30008 , CG18609 and CG9458 . In males , CG30008 may play a role in the elongation of precursors of n-alkanes and monoenes , and CG18609 may elongate precursors of longer-chain n-alkanes and 2-methlyalkanes ( Wicker-Thomas and Chertemps , 2010 ) . Disruption of CG30008 expression in males resulted in greatly reduced amounts of total CHCs ( ~2-fold ) . The effect was less severe in females , but the overall trend was also towards reduced CHCs ( Figure 10 ) . Disruption of CG18609 decreased longer-chain CHCs in both sexes but increased 2-Me-C24 in females ( Figure 7 ) ( Wicker-Thomas and Chertemps , 2010 ) . Males also had increased total CHCs and large increases in shorter-chain CHCs ( Figure 8 ) . This effect was mimicked by interference with expression of Cyp9f2 , which also resulted in fewer longer-chain monoenes , alkanes , and many methyl-branched CHCs . However , as for CG18609 , 2-Me-C24 increased in females and other shorter-chain methyl-branched CHCs were trending upward . In males , there were overall increases , but especially in the shorter-chain CHCs . These observations suggest that CG18609 may function in the biosynthesis of CHCs in coordination with Cyp9f2 . Interestingly , the CG9458 knockdown had sexually dimorphic effects on CHC production , increasing male CHCs and decreasing nearly all n-alkanes and monoenes in females . This suggests that CG9458 is critical for the biosynthesis of n-alkanes and monoenes in females . 10 . 7554/eLife . 09861 . 039Figure 10 . Example chromatograms of oenocyte-specific RNAi knockdowns – CG8680 and CG30008 . ( A ) and ( B ) PromE ( 800 ) -GAL4 x UAS-CG8680 . ( C ) and ( D ) PromE ( 800 ) -GAL4 x UAS-CG30008 . pA = picoAmperes , IS = internal standard , ↑ CHCs significantly increased or ↓ decreased according to the individual t-tests . DOI: http://dx . doi . org/10 . 7554/eLife . 09861 . 039 The three cytochrome P450s that we tested ( Cyp49a1 , Cyp9f2 , Cyp4s3 ) all affected overall amounts of CHCs . RNAi knockdown of Cyp49a1 and Cyp9f2 in males led to general increases in CHCs , with a few exceptions . The increase in CHCs in males with compromised Cyp49a1 function was accompanied by a decrease in many female CHCs . Reduced expression of Cyp4s3 resulted in increased CHC abundance in both sexes . The link between these CYPs and CHC production and maintenance is unclear . However , we speculate that oxidation reactions mediated by these CYPs may regulate CHC degradation and turnover . The specific functions of most insect CYPs are still unknown ( Chung et al . , 2009 ) . Given the role of CYP4G1 in CHC production ( Qiu et al . , 2012 ) , we believe it is possible for these CYPs to have similar functions that are perhaps specific to particular subsets of CHCs . The complex interrelationships that give rise to variation in sexually dimorphic CHC profiles is further illustrated by RNAi interference of Irc , Pxd , and Pxn , which all have corresponding alleles associated with variation in CHC composition in the DGRP and encode peroxidases . However , interference with their expression through targeted RNAi resulted in different shifts in CHC composition . Disruption of Irc reduced the amounts of monoenes and increased the amounts of dienes in females , while increasing male CHCs , reminiscent of the effects of disruption of Cyp49a1 . The effects of interference with Pxd were more complex . In females , nearly all CHCs increased , while in males many odd-chain CHCs increased , but there were decreases only in even-chain CHCs . A similar phenomenon was observed with disruption of Cyp4s3 . Pxn may be important for diene synthesis because the knockdown females had decreased dienes and corresponding increased levels of longer-chain monoenes . However , in males there were decreased longer-chain monoenes , methyl-branched alkanes , and n-alkanes but increased shorter-chain CHCs . CG7724 is inferred to contribute to oxidation-reduction processes and steroid synthesis . Disruption of CG7724 expression in females resulted in a decrease in longer-chain monoenes and methyl-branched CHCs and an increase in 2-Me-C22 , 2-Me-C24 , 2-Me-C25 , 2-Me-C26 and 7 , 11-C27:2 , x , y-C24:2 and x , y-C26:2 . Males also showed increases in 2-methyl alkanes but also in shorter-chain alkanes and monoenes . These results reveal a complex and dynamic network of oxidative enzymes of which the summed activity determines the sexually dimorphic composition of CHCs . Disruption of the NADH dehydrogenase CG8680 and of CG5599 , which is predicted to have dihydrolipoamide branched chain acyltransferase activity , resulted in remarkable increases in the total amount of CHCs in both males and females ( Figures 7 , 8 and 10 , Figure 7—figure supplements 2 , 4 , Supplementary file 9 ) . Individual control female and male flies produced ~1 . 5–2 µg and ~1–1 . 5 µg of CHCs , respectively; in contrast , the RNAi-CG8680 females and males produced ~5 . 5 µg and ~3 µg , respectively , of CHCs per fly ( Figure 7—figure supplement 4 ) . The CG5599 knockdown caused a >4 µg increase per fly in each sex , representing ~3-fold ( ~7 µg/fly ) and 4-fold ( ~5 . 5 µg/fly ) increase for females and males , respectively . Thus , it is possible that these genes play roles in catabolic pathways thus resulting in an increase in CHCs . Alternatively , RNAi knockdown of these genes may lead to reduced or abnormal grooming behavior leading to an accumulation of CHCs ( Böröczky et al . , 2013 ) . Inhibition of expression of Pxd , CG8814 and Cyp4s3 also resulted in increased levels of total CHCs in both sexes , with notable decreases in 2-Me-C26 in Pxd and Cyp4s3 males ( Figures 7 and 8 , Figure 7—figure supplements 5 , 14 , 23 ) . The gene products of app , PHGPx , and Prx6005 may be involved in the release of fatty acids from the fatty acid synthase complex . The app palmitoyltransferase may be specific to methyl-branched fatty acids since we observe decreases in methyl-branched CHCs in males and females when its expression is disrupted . Male and female PCAs were here clearly separated from the controls and in both sexes the total amounts of CHCs were slightly increased ( Figures 7 and 8 , Figure 7—figure supplement 1 ) . Additionally , males had elevated longer-chain methyl-branched alkanes , n-alkanes and monoenes , while females had increased abundances of n-alkanes , methyl-branched alkanes , monoenes and dienes of both longer- and shorter-chain lengths . RNAi targeting of Prx6005 increased shorter-chain monoenes and alkanes in both sexes . In females , longer-chain dienes and methyl-branched CHCs were trending downward , and in males levels of 2-Me-C26 and 2-Me-C24 decreased . Disruption of PHGPx increased the dienes 7 , 11-C25:2 and 9 , 13-C25:2 , and some longer-chain n-alkanes in females . Levels of many short-chain monoenes and nearly all 2-methyl alkanes were elevated in males . Furthermore , disruption of CG16979 , which encodes a gene product annotated as having thiolesterase activity , had no effect in females but caused many CHC increases in males , predominantly in monoenes . Total CHCs were also elevated ( Figures 7 and 8 ) . Thus , there appears to be functional specialization among thiolesterases in the biosynthesis of CHCs . We also assessed the effects of two candidate genes , Desi and Lip2 , associated with variation in CHC composition in the DGRP that affect desiccation resistance . Expression of Desi fluctuates in wandering D . melanogaster larvae in response to environmental conditions and RNAi knockdown of Desi in larvae leads to higher mortality ( Kawano et al . , 2010 ) . However , the nature of the resistance to desiccation is unknown and CHCs were not phenotyped in the larvae . In contrast to the GWA results , knocking down Desi had no effect on female CHCs , while the male CHCs clearly separated from the controls in the PCA ( Figure 7—figure supplement 17 ) . Males had increased alkanes , monoenes , and 2-Me-C28 ( Figure 8 ) . Lip2 has been associated with clinal variation of life history traits in D . melanogaster populations in the eastern United States ( Fabian et al . , 2012 ) , and has triglyceride lipase activity that could regulate release of free fatty acids for CHC synthesis . Females in which expression of Lip2 was targeted with RNAi had increased n-alkanes and increased total CHCs , whereas the males had increased 2- and 3-methyl alkanes ( Figure 7 ) . Finally , we examined the effects of CG14688 and CG9801; little is known about the function of both of these genes ( St . Pierre et al . , 2014 ) . Suppression of gene expression resulted in sexually dimorphic increases and decreases in overall CHCs . RNAi targeting of CG14688 increased female CHCs but had more complex effects in the males , increasing shorter-chain n-alkanes and methyl-branched alkanes and decreasing longer-chain monoenes . Males expressing RNAi against CG9801 had increased total CHC levels , while females of these lines had overall decreases in CHCs . In mammals , alpha-oxidation is used to chain-shorten 3-methyl-fatty acyl-CoAs by one carbon . We are not aware of any examples of alpha-oxidation in insects , but it could be involved in CHC metabolism . We found substantial heritable natural variation in CHC composition for males and females of the inbred , sequenced DGRP lines . Several of the epicuticular compounds identified in this study have not been reported previously for D . melanogaster ( Jallon and David , 1987; Foley et al . , 2007; Everaerts et al . , 2010; Dweck et al . , 2015 ) . These compounds separated in our GC analyses because we used a thin high-resolution column and a relatively long temperature program . Several of the newly identified monoenes and dienes had double bonds in an even carbon position , an unusual configuration in insects . While most of these new compounds represented a very small fraction of the total CHCs , two ( peak 51 = 8 , 12-C28:2 and peak 53 = 6 , 10-C28:2 , 9-C28:1 and 3-Me-C27 ) were female-specific and could potentially play a role in sexual communication . The African CHC phenotype has an abundance of 5 , 9-C27:2 and lower levels of 7 , 11-C27:2 , and is present only in populations from Sub-Saharan Africa and the Caribbean ( Coyne et al . , 1999 ) . Based on genotyping the 16-bp deletion/ancestral Desat2 allele , Caribbean populations are known to have spread northward into the southern United States; however , populations north of Alabama and Mississippi were thought to be nearly fixed for the Cosmopolitan deletion ( Yukilevich and True , 2008 ) . Surprisingly , we found that the DGRP is segregating for the African CHC phenotype . Thus , to the best of our knowledge the DGRP progenitor population at the North Carolina Farmers Market represents the northern most population of D . melanogaster with the African Desat2 allele documented to date . While 17 DGRP lines contain the ancestral allele that confers a functional Desat2 , only 15 of these lines exhibited the African phenotype . On average , females of DGRP lines that were heterozygous or homozygous for the functional allele , regardless of inversion status , had intermediate amounts of the sex pheromone CHC peaks relative to the African lines or DGRP lines with the deletion . However , one DGRP line , DGRP_367 , had more 5 , 9-C27:2 than either African line . Two DGRP lines ( DGRP_105 , ins/ins , INV/ INV and DGRP_551 , ins/del , ST/ ST ) had the functional Desat2 allele , yet exhibited the Cosmopolitan phenotype . These results are in contrast to previous reports of complete association of the ancestral allele with the African female CHC phenotype ( Dallerac et al . , 2000; Takahashi et al . , 2001 ) . Further analyses showed that other Desat2 polymorphisms in DGRP_105 likely result in a non-functional protein . However , the other Cosmopolitan-like line , DGRP_551 , is puzzling because it is heterozygous for the functional allele , but homozygous for the standard karyotype . It is possible that this line is segregating for the In ( 3R ) K inversion at low frequency and individuals with the ST/INV karyotype were not detected , but the lack of correspondence between the functional allele and CHC status remains unexplained since Desat2 itself does not harbor a potentially damaging mutation in this line . Perhaps a polymorphism in an unknown gene in DGRP_551 interacts with the Desat2 functional allele to suppress its effects . Finally , there were four DGRP lines that exhibited the African phenotype and were homozygous for the ancestral allele , but they were ST/ST in karyotype . One possible explanation for this 'mismatching' of phenotype and genotype with the karyotype is that the Desat2 16-bp deletion occurred prior to the inversion event . This would mean that the inversion may be segregating for the Desat2 16-bp allele . We used principal component analysis to reduce dimensionality of the data , which was motivated by the high co-linearity among the CHCs . However , PCA may cause genetic variation for a certain CHC to be distributed among multiple PCs and therefore dilute its association with QTLs . Surprisingly , we did not detect variants in the DGRP in previously identified CHC biosynthesis genes ( FASN1 , Desat1 , eloF , DesatF , Cyp4G1; and several genes reported in Wicker-Thomas et al . , 2015 ) associated with CHC variation . However , we did find many novel candidate genes , and functional tests showed that disruption of expression of all tested candidate genes had significant effects on the amount of CHC on the cuticular surface . While the mechanistic relationships between any of these genes are unknown , some share commonalities in their phenotypes when their expression is disrupted with RNAi . However , the majority of genes which encode gene products with similar molecular functions result in different shifts in CHC profiles when disrupted . Thus , variation in the CHC profile arises as an emergent phenotype from the dynamics of complex interrelated biosynthetic and catabolic pathways . The RNAi-induced shifts in CHC profiles are frequently sexually dimorphic . This could reflect different expression levels of metabolic enzymes associated with CHC production . However , we cannot exclude the possibility that differential effectiveness of RNAi in males and females may contribute to apparent sexual dimorphism . In addition , we note that phenotypic changes associated with knocking down expression of a target gene with RNAi may not be causal , but a consequence of off-target effects of RNAi or the GAL4 driver . Further studies are needed to clarify the effects on CHC production of these genes and their interactions , and to test specific mechanisms and enzymatic activities through which they exert these effects . Furthermore , it is important to note that the RNAi and mutant experiments test for effects at the level of genes and are only proxies for the effects of segregating natural variants in these genes . We recognize that the 24 candidate genes on which we focused represent only a subset of all candidate genes associated with variation in CHC composition . Many of these genes may directly or indirectly affect CHC composition through as yet unknown mechanisms . GWA studies of glucosinolates in Arabidopsis thaliana ( Chan et al . , 2011 ) , flowering time in maize ( Buckler et al . , 2009 ) , human height ( Yang et al . , 2010 ) , and other traits in D . melanogaster such as body pigmentation ( Dembeck et al . , 2015 ) have also shown that the genetic architecture underlying variation in potentially adaptive traits includes many polymorphic loci with small effect sizes . Thus there is no reason to assume that variation in adaptive traits is controlled by few , large effect loci . Our results provide a framework for future studies of the mechanisms that regulate CHC composition and their adaptive potential regarding cold/heat tolerance and desiccation resistance , and pleiotropic effects on chemical communication and mate choice . DGRP and African ( Z30 and Z53 ) lines were reared in vials containing cornmeal-molasses-agar medium at 25°C , 75% relative humidity , a 12:12-h light-dark cycle , and a controlled adult density of 10 males and 10 females . The parental generation was allowed to lay eggs for three days . Upon eclosion , virgin males and females were separated and placed into new vials containing the same medium and aged for four days . The flies from each line were separated into at least two samples of five flies each per sex . On average , three samples were collected for each . To avoid cross-contamination of cuticular lipids a fresh tissue paper was placed on the carbon dioxide pad and the flies were handled with acetone-washed titanium forceps at each round of sorting . Flies were placed in 2 mL glass auto-injection vials with a Teflon cap and were flash frozen . All samples were stored at -30°C until CHC extraction . We collected samples from 169 and 157 DGRP lines for females and males , respectively ( 1 , 078 total samples ) . All lines were reared simultaneously and DGRP lines that did not produce sufficient offspring for CHC analysis were excluded to avoid any block effects of rearing . For the two African lines , Z30 and Z53 , we reared and collected 5 samples for each sex . Cuticular lipids were extracted from each sample using 200 µl of hexane containing an internal standard ( IS , 1 µg n-C32 ) with gentle swirling for five minutes . The flies were briefly extracted a second time with 100 µl of hexane ( free of internal standard ) . After each wash the extract was transferred to a 300 µl conical glass insert . The extract was dried using a gentle stream of high-purity N2 and re-suspended in 50 µl of hexane . The samples were immediately processed using gas chromatography or stored at 4°C ( no longer than one day ) until processing . The cuticular lipid extracts were analyzed using an Agilent 7890A gas chromatograph with a DB-5 Agilent capillary column ( 20 m x 0 . 18 mm x 0 . 18 µm ) and a flame ionization detector ( FID ) for quantification . We introduced 1 µl of sample using an Agilent 7683B auto injector into a 290°C inlet operated in splitless mode . The split valve was turned on after 1 min . The oven temperature program was as follows: 50°C for one min , increased at 20°C/min to 150°C , and increased at 5°C/min to 300°C followed by a 10 min hold . Hydrogen was used as the carrier gas at constant flow ( average linear velocity = 35 cm/sec ) and the FID was set at 300°C . Selected samples were analyzed for chemical identification in a 6890N GC system ( Agilent ) coupled with a 5975 mass selective detector ( MSD ) ( Agilent ) and equipped with a DB-5 ( 20 m x 0 . 18 mm x 0 . 18 μm ) column ( Agilent ) . Helium was used as carrier gas at 33 cm/s average linear velocity . Injection and temperature settings were identical to the settings described above , and the transfer line was maintained at 300°C . Positive electron ionization at 70 eV with default temperature settings ( ion source at 150°C , quadrupole at 230°C ) were used for the MSD . Ions were detected in scan mode in the range of 33–650 m/z at 1 . 23 scan/s scan rate . Compounds were identified based on their mass spectra in comparison to those in the reference library ( Wiley 7th/NIST 05 ) and based on comparison of their retention indices and fragmentation patterns to already published Drosophila CHCs ( Howard et al . , 2003; Everaerts et al . , 2010; Dweck et al . , 2015 ) . The position of the double bonds was not confirmed by performing microderivatization reactions and chirality was not determined for any of the CHCs . More details on how the putative structure of previously unpublished D . melanogaster CHCs was deduced ( based on mass spectra and retention index data ) are given in Supplementary file 10 . All chromatograms were analyzed using Agilent ChemStation software . For quantification individual peak areas were obtained for 42 and 60 male and female CHC containing peaks , respectively ( some of the peaks contained multiple CHCs ) . Response factors were not determined for individual components . To account for natural variation in body size and absolute amounts of CHCs between the lines , the data were represented as proportions by dividing each peak area by the sum of all integrated peaks . We partitioned variation of each CHC peak into genetic and environmental components using an ANOVA model of form Y = µ + L + ε , where Y is phenotype , µ is the overall mean , L is the random effect of line , and ε is the residual . We estimated variance components using restricted maximum likelihood and computed the broad-sense heritability ( H2 ) of each CHC peak as H2 = σ2L/ ( σ2L + σ2ε ) , where σ2L is the among-line variance component and σ2ε is the error variance . All analyses were performed with version 9 . 3 of the SAS System for Windows ( 2013 SAS Institute Inc . ) . A majority of CHCs belong to homologous series in which the chain length increases by two carbons; thus these compounds may be genetically correlated due to shared biosynthetic pathways and the data may be confounded with multi-co-linearity ( Martin and Drijfhout , 2009 ) . We visualized the correlations between CHCs using modulated modularity clustering ( MMC ) ( Stone and Ayroles , 2009 ) . The MMC algorithm clusters highly correlated variables based on the Spearman's rank correlation coefficients ( ρ ) . In order to take these correlations into account , we conducted principal components ( PC ) analysis on the variance-covariance matrices for the male and female CHC line means . For each analysis we included only CHC peaks that had an estimated H2 ≥ 0 . 25 . We retained PCs explaining greater than 1% of the variation for subsequent GWA analysis . PCA was conducted in JMP Pro10 ( 2013 SAS Institute Inc . ) . We conducted a GWA analysis for each CHC PC , separately for males and females . The DGRP lines are segregating for Wolbachia infection status and for the following common inversions: In ( 2L ) t , In ( 2R ) NS , In ( 3R ) P , In ( 3R ) K , and In ( 3R ) Mo . We performed GWA studies in two stages . In the first stage , we adjusted the line means for the effects of Wolbachia infection and major inversions . We then used the adjusted line means to fit a linear mixed model in the form of Y = Xb + Zu + ε , where Y is the adjusted phenotypic value , X is the design matrix for the fixed SNP effect b , Z is the incidence matrix for the random polygenic effect u , and ε is the residual . The vector of polygenic effects u has a covariance matrix in the form of Aσ2 , where σ2 is the polygenic variance component . We fitted this linear mixed model using the FastLMM program ( version 1 . 09 ) ( Lippert et al . , 2011 ) . We performed these single marker analyses for the 1 , 883 , 938 ( females ) and 1 , 912 , 894 ( males ) biallelic variants ( SNPs and indels ) with minor allele frequencies ≥ 0 . 05 whose Phred scale quality scores were at least 500 and genotypes whose sequencing depths were at least one and genotype quality scores at least 20 ( Huang et al . , 2014 ) . All segregating sites within lines were treated as missing data . We selected candidate genes with available mutations and RNAi knockdown constructs to test for effects on CHC production based on FlyBase annotations . We obtained lines with RNAi knockdown constructs from the Vienna Drosophila RNAi Center ( VDRC ) and crossed them to the oenocyte-specific GAL4 driver , PromE ( 800 ) -GAL4 ( Dietzl et al . , 2007; Billeter et al . , 2009 ) . We tested knockdown constructs and their co-isogenic controls ( F1 individuals from crosses of the empty vector strain to PromE ( 800 ) -GAL4 ) for 23 genes ( Supplementary file 8 ) . Since no RNAi knockdown line was available from VDRC for CG10097 identified in the male GWA analysis , we obtained and tested for this gene a PiggyBac insertion mutant from the Harvard Exelixis Collection ( Thibault et al . , 2004 ) along with the w1118 control line . From each cross and mutant line , we collected and aged both male and female virgins and analyzed the CHCs in the same manner as described for the DGRP flies . The analysis of CHCs using GC was also the same . However , for these lines , instead of calculating the proportion that each peak contributed to the total chromatogram , we used the internal standard to calculate the amount of each CHC present in the sample ( ng/fly ) with the assumption that body sizes between the control and RNAi knockdown or mutant were not significantly different . These measures provide a more quantitative measure of CHCs and capture differences that proportion data may not resolve . PC analyses and t-tests pairing the test lines with the controls were conducted on these data . We also calculated the mean total amount of CHCs ( ng/fly ) ; we used the more conservative Cochran and Cox test , which assumes unequal group variances to assess statistically significant effects on CHC composition . We also present these data as µg/fly in Figure 7—figure supplement 1–24 . t-tests were conducted in SAS v . 9 . 4 ( 2013 SAS Institute Inc . ) .	The outermost layer of an insect’s body is called the epicuticle and is made of a blend of fat molecules . “Cuticular hydrocarbons” ( or CHCs ) are the most common fat molecules in the epicuticle , and play an important role in protecting the insect’s body from harsh , dry habitats . CHCs also have other roles in insect behavior . For example , these molecules act as chemical cues when insects search for mates ( i . e . pheromones ) , and they can even contribute to camouflage . Insects are amongst the most diverse groups of animals on Earth , and different species have different blends of CHC molecules in their epicuticles . Fruit flies are a useful model to understand the genetics of CHC production , including CHCs that act as sex pheromones . Previous research has analyzed the CHCs made by both sexes in several fruit fly strains . However this work was unable to uncover which genes influence how much of a given CHC an individual fly will make . Dembeck et al . have now looked into CHC production in a collection of 205 different fly strains , all of which have already had their total genetic material sequenced and studied . Comparing these known sequences and looking for associations between genetic differences and particular CHCs uncovered 24 genes that may be involved in CHC manufacture . Only six of the genes had been identified previously . Dembeck et al . found that interfering with the activity of any of the 24 genes had a knock-on effect on many other CHCs present in the flies’ epicuticle . These 24 genes could to be pieced together in a network that is needed to make and recycle CHCs . The complexity and flexibility of this network can explain in part how insects have been able to build epicuticles for almost every environment . These data set the stage for future work directed towards understanding the evolutionary significance of variation in CHC composition in many fruit fly populations .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "genetics", "and", "genomics" ]	2015	Genetic architecture of natural variation in cuticular hydrocarbon composition in Drosophila melanogaster
The common human pathogen Salmonella enterica takes up citrate as a nutrient via the sodium symporter SeCitS . Uniquely , our 2 . 5 Å x-ray structure of the SeCitS dimer shows three different conformations of the active protomer . One protomer is in the outside-facing state . Two are in different inside-facing states . All three states resolve the substrates in their respective binding environments . Together with comprehensive functional studies on reconstituted proteoliposomes , the structures explain the transport mechanism in detail . Our results indicate a six-step process , with a rigid-body 31° rotation of a helix bundle that translocates the bound substrates by 16 Å across the membrane . Similar transport mechanisms may apply to a wide variety of related and unrelated secondary transporters , including important drug targets . Citrate transporters are found in a wide range of bacteria , archaea and eukaryotes . Bacteria use specific transporters ( Sobczak and Lolkema , 2005 ) to take up di- and tricarboxylates as a carbon source ( Mulligan et al . , 2014; Pos et al . , 1998 ) . The human citrate transporter NaCT plays a central role in fatty acid synthesis and glycolysis ( Gopal et al . , 2007 ) , and is a potential drug target against obesity and diabetes ( Liang et al . , 2015 ) . The Drosophila INDY gene encodes a related dicarboxylate transporter implicated in fat storage ( Rogina et al . , 2000 ) . The x-ray structure of VcINDY , a homologous dicarboxylate transporter from Vibrio cholerae is known in the inward-facing state ( Mancusso et al . , 2012 ) . Unexpectedly , a recent cryo-EM structure of the citrate transporter KpCitS from Klebsiella pneumoniae ( Kebbel et al . , 2013 ) revealed a similar overall domain architecture to VcINDY ( Mancusso et al . , 2012 ) and to archaeal Na+/H+ antiporters of the NhaP family ( Goswami et al . , 2011; Paulino et al . , 2014; Wöhlert et al . , 2014 ) , in both cases without detectable sequence homology . CitS , VcINDY and the NhaP antiporters all form homodimers of two protomers , each organized in a helix bundle and a dimer contact domain ( Kebbel et al . , 2013; Mancusso et al . , 2012; Vinothkumar et al . , 2005 ) , which suggests similar transport mechanisms . In membrane vesicles ( Lolkema et al . , 1994; van der Rest et al . , 1992 ) and proteoliposomes ( Pos and Dimroth , 1996 ) , CitS from Klebsiella pneumoniae ( KpCitS ) was previously shown to transport citrate as HCit2- in a sodium-dependent manner . We observed similar transport properties for CitS from Salmonella enterica ( SeCitS ) , which is closely related to KpCitS . The two homologues share a remarkably high sequence identity of 92% ( Figure 1 ) , indicating that their transport mechanisms must be very similar . Iso-citrate and , to a lesser extent , malate inhibit Na+-dependent 14C-citrate uptake by SeCitS into proteoliposomes . Succinate , α-ketoglutarate , and glutaric acid reduce uptake slightly , whereas tricarballylic acid , which lacks the citrate hydroxyl group , has no effect ( Figure 2A ) . This demonstrates the specificity of the CitS binding site for 2-hydroxycarboxylates . Malate , which is smaller than citrate , inhibits citrate uptake by SeCitS but is not transported ( Figure 2B ) . Citrate symport is driven by Na+ but not by K+ or Li+ ( Figure 2C , D ) , demonstrating the exquisite specificity of SeCitS for Na+ ions . Sodium transport is cooperative with a Hill coefficient of 1 . 89 , whereas citrate is not , suggesting that citrate transport is coupled to at least two Na+ ions ( Figure 3 ) . SeCitS is active between pH5 and pH8 with an optimum at pH7 , resulting in a roughly bell-shaped pH profile ( Figure 3—figure supplement 1 ) . Down-regulation of transport at low pH can be attributed to a limitation in sodium binding , while at elevated pH the availability of the HCit2- citrate species is limiting . Citrate uptake is enhanced at lower outside pH ( Figure 3—figure supplement 2A ) . Under these conditions , transport by SeCitS is electroneutral , since valinomycin has no effect ( Figure 3—figure supplement 2B ) . A lower outside pH would shift the citrate buffer equilibrium towards HCit2- . Therefore , a low outside pH increases the local substrate concentration , while a high inside pH tends to deprotonate the HCit2- substrate and thus removes it from the transport equilibrium . We conclude that protons do not participate directly in the transport mechanism . This conclusion is substantiated by the observation that an increase in the internal Na+ concentration does not stimulate citrate uptake ( Figure 3—figure supplement 2C ) , which argues against a previously proposed citrate/proton symport ( or citrate/hydroxide antiport ) in exchange for internal sodium ( Pos and Dimroth , 1996 ) . 10 . 7554/eLife . 09375 . 003Figure 1 . Sequence alignment . Sequence alignment of 2-hydroxycarboxylate transporters . The secondary structure of SeCitS is shown above the alignment . R402 and R428 of the citrate-binding site are outlined in red . Symbols above the sequence indicate residues involved in sodium binding . A hashtag ( # ) marks the residues that form the Na1 site . Residues with sidechains coordinating Na2 are marked with a diamond ( ♦ ) , and those that coordinate Na2 with backbone carbonyls with an open circle ( ○ ) . Most of the conserved residues ( * ) are found in the two helix hairpins H6 and H12 , and in transmembrane helix H13 . SeCitS: Citrate/sodium symporter from Salmonella Enterica ( WP_024797394 . 1 ) KpCitS: Citrate/sodium symporter from Klebsiella pneumoniae ( WP_025860623 . 1 ) VcCitS: Citrate/sodium symporter from Vibrio_cholerae ( WP_001003397 . 1 ) BsCimH: Citrate/malate transporter from Bacillus_subtillis , ( P94363 . 1 ) KpCitW: Citrate/acetate transporter from Klebsiella_pneumoniae , ( AF411142 . 1 ) LmCitP: Citrate transporter from Leuconostoc_mesenteroides ( AAA60396 . 1 ) LlMleP: Malate transporter from Lactococcus lactis , ( CAA53590 . 1 ) BsMaeN: Malate/sodium symporter from Bacillus_subtilis , ( AFQ59004 . 1 ) DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 00310 . 7554/eLife . 09375 . 004Figure 2 . Substrate specificity of SeCitS . ( A ) The substrate specificity of SeCitS was established by a proteoliposome uptake inhibition assay . Potential substrates or competitors were added in thousandfold excess of 14C-citrate ( 5 µM ) and transport was measured . The 2-hydroxycarboxylates malate and iso-citrate inhibit 14C-citrate uptake completely . α-Ketoglutarate , which has a carbonyl instead of the citrate hydroxyl group , inhibits less strongly . Succinate and glutarate inhibit transport only slightly . Tricarballate has no effect . ( B ) While malate inhibits citrate uptake , it is not a substrate for SeCitS , as uptake of 14C-malate ( 43 µM ) is not detectable . ( C , D ) SeCitS is highly specific for Na+ . Neither Li+ nor K+ drive ( C ) or inhibit ( D ) citrate uptake . Choline was used as a negative control in both assays . Initial uptake rates were plotted relative to ( A ) absence of competitor , ( B ) citrate transport or ( C , D ) sodium-driven transport . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 00410 . 7554/eLife . 09375 . 005Figure 3 . Citrate and sodium transport kinetics . ( A ) Citrate uptake by SeCitS containing proteoliposomes in presence of 25 mM Na+ is non-cooperative and follows Michealis-Menten kinetics with a Km of 4 . 1 µM and a vmax of 23 . 1 nmol · min-1 · mg-1 . ( B ) Na+ transport in presence of 5 µM citrate is cooperative , with a Hill coefficient of 1 . 89 . The affinity of SeCitS for Na+ is lower than for citrate , as demonstrated by a Km of 3 . 3 mM . The vmax of 24 . 9 nmol · min-1 · mg-1 indicates a turnover rate of 1 . 2 citrate molecules per protomer per minute at room temperature . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 00510 . 7554/eLife . 09375 . 006Figure 3—figure supplement 1 . pH-dependence of SeCitS transport . pH-dependent citrate uptake in proteoliposomes measured under symmetrical conditions ( same pH inside and outside ) . SeCitS activity is maximal at pH7 ( 100% ) and decreases to 55% at pH6 . Down-regulation to 10% activity at pH5 or 20% at pH8 results in a roughly bell-shaped pH profile . Initial uptake rates were plotted relative to the activity maximum at pH7 . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 00610 . 7554/eLife . 09375 . 007Figure 3—figure supplement 2 . Driving force , electrogenicity and effect of internal salt concentration . ( A ) Activity of SeCitS reconstituted into proteoliposomes with an inside pH of 7 . 0 and variable outside pH . A ΔpH increases the transport rate up to twofold compared to transport driven by sodium only . ( B ) Transport activity of SeCitS is electroneutral , as the addition of 1 µM valinomycin ( pH 7 . 0 , 5 mM KCl inside and outside ) has no effect . ( C ) To investigate the influence of internal salt on the transport activity , SeCitS was reconstituted with either 1 mM Na+ , Li+ , K+ , or choline . Transport is slightly inhibited by inside Na+ , Li+ , or K+ . However , there is no difference between Na+ , Li+ , or K+ , indicating that internal sodium does not favour transport . Transport rates were plotted relative to ( A ) transport under symmetrical pH , ( B ) transport without valinomycin or ( C ) transport in the absence of additional , intraliposomal Na+ , K+ or Li+ . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 007 To understand the mechanism in detail we determined the structure of CitS from the human pathogen Salmonella enterica ( Figure 4 ) by single-wavelength anomalous dispersion with crystals of seleno-methionine derivatized protein ( Figure 4—figure supplement 1 ) . Phases were extended to the 2 . 5 Å diffraction limit of native crystals ( Table 1 ) . The asymmetric unit contains two homodimers of two protomers in different conformations ( Figure 4A , B ) . Each protomer has 13 helix elements ( H1–H13 ) , including eleven transmembrane helices ( TMH ) and two helix hairpins ( H6 , H12 ) , with the N-terminus on the cytoplasmic side and the C-terminus on the outside . Helices H2-–7 and H8–13 are organized in two repeats with inverted topology ( Figure 4C ) , connected by a flexible cytoplasmic loop ( Figure 4B ) . Together , helices H5–7 of repeat 1 and H11–13 of repeat 2 form a bundle on either side of the central contact domains , which hold the dimer together through extensive hydrophobic interactions of H2 , 4 , 8 and 10 . A 16 Å-deep hydrophobic cavity on the cytoplasmic side of the dimer interface contains the hydrophobic tail of a detergent or lipid molecule ( Figure 5A ) . 10 . 7554/eLife . 09375 . 008Figure 4 . Overall structure of SeCitS and topology . Side view ( A ) and cytoplasmic view ( B ) of the SeCitS homodimer . The dimer is oval , with a long axis of 96 Å and a short axis of 60 Å . Each protomer consists of eleven transmembrane helices and two helix hairpins ( yellow and pink ) . ( C ) SeCitS consists of two inverted 5-TMH repeats connected by a long cytoplasmic loop plus an additional N-terminal helix . Each repeat contains one hairpin . Helices belonging to the helix bundle are shown on blue background , while helices of the dimer contact domain are shown on grey background . The extended flexible link between the two inverted repeats is completely resolved in protomer A ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 00810 . 7554/eLife . 09375 . 009Figure 4—figure supplement 1 . SeMet phasing . Top view ( A ) and side view ( B ) of electron density ( blue , 1 . 5 σ ) of one dimer in the asymmetric unit after phasing and density modification . Strong Se peaks in the anomalous difference map contoured at 5 σ ( magenta ) indicate SeMet positions . Out of 72 potential selenium sites in the asymmetric unit , 53 were found in the substructure with an occupancy > 20% . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 00910 . 7554/eLife . 09375 . 010Figure 5 . Two different states of the asymmetrical SeCitS dimer . ( A ) The outward-facing protomers A and A' bind citrate in a shallow , positively charged cavity between the helix bundle and dimer contact domain . In the inward-facing protomers B and B' , citrate binds in a deep cytoplasmic cavity . In B' , two citrate molecules are resolved . ( B ) In protomers A , A’ and B , two Na+are occluded in the helix bundle , while in B' only one Na+ is present . The substrates are translocated 16 Å across the membrane by a 31° rotation of the helix bundle relative to the static dimer contact domain . ( C ) In the outward-facing protomers , citrate is closely coordinated by sidechains of both hairpins and H13 . Neither Na+ participates directly in citrate coordination . ( D ) In the inward-facing protomer B , citrate is hydrated and attached weakly to the glycine-rich loop of H12 . The Na1 and Na2 sites in ( C ) and ( D ) are virtually identical , indicating that the transition from the outward-facing to the inward-facing state does not affect Na+-coordination geometry . ( E ) In protomer B' , only the Na1 site is occupied . Two citrate molecules are resolved , outlining a likely trajectory for citrate release ( Video 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 01010 . 7554/eLife . 09375 . 011Video 1 . Schematic representation of SeCitS transport . The movie shows a morph from the outward-facing to the inward-facing conformation for one protomer of the SeCitS dimer . Arg402 , Arg428 and Tyr348 , which coordinate citrate in the outward-facing conformation , are drawn as stick models , while the Na+ ions are represented as grey spheres . Na+ ions bind to their respective sites in the helix bundle , followed by citrate binding between helix bundle and dimer contact domain . Subsequently , the substrates are translocated by a rotation of the bundle . Citrate release is independent from the release of either Na+ ion . Due to the empty Na2 binding site in protomer B’ we assume that this ion is released immediately after the citrate . After substrate release the empty transporter changes its conformation back to the outward-facing state to repeat the cycle . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 011 The two dimers in the asymmetric unit are similar , with an overall rmsd of 0 . 5 Å , whereas the protomers within one dimer differ substantially by an rmsd of 8 . 4 Å . The most conspicuous differences are manifest in the vertical positions of the two hairpins H6 and H12 in the helix bundle ( Figure 5A ) . Comparing the two protomers of dimer 1 , the C-terminal end of H6 projects 16 Å above the outer membrane surface in protomer A , while it hardly protrudes in protomer B . Conversely , the cytoplasmic H12 ends roughly at the inner membrane surface in protomer A , but extends 13 Å above it in protomer B . The relative position of helices and hairpins within each bundle is unchanged . Evidently , the whole bundle moves as a rigid body from its position in protomer A to that in protomer B , while the central dimer contact domain remains static . The crystal contacts of both dimers in the asymmetric unit are different . Since the polyptide structures of the two dimers are almost identical , the observed asymmetry cannot be attributed to crystal packing . Dimer asymmetry is equally striking with respect to surface structure and electrostatic potential distribution . Protomer A has more positive charges on the periplasmic side than protomer B ( Figure 6A ) . On the cytoplasmic side , positive charges predominate on the surface of protomer B , while positive and negative charges are roughly evenly distributed on protomer A ( Figure 6B ) . Overall , positive charges dominate on the cytoplasmic side of the dimer ( Figure 6C , 6D ) , in line with the positive-inside rule for membrane proteins ( Nilsson and von Heijne , 1990; von Heijne , 1992 ) . 10 . 7554/eLife . 09375 . 012Figure 6 . Electrostatic surface potential and bound detergent/lipid molecules . Exterior ( A ) and cytoplasmic views ( B ) of the electrostatic surface potential of SeCitS accentuates the dimer asymmetry . The binding sites for the citrate di-anion ( green ) on the exterior surface of protomer A and the cytoplasmic side of protomer B are strongly positively charged ( dark blue ) . ( C , D ) Positions of bound detergent and lipid molecules ( yellow ) are shown in the side view of the electrostatic surface . Apart from the aliphatic chain in the hydrophobic cavity of the dimer interface ( Figure 5 ) , they are positioned close to the helix bundle . ( E ) In the outward-facing protomers , a hydrophobic cavity between H5 , H13 and the dimer contact domain is filled by a detergent molecule . This cavity is closed in the inward-facing protomers ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 012 Short stretches of glycine-rich unwound polypeptide link the two halves of hairpins H6 and H12 . Together they define the substrate-binding site ( Figure 5C–E ) at the interface between the helix bundle and the dimer contact domain . On the extracellular surface , the binding site is found in a ∼6 Å deep cavity of protomer A ( Figure 6A ) , while in protomer B it is located at the bottom of a ∼13 Å-deep channel on the cytoplasmic side ( Figure 6B ) . We conclude that protomer A is outward-facing and that protomer B faces inward . The binding sites in both protomers are strongly positively charged ( Figure 6A , B ) . Two detergents and one lipid molecule were identified on the periphery of the dimer . A further detergent molecule was situated in a hydrophobic cavity between the central dimer contact domain and the six-helix bundle of the outward-facing protomer A ( Figure 6C–E ) . H5 in this bundle is straight in protomer A but kinked near its cytoplasmic end in protomer B , to accommodate the movement of the helix bundle ( Figure 6E , F ) . All four protomers show clear electron density for citrate in the binding site ( Figure 7 ) . In the outward-facing protomers , the citrate is closely coordinated by two arginines ( Arg402 , Arg428 ) , two polar sidechains ( Asn186 , Ser405 ) and the protein backbone of both hairpins ( Figure 5 and 7A ) . The only residue in the static contact domain involved in substrate coordination is Tyr348 in H10 , which forms a π-π-interaction with a citrate carboxyl . One ordered water molecule participates directly in citrate binding . Its trigonal-bipyramidal coordination geometry ( Figure 7 ) might suggest a Na+ ion rather than water , which would imply that the transported entity is NaCit2- rather than HCit2- . Because the electron density is weak and the coordination distance of >2 . 8 Å is longer than would be expected for Na+ , we interpret this density as a water molecule . In both outward-facing protomers , two Na+ ions are clearly resolved next to the citrate ( Figures 5C and 7A ) . In the Na1 site , four backbone carbonyls in the unwound hairpin stretches coordinate one Na+ . In the Na2 site , the carboxyl group of Asp407 , the polar sidechains of Asn401 , Ser427 and the backbone carbonyls of Cys398 and Gly403 coordinate the ion . Two ordered water molecules participate in Na+binding , one of them suspended between the two Na+ ions ( Figures 5C and 7A ) , accounting for the observed cooperativity of Na+ transport ( Figure 3B ) . Asn401 , which coordinates Na1 with its backbone carbonyl and Na2 via its side chain , may contribute to this effect . 10 . 7554/eLife . 09375 . 013Figure 7 . Binding sites and Fo-Fc ligand density . ( A ) Stereo view of the outward-facing substrate-binding site of protomer A with an extensively coordinated citrate molecule . ( B ) In the inward-facing binding site of protomer B the citrate is attached less strongly . In ( A ) and ( B ) the Fo-Fc density ( blue mesh ) is contoured at 3σ for citrate and at 5σ for the two bound Na+ ions and the water molecule between them . ( C ) In the inward-facing protomer B’ , the Fo-Fc map contoured at 4σ shows an occupied Na1 site , while the Na2 site is empty . The Fo-Fc omit map contoured at 2 . 5 clearly shows two citrate molecules . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 013 There is no difference in substrate coordination or in main-chain conformation between the two outward-facing protomers A and A’ ( rmsd 0 . 5 Å ) . A and A’ can therefore be considered as identical . Interestingly , the main chain conformations of the two inward-facing protomers B and B’ are likewise practically identical ( rmsd 0 . 6 Å ) , but the citrate and Na+ coordination in B and B’ is clearly different . In protomer B , citrate is partially hydrated and coordinated by the hydroxyl of Ser405 and the backbone carbonyl of Gly404 in the conserved GGXG motif of H12 ( Figures 1 , 5D and 7B ) . Both Na+ sites are occupied and take up the same position relative to the citrate as in the outward-facing state . In protomer B' , the Na2 site is empty , even though the structure of the ion-coordinating hairpin hardly changes ( Figures 5E and 7C ) . The citrate is fully hydrated and not directly attached to a sidechain , and a second citrate is present near Gln424 in H7 . The rigid-body movement of the helix bundle from its position in protomers A and A’ to that in protomers B or B' can be described as a 31° arc-like rotation around an axis roughly parallel to the membrane and perpendicular to the long dimer axis ( Figures 5 , 7 and Video 1 ) . The rotation is facilitated by the greasy interface between the helix bundle and the static dimer contact domain . The greasy interface consists almost entirely of small hydrophobic sidechains and a bound detergent molecule that may take the place of a membrane lipid alkyl chain ( Figures 5 and 8 A-D ) . During the bundle rotation the detergent molecule is displaced by H13 . As the helix bundle reaches the inward-facing position , the straight , hydrophobic helix H5 kinks at Gly143 , thus preventing its partial exposure to the cytoplasm , and an ion bridge forms between Asp112 and Arg205 in H7 ( Figure 8E , F; Video 2 ) . As a result of the helix bundle rotation , the binding site with the bound citrate moves by 16 Å from the external membrane surface in the outward-facing state to a position where it is accessible from the cytoplasmic membrane surface in the inward-facing state ( Figure 5 ) . Since transport is non-cooperative with respect to citrate ( Figure 3A ) , we conclude that the two binding sites in the dimer act independently of one another . 10 . 7554/eLife . 09375 . 014Figure 8 . Hydrophobic interface between helix bundle and dimer contact domain . ( A , C ) In the outward-facing protomers A and A’ , a hydrophobic pocket between helix H5 , H13 and the dimer contact domain harbors a detergent molecule that apparently replaces a membrane lipid . ( B , D ) In the inward-facing protomers B and B' H5 kinks at Gly143 and shifts towards the cytoplasm . We assume that H13 fills this hydrophobic cavity in the inward-facing state . ( E ) In the outward-facing protomers , Tyr348 coordinates the citrate by π-π-interactions . As a result of the arc-like helix bundle rotation , an ion bridge forms between Asp112 and Arg205 ( H7 ) in the inward-facing protomers ( F ) . Arg205 moves by more than 20 Å from its position in the outward-facing conformation ( E ) . The sidechain of Tyr348 rotates by 90° , blocking the entrance to the substrate binding site ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 014 Unlike the Na+ ions , the citrate di-anion is not occluded by the hairpin loops in SeCitS . Similarly , the dicarboxylate substrate is not occluded in VcINDY ( Mancusso et al . , 2012 ) , whereas the corresponding substrate is occluded within the helix bundle of GltPh ( Boudker et al . , 2007 ) . SeCitS may lack a well-defined substrate-occluded state , but the citrate would effectively be occluded during the transition from the outward-facing to the inward-facing state , while the occupied binding site rotates past the hydrophobic surface of the dimer interface domain ( Video 2 ) . 10 . 7554/eLife . 09375 . 015Video 2 . Schematic representation of domain , helix and sidechain movements . Three synchronized movies show different views of one SeCitS protomer during the transport cycle: ( A ) from the membrane plane , ( B ) in the perpendicular direction from the cell exterior and ( C ) a detailed view of the substrate-binding site and the detergent/lipid binding pocket . Helices of the rotating bundle domain are coloured , while helices in the static dimer contact domain are shown in grey behind their corresponding transparent electrostatic surface . The negatively charged periplasmic surface of SeCitS ( transparent red ) attracts Arg205 of H7 ( green ) , which , in the inward-facing state , forms an ion bridge to Asp112 in H4 and a hydrogen bond to Tyr348 ( lavender ) in H10 of the dimer contact domain . In the outward-facing state , Asp112 interacts with Tyr348 , which rotates to block access to the substrate-binding site in the inward-facing state . A detergent molecule ( yellow ) in the hydrophobic pocket between H5 ( purple ) , H13 ( cyan ) and the dimer contact domain , is displaced in the inward-facing state by the movement of H13 . H5 , which is straight in the outward-facing state , kinks during the bundle rotation to prevent its partial exposure to the cytoplasm . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 015 In the outward-facing state , strong polar and ionic interactions facilitate citrate binding at low ambient substrate concentrations ( Figure 3A and 5C ) . In the inward-facing state , the binding affinity for the substrate is reduced ( Pos and Dimroth , 1996 ) , so the citrate can detach . We propose that the three citrate positions we observe in the two inward-facing protomers mark the path of the substrate during its release from the binding site past the highly conserved Arg428 ( Figures 1 , 5 and 7 ) , along a trajectory that guides the negatively charged substrate towards the cytoplasm , where it is metabolized . Partial release of the citrate di-anion would weaken Na+ binding , which explains why only one Na site is occupied in B’ . Since transport is electroneutral , both Na+ ions must dissociate from the inward-facing state . MD simulations suggest that in other Na-dependent transporters such as LeuT ( Grouleff et al . , 2015 ) , GltPh ( Zomot and Bahar , 2013 ) , vSGLT ( Watanabe et al . , 2010 ) , at least one of the Na+ ions is released before the main substrate . In the case of SeCitS , comparison of the inward-facing protomers B and B’ indicates unambiguously that citrate is released before Na+ , and that Na2 is released before Na1 ( Figure 5D , E ) . Once the citrate has left the binding site , the helix hairpins or H13 would need to rearrange to release Na1 , while a minor reorientation of the Asp407 or Ser427 sidechains is sufficient to release Na2 . Comparison of the binding sites in the outward-facing protomers indicates that both Na+ ions have to be in place before citrate can bind . A cryo-EM structure from 2D crystals of the closely related KpCitS from Klebsiella pneumoniae found that sodium citrate induced a major conformational change in the helix bundle , whereas potassium citrate did not ( Kebbel et al . , 2013 ) , supporting the proposed binding order . Therefore the complete transport mechanism entails the following six steps: ( 1 ) The Na sites are occupied by Na+ in the outward-facing state; ( 2 ) a citrate binds from the external medium; ( 3 ) citrate binding triggers the arc-like rotation of the helix bundle in the transition from the outward-facing to the inward-facing state; ( 4 ) in the inward-facing state , the citrate becomes hydrated and diffuses into the cytoplasm; ( 5 ) the sodium ions come off; ( 6 ) the release of all substrates enables the reverse arc-like rotation of the helix bundle to expose the empty binding site again to the cell exterior , and the cycle repeats ( Figure 9; Videos 1 and 2 ) . 10 . 7554/eLife . 09375 . 016Table 1 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 016Native SeCitSSeMet SeCitSData collection SLS PXIIWavelength ( Å ) 0 . 9790 . 980Space groupP1P21Cell dimensionsa , b , c ( Å ) 86 . 4 , 89 . 9 , 91 . 890 . 9 , 168 . 8 , 97 . 9α , β , γ ( ° ) 90 . 4 , 113 . 8 , 99 . 590 . 0 , 91 . 0 , 90 . 0Resolution ( Å ) 47 . 98 – 2 . 5 ( 2 . 6 – 2 . 5 ) 48 . 95 – 3 . 9 ( 4 . 0 -– 3 . 9 ) Rpim0 . 052 ( 0 . 872 ) 0 . 038 ( 0 . 539 ) I / σI 8 . 9 ( 1 . 3 ) 16 . 8 ( 2 . 2 ) CC*0 . 999 ( 0 . 828 ) 1 . 000 ( 0 . 944 ) Completeness ( % ) 98 . 8 ( 98 . 1 ) 100 ( 100 ) Multiplicity8 . 2 ( 8 . 1 ) 41 . 4 ( 40 . 9 ) Refinement Resolution ( Å ) 47 . 98 – 2 . 5 ( 2 . 6 – 2 . 5 ) Unique reflections84765Reflections in test set4193Rwork/Rfree ( % ) 21 . 0/24 . 8 ( 33 . 6/36 . 3 ) CC ( work ) /CC ( free ) 0 . 848/0 . 742 ( 0 . 796/0 . 773 ) Average B-Factor ( Å2 ) 70No . atoms in AU13270Protein12916Ligands285Water69r . m . s . deviations:Bond lengths ( Å ) 0 . 003Bond angles ( ° ) 0 . 762Values for the highest resolution shell are shown in parentheses10 . 7554/eLife . 09375 . 017Figure 9 . Six-step mechanism of Na+-dependent citrate uptake by SeCitS . ( 1 ) Two Na+ bind to the empty transporter; ( 2 ) citrate from the external medium attaches to the binding site; ( 3 ) the substrates are translocated across the membrane through a rigid-body 31° rotation of the helix bundle domain; ( 4 ) first the citrate and then ( 5 ) the Na+ ions are released to the cytoplasm; ( 6 ) the unloaded protomer changes its conformation back to the outward-facing state and the cycle restarts . In the cell , the inward-directed Na+gradient drives citrate uptake , but all steps are in principle reversible . The approximate position of the rotation axis parallel to the membrane and perpendicular to the long dimer axis is indicated in ( 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09375 . 017 Notwithstanding the large domain movements associated with substrate translocation , citrate exchange rates are high , with a turnover of up to 137 s-1 reported for the closely related KpCitS ( Pos and Dimroth , 1996 ) . Citrate uptake by SeCitS is substantially slower at 1 . 2 molecules per minute ( Figure 3 ) . Therefore , the arc-like rotation of the helix bundle that translocates the bound substrate across the membrane is not rate-limiting . The same seems to hold true for GltPh , which shows a slow substrate uptake rate of 0 . 29 molecules per minute ( Ryan et al . , 2009 ) , while crosslinking experiments show that the conformational change happens within seconds ( Reyes et al . , 2009 ) . Assuming that the Na+ concentration does not limit substrate binding or release under physiological conditions , the rate-limiting step in SeCitS is most likely the reverse rotation of the helix bundle with the binding site empty . An influence of lipids on the conformational dynamics in GltPh by inserting a lipid molecule between both domains was recently proposed by MD simulations ( Akyuz et al . , 2015 ) . The structure of SeCitS offers experimental evidence for the existence of a hydrophobic pocket at the interface of both domains and highlights the importance of the bilayer for the activity of membrane transporters . It remains to be seen whether any of the lipid-binding sites in these transporters are structurally conserved . The arc-like rotation of the helix hairpins in SeCitS is reminiscent of the recently proposed conformational change for substrate translocation in GltPh ( Crisman et al . , 2009; Reyes et al . , 2009; Verdon et al . , 2014 ) . In the Na+/H+ antiporters ( Lee et al . , 2013; Paulino et al . , 2014 ) or the bile acid transporter ASBT ( Zhou et al . , 2014 ) , where the binding site is defined by unwound stretches of two trans-membrane helices in a structurally homologous bundle , this process also involves a rotation of the bundle around a similar axis as in SeCitS , although the movement is significantly smaller . The domain structure of the unrelated transporter YdaH ( Bolla et al . , 2015 ) bears a striking resemblance to that of SeCitS , suggesting that it may work in the same way . The rotating arc mechanism described here for SeCitS thus seems to apply to a large class of secondary membrane transporters with unwound helix elements or hairpins that were previously thought to be unrelated . A gene coding for CitS from Salmonella enterica ( WP_000183608 ) was cloned into a pET21d plasmid harboring an N-terminal His10-Tag and a thrombin cleavage site between tag and target protein . The resulting plasmid was used to transform E . coli C41- ( DE3 ) cells . After expression for 10 h at 37°C in ZYM-5052 autoinduction medium ( Studier , 2005 ) cells were harvested , resuspended in 20 mM Tris/HCl pH7 . 4 , 150 mM NaCl , 5 mM EDTA , 5 mM β-mercaptoethanol ( β-ME ) and broken using a microfluidizer ( M-110L , Microfluidics ) . Unbroken cells and cell debris were removed by centrifugation at 18 , 000 g for 30 min . Membranes were isolated by centrifugation at 100 , 000 g for 1 h and resuspended at a total protein concentration of 15 mg/ml in 20 mM Tris/HCl , 140 mM choline chloride , 250 mM sucrose , 1 mM Na-citrate , 5 mM β-ME . SeCitS was solubilized by 1:1 dilution of membranes with 20 mM Tris/HCl pH7 . 4 , 150 mM NaCl , 3% n-decyl-β-D-maltopyranoside ( DM ) , 1 mM Na-citrate , 5 mM β-ME . Unsolubilized material was removed by ultracentrifugation at 100 , 000 g for 1h . The supernatant was supplemented with 45 mM imidazole and incubated with Ni-NTA beads equilibrated with 20 mM Tris/HCl pH7 . 4 , 300 mM NaCl , 45 mM imidazole , 1 mM Na-citrate , 0 . 15% DM , 5 mM β-ME for 2h at 4°C . The mixture was loaded on a column and washed with equilibration buffer to remove unspecifically bound protein . For on-column cleavage the buffer was changed to 10 mM Tris/HCl pH8 . 2 , 150 mM NaCl , 2 . 5 mM CaCl2 , 1 mM Na-citrate , 0 . 15% DM . Thrombin was added to the beads to a concentration of 1 U/mg protein and incubated overnight under constant mixing . The beads were washed with exchange buffer to recover tag-free SeCitS and the protein was concentrated to 5 mg/ml ( 50 kDa cut-off ) . The concentrated protein was applied to a Superdex-200 size exclusion column equilibrated with 20 mM Tris/HCl pH8 . 2 , 150 mM NaCl , 1 mM Na-citrate , 0 . 15% DM , 1 mM TCEP ( Tris- ( 2-carboxyethyl ) phosphine ) . Fractions containing SeCitS were pooled , concentrated as above , frozen in liquid nitrogen and stored at -80°C . Selenomethionine ( SeMet ) -substituted protein was expressed in a defined medium by methionine biosynthesis inhibition ( Doublie , 2007 ) . Expression cultures were directly inoculated with pre-cultures grown in non-inducing PA-0 . 5G medium ( Studier , 2005 ) . The main culture was grown at 37°C , induced at an OD600 of 0 . 5 and harvested after 4 h . Purification of SeMet SeCitS was performed as described for the native protein . For crystallization , native SeCitS was supplemented with n-octyl-β-D-glucopyranoside ( OG ) to a concentration of 1% . The protein was mixed 1:1 with reservoir solution ( 100 mM MES pH6 . 5 , 200 mM NaCl , 29% PEG400 ) and crystallized in 24-well hanging drop plates . Rhombic crystals appeared within 3 days and grew to a size of 150 µm within a week . Crystals were harvested and vitrified in liquid nitrogen using Al´’s oil ( D'arcy et al . , 2003 ) as cryo-protectant . SeMet-derivatized SeCitS was supplemented with 2% n-heptyl-β-D-glucopyranoside ( HG ) and mixed 1:1 with reservoir solution ( 100 mM MES pH6 . 5 , 250 mM NaCl , 30% PEG400 ) . Thin needle-like crystals grew to 400 µm within a week and were vitrified in liquid nitrogen directly . All datasets were collected on beamline X10SA ( PXII ) at the SLS ( Villigen , Switzerland ) . All datasets were processed with XDS ( Kabsch , 1993 ) and scaled with AIMLESS ( Evans , 2006 ) from the CCP4 package ( Winn et al . , 2011 ) . Resolution limits were based on I/σ ( I ) -values , completeness and cross correlation of half datasets ( Karplus and Diederichs , 2012 ) in the high-resolution shells . PHENIX ( Adams et al . , 2010 ) and Coot ( Emsley et al . , 2010 ) were used for refinement and model building , respectively . Experimental phases were obtained by single-wavelength anomalous dispersion ( SAD ) from SeMet-derivatized SeCitS . Initial SeMet positions were determined by SHELXD ( Schneider and Sheldrick , 2002 ) through the HKL2MAP ( Pape and Schneider , 2004 ) interface and fed into Crank2 ( Skubak and Pannu , 2013 ) for substructure refinement , phasing with Refmac ( Murshudov et al . , 1997 ) , hand determination , initial density modification with Parrot ( Zhang et al . , 1997 ) and model building using Buccaneer ( Cowtan , 2006 ) . An initial backbone model of SeCitS was created for phasing of the native high-resolution data by molecular replacement with PHASER ( McCoy et al . , 2007 ) . Model building was performed by PHENIX autobuild ( Terwilliger et al . , 2008 ) , followed by cycles of manual model building and refinement . Superimpositions were performed with GESAMT ( Winn et al . , 2011 ) . Figures were drawn and rmsd values were calculated with PyMOL ( DeLano and Lam , 2005 ) . Electrostatic surfaces were calculated with PDB2PQR ( Dolinsky et al . , 2004 ) and APBS ( Baker et al . , 2001 ) . E . coli polar lipids in chloroform ( Avanti Polar Lipids ) were dried under nitrogen and resuspended in reconstitution buffer ( 20 mM Tris/BisTris/Acetate pH 4-–8 , 50 mM choline chloride ) , supplemented with 15 mM β-ME . Unilamellar ∼400 nm vesicles were prepared using polycarbonate filters in an extruder ( Avestin ) . Preformed liposomes were diluted to 5 mg/ml in reconstitution buffer and destabilized by addition of 1% OG . SeCitS was added at a lipid-to-protein ratio of 50 and incubated for 1 h . The protein/lipid mixture was filled into dialysis bags ( 14 kDa cutoff ) and dialyzed against detergent-free reconstitution buffer overnight . Biobeads were added to the dialysis buffer to facilitate complete detergent removal . The proteoliposomes were centrifuged for 25 min at 300 , 000 g and resuspended in fresh reconstitution buffer . Transport activity was measured with [1 , 5]14C-citrate or 1 , 4 ( 2 , 3 ) -14C-malate as a reporter molecule . Measurements were started by dilution of 2 µl freshly prepared proteoliposomes into 200 µl reaction buffer ( 20 mM Tris/BisTris/acetate pH 5–8 , 50 mM NaCl , 5 µM [1 , 5]14C-citrate or 43 µM 1 , 4 ( 2 , 3 ) -14C-malate ) . Within the linear range of uptake , 200 µl samples were transferred on 0 . 2 µm nitrocellulose filters that were subsequently washed with 3 ml of reaction buffer . Filters were transferred into counting tubes and filled with 4 ml liquid scintillation cocktail ( Rotiszint ) before evaluation . All measurements were performed in triplicates . In all experiments initial rates within the linear range of uptake were recorded over a total of 4 time points . Kinetic measurements were performed at pH6 by varying the concentration of one substrate while keeping the other constant . Ion specificity of SeCitS was determined by changing the co-substrate to LiCl , KCl or choline chloride , which is not transported . Specificity for citrate was established with a competition assay . Potential substrates were added to the reaction buffer at a concentration of 5 mM ( 1000x excess ) to compete with 14C-citrate uptake . The effect of ΔpH on the transport activity was measured by changing the pH of the reaction buffer while keeping the inside pH constant .	Cells have specialized proteins known as transporters in their surface membranes that move molecules into or out of the cell . Transporters pass their cargo through the membrane by changing shape . This process requires energy and is sometimes driven by simultaneously transporting a charged ion such as sodium . There are different classes of transporters and researchers have described a range of structural changes , and therefore transport mechanisms , that different transporters use . Citrate transporters are found in a wide range of organisms . In bacteria , they bring the citrate substrate molecule into the cell to be used as a nutrient . In humans , citrate transporters are important in metabolism , and are of interest as targets for drugs that could potentially treat obesity and diabetes . This requires an understanding of the atomic structure and the transport mechanisms used by citrate transporters , which were not known . Wöhlert et al . now use a technique called X-ray crystallography to uncover the structure of a citrate transporter called SeCitS in high detail . This transporter is found in a bacterium called Salmonella enterica , a well-known human pathogen that causes typhoid . The crystallized protein simultaneously showed three different structural states – one where the citrate binding site faces the outside of the cell , and two where the binding site faces the inside of the cell . The simultaneous occurrence of different functional states in one and the same crystal structure of a membrane transporter is so far unique . Combining the detailed structures of SeCitS with biochemical studies allowed Wöhlert et al . to deduce that citrate is transported in a six-step process . Sodium ions attach to SeCitS , and then citrate binds to the transporter from outside the cell . This binding causes part of the protein to undergo a substantial rotation , shifting it to an inward-facing state and moving the citrate and sodium ions inside the cell . The release of the citrate and sodium ions then triggers the reverse rotation of the transporter , bringing the empty binding site back to the outside of the cell for a repeat of the cycle . After working out the mechanisms of a bacterial citrate transporter , the next challenge is to extend the analysis to the structure of similar transporters in more complex organisms , including human cells . This could provide an accurate basis for drug development .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	Mechanism of Na+-dependent citrate transport from the structure of an asymmetrical CitS dimer
The ring-shaped MCM helicase is essential to all phases of DNA replication . The complex loads at replication origins as an inactive double-hexamer encircling duplex DNA . Helicase activation converts this species to two active single hexamers that encircle single-stranded DNA ( ssDNA ) . The molecular details of MCM DNA interactions during these events are unknown . We determined the crystal structure of the Pyrococcus furiosus MCM N-terminal domain hexamer bound to ssDNA and define a conserved MCM-ssDNA binding motif ( MSSB ) . Intriguingly , ssDNA binds the MCM ring interior perpendicular to the central channel with defined polarity . In eukaryotes , the MSSB is conserved in several Mcm2-7 subunits , and MSSB mutant combinations in S . cerevisiae Mcm2-7 are not viable . Mutant Mcm2-7 complexes assemble and are recruited to replication origins , but are defective in helicase loading and activation . Our findings identify an important MCM-ssDNA interaction and suggest it functions during helicase activation to select the strand for translocation . Mcm proteins were first identified in yeast when mutations in their genes were defective for minichromosome maintenance ( Maiorano et al . , 2006 ) . In eukaryotic cells , six related Mcm proteins ( Mcm2-7 ) form a ring-shaped heterohexamer , the Mcm2-7 complex . Hexameric MCM rings act as the replicative DNA helicase ( Bochman and Schwacha , 2008; Ilves et al . , 2010 ) , encircling the leading strand DNA template at the replication fork ( Fu et al . , 2011 ) . Replication forks are established in a cell-cycle-regulated manner at specific regions of DNA called replication origins ( Bell and Dutta , 2002 ) . Mcm2-7 complexes are loaded onto double-stranded DNA at each replication origin by the Origin Recognition Complex ( ORC ) , Cdc6 , and Cdt1 ( Remus and Diffley , 2009 ) . Because replication origins are located far from the DNA ends , loading of Mcm2-7 hexamers such that they encircle double-stranded DNA requires opening of the Mcm2-7 ring . A ‘gate’ between the Mcm2 and Mcm5 subunits has been identified and is the likely site of ring opening and closing ( Bochman and Schwacha , 2007 , 2008; Costa et al . , 2011 ) . After helicase loading , the two Mcm2-7 complexes encircle double-stranded DNA ( dsDNA ) as a head-to-head double hexamer ( Evrin et al . , 2009; Remus et al . , 2009 ) that is inactive as a helicase . Helicase activation requires substantial remodeling of the initially loaded Mcm2-7 double hexamer . The Dbf4-dependent Cdc7 kinase ( DDK ) and cyclin-dependent kinases ( CDKs ) drive recruitment of two Mcm2-7 activating proteins , Cdc45 and the tetrameric GINS complex ( Labib , 2010 ) . These proteins together stimulate the Mcm2-7 ATPase and helicase ( Ilves et al . , 2010 ) and with Mcm2-7 form the active replicative DNA helicase , the CMG complex ( Cdc45-Mcm2-7-GINS ) ( Moyer et al . , 2006; Bochman and Schwacha , 2008; Ilves et al . , 2010 ) . The initially loaded double-hexamer has the capacity to passively slide over dsDNA ( Evrin et al . , 2009; Remus et al . , 2009 ) , suggesting MCM DNA interactions are not fixed at this stage . Upon activation , the two Mcm2-7 helicases translocate independently ( Yardimci et al . , 2010 ) in a 3′→5′ direction on the single-stranded leading strand DNA template ( Fu et al . , 2011 ) . This transformation necessitates two structural changes in the initially loaded double-hexamer that are poorly understood: ( i ) the double-hexamer interface must be broken to allow independent replisome movement; ( ii ) the dsDNA at the origin must be melted and the lagging strand DNA template excluded from the central channel of each MCM hexamer . How Mcm2-7 retains one strand in its central channel while excluding the other during this transition is unknown . Each Mcm subunit contains three domains . The N-terminal domain ( MCMN ) possesses an OB ( oligonucleotide/oligosaccharide binding ) -fold and usually a zinc-binding motif ( Fletcher et al . , 2003 ) . This domain mediates the head-to-head interaction of the two hexamers ( Gomez-Llorente et al . , 2005; Evrin et al . , 2009; Remus et al . , 2009 ) . The second domain contains a conserved ATPase AAA+ fold ( Neuwald et al . , 1999 ) , which binds and hydrolyzes ATP at subunit interfaces around the hexameric ring ( Schwacha and Bell , 2001; Davey et al . , 2003 ) and is required for DNA unwinding ( Bochman and Schwacha , 2008; Ilves et al . , 2010 ) . A short domain at the C-terminus includes a helix-turn-helix fold ( Aravind and Koonin , 1999 ) , one of which ( Mcm6 ) interacts with Cdt1 ( Wei et al . , 2010 ) . MCM hexamers demonstrate a two-tiered ring architecture in electron microscopy studies with an N-terminal domain tier and an ATPase domain tier ( Chong et al . , 2000; Pape et al . , 2003; Gomez-Llorente et al . , 2005; Costa et al . , 2006; Bochman and Schwacha , 2007; Remus et al . , 2009; Costa et al . , 2011 ) . The MCM complexes of several archaeal organisms consist of six identical subunits and have provided powerful models to investigate the atomic details of MCM structure . Crystal structures have identified a consistent hexameric arrangement for MCMN of Methanothermobacter thermautotrophicus ( Mt ) ( Fletcher et al . , 2003 ) and Sulfolobus solfataricus ( Sso ) ( Liu et al . , 2008 ) that correspond to the smaller tier observed by electron microscopy ( Remus et al . , 2009; Costa et al . , 2011 ) . Although no atomic structure has been determined for the complete archaeal or eukaryotic Mcm hexamer , hypothetical atomic models for full-length archaeal MCM hexamers have been generated by superimposition of six copies of a monomeric crystal structure of nearly full-length MCM onto the hexameric structure of MtMCMN ( Brewster et al . , 2008; Bae et al . , 2009 ) . Despite a growing understanding of the overall structure of the MCM complex , its multiple interactions with DNA during helicase loading , activation and elongation remain mysterious . Atomic structures of MCM bound to DNA have not been reported . Given the different forms of DNA that are bound to the MCM complex during the steps of the initiation pathway , the MCM proteins must transition between different DNA interactions during this process . To investigate the interactions after origin melting and how the MCM hexamer selectively encircles the leading strand template , we determined the crystal structure of the MCMN hexamer of Pyrococcus furiosus bound to ssDNA . We present an analysis of this the structure and biochemical and genetic characterizations of archaeal and S . cerevisiae proteins with mutations in the identified ssDNA binding region . These findings reveal two residues on the surface of the MCM OB-fold that are critical for MCM DNA-binding and contribute to multiple Mcm2-7 functions during replication initiation . Our findings support a model in which the identified MCM-ssDNA interactions contribute to the selection of the leading strand DNA template during helicase activation . The asymmetric unit of the crystal of PfMCMN:ssDNA contains two independent hexamers , each bound to ssDNA ( Figure 1 , Figure 1—figure supplements 1 , 2; Video 1 ) . The subunits are referred to as A through F ( hexamer 1 ) and G through L ( hexamer 2 ) . Like SsoMCMN ( Pucci et al . , 2007; Liu et al . , 2008 ) , PfMCMN elutes as a monomer by size-exclusion chromatography ( data not shown ) but adopts a hexameric arrangement in the crystal structure . The structure is similar to those of MtMCMN ( Fletcher et al . , 2003 ) and SsoMCMN ( Liu et al . , 2008 ) with three subdomains ( Figure 1—figure supplement 3 ) : a largely helical subdomain A; a Zn-binding subdomain B; and an OB-fold subdomain C . The central pore of the PfMCMN hexameric ring is oval-shaped with a variable diameter around the ring reflecting a significant deviation from pure sixfold symmetry . The RMSD of the C-subdomain Cα-positions from the sixfold permutation is 3 . 03 Å and 1 . 45 Å for hexamers 1 and 2 , respectively . In contrast , PfMCMN without DNA bound is highly symmetric and shows minimal RMSD from sixfold symmetry ( Figure 1—figure supplements 4–6 , RMSD = 0 . 33 Å ) , indicating that DNA induces asymmetry in the MCM ring . The narrowest diameter of the channel is at the β-turn of the C-subdomain ( Figure 1—figure supplement 3 ) , consistent with previous structures of MCMN ( Fletcher et al . , 2003; Liu et al . , 2008 ) . 10 . 7554/eLife . 01993 . 004Figure 1 . One crystallographically unique hexamer viewed parallel ( A ) and perpendicular ( B ) to the channel . The ssDNA is colored cyan . ( A ) Each subunit is uniquely colored and labeled . The side-chains of the two MSSB arginine residues that bind ssDNA are represented in stick . The Zn-binding domains project into the page . The ATPase domains , not present in the crystal structure , would project out of the page . ( B ) The protein is represented in transparent grey to highlight that the ssDNA runs perpendicular to the channel . The Zn-binding domains are at the bottom , and the ATPase domains would be located at the top . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 00410 . 7554/eLife . 01993 . 005Figure 1—figure supplement 1 . Views of the two hexamers of the crystallographic asymmetric unit parallel ( A ) and perpendicular ( B ) to the channel . The ssDNA is colored cyan . ( A ) Each subunit is uniquely colored and labeled . For hexamer 1 , an example MSSB and β-turn are labeled . The Zn-binding domains are projected into the page . The ATPase domains ( not present in the crystal structure ) would project out of the page . The 5′ and 3′ ends of the ssDNA are marked . ( B ) The protein is represented in transparent grey to highlight that the ssDNA runs perpendicular to the channel . The Zn-binding domains are at the bottom , and the ATPase domains ( not present ) would be at the top . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 00510 . 7554/eLife . 01993 . 006Figure 1—figure supplement 2 . Stereoimages of one ssDNA binding PfMCMN subunit interface of each hexamer with Fo-Fc electron density calculated prior to including any DNA in the model . The final model is displayed with the 2 subunits colored and labeled in yellow and cyan and the DNA colored blue . The Fo-Fc electron density is contoured at 3-sigma ( red ) and 5-sigma ( green ) . The DNA backbone is visible at 3-sigma , and the phosphates are visible at 5-sigma . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 00610 . 7554/eLife . 01993 . 007Figure 1—figure supplement 3 . The ssDNA binds to the OB-fold subdomain . ( A ) The individual subdomains are color-coded with the helical bundle in blue , the Zn-binding subdomain in green , the OB-fold subdomain in magenta , and the ssDNA in cyan . ( B ) Cylindrical merge showing how closely MCMN approaches the channel center at each position along the channel axis , and that the greatest available volume in the MCMN channel is at the OB-fold above the β-turn . The hexamer was rotated 360° about the channel axis in 5° increments . All of the models were superimposed , and the Cα positions of each subdomain were used to generate surfaces with MSMS ( Sanner et al . , 1996 ) . The surfaces were uniquely colored as in ( A ) , rendered simultaneously with Raster3D ( Merritt and Bacon , 1997 ) , and clipped with a vertical plane through the center to show the extent of projection into the channel for each part of the hexamer . A grey cylinder ( unclipped ) with 20 Å diameter was placed in the center to indicate the volume for a hypothetical B-form DNA . A similar 360° cylindrical merge was constructed for one of the contiguous ssDNA molecules , and a surface was constructed over all ssDNA atoms . The ssDNA surface was clipped with a vertical plane through the center , and is represented in cyan ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 00710 . 7554/eLife . 01993 . 008Figure 1—figure supplement 4 . Crystal structure of PfMCMN in the absence of DNA viewed parallel ( A ) and perpendicular ( B ) to the channel . Each subunit is uniquely colored and labeled . ( A ) The side-chains of the two arginine residues of the MSSB are represented in stick , and the Zn-binding domains are projected into the page . The ATPase domains , not present in the crystal structure , would project out of the page . ( B ) The Zn-binding domains are at the bottom , and the ATPase domains would be located at the top . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 00810 . 7554/eLife . 01993 . 009Figure 1—figure supplement 5 . Comparison of the crystal structures of PfMCMN bound to ssDNA ( left , in color ) and in the absence of DNA ( right , transparent grey ) . The MSSB arginines are shown in stick representation . The two hexamers are superimposed based upon least-squares alignment of the six C-subdomains ( middle ) . The oval shape of the ssDNA-bound ring is apparent at the red ( chain A ) and green ( chain D ) subunits , which are further from the channel center than in the DNA-free structure . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 00910 . 7554/eLife . 01993 . 010Figure 1—figure supplement 6 . RMSD from sixfold symmetry for each crystallographic hexamer . For each hexamer , the least-squares superposition of all six subunits upon the permuted configuration ( chains ABCDEF superimposed upon BCDEFA ) was calculated based upon the C-subdomains . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 01010 . 7554/eLife . 01993 . 011Figure 1—figure supplement 7 . Comparison of ssDNA binding by the PfMCMN OB-fold subdomain C and by a prototypical OB-fold protein , SSB . Left panels show one monomer of PfMCMN ( chain F ) colored yellow , and the other subunits of the hexamer colored grey . The ssDNA bound by PfMCMN is in cyan . ( A ) One monomer of E . coli SSB ( Raghunathan et al . , 2000 ) is shown in magenta and its associated ssDNA in blue in the right panel . An overlay with ssDNA bound PfMCMN is shown in the middle . ( B ) Comparison of PfMCMN:ssDNA with one monomer of H . pylori SSB ( Chan et al . , 2009 ) in magenta and its associated ssDNA in blue . Note the ∼90° change in direction of ssDNA for the PfMCM compared to the SSB structures . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 01110 . 7554/eLife . 01993 . 012Video 1 . Crystal structure details for PfMCMN:dT30 . The video illustrates the asymmetric unit , which includes two MCM hexamers in a side-by-side orientation . Each subdomain is illustrated in Hexamer 1 to show that the ssDNA interacts with the OB-fold subdomain C . Finally , detailed views of the β-turn and the MCM Single-Stranded DNA binding motif ( MSSB ) are illustrated . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 012 The ssDNA binds inside the central channel of the hexameric ring in an intriguing configuration . The ssDNA circles the interior of the PfMCMN ring in a plane perpendicular to the central channel ( Figure 1 , Figure 1—figure supplement 1 ) . This is in contrast to the ssDNA passing through the central channel , as observed in the structures of the nucleic acid complexes of the motor domains of the hexameric helicases E1 ( Enemark and Joshua-Tor , 2006 ) , Rho ( Thomsen and Berger , 2009 ) , and DnaB ( Itsathitphaisarn et al . , 2012 ) . This distinction suggests that the newly identified MCM-ssDNA interactions might serve a function distinct from motor-driven helicase and translocase activities . The ssDNA binds to the MCMN OB-fold subdomain C at a region consistent with that of the prototype OB-fold protein SSB , but the ssDNA is oriented approximately perpendicular to that seen in SSB-ssDNA structures ( Figure 1—figure supplement 7 , Raghunathan et al . , 2000; Chan et al . , 2009 ) . The ssDNA does not progress towards a specific end of the channel; therefore , the ssDNA does not have an assignable entry or exit direction from the ring . Instead , the ssDNA has a defined polarity relative to the MCM ring . When viewed from the C-terminal side of the complex ( as shown in Figure 1A ) , the 5′ to 3′ direction of the bound ssDNA proceeds clockwise around the channel . This polarity is observed for both ssDNAs in each hexamer of the asymmetric unit . The structure reveals that the individual MCM subunits do not all simultaneously participate in ssDNA binding . In each hexamer , the bound nucleotides are not continuous but are separated into two stretches . Overall , two 7-mer stretches are observed in hexamer 1 , and 11-mer and 4-mer stretches are observed in hexamer 2 . The subunits that interact with DNA use a consistent binding mode with four nucleotides per subunit ( Figure 1 , Figure 2 , Figure 2—figure supplement 1 ) . The fourth nucleotide from the 5′-end of this binding mode is visible in the cases where it spans binding at adjacent subunits , but it is often disordered at the 3′-end of a ssDNA stretch . The four nucleotide per subunit binding increment contrasts with the motor domains of other hexameric helicases that bind either one ( E1 , Enemark and Joshua-Tor , 2006; Rho , Thomsen and Berger , 2009 ) or two ( DnaB , Itsathitphaisarn et al . , 2012 ) nucleotides per subunit and indicates that 24 nucleotides can bind if all the subunits simultaneously engage the ssDNA . The absence of ssDNA binding at some subunits is not due to insufficient DNA length because a 30-mer oligonucleotide was used for crystallization . The discontinuous DNA could result from the hexamer binding two separate 30-mer strands simultaneously or from the hexamer tightly binding one 30-mer ssDNA strand at two regions with the intervening nucleotides binding either weakly or not at all . We consider the latter to be more likely because binding of two parts of the same strand is anticipated to be cooperative . 10 . 7554/eLife . 01993 . 013Figure 2 . Stereoviews of the protein-DNA interaction details for two subunit interfaces . The binding predominantly involves residues on the face of the OB-fold of one subunit , yellow , including an interaction between a thymidine base and main-chain atoms of the β-strand . This thymidine is sandwiched between F202 of one subunit and E127 of the adjacent subunit in cyan . Lysine 129 of the neighboring subunit ( cyan ) interacts with both the DNA and the yellow subunit . The specific interfaces depicted are ( top ) between chains F ( yellow ) and A ( cyan ) and ( bottom ) between chains A ( yellow ) and B ( cyan ) . The structural details of DNA-binding appear highly similar at the other interfaces where DNA is observed ( see Figure 2—figure supplement 1 ) . The main interactions involve R124 and R186 . The presence of ssDNA correlates with the proximity of the two subunits as defined by the distance between the R201 Cα and E127 Cα positions ( magenta arrow ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 01310 . 7554/eLife . 01993 . 014Figure 2—figure supplement 1 . 12 stereoimages of the PfMCM interfaces sorted by intersubunit distance to emphasize the correlation with DNA-binding . Electron density following refmac refinement ( refmac FWT map ) is displayed around DNA in green , and around the protein in blue . The adjacent subunits are colored yellow and cyan , and the specific chains are noted with the same color scheme . The distance between the R201 Cα atom of the yellow subunit and the E127 Cα atom of the cyan subunit is displayed in red . Electron density for ssDNA is observed for each interface where the distance is less than 7 . 5 Å . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 014 The capacity of a subunit to bind ssDNA is determined by intersubunit distance ( Figure 1 , Figure 2 , Figure 2—figure supplement 1 ) . To compare the distance between different subunit pairs , we measured the distance between the R201 Cα atom of one subunit and the E127 Cα atom of the counterclockwise subunit as viewed in Figure 1 ( Magenta arrow , Figure 2 ) . DNA-binding is consistently observed at the first subunit if this distance is less than 7 . 5 Å , and it is not observed if this distance exceeds 8 . 4 Å . The interface between subunits J and K shows an intermediate ( 7 . 6 Å ) distance , and the electron density between F202 ( subunit J ) and E127 ( subunit K ) is much weaker than at the interfaces where DNA has been modeled ( Figure 2—figure supplement 1 ) . The correlation of ssDNA binding with intersubunit configuration is conceptually similar to multi-subunit ATPase sites where different intersubunit configurations determine the ability to bind or hydrolyze ATP ( Abrahams et al . , 1994; Enemark and Joshua-Tor , 2008 ) . In MCMN , changes to the intersubunit configuration dictate binding to ssDNA . The most significant interactions between PfMCMN and ssDNA involve two adjacent arginines , R124 and R186 , that project from the β-barrel of the OB-fold towards the ring interior ( Figures 1 and 2 ) . These residues interact with oxygen atoms of the sugars and bases of the ssDNA ( Figure 2 ) and are highly conserved in other MCM proteins ( Figure 3 ) . We refer to this conserved region as the MCM Single-Stranded DNA Binding motif ( MSSB ) . Interestingly , one thymidine base projects towards the β-barrel of the OB-fold ( Figure 2 ) and makes two hydrogen bonds to main-chain atoms of one strand of the β-barrel . This base also sits at the subunit interface , between the side-chains of phenylalanine 202 of one subunit and glutamic acid 127 of the adjacent subunit . The β-turn residues R234 and K236 do not interact with ssDNA in the structure . The DNA-binding consists predominantly of interactions with the sugars and bases rather than the backbone phosphates . In contrast , the hexameric helicases E1 ( Enemark and Joshua-Tor , 2006 ) ; Rho ( Thomsen and Berger , 2009 ) ; and DnaB ( Itsathitphaisarn et al . , 2012 ) bind nucleic acid mainly through interactions with backbone phosphates . 10 . 7554/eLife . 01993 . 015Figure 3 . MCM family-specific sequence-alignment in the regions where the strongest interactions with ssDNA are observed . Globally conserved residues are shaded dark blue , and family-specific conserved residues are shaded light blue . Residues identified to participate in DNA-binding from our structure ( red dot ) and prior work ( Pucci et al . , 2004 ) ( lavendar dot ) are noted above the sequences . Conserved residue positions for ssDNA binding are shaded red and correspond to R124 and R186 in PfMCM ( Figure 2 ) . pf = Pyrococcus furiosus; mt = Methanothermobacter thermautotrophicus; sso = Sulfolobus solfataricus; ap = Aeropyrum pernix; gi = Giardia lamblia; aq = Amphimedon queenslandica; cr = Chlamydomonas reinhardtii; sc = Saccharomyces cerevisiae; sp = Schizosaccharomyces pombe; at = Arabidopsis thaliana; ce = Caenorhabditis elegans; dm = Drosophila melanogaster; xl = Xenopus laevis; dr = Danio rerio; gg = Gallus gallus; hs = Homo sapiens . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 015 We investigated the role of the identified residues in MCM DNA binding using mutational analysis and electrophoretic mobility shift assays . As expected , wild-type PfMCMN binds single-stranded ( Figure 4 , Khalf = 6 . 8 μM ) and double-stranded ( Figure 4—figure supplement 1 , Khalf = 7 . 0 μM ) oligonucleotides . The arginine residues R124 and R186 make the most significant ssDNA interactions in the structure . R124A and R186A mutants each show a significant decrease in ssDNA binding ( 7- and 6-fold reduction , respectively ) . Simultaneous mutation of both arginines showed even stronger defects ( 25-fold reduction ) , with no detectable ssDNA binding unless the protein concentration was increased dramatically ( Figure 4 ) . The K129A mutant is modestly defective in binding ssDNA ( fourfold reduction , Figure 4 ) . The individual R124A , R186A , and K129A mutants bind dsDNA with comparable affinity to wild-type ( Figure 4—figure supplement 1 ) . The R124A/R186A double mutant shows only modest defects in dsDNA binding ( threefold reduction ) . Alanine mutants of other less-conserved residues did not significantly impair ssDNA- or dsDNA-binding . For example , consistent with the involvement of its main chain amide rather than its side chain in ssDNA binding , the β-turn K233A mutant does not significantly impair ssDNA binding . Similarly , the F202 side-chain interacts with a thymidine base , but it is offset from an ideal stacking interaction ( Figure 2 ) . The corresponding F202A mutant is not impaired in ssDNA binding and is not conserved as aromatic in other Mcm proteins ( Figure 3 ) . 10 . 7554/eLife . 01993 . 016Figure 4 . Electrophoretic mobility shift of 40-mer oligo-dT in the presence of PfMCMN . The ssDNA , 160 nM with a 5′-fluorescein-label , was titrated with increasing concentrations ( 1 . 4 , 2 , 2 . 7 , 6 . 8 , 13 . 5 , 20 . 3 , 27 , 40 . 5 , 54 μM ) of PfMCMN . The lane marked ‘−’ is loaded with control sample lacking protein . Mutation of residues R124 and R186 significantly impairs binding to ssDNA . The R124A/R186A double mutant was titrated with larger concentrations ( 54 , 81 , 108 , 135 , 162 , 189 , 216 , 243 , 270 μM ) of PfMCMN in order to detect binding . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 01610 . 7554/eLife . 01993 . 017Figure 4—figure supplement 1 . Electrophoretic mobility shift assay of a 26-mer dsDNA substrate in the presence of PfMCMN . The dsDNA , 160 nM with a 5′-fluorescein-label , was titrated with increasing concentrations ( 2 , 3 , 4 , 5 , 7 . 5 , 10 , 12 . 5 , 15 , 20 μM ) of PfMCMN . The lane marked ‘−’ is loaded with control sample lacking protein . The R124A/R186A double mutant was slightly impaired in binding dsDNA and was titrated with larger concentrations ( 10 , 12 . 5 , 15 , 17 . 5 , 20 , 25 , 30 , 35 , 40 μM ) of PfMCMN . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 017 In S . cerevisiae ( Sc ) , the PfMCM R124 and R186 amino acids within the MSSB motif are both conserved as arginine or lysine in Mcm4 , Mcm6 and Mcm7 whereas Mcm2 , Mcm3 and Mcm5 show a positively charged residue at only one of the two sites ( Figure 3 ) . To test the role of the MSSB motif in S . cerevisiae DNA replication , we constructed double-alanine mutants in ScMCM4 ( mcm4-R334A/K398A = mcm4D ) , ScMCM6 ( mcm6-R296A/R360A = mcm6D ) and ScMCM7 ( mcm7-R247A/K314A = mcm7D ) as these subunits showed the most similarity to PfMCM in the MSSB . We tested the ability of these mutations to replace the corresponding wild-type Mcm subunit in S . cerevisiae cells . When present as the only mutant Mcm subunit in the cell , mutations in the ScMCM4 , ScMCM6 or ScMCM7 MSSB complemented deletion of the corresponding gene ( Figure 5A , Figure 5—figure supplement 1 ) . Because the DNA binding defects observed for the mutant PfMCM complexes altered all six subunits , we tested the ability of pairwise combinations of the ScMCM MSSB mutations to function in place of their wild-type counterparts . In contrast to the single mutations , all three double-mutant combinations did not support cell division . The dramatic phenotypic difference between the double and single mutations may be due to a requirement for two adjacent subunits to create a productive ssDNA interaction . Because the Mcm4 , 6 and 7 subunits are adjacent to one another in the Mcm2-7 complex , each pairwise combination would be expected to interrupt at least three possible subunit pairs for binding ( e . g . , the Mcm4/6 double mutant would interfere with Mcm2/6 , Mcm6/4 and Mcm4/7 subunit pairs for ssDNA binding ) . 10 . 7554/eLife . 01993 . 018Figure 5 . Mutation of two MSSB motifs is lethal and causes helicase loading defects . ( A ) Mutation of two Mcm4 , 6 , 7 MSSB motifs is lethal . Subunit arrangement in the Mcm2-7 ring viewed from the C-terminal side . The Mcm4 , 6 , and 7 subunits are adjacent to each other across from the Mcm2/5 gate . All pairwise combinations of the Mcm4 , 6 and 7 MSSB mutants are lethal whereas the individual MSSB mutants are viable . ( B ) Helicase loading with the indicated MSSB double mutant Mcm2-7 complexes . Three forms of the assay are shown: following a high-salt wash to monitor completion of loading ( top panel ) ; in the presence of ATPγS instead of ATP to monitor the initial association of the helicase and all of the helicase loading proteins ( ORC , Cdc6 and Cdt1 , middle panel ) ; and with ATP following a low salt-wash , allowing bound helicase loading proteins to be maintained ( bottom panel ) . All loading was dependent on Cdc6 and proteins are detected after SDS-PAGE and fluorescent protein staining . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 01810 . 7554/eLife . 01993 . 019Figure 5—figure supplement 1 . All pairwise combinations of mcm4D , mcm6D and mcm7D mutants were not viable . The parent strains have an MSSB mutation in the indicated MCM gene . They are also deleted for the indicated MCM gene and depend on a URA+ plasmid expressing a wild-type copy of the same gene for viability . These strains were transformed with the indicated ( in the center of each plate ) TRP+ plasmid expressing wild-type MCM gene ( left ) or the MSSB mutant MCM gene ( right ) . Complementation in the absence of the URA3/MCMX-WT plasmid was tested by growth on plates containing 5-FOA ( which selects against cells containing the URA3 plasmid ) . Consistent with the single mcm4 , mcm6 or mcm7-MSSB mutations being viable , we observe growth in the presence of the pTRP/MCMX-WT plasmid but no growth when the TRP plasmid contains an MSSB allele in the second MCM gene ( creating MSSB mutations in two of the MCM4/6/7 genes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 01910 . 7554/eLife . 01993 . 020Figure 5—figure supplement 2 . Comparison of wild-type and MSSB double- and triple-mutant Mcm2-7/Cdt1 complexes . ( A ) Wild-type and MSSB mutant Mcm2-7/Cdt1 complexes have similar subunit composition . Purified Mcm2-7/Cdt1 complexes were separated by SDS-PAGE and stained with coomassie blue . Mcm2-7 and Cdt1 proteins are indicated . ( B ) Wild-type and MSSB mutant complexes have similar Stokes radii . Wild-type and mutant Mcm2-7/Cdt1 complexes were separated on a Superdex 200 gel filtration chromatography . Fractions 16–19 of each separation are shown after SDS-PAGE and coomassie blue staining . ( C ) Helicase loading with the indicated MSSB mutant Mcm2-7 complexes . Three forms of the assay are shown: following a high-salt wash to monitor completion of loading ( top panel ) ; in the presence of ATPγS instead of ATP to monitor the initial association of all of the helicase and all of the helicase loading proteins ( ORC , Cdc6 and Cdt1 , third panel ) ; and with ATP following a low salt-wash , allowing bound helicase loading proteins to be maintained ( fourth panel ) . The relative loading of the Mcm mutants compared to wild-type Mcm2-7 was measured based on three independent loading ( high-salt wash ) experiments ( second panel ) . Error bars indicate the standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 020 To define further the molecular defects of the mutant S . cerevisiae Mcm2-7 complexes , we purified Mcm2-7 complexes containing the lethal double mutants ( Mcm4D/6D , Mcm4D/7D and Mcm6D/7D ) along with the associated Cdt1 protein . We also purified the Mcm4/6/7 triple mutant ( Mcm4D/6D/7D ) with associated Cdt1 . After purification , all of the mutant complexes showed similar subunit composition and migration in gel filtration columns as wild-type Mcm2-7/Cdt1 ( Figure 5—figure supplement 2 ) . Thus , these mutations do not inhibit the initial assembly of the Mcm2-7/Cdt1 complex . We tested each of the mutant complexes for their ability to be loaded onto origin DNA using a reconstituted helicase-loading assay ( Evrin et al . , 2009; Remus et al . , 2009; Figure 5B ) . To ensure that all of the Mcm2-7 hexamers retained on the DNA were loaded , we washed the final DNA-associated proteins with high salt . This treatment removes all of the helicase loading proteins ( ORC , Cdc6 and Cdt1 ) from the DNA but leaves loaded Mcm2-7 complexes ( Figure 5B , top panel ) ( Randell et al . , 2006 ) . Wild-type protein showed robust , Cdc6-dependent loading onto origin DNA . In contrast , each of the double mutant Mcm2-7 complexes showed reduced Mcm2-7 loading . The Mcm4D/6D and Mcm6D/7D complexes showed only modest defects ( less than ∼ two-fold , Figure 5—figure supplement 2 ) . The Mcm4D/7D complex showed a stronger defect ( ∼10-fold ) , and the Mcm4/6/7 triple mutant showed the most severe defect in helicase loading ( ∼20-fold reduction , Figure 5—figure supplement 2 ) . To establish at what step in the helicase loading process these defects occurred , we studied the initial recruitment of the complexes to origin DNA . To this end , we replaced ATP with the poorly hydrolyzable ATP-γS in the assay . In the presence of ATPγS , all of the proteins required for helicase loading are recruited to the origin , but no loading occurs ( Randell et al . , 2006 ) . Under these conditions , we observed a similar pattern of Mcm2-7/Cdt1 and ORC association for wild-type and the mutant Mcm2-7 complexes ( Figure 5B , middle panel , Figure 5—figure supplement 2 ) . Thus , mutating two or three MSSB motifs did not alter the initial recruitment of the Mcm2-7/Cdt1 complex to the origin DNA . We also examined the DNA-associated proteins when ATP-containing reactions were washed with low-salt ( Figure 5B , bottom panel , Figure 5—figure supplement 2 ) , a condition that retains helicase-loading proteins on DNA . Under these conditions , the mutant complexes showed a similar pattern of reduced Mcm2-7 DNA association as seen for the high-salt wash experiments . Cdt1 was not retained on the DNA under these conditions for mutant Mcm2-7 complexes , indicating that the MSSB mutations did not interfere with the release of Cdt1 from the Mcm2-7 complex during loading . Together , these data indicate that the loading defect for these Mcm2-7 mutants occurs after their initial recruitment to origin DNA but before the establishment of the ORC-independent association of Mcm2-7 with origin DNA . We looked for additional replication initiation defects for the Mcm2-7 mutants that showed detectable loading using a modified in vitro replication assay that recapitulates origin-dependent DNA replication initiation and elongation ( Heller et al . , 2011 ) . In contrast to our original studies , helicase loading in these assays was performed using purified proteins . In addition to measuring new DNA synthesis , we monitored association of Mcm2-7 , the helicase activation proteins Cdc45 and GINS and the ssDNA binding protein , RPA , with the origin DNA during the reaction . The analysis of protein associations provided insights into the step during replication initiation during which the mutant Mcm2-7 complexes fail . Consistent with their inability to support cell growth , none of the mutant complexes supported significant DNA synthesis ( Figure 6 ) . Analysis of FLAG-Mcm3 DNA association showed that , as in the loading assays , the Mcm4D/6D and Mcm6D/7D complexes are retained on the DNA more strongly than the Mcm4D/7D complex . Cdc45 association mirrored the level of FLAG-Mcm3 association with the DNA , suggesting Cdc45 recruitment is independent of the MSSB ( Figure 6—figure supplement 1 ) . In contrast , all of the Mcm2-7 double mutants showed similarly strong defects ( ≥10-fold ) in both GINS and RPA DNA association . In the case of Mcm4D/7D mutant , the DNA replication , GINS and RPA DNA association defects are consistent with its helicase-loading defect . In contrast , for Mcm4D/6D and Mcm6D/7D , the extent of helicase loading and Cdc45 DNA association is distinct from the much larger losses in GINS and RPA DNA association and DNA replication ( Figure 6—figure supplement 1 ) . These data strongly suggest that an inability to recruit or maintain GINS and/or RPA is responsible for the replication defects exhibited by these mutants . Because RPA DNA binding is a readout for ssDNA formation and GINS is required to activate the Mcm2-7 helicase , both of these defects indicate that the Mcm4/6 and Mcm6/7 MSSB mutants are defective for helicase activation . 10 . 7554/eLife . 01993 . 021Figure 6 . The Mcm2-7 MSSB double mutants are severely defective for in vitro DNA replication . Proteins associated with the DNA template during DNA replication were analyzed by immunoblotting ( top panels ) and radiolabeled DNA replication products were analyzed by alkaline agarose electrophoresis ( bottom panel ) . All of the mutants are strongly defective for DNA replication and GINS and RPA DNA template association relative to wild-type Mcm2-7 . The levels of Cdc45 and Mcm2-7 ( FLAG-Mcm3 ) association reflected the levels of helicase loading by the same MSSB double mutant Mcm2-7 complexes . Quantitation of these data is shown in Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 02110 . 7554/eLife . 01993 . 022Figure 6—figure supplement 1 . Quantitation of DNA template association of Mcm3 , Cdc45 , GINS and RPA and DNA replication products for the Mcm2-7 mutants relative to wild-type . The level of Cdc45 DNA-association mirrored the level of Mcm3 DNA-association . All of the mutants were severely defective for GINS and RPA template association and in vitro replicaiton ( bottom right ) . The levels of GINS and RPA template association ( rather than Cdc45 association ) correlate with the levels of replication observed . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 022 Here we show how the PfMCM N-terminal domain interacts with single-stranded DNA and identify a critical set of interacting residues that we define as the MSSB . These residues are important for binding ssDNA and , to a lesser extent , dsDNA . A DNA-binding role for positively charged residues in this region is consistent with previous mutational analysis of SsoMCM showing that K129A ( equivalent to PfMCM R124 ) displays very little binding to ssDNA , blunt duplex DNA , and bubble-DNA substrates ( Pucci et al . , 2004 ) . Although a positive residue equivalent to PfMCM R186 is not conserved in SsoMCM , mutation of an adjacent residue , K194A also displays very little binding to these DNA substrates ( Pucci et al . , 2004 ) . As previously noted in overall sequence comparisons ( Pucci et al . , 2004 ) , residues in the MSSB motif are conserved in specific families in eukaryotic Mcm2-7 . Importantly , we show that conserved residues within this motif are critical for S . cerevisiae cell division and multiple Mcm2-7 functions during replication initiation . Biochemical analysis of the S . cerevisiae mutant complexes reveals multiple defects during replication initiation . Two mutant complexes ( Mcm4D/7D and Mcm4D/6D/7D ) show strong defects in Mcm2-7 loading . This is unexpected because Mcm2-7 proteins are loaded around dsDNA and there is no evidence for ssDNA at this stage of replication ( Evrin et al . , 2009; Remus et al . , 2009 ) . It is possible that one or more MSSB motifs interact with dsDNA prior to ssDNA formation at the origin and that these interactions stabilize loaded Mcm2-7 . This would be consistent with the dsDNA binding defects observed for the PfMCMN R124A/R186A double mutant ( Figure 4—figure supplement 1 ) and also the ( R124-equivalent ) K129A mutant of SsoMCM . Alternatively , elimination of positive charges in the central channel could alter the opening and closing of the Mcm2-7 ring . The abundance of positive charges in the Mcm2-7 ring could predispose the ring to remain open prior to DNA binding . Encircling dsDNA could neutralize the repulsion and favor ring closing . It is possible that a reduction in positive charge in the mutant complexes leads to the Mcm2-7 ring spending more time in the closed state , inhibiting entry of the dsDNA during loading . Analogously , the reduced positive charge of the MSSB mutants could destabilize ring closure around dsDNA during loading . Consistent with this model , the Mcm2-7 complex appears as a cracked-ring in solution ( Costa et al . , 2011 ) . As we observe , both scenarios predict that the strongest loading defects would be observed for the Mcm4D/6D/7D mutant that eliminates the greatest number of positive charges . Among the double mutants , the strongest loading defect is observed when the Mcm4 and Mcm7 subunits are mutated , which are across from the Mcm2/5 gate and could influence opening and closing more than other subunits . Several lines of evidence suggest that the MCM-ssDNA interactions that we have identified have a role during dsDNA melting . First , the MCM-ssDNA interactions identified in our structure predominantly involve the sugars and bases of the ssDNA , ideally suited to bind and shield one strand from its complement during melting . Also consistent with a role in dsDNA melting , the Mcm2-7 MSSB mutant complexes showed strong defects in events linked to helicase activation . The MSSB mutations did not alter Cdc45 recruitment , consistent with the observation that this event can occur in G1 phase prior to ssDNA formation ( Aparicio et al . , 1999; Heller et al . , 2011; Tanaka et al . , 2011 ) . In contrast , the levels of GINS and RPA DNA association by each of the MSSB mutant complexes were strongly defective . The defect in RPA DNA binding is almost certainly due to reduced ssDNA generation by the mutant complexes . The reduction in DNA-associated GINS could be the result of a defect in recruitment or retention of GINS . Unlike Cdc45 , GINS recruitment does not occur until entry into S phase ( Kanemaki et al . , 2003; Takayama et al . , 2003 ) and , therefore , could require ssDNA formation . Alternatively , it is possible that the defect in ssDNA binding prevents the CMG complex from attaining a particular DNA binding state and this destabilizes GINS binding . Interactions between the MSSB and ssDNA could also occur during elongation . Consistent with a role for the MSSB in unwinding , the SsoMCM K129A mutant ( PfMCM R124 equivalent ) is defective for helicase activity ( Pucci et al . , 2004 ) . Although the MCM ATPase domain alone is sufficient to produce unwinding activity in SsoMCM ( Barry et al . , 2007; Pucci et al . , 2007 ) and in Aeropyrum pernix MCM ( Atanassova and Grainge , 2008 ) , unwinding displays greater processivity in the presence of the N-terminal domain for SsoMCM ( Barry et al . , 2007 ) . Thus , although the N-terminal domain and the residues of the MSSB are not intrinsically required to produce an unwinding activity , the N-terminal domain can regulate and enhance MCM unwinding activity ( Barry et al . , 2007 ) . The positively charged residues of the MSSB could help maintain a closed MCM ring as described above for loading , and thus contribute to the enhanced processivity afforded by the N-terminal domain . It is also possible that ssDNA binding by the MSSB has a more direct impact on DNA unwinding . For example , the directional ssDNA:MSSB interactions observed here could influence the polarity of unwinding either during initiation ( see below ) or elongation . To permit the ssDNA:MCMN interactions that we observe , the ssDNA would need to alter its trajectory as it passes through the MCM central channel . Alternatively , the MSSB could bind ssDNA differently during unwinding . An interesting possibility is that during elongation the MCM OB-fold binds ssDNA similar to the OB-fold prototype SSB ( Raghunathan et al . , 2000; Chan et al . , 2009 ) . This mode of binding would place the ssDNA approximately parallel to the central channel ( Figure 1—figure supplement 7 ) , a position consistent with the expected ssDNA trajectory during unwinding . Different modes of interaction between the MSSB and ssDNA could be modulated by the AAA+ domain of MCM and a conserved ‘allosteric communication loop’ ( ACL , Sakakibara et al . , 2008 , Barry et al . , 2009 ) that projects from the N-terminal domain towards the anticipated position of the ATPase domain . The ACL directly follows the β-strand that contains the second positively charged MSSB residue ( PfMCM residue R186 ) and thus could couple the MSSB to the ATPase domains . The polarity of ssDNA bound to MCMN observed in our structure has important implications for the transition between MCM dsDNA and ssDNA binding . In the view shown in Figure 7A , the AAA+ motors are located above the MCMN domain , and the corresponding Mcm2-7 subunits occur clockwise in the order Mcm5 , 3 , 7 , 4 , 6 , 2 . Given that the Mcm2-7 complex is initially loaded around dsDNA , only one of the two strands of dsDNA can easily attain the 5′→3′ coplanar clockwise configuration observed in our structure: the DNA strand that passes from the C- to N-terminus of the MCM complex in a 5′→3′ direction ( Figure 7A ) . Intriguingly , this strand corresponds to the leading strand DNA template that is encircled by the MCM complex during translocation/DNA unwinding . For the opposite strand to interact with the MCMN with the observed polarity , it would either need to pass through the other strand or dramatically re-orient . Thus , if ssDNA is formed within the MCM ring during origin melting ( see below ) , our structure predicts that MCMN would preferentially bind to the translocating strand ( i . e . , the leading strand DNA template ) . Consistent with this model , the 3′→5′ helicase polarity of SsoMCM is only observed when the N-terminal domain is present , implicating this domain in substrate selection ( Barry et al . , 2007 ) . 10 . 7554/eLife . 01993 . 023Figure 7 . A model for MSSB-dependent selection of the translocating DNA strand during helicase activation . ( A ) The defined polarity of ssDNA binding by the MCMN would preferentially bind the leading-strand DNA template . The Mcm2-7 complex N-terminus is shown from the C-terminal side of the complex . This is the side where DNA is expected to enter during translocation . Duplex DNA is first encircled by the ring ( left ) . Only the red strand can readily attain the 5′→3′ clockwise polarity observed in the crystal structure . This strand passes through the ring 5′→3′ from the C- to the N-terminal side and thus is the correct polarity to serve as the translocating strand . We propose the grey , lagging strand DNA template will exit through the Mcm2/5 gate , possibly as a result of accumulation of ssDNA in the central channel ( right ) . ( B ) A model for selecting the translocating strand during origin melting . Symmetric surfaces in different shades of green represent the two MCMN portions of a double hexamer . The dsDNA is first encircled by the MCM double hexamer ( left panel ) . The dsDNA is driven toward the double hexamer interface by the dsDNA translocase activity of the AAA+ ATPase domains ( not shown ) , which would be located above the light green surface and below the dark green surface . The dsDNA translocation creates strand separation where volume is available , enabling the MSSB to preferentially bind the strand with 5′→3′ clockwise polarity when viewed from the ATPase domain ( middle panel ) . Importantly , the MSSB-bound strand corresponds to the strand upon which the MCM helicase will translocate during unwinding ( right panel , magenta at top , cyan at bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 023 The MCM helicase is conceptually similar to the Rho hexameric helicase because both possess an N-terminal OB-fold linked to a C-terminal ATPase . This analogy further supports a role for the MCM OB-fold during helicase activation prior to unwinding . The crystal structure of Rho with RNA bound at the OB-fold ( Skordalakes and Berger , 2003 ) suggests that 70–80 nucleotides of RNA would adopt a circular path around the ring ( Skordalakes and Berger , 2003 ) that is roughly perpendicular to the hexameric channel . This arrangement is conceptually similar to our PfMCMN:ssDNA structure . The Rho OB-fold is believed to bind RNA and facilitate encircling of single-stranded RNA during ring closure by the ATPase domains ( Skordalakes and Berger , 2003 ) , a prerequisite for establishing an activated helicase . Subsequently , the proposed unwinding mechanism for Rho exclusively involves distinct interactions between the ATPase motor domain and RNA ( Thomsen and Berger , 2009 ) . The MCM N-terminal domain may also function to enable the ATPase domains to select and encircle one strand of DNA during ring closure . A key difference between MCM and Rho is that the Rho helicase ring is loaded on a species that is already single-stranded , whereas the MCM hexamer is first loaded onto double-stranded DNA that must somehow be converted to single-stranded DNA ( Evrin et al . , 2009; Remus et al . , 2009 ) . Combining the features of eukaryotic MCMs with our new structural information , we suggest the following model for MSSB function during helicase activation . After helicase loading , we propose that DNA melting is initiated by activating the ATPase domains of the double-hexamer to pump dsDNA from the C-terminal lobe side towards the double-hexamer interface ( Figure 7B; Video 2 ) . This is consistent with the known direction of MCM DNA translocation ( McGeoch et al . , 2005 ) as well as observations that Mcm complexes can translocate on dsDNA ( Kaplan et al . , 2003 ) . As additional nucleotides of DNA are forced to occupy the same distance along the DNA helical axis , a B-form structure can no longer be maintained . We predict that the DNA strands would be forced apart at the site where the diameter of the MCM central channel is largest . Intriguingly , the MSSB is on the surface of the largest diameter of the MCMN central channel ( Figure 1—figure supplement 3 ) , putting the MSSB in a prime position to bind the leading strand ssDNA upon DNA melting . The channel diameter elsewhere in the MCMN is too narrow to permit B-form DNA strands to separate . Such melting activity requires that the two hexamers are anchored to one another because the two hexamers would otherwise simply translocate away from one another without melting the DNA . Further dsDNA pumping after the volume around the MSSB has been filled would require the MCM ring to open and allow the unbound lagging strand DNA template to exit ( Figure 7A ) . The presumed exit site would be through the Mcm2/5 gate ( Bochman and Schwacha , 2007; Costa et al . , 2011 ) . Intriguingly , Mcm2 and Mcm5 are the only two subunits that lack a conserved positive residue at the PfMCM R124 position , reducing ssDNA affinity and potentially facilitating strand exit . Following strand exit and extrusion of additional lengths of ssDNA , ring closure would poise each isolated hexamer to unwind DNA using a strand exclusion mechanism ( Fu et al . , 2011 ) . The event that would drive double hexamer separation is unclear but could be facilitated by the change from encircling dsDNA to ssDNA , binding of additional factors ( e . g . , Mcm10 ) or modification of the helicase . A definitive test for this model awaits the development of assays that directly monitor the events of origin DNA melting and strand exclusion . Nevertheless , our studies provide structural and biochemical evidence that the MSSB is a critical ssDNA binding domain that functions during helicase loading and activation and provide initial insights into how ssDNA binding by MCM complex could facilitate selection of one strand during helicase activation . 10 . 7554/eLife . 01993 . 024Video 2 . Animation of a model for MCM to select the translocating strand during origin melting . Symmetric surfaces in different shades of green represent the two MCMN portions of a double hexamer . The dsDNA is first encircled by the MCM double hexamer . The dsDNA is driven toward the double hexamer interface by the dsDNA translocase activity of the AAA+ ATPase domains ( not shown ) , which would be located above the light green surface and below the dark green surface . The dsDNA translocation creates strand separation where volume is available , enabling the MSSB to preferentially bind the strand with 5′→3′ clockwise polarity when viewed from the ATPase domain . Importantly , the MSSB-bound strand corresponds to the strand upon which the MCM helicase will translocate ( magenta at top , cyan at bottom ) , as shown in Figure 7B , right panel . DOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 024 An N-terminal His6-SUMO-PfMCM1-256 expression construct was prepared . The original SUMO vector was the generous gift of Dr Christopher D Lima ( Mossessova and Lima , 2000 ) . An existing His6-SUMO-tagged-fusion protein expression construct in a pRSFduet ( EMD Millipore , Darmstadt , Germany ) plasmid was treated with BamHI and XhoI to completely excise the original fusion partner to generate a BamHI site in-frame with the SUMO tag . This digested species was treated with phosphatase and gel-purified . A DNA fragment encoding the first 256 amino acids of Pyrococcus furiosus MCM was amplified by PCR with primers flanked by BamHI and SalI restriction sites . This fragment was digested with BamHI/SalI , ligated into the BamHI/XhoI-prepared vector , and was transformed into DH5α cells . The integrity of a single colony clone was verified by restriction digest pattern and by DNA sequencing ( pLE009 . 3 ) . Mutants were prepared by site-directed mutagenesis , and the sequences were verified by the Hartwell Center DNA Sequencing Facility ( St . Jude Children’s Research Hospital ) . Expression plasmid pLE009 . 3 ( WT ) , pCF001 . 1 ( R124A ) , pCF002 . 1 ( K129A ) , pCF003 . 1 ( R186A ) , pCF004 . 1 ( F202A ) , pCF0027 . 1 ( K233A ) , or pCF009 . 1 ( R124A/R186A ) was transformed into BL21 ( DE3 ) -RIPL ( Agilent Technologies , Santa Clara , CA ) chemically competent cells and grown overnight in a 100 ml starter culture containing 30 mg/l kanamycin . The starter culture was distributed among 6 l of LB media containing 0 . 4% glucose and 30 mg/l kanamycin and grown to an O . D . of 0 . 3 at 37°C when the temperature was lowered to 18°C . When the O . D . had reached 0 . 7 , expression was induced by 0 . 5 mM IPTG , and the cells were grown for 16 hr at 18°C and harvested by centrifugation . The cells were lysed with a microfluidizer , and the soluble fraction was isolated by centrifugation and ammonium sulfate was added to 70% saturation . The precipitate was isolated by centrifugation , resuspended , and purified by Ni-NTA ( Qiagen , Venlo , the Netherlands ) chromatography . The elution was further purified by anion exchange , and the SUMO tag was removed by overnight digestion with Ulp1 protease ( the Ulp1 protease plasmid was the generous gift of Dr Christopher D Lima , Mossessova and Lima , 2000 ) . The NaCl concentration was raised to 1M , and the sample was passed over Ni-NTA resin , and the flowthrough was purified by anion exchange followed by gel filtration chromatography . The protein elutes at a volume consistent with a monomer . Pooled fractions were concentrated to 10–20 mg/ml . SDS-PAGE was used to assess the purity , and the protein concentration was determined by A280 measurements ( ε = 11 , 460 M−1 cm−1 as determined by the ExPASy ProtParam tool ) . Purified PfMCMN variants were stored at 4°C in buffer containing 20 mM HEPES , pH 7 . 6 , 200 mM NaCl , 5 mM β-mercaptoethanol . Crystals of PfMCMN in complex with a 30-mer poly-dT oligonucleotide were grown at 18°C in a hanging drop containing 1 μl of protein solution pre-mixed with a 30-mer poly-dT oligonucleotide ( 13 . 2 mg/ml protein; 120 μM poly-dT ) and 2 μl of well solution ( 50 mM MES , pH 6 . 0 , 10 mM Mg ( OAc ) 2 , 28 . 5% PEG 3350 ) . Data were collected at SER-CAT beamline 22-ID at the Advanced Photon Source at Argonne National Lab . Data were collected at 1 . 0 Å wavelength in 0 . 5° oscillations for a total of 190° of crystal rotation at 100 K . Data were integrated and scaled with the HKL-2000 package ( Otwinowski and Minor , 1997 ) to 3 . 2 Å resolution . Initial phases were determined by molecular replacement by the program Phaser ( McCoy et al . , 2007 ) that placed 12 copies of a monomer of PfMCMN ( see below ) in two hexamers . Following this placement , difference maps revealed strong electron density within the hexameric channels of both hexamers . The protein model was iteratively refined and manually improved until advancement ceased . At this stage , the difference electron density within the channel was observed at the 5-sigma level ( Figure 1—figure supplement 2 ) , and it was assigned as single-stranded DNA . The model was refined at various stages with CNS ( Brunger et al . , 1998; Brunger , 2007 ) , phenix ( Afonine et al . , 2012 ) , and refmac5 ( Vagin et al . , 2004 ) . The final refinement was carried out with refmac5 using 3 TLS ( Winn et al . , 2003 ) groups for each protein monomer ( one per subdomain ) . A Ramachandran plot calculated by Procheck ( Laskowski et al . , 1993 ) indicated the following statistics: core: 2244 ( 82 . 7% ) ; allowed: 423 ( 15 . 6% ) ; generously allowed: 48 ( 1 . 8% ) ; disallowed: 0 ( 0% ) . Figures were prepared with the program Bobscript ( Esnouf , 1997 ) and rendered with the Raster3D ( Merritt and Bacon , 1997 ) package or prepared with the program PyMOL ( Schrodinger , 2010 ) . Crystals of PfMCMN without DNA were grown at 18°C in a sitting drop containing 200 nl of protein solution ( 10 mg/ml ) and 200 nl of well solution ( 0 . 2 M sodium malonate , pH 7 . 0 , 20% PEG 3350 ) . A plate crystal was cryoprotected by quickly passing it through well solution containing 15% ethylene glycol and flash frozen in liquid nitrogen . Data were collected at SER-CAT beamline 22-ID at the Advanced Photon Source at Argonne National Lab . Data were collected at 1 . 0 Å wavelength in 0 . 5° oscillations with two different segments of the same crystal . A total of 450 images were integrated and scaled with the HKL-2000 package ( Otwinowski and Minor , 1997 ) to 2 . 65 Å resolution . The unit cell parameters are very close to hexagonal , but initial data merging showed the presence of a crystallographic twofold axis and a clear absence of a crystallographic threefold axis , indicating a monoclinic lattice . Initial phases were determined by molecular replacement by the program Molrep ( Vagin and Teplyakov , 1997 ) by including a locked rotation and pseudo-translation . The program placed 6 copies of a monomer of MtMCMN ( Fletcher et al . , 2003 ) as a single hexamer in the asymmetric unit in space group P21 . The hexamers pack in layers with the hexameric axes mutually aligned parallel to the crystallographic 21 axis . Individual layers are highly sixfold symmetric , but a crystallographic 6-fold symmetry is precluded because the NCS 6-fold axes of successive layers are not mutually compatible . The model was refined at various stages with CNS ( Brunger et al . , 1998; Brunger , 2007 ) , phenix ( Afonine et al . , 2012 ) , and refmac5 ( Vagin et al . , 2004 ) . The final refinement was carried out with refmac5 using 3 TLS ( Winn et al . , 2003 ) groups for each protein monomer ( one per subdomain ) . A Ramachandran plot calculated by Procheck ( Laskowski et al . , 1993 ) indicated the following statistics: core: 1168 ( 85 . 8% ) ; allowed: 183 ( 13 . 4% ) ; generously allowed: 11 ( 0 . 8% ) ; disallowed: 0 ( 0% ) . Figures were prepared with the program Bobscript ( Esnouf , 1997 ) and rendered with the Raster3D ( Merritt and Bacon , 1997 ) package . DNA-binding reactions were set up in 20 μl with varying concentrations of PfMCMN ( 0–54 μM ) and 160 nM 5′-fluorescein-labeled T40 ssDNA ( Sigma-Aldrich , St . Louis , MO ) in 20 mM HEPES , pH 7 . 6 , 200 mM NaCl , 5 mM MgCl2 , and 5 mM βME . Reactions were incubated at 25°C in a BioRad DNA Engine thermocycler for 30 min . Loading buffer ( 2 . 5 mg/ml bromophenol blue and 40% sucrose; 5 μl ) was added , and 5 μl were loaded in a 4–20% 1X TBE gradient PAGE gel ( BioRad , Berkeley , CA ) and run at 100 V for 105 min . Gels were imaged by a Fuji LAS-4000 with an 8 s exposure and a SYBR-Green filter . The fluorescence intensities of bands for the free and bound species were quantified with MultiGauge ( GE Healthcare , Piscataway , NJ ) and fit to two simultaneous equations with Prism ( GraphPad Software , La Jolla , CA ) :I ( free ) /I0=Khalfh/ ( Khalfh+[MCMN]h ) ; I ( bound ) /I0=[MCMN]h/ ( Khalfh+[MCMN]h ) to determine the concentration of half-binding ( Khalf ) and a hill coefficient ( h ) . The dsDNA EMSAs were identical except that they included a 26-mer dsDNA substrate and a different concentration range of PfMCMN ( 0–20 μM ) . The dsDNA substrate was prepared by annealing two oligos ( 5′-[Fluorescein]-ATGGCAGATCTCAATTGGATATCGGC-3′ and 5′-GCCGATATCCAATTGAGATCTGCCAT-3′ , Sigma-Aldrich ) followed by purification on a gel filtration column ( GE Healthcare Superose 12 10/300 ) . Mcm2-7/Cdt1 , Mcm4D6D/Cdt1 , Mcm4D7D/Cdt1 , Mcm6D7D/Cdt1 and Mcm4D6D7D/Cdt1complexes were purified from 2 L cultures of ySKM01 , ySKM02 , ySKM03 , ySKM04 and ySKM05 , respectively . Cultures were grown to O . D . = 0 . 8 and arrested at G1 phase by addition of alpha factor ( 200 ng/ml ) for two hours followed by induction of Mcm2-7/Cdt1 expression by addition of galactose to 2% for 4 hr . Harvested cell pellets were re-suspended in 1/3 pellet volume of cell lysis buffer ( 100 mM HEPES-KOH ( pH 7 . 6 ) , 1 . 5 M potassium glutamate , 0 . 8 M sorbitol , 10 mM magnesium acetate , 1 mM dithiothreitol and 1X Complete Protease Inhibitor Cocktail [Roche Diagnostics , Indianapolis , IN] ) and frozen in liquid nitrogen . The frozen cell pellets were broken using a SPEX SamplePrep Freezer/Mill . After thawing , 15 ml of Buffer H ( 25 mM HEPES-KOH ( pH 7 . 6 ) , 1 mM EDTA , 1 mM EGTA , 5 mM magnesium acetate , 10% glycerol , 0 . 02% NP40 ) containing 0 . 5 M potassium glutamate , 3 mM ATP and 1X Complete Protease Inhibitor Cocktail was added to the broken cells . The cell lysate was centrifuged at 45 , 000×g rpm for 90 min ( Ti70 Rotor , Beckman ) and the supernatant was mixed with 0 . 6 ml anti-Flag Agarose ( Sigma-Aldrich ) equilibrated with Buffer H containing 0 . 5 M potassium glutamate . The mix was incubated for 4 hr at 4°C . The resin was washed and Mcm2-7/Cdt1 complexes were eluted with Buffer H containing 0 . 3 M potassium glutamate , 3 mM ATP and 0 . 15 mg/ml 3xFlag peptides . The eluted fractions were concentrated using Vivaspin 6 ( Mw . cutoff 100 KDa , Sartorius ) to 500 μl and applied to Superdex 200 HR 10/30 gel filtration column ( GE Healthcare ) . For each mutant complex , the corresponding wild-type proteins were epitope-tagged with V5 ( e . g . , in the strain expressing the Mcm4D7D/Cdt1 the wild-type MCM4 and MCM7 genes were tagged with V5 ) . This allowed the endogenous V5-tagged Mcm4 , 6 or 7 subunits to be depleted by incubating with anti-V5 agarose ( Sigma ) before applying the MSSB mutant complexes to the gel filtration column . 2 pmole ORC , 3 pmole Cdc6 and 6 pmole Mcm2-7/Cdt1 were sequentially added to the 40 μl reaction solution containing 1 pmole of bead-coupled 1 . 3 Kbps ARS1 DNA in helicase loading buffer ( 25 mM HEPES-KOH ( pH7 . 6 ) , 12 . 5 mM magnesium acetate , 0 . 1 mM zinc acetate , 300 mM potassium glutamate , 20 μM creatine phosphate , 0 . 02% NP40 , 10% glycerol , 3 mM ATP , 1 mM dithiothreitol and 2 μg creatine kinase ) . The reaction mix was incubated at 25°C at 1200 rpm for 30 min in a thermomixer ( Eppendorf ) . Beads were washed three times with Buffer H containing 0 . 3 M potassium glutamate and DNA bound proteins were eluted from the beads using DNase I . Eluted proteins were separated by SDS-PAGE and stained with a fluorescent protein stain ( Krypton , Thermo Scientific ) . For high salt wash experiments , Buffer H containing 0 . 5 M NaCl was used at the second wash step . In ATPγS experiments , 6 mM ATPγS was used instead of ATP . Helicase loading reactions were performed using 0 . 5 pmole ORC , 0 . 75 pmole Cdc6 and 2 pmole MCM/Cdt1 and 250 fmole bead-coupled 3 . 6 Kbps circular pUC19-ARS1 plasmid DNA ( Heller et al . , 2011 ) . Origin-loaded MCM complexes were phosphorylated with 450 μg purified DDK in DDK reaction buffer ( 50 mM HEPES-KOH ( pH7 . 6 ) , 3 . 5 mM magnesium acetate , 0 . 1 mM zinc acetate , 150 mM potassium glutamate , 0 . 02% NP40 , 10% glycerol , 1 mM spermine , 1 mM ATP and 1 mM dithiothreitol , 30 μl ) . Phosphorylated MCM complexes were then activated with 750 μg S phase extract in the replication reaction buffer ( 25 mM HEPES-KOH ( pH7 . 6 ) , 12 . 5 mM magnesium acetate , 0 . 1 mM zinc acetate , 300 mM potassium glutamate , 20 μM creatine phosphate , 0 . 02% NP40 , 10% Glycerol , 3 mM ATP , 40 μM dNTPs , 200 μM CTP/UTP/GTP , 1 mM dithiothreitol , 10 μCi [α-P32] dCTP and 2 μg creatine kinase , 40 μl ) for 1 hr at 25°C and 1200 RPM in a Thermomixer ( Eppendorf ) . After the reaction , DNA synthesis was monitored using alkaline agarose gel . DNA bound proteins were released from the beads by DNase I treatment and analyzed by immunoblot . S phase extracts were prepared from ySKS10 and ySKS11 as described previously ( Heller et al . , 2011 ) . MSSB mutations were introduced into TRP + ARS/CEN plasmids containing MCM4 , MCM6 , or MCM7 under the control of the MCM5 promoter . The resultant constructs were tested for MCM4 , MCM6 , or MCM7 function by a plasmid shuffle assay ( Schwacha and Bell , 2001 ) . To test the double mutant complementation , one MSSB mutant Mcm subunit ( either MCM4 or MCM6 ) was integrated into a plasmid shuffle strain for a second subunit . To integrate MSSB mutations into the chromosomal locus , we constructed plasmids containing the MCM4 or MCM6 promoter upstream of a NatMX4 ( for MCM4 ) or LEU2 ( for MCM6 ) marker cassette , with the Mcm5 promoter plus the MCM4 or MCM6 gene downstream of the marker and restriction enzyme sites flanking the entire integration unit ( pSKC04 and pSKC05 , respectively ) . Proper integration was confirmed by PCR followed by sequencing . To create strains carrying MSSB mutations in MCM4 and MCM6 or MCM6 and MCM7 , we began with strains carrying mcm4 or mcm7 deletion and the wild-type copy of MCM4 or MCM7 on URA+ ARS/CEN constructs , respectively . MCM6 MSSB mutation was integrated into these strains using the LEU+ integrating construct described above . For a strain carrying MSSB mutations in MCM4 and MCM7 , MCM4 MSSB mutations were incorporated in to a strain carrying mcm7 deletion and wild-type copy of MCM7 on URA+ ARS/CEN constructs , using NAT+ integrating construct . Then TRP+ ARS/CEN plasmids carrying MCM4 or MCM7 MSSB mutant allele were transformed above strains . Proliferation of double-mutant strains was analyzed using FOA counter-selection against the URA+ wild-type MCM4 or MCM7 plasmid . Yeast strains and plasmids of this study are listed in Tables 2 and 3 . 10 . 7554/eLife . 01993 . 025Table 2 . Yeast strains used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 025StrainsGenotypeSourceySKM01ade2-1 trp1-1 leu2-3 , 112 his3-11 , 15 ura3-1 can1-100 bar1::HisG lys2::HisG pep4Δ::unmarkedThis studyhis3::pSKM004 ( GAL1 , 10-MCM2 , Flag-MCM3 ) leu::pSKM007 ( GAL1 , 10-Cdt1-Strep , GAL4 ) lys::pSKM002 ( GAL1 , 10-MCM4 , MCM5 ) trp::pSKM003 ( GAL1 , 10-MCM6 , MCM7 ) ySKM02ade2-1 trp1-1 leu2-3 , 112 his3-11 , 15 ura3-1 can1-100 bar1::HisG lys2::HisG pep4Δ::KanMX6This studyMCM4-V5 ( NatMX4 ) MCM6-V5 ( CaURA3MX4 ) MCM7-V5 ( HphMX4 ) his3::pSKM004 ( GAL1 , 10-MCM2 , Flag-MCM3 ) leu::pSKM007 ( GAL1 , 10-Cdt1-Strep , GAL4 ) lys::pSKM008 ( GAL1 , 10-mcm4[R334A/K398A] , MCM5 ) trp::pSKM009 ( GAL1 , 10-mcm6[R296A/R360A] , MCM7 ) ySKM03ade2-1 trp1-1 leu2-3 , 112 his3-11 , 15 ura3-1 can1-100 bar1::HisG lys2::HisG pep4Δ::KanMX6This studyMCM4-V5 ( NatMX4 ) MCM6-V5 ( CaURA3MX4 ) MCM7-V5 ( HphMX4 ) his3::pSKM004 ( GAL1 , 10-MCM2 , Flag-MCM3 ) leu::pSKM007 ( GAL1 , 10-Cdt1-Strep , GAL4 ) lys::pSKM008 ( GAL1 , 10-mcm4[R334A/K398A] , MCM5 ) trp::pSKM010 ( GAL1 , 10-MCM6 , mcm7[R247A/K314A] ) ySKM04ade2-1 trp1-1 leu2-3 , 112 his3-11 , 15 ura3-1 can1-100 bar1::HisG lys2::HisG pep4Δ::KanMX6This studyMCM4-V5 ( NatMX4 ) MCM6-V5 ( CaURA3MX4 ) MCM7-V5 ( HphMX4 ) his3::pSKM004 ( GAL1 , 10-MCM2 , Flag-MCM3 ) leu::pSKM007 ( GAL1 , 10-Cdt1-Strep , GAL4 ) lys::pSKM002 ( GAL1 , 10-MCM4 , MCM5 ) trp::pSKM011 ( GAL1 , 10-mcm6[R296A/R360A] , mcm7[R247A/K314A] ) ySKM05ade2-1 trp1-1 leu2-3 , 112 his3-11 , 15 ura3-1 can1-100 bar1::HisG lys2::HisG pep4Δ::KanMX6This studyMCM4-V5 ( NatMX4 ) MCM6-V5 ( CaURA3MX4 ) MCM7-V5 ( HphMX4 ) his3::pSKM004 ( GAL1 , 10-MCM2 , Flag-MCM3 ) leu::pSKM007 ( GAL1 , 10-Cdt1-Strep , GAL4 ) lys::pSKM008 ( GAL1 , 10-mcm4[R334A/K398A] , MCM5 ) trp::pSKM011 ( GAL1 , 10-mcm6[R296A/R360A] , mcm7[R247A/K314A] ) ySKS10ade2-1 trp1-1 leu2-3 , 112 his3-11 , 15 ura3-1 can1-100 lys2::HisG pep4Δ::Hph cdc7-4This studypol1-5xFlag ( KanMX4 ) leu::pSKS001 ( GAL1 , 10-Cdc45-V5 , Sld3-S ) lys::pSKS002 ( GAL1 , 10-Dpb11-VSVG , Sld2-HSV ) ura::pSKS003 ( Gal1 , 10-Cdc28 , Clb5 ) his::pSKS004 ( Gal1 , 10-Sld7 ) ySKS11ade2-1 trp1-1 leu2-3 , 112 his3-11 , 15 ura3-1 can1-100 lys2::HisG pep4Δ::Hph cdc7-4This studypol2-5xFlag ( KanMX4 ) leu::pSKS001 ( GAL1 , 10-Cdc45-V5 , Sld3-S ) lys::pSKS002 ( GAL1 , 10-Dpb11-VSVG , Sld2-HSV ) ura::pSKS003 ( Gal1 , 10-Cdc28 , Clb5 ) his::pSKS004 ( Gal1 , 10-Sld7 ) ASY1059 . 1MatA , ade2-1 , ura3-11 , his3-11 , 15 , leu2-3 , 12 , can-100 , trp1-1 ( Schwacha and Bell , 2001 ) mcm4 Δ::hisG/pAS412 ( ARS/CEN URA+ PMCM5-MCM4-HA/HIS ) ASY2157MatA , ade2-1 , ura3-11 , his3-11 , 15 , leu2-3 , 12 , can-100 , trp1-1 , lys2::hisG , bar1::hisG , PEP4 Δ::KANMX4 , ( Schwacha and Bell , 2001 ) MCM6 Δ::HISMX6/pAS452 ( ARS/CEN URA+ PMCM5-MCM6-HA/HIS ) ASY1050 . 1MatA , ade2-1 , ura3-11 , his3-11 , 15 , leu2-3 , 12 , can-100 , trp1-1 ( Schwacha and Bell , 2001 ) mcm7Δ::hisG/pGEMCDC47 ( ARS/CEN URA+ MCM7 ) ySKC01MatA , ade2-1 , ura3-11 , his3-11 , 15 , leu2-3 , 12 , can-100 , trp1-1This studymcm4 Δ::hisG/pAS412 ( ARS/CEN URA+ PMCM5-MCM4-HA/HIS ) mcm6::LEU2-PMCM5-mcm6[R296A/R360A]ySKC02MatA , ade2-1 , ura3-11 , his3-11 , 15 , leu2-3 , 12 , can-100 , trp1-1This studymcm7Δ::hisG/pGEMCDC47 ( ARS/CEN URA+ MCM7 ) mcm6::LEU2-PMCM5-mcm6[R296A/R360A]ySKC03MatA , ade2-1 , ura3-11 , his3-11 , 15 , leu2-3 , 12 , can-100 , trp1-1This studymcm7Δ::hisG/pGEMCDC47 ( ARS/CEN URA+ MCM7 ) mcm4::NatMX4-PMCM5-mcm4[R334A/K398A]10 . 7554/eLife . 01993 . 026Table 3 . Yeast plasmids used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 01993 . 026PlasmidsDescriptionSourcepSKM002pRS307 ( GAL1 , 10-MCM4 , MCM5 ) This studypSKM003pRS404 ( GAL1 , 10-MCM6 , MCM7 ) This studypSKM004pRS403 ( GAL1 , 10-MCM2 , Flag-MCM3 ) This studypSKM007pRS305 ( GAL1 , 10-Cdt1-Strep , GAL4 ) This studypSKM008pRS307 ( GAL1 , 10-mcm4[R334A/K398A] , MCM5 ) This studypSKM009pRS404 ( GAL1 , 10-mcm6[R296A/R360A] , MCM7 ) This studypSKM010pRS404 ( GAL1 , 10-MCM6 , mcm7[R247A/K314A] ) This studypSKM011pRS404 ( GAL1 , 10-mcm6[R296A/R360A] , mcm7[R247A/K314A] ) This studypSKS001pRS305 ( GAL1 , 10-Cdc45-V5 , Sld3-S ) This studypSKS002pRS307 ( GAL1 , 10-Dpb11-VSVG , Sld2-HSV ) This studypSKS003pRS306 ( Gal1 , 10-Cdc28 , Clb5 ) This studypSKS004pRS403 ( Gal1 , 10-Sld7 ) This studypSKC001pRS414 ( PMCM5-mcm4[R334A/K398A] ) This studypSKC002pRS414 ( PMCM5- mcm6[R296A/R360A] ) This studypSKC003pRS414 ( PMCM5-mcm7[R247A/K314A] ) This studypSKC004pRS414 ( PMCM4-NatMX4-PMCM5- mcm4[R334A/K398A] ) This studypSKC005pRS414 ( PMCM6-LEU2-PMCM5- mcm6[R296A/R360A] ) This study	When DNA was first recognised to be a double helix , it was clear that this structure could easily explain how DNA could be replicated . Each strand was made of bases—represented by the letters ‘A’ , ‘C’ , ‘G’ and ‘T’—and the two strands were held together by bonds between pairs of bases , one from each strand . Moreover , ‘A’ always paired with ‘T’ , and ‘C’ always paired with ‘G’ . Therefore , if the two strands were separated , each could be used as a template to guide the synthesis of a new complementary strand and thus create two copies of the original double-stranded molecule . One of the first steps in this replication process involves a ring-shaped complex of six proteins , called an MCM helicase , separating the two strands . To prepare for DNA replication , two MCM helicase rings wrap around the double-stranded DNA . Then , after the helicase has been activated , the bonds between the DNA base pairs break , and the two rings separate with one ring encircling each DNA strand . However , the details of the interactions between the helicase and the DNA during these events are not fully understood . Now Froelich , Kang et al . have solved the three-dimensional structure of an MCM helicase ring—taken from a microbe originally found at deep ocean vents—on its own and also when bound to a short piece of single-stranded DNA . The helicase ring becomes more oval when the DNA binds to it . Moreover , rather than passing straight through the ring , the DNA wraps part of the way around the inside of the ring . Specific amino acids—the building blocks of proteins—on the inside of the ring interact with the single-stranded DNA , and these amino acids are also found in MCM proteins in many other organisms . Furthermore , swapping these amino acids for different amino acids significantly reduced the ability of the ring to bind to single-stranded DNA , but its ability to bind to double-stranded DNA was only slightly affected . Engineering similar changes into the ring complexes of yeast cells was lethal , and the mutant complexes were less able to be loaded onto the DNA , or to be activated and separate the two strands ready for replication . These insights into how helicases are loaded onto double-stranded DNA , and select one DNA strand to encircle , have improved our understanding of how DNA replication is initiated: a process that is vital for living things .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2014	A conserved MCM single-stranded DNA binding element is essential for replication initiation
Protein biomarkers have been identified across many age-related morbidities . However , characterising epigenetic influences could further inform disease predictions . Here , we leverage epigenome-wide data to study links between the DNA methylation ( DNAm ) signatures of the circulating proteome and incident diseases . Using data from four cohorts , we trained and tested epigenetic scores ( EpiScores ) for 953 plasma proteins , identifying 109 scores that explained between 1% and 58% of the variance in protein levels after adjusting for known protein quantitative trait loci ( pQTL ) genetic effects . By projecting these EpiScores into an independent sample ( Generation Scotland; n = 9537 ) and relating them to incident morbidities over a follow-up of 14 years , we uncovered 137 EpiScore-disease associations . These associations were largely independent of immune cell proportions , common lifestyle and health factors , and biological aging . Notably , we found that our diabetes-associated EpiScores highlighted previous top biomarker associations from proteome-wide assessments of diabetes . These EpiScores for protein levels can therefore be a valuable resource for disease prediction and risk stratification . Chronic morbidities place longstanding burdens on our health as we age . Stratifying an individual’s risk prior to symptom presentation is therefore critical ( NHS England , 2016 ) . Though complex morbidities are partially driven by genetic factors ( Fuchsberger et al . , 2016; Yao et al . , 2018 ) , epigenetic modifications have also been associated with disease ( Lord and Cruchaga , 2014 ) . DNA methylation ( DNAm ) encodes information on the epigenetic landscape of an individual and blood-based DNAm signatures have been found to predict all-cause mortality and disease onset , providing strong evidence to suggest that methylation is an important measure of disease risk ( Hillary et al . , 2020a; Lu et al . , 2019; Zhang et al . , 2017 ) . DNAm can regulate gene transcription ( Lea et al . , 2018 ) , and epigenetic differences can be reflected in the variability of the proteome ( Hillary et al . , 2019; Hillary et al . , 2020b; Zaghlool et al . , 2020 ) . Low-grade inflammation , which is thought to exacerbate many age-related morbidities , is particularly well captured through DNAm studies of plasma protein levels ( Zaghlool et al . , 2020 ) . As proteins are the primary effectors of disease , connecting the epigenome , proteome , and time to disease onset may help to resolve predictive biological signatures . Epigenetic predictors have utilised DNAm from the blood to estimate a person’s ‘biological age’ ( Lu et al . , 2019 ) , measure their exposure to lifestyle and environmental factors ( McCartney et al . , 2018c; McCartney et al . , 2018a; Peters et al . , 2021 ) , and predict circulating levels of inflammatory proteins ( Stevenson et al . , 2020; Stevenson et al . , 2021 ) . A leading epigenetic predictor of biological aging , the GrimAge epigenetic clock incorporates methylation scores for seven proteins along with smoking and chronological age , and is associated with numerous incident disease outcomes independently of smoking ( Hillary et al . , 2020a; Lu et al . , 2019 ) . This suggests there is predictive value gained in utilising DNAm scores relevant to protein levels as intermediaries for predictions . Methylation scores also point towards the pathways that may act on health beyond the protein biomarker that they are trained on . A portfolio of methylation scores for proteins across the circulating proteome could therefore aid in the prediction of disease and offer a different , but additive signal beyond methylation or protein data alone . Generation of an extensive range of epigenetic scores for protein levels has not been attempted to date . The capability of specific protein scores to predict a range of morbidities has also not been tested . However , DNAm scores for interleukin-6 ( IL-6 ) and C-reactive protein ( CRP ) have been found to associate with a range of phenotypes independently of measured protein levels , show more stable longitudinal trajectories than repeated protein measurements , and , in some cases , outperform blood-based proteomic associations with brain morphology ( Stevenson et al . , 2021; Conole et al . , 2021 ) . This is likely due to DNAm representing the accumulation of more sustained effects over a longer period of time than protein measurements , which have often been shown to be variable in their levels when measured at multiple time points ( Koenig et al . , 2003; Liu et al . , 2015; Moldoveanu et al . , 2000; Shah et al . , 2014 ) . DNAm scores for proteins could therefore be used to alert clinicians to individuals with high-risk biological signatures , many years prior to disease onset . Here , we report a comprehensive association study of blood-based DNAm with proteomics and disease ( Figure 1 ) . We trained epigenetic scores – referred to as EpiScores – for 953 plasma proteins ( with sample size ranging from 706 to 944 individuals ) and validated them using two independent cohorts with 778 and 162 participants . We regressed out known genetic pQTL effects from the protein levels prior to generating the EpiScores to preclude the signatures being driven by common SNP data that are invariant across the lifespan . We then examined whether the most robust predictors ( n = 109 EpiScores ) associated with the incidence of 12 major morbidities ( Table 1 ) , over a follow-up period of up to 14 years in the Generation Scotland cohort ( n = 9537 ) . We also tested for associations between EpiScore levels and COVID-19 disease outcomes . We regressed out the effects of age on protein levels prior to training and testing; age was also included as a covariate in disease prediction models . We controlled for common risk factors for disease and assessed the capacity of EpiScores to identify previously reported protein-disease associations . A video summarising this study and detailing how the 109 EpiScores can be calculated in cohorts with DNAm data is available at: https://youtu . be/xKDWg0Wzvrg . Our MethylDetectR shiny app ( Hillary and Marioni , 2020 ) has CpG weights for the 109 EpiScores integrated such that it automates the process of score generation for any DNAm dataset and is available at: https://www . ed . ac . uk/centre-genomic-medicine/research-groups/marioni-group/methyldetectr . A video on how to use the MethylDetectR shiny app to generate EpiScores is available at: https://youtu . be/65Y2Rv-4tPU . To generate epigenetic scores for a comprehensive set of plasma proteins , we ran elastic net penalised regression models using protein measurements from the SOMAscan ( aptamer-based ) and Olink ( antibody-based ) platforms . We used two cohorts: the German population-based study KORA ( n = 944 , mean age 59 years [SD 7 . 8] , with 793 SOMAscan proteins ) and the Scottish Lothian Birth Cohort 1936 ( LBC1936 ) study ( between 706 and 875 individuals in the training cohort , with a total of 160 Olink neurology and inflammatory panel proteins ) . The mean age of the LBC1936 participants at sampling was 70 ( SD 0 . 8 ) for inflammatory and 73 ( SD 0 . 7 ) for neurology proteins . Full demographic information is available for all cohorts in Supplementary file 1A . Prior to running the elastic net models , we rank-based inverse normalised protein levels and adjusted for age , sex , cohort-specific variables and , where present , cis and trans pQTL effects identified from previous analyses ( Hillary et al . , 2019; Hillary et al . , 2020b; Suhre et al . , 2017 ) ( Materials and methods ) . Of a possible 793 proteins in KORA , 84 EpiScores had Pearson r > 0 . 1 and p < 0 . 05 when tested in an independent subset of Generation Scotland ( The Stratifying Resilience and Depression Longitudinally [STRADL] study , n = 778 ) ( Supplementary file 1B ) . These EpiScores were selected for EpiScore-disease analyses . Of the 160 Olink proteins trained in LBC1936 , there were 21 with r > 0 . 1 and p < 0 . 05 in independent test sets ( STRADL , n = 778 , Lothian Birth Cohort 1921: LBC1921 , n = 162 ) ( Supplementary file 1C ) . Independent test set data were not available for four Olink proteins . However , they were included based on their performance ( r > 0 . 1 and p < 0 . 05 ) in a holdout sample of 150 individuals who were left out of the training set . We then retrained all selected predictors on the full training samples . A total of 109 EpiScores ( 84 SOMAscan-based and 25 Olink-based ) were brought forward ( r > 0 . 1 and p < 0 . 05 ) to EpiScore-disease analyses ( Figure 2 and Supplementary file 1D ) . There were five EpiScores for proteins common to both Olink and SOMAscan panels , which had variable correlation strength ( GZMA r = 0 . 71 , MMP . 1 r = 0 . 46 , CXCL10 r = 0 . 35 , NTRK3 r = 0 . 26 , and CXCL11 r = 0 . 09 ) . Predictor weights , positional information , and cis/trans status for CpG sites contributing to these EpiScores are available in Supplementary file 1E . The number of CpG features selected for EpiScores ranged from 1 ( lyzozyme ) to 395 ( aminoacylase-1 [ACY-1] ) , with a mean of 96 ( Supplementary file 1F ) . The most frequently selected CpG was the smoking-related site cg05575921 ( mapping to the AHRR gene ) , which was included in 25 EpiScores . Counts for each CpG site are summarised in Supplementary file 1G . This table includes the set of protein EpiScores that each CpG contributes to , along with phenotypic annotations ( traits ) from the MRC-IEU EWAS catalog ( MRC-IEU , 2021 ) for each CpG site having genome-wide significance ( p < 3 . 6 × 10–8 ) ( Saffari et al . , 2018 ) . GeneSet enrichment analysis of the original proteins used to train the 109 EpiScores highlighted pathways associated with immune response and cell remodelling , adhesion , and extracellular matrix function ( Supplementary file 1H ) . The Generation Scotland dataset contains extensive electronic health data from GP and hospital records as well as DNAm data for 9537 individuals . This makes it uniquely positioned to test whether EpiScore signals can predict disease onset . We ran nested mixed effects Cox proportional hazards models ( Figure 3 ) to determine whether the levels of each EpiScore at baseline associated with the incidence of 12 morbidities over a maximum of 14 years of follow-up . The correlation structures for the 109 EpiScore measures used for Cox modelling are presented in Figure 2—figure supplement 1 . The Cox proportional hazards assumption was assessed through the Schoenfeld residuals from the models . Two associations in the basic model adjusting for age and sex failed to satisfy the global assumption ( across all covariates ) and were excluded . There were 294 remaining EpiScore-disease associations with a false discovery rate ( FDR ) -adjusted p < 0 . 05 in the basic model . After further adjustment for common risk factor covariates ( smoking , social deprivation status , educational attainment , body mass index [BMI] , and alcohol consumption ) , 137 of the 294 EpiScore-disease associations from the basic model had p < 0 . 05 in the fully adjusted model ( Supplementary file 1I-J ) . Eleven of the 137 fully adjusted associations failed the Cox proportional hazards assumption for the EpiScore variable ( p < 0 . 05 for the association between the Schoenfeld residuals and time; Supplementary file 1K ) . When we restricted the time-to-event/censor period by each year of possible follow-up , there were minimal differences in the EpiScore-disease hazard ratios between follow-up periods that did not violate the assumption and those that did ( Supplementary file 1L ) . The 137 associations were therefore retained as the primary results . The 137 associations found in the fully adjusted model comprised 78 unique EpiScores that were related to the incidence of 11 of the 12 morbidities studied . Diabetes and chronic obstructive pulmonary disease ( COPD ) had the greatest number of associations , with 33 and 41 , respectively . Figure 4 presents the EpiScore-disease relationships for COPD and the remaining nine morbidities: stroke , lung cancer , ischaemic heart disease ( IHD ) , inflammatory bowel disease ( IBD ) , rheumatoid arthritis ( RA ) , depression , bowel cancer , pain ( back/neck ) , and Alzheimer’s dementia . There were 13 EpiScores that associated with the onset of three or more morbidities . Figure 5 presents relationships for these 13 EpiScores in the fully adjusted Cox model results . Of note is the EpiScore for Complement 5 ( C5 ) , which associated with five outcomes: stroke , diabetes , IHD , RA , and COPD . Of the 29 SOMAscan-derived EpiScore associations with incident diabetes , 23 replicated previously reported SOMAscan protein associations ( Elhadad et al . , 2020; Gudmundsdottir et al . , 2020; Ngo et al . , 2021 ) with incident or prevalent diabetes in one or more cohorts ( Figure 6 and Supplementary file 1M ) . Correlations of the 78 EpiScores that were associated with incident disease ( P < 0 . 05 in the fully-adjusted cox proportional hazards models ) with covariates suggested interlinked relationships with both estimated white blood cell proportions and GrimAge acceleration ( Figure 3—figure supplement 1 ) . These covariates were therefore added incrementally to the fully-adjusted Cox models ( Figure 3 ) . There were 103 associations that remained statistically significant ( FDR p < 0 . 05 in the basic model and p < 0 . 05 in the fully adjusted model ) after adjustment for immune cell proportions , of which 81 remained significant when GrimAge acceleration scores were added to this model ( Supplementary file 1J ) . In a further sensitivity analysis , relationships between both estimated white blood cell ( WBC ) proportions and GrimAge acceleration scores with incident diseases were assessed in the Cox model structure independently of EpiScores . Of the 60 possible relationships between WBC measures and the morbidities assessed , four were statistically significant ( FDR-adjusted p < 0 . 05 ) in the basic model and remained significant with p < 0 . 05 in the fully adjusted model ( Supplementary file 1N ) . A higher proportion of natural killer cells was linked to decreased risk of incident COPD , RA , diabetes , and pain ( back/neck ) . The GrimAge acceleration composite score was associated with COPD , IHD , diabetes , and pain ( back/neck ) in the fully adjusted models ( p < 0 . 05 ) ( Supplementary file 1O ) . The magnitude of the GrimAge effect sizes was comparable to the EpiScore findings . Two previous studies including pilot proteomic measurements from the Generation Scotland cohort ( N = 199 controls ) as part of wider analyses found that several proteins corresponding to our EpiScores were associated with COVID-19 outcomes ( Demichev et al . , 2021; Messner et al . , 2020 ) . These included proteins such as CRP , C9 , SELL , and SHBG , all of which were associated with one or more incident diseases in this study . Two subsets ( N = 268 and N = 173 ) of the Generation Scotland sample who contracted COVID-19 were therefore used to test the hypothesis that EpiScores would associate with COVID-19 outcomes ( acquired >9 years after the blood draw for DNAm analyses ) . No significant associations were identified that delineated differences between the development of long-covid ( duration >4 weeks ) or hospitalisation from COVID-19 ( associations that had p < 0 . 05 did not withstand Bonferroni adjustment for multiple testing ) ( Supplementary file 1P ) . Here , we report a comprehensive DNAm scoring study of 953 circulating proteins . We define 109 robust EpiScores for plasma protein levels that are independent of known pQTL effects . By projecting these EpiScores into a large cohort with extant data linkage , we show that 78 EpiScores associate with the incidence of 11 leading causes of morbidity ( 137 EpiScore-disease associations in total ) , but do not associate with COVID-19 outcomes . Finally , we show that EpiScore-diabetes associations highlight previously measured protein-diabetes relationships . The bulk of EpiScore-disease associations are independent of common lifestyle and health factors , differences in immune cell composition and GrimAge acceleration . EpiScores therefore provide methylation-proteomic signatures for disease prediction and risk stratification . The consistency between our EpiScore-diabetes associations and previously identified protein-diabetes relationships ( Elhadad et al . , 2020; Gudmundsdottir et al . , 2020; Ngo et al . , 2021 ) suggests that epigenetic scores identify disease-relevant biological signals . In addition to the comprehensive lookup of SOMAscan proteins with diabetes , several of the markers we identified for COPD and IHD also reflect previous associations with measured proteins ( Ganz et al . , 2016; Serban et al . , 2021 ) . The three studies used for the diabetes comparison represent the largest candidate protein characterisations of type 2 diabetes to date and the top markers identified included aminoacylase-1 ( ACY-1 ) , sex hormone-binding globulin ( SHBG ) , growth hormone receptor ( GHR ) , and insulin-like growth factor-binding protein 2 ( IFGBP-2 ) ( Elhadad et al . , 2020; Gudmundsdottir et al . , 2020; Ngo et al . , 2021 ) . Our EpiScores for these top markers were also associated with diabetes , in addition to EpiScores for several other protein markers reported in these studies . A growing body of evidence suggests that type 2 diabetes is mediated by genetic and epigenetic regulators ( Kwak and Park , 2016 ) and proteins such as ACY-1 and GHR are thought to influence a range of diabetes-associated metabolic mechanisms ( Kim and Park , 2017; Pérez-Pérez et al . , 2012 ) . Proteins that we identify through EpiScore associations , such as NTR domain-containing protein 2 ( WFIKKN2 ) , have also been causally implicated in type 2 diabetes onset through Mendelian randomisation analysis ( Ngo et al . , 2021 ) . In the case of diabetes , EpiScores may therefore be used as disease-relevant risk biomarkers , many years prior to onset . Validation should be tested when sufficient data become available for the remaining morbidities . With modest test set performances ( e . g . , SHBG r = 0 . 18 and ACY-1 r = 0 . 25 ) , it is perhaps surprising that such strong synergy is observed between EpiScores for proteins that associated with diabetes and the trends seen with measured proteins . Nonetheless , DNAm scores for CRP and IL-6 have previously been shown to perform modestly in test sets ( r ~ 0 . 2 , equivalent to ~4% explained variance in protein level ) , but augment and often outperform the measured protein related to a range of phenotypes ( Stevenson et al . , 2020; Stevenson et al . , 2021 ) . Compared to scores utilising DNAm for the prediction of singular diseases , our EpiScores enable the granular study of individual protein predictor signatures with clinical outcomes . For example , levels of the acid sphingomyelinase ( ASM ) EpiScore predicted onset of Alzheimer’s dementia , several years prior to diagnosis . ASM ( encoded by SMPD1 ) has been discussed as a therapeutic candidate for Alzheimer’s disease ( Cataldo et al . , 2004; Lee et al . , 2014; Park et al . , 2020 ) and has been shown to disrupt autophagic protein degradation and associate with accumulation of amyloid-beta in murine models of Alzheimer’s pathology ( Lee et al . , 2014; Park et al . , 2020 ) . Our large-scale assessment of EpiScores provides a platform for future studies , as composite predictors for traits may be created using our EpiScore database . These should be tested in incident disease predictions when sufficient case data are available . Our results indicated that the set of 109 EpiScores are likely to be heavily enriched for inflammatory , complement system and innate immune system pathways , in addition to extracellular matrix , cell remodelling , and cell adhesion pathways . This reinforces previous work linking chronic inflammation and the epigenome ( Zaghlool et al . , 2020 ) . It also suggests that EpiScores could be useful in the prediction of morbidities that are characterised by differential inflammatory states . An example of this is the EpiScore for Complement Component 5 ( C5 ) , which was associated with the onset of five morbidities , the highest number for any EpiScore ( Figure 5 ) . The EpiScore for C5 is likely to reflect the biological pathways occurring in individuals with heightened complement cascade activity and could be utilised to alert clinicians to individuals at high risk of multimorbidity . Elevated levels of C5 peptides have been associated with severe inflammatory , autoimmune , and neurodegenerative states ( Ma et al . , 2019; Mantovani et al . , 2014; Morgan and Harris , 2015 ) and a range of C5-targetting therapeutic approaches are in development ( Alawieh et al . , 2018; Brandolini et al . , 2019; Hawksworth et al . , 2017; Hernandez et al . , 2017; Morgan and Harris , 2015; Ort et al . , 2020 ) . Though EpiScores such as C5 – which occupy central hubs in the disease prediction framework – may provide evidence of early methylation signatures common to the onset of multiple diseases , we did not observe associations between EpiScores and COVID-19 hospitalisation or long-COVID status . This is perhaps surprising , given that many of the morbidities that our EpiScores predicted are also known risk factors for increased risk of death due to COVID-19 ( Williamson et al . , 2020 ) . Many of the proteins corresponding to EpiScores in our study were also associated with COVID-19 severity and progression in two previous studies that included a pilot sample ( N = 199 ) from the Generation Scotland cohort at baseline as control data ( Demichev et al . , 2021; Messner et al . , 2020 ) . COVID-19 likely has multiple intersecting risk factors that impact severity and recovery , and the lack of associations we observe is likely to be in part due to the limited number of COVID-19 cases available in Generation Scotland . Additionally , there is a large lag time between baseline biological measurement and COVID-19 in our analyses , whereas the two studies that found protein marker associations integrated protein measurements longitudinally and from samples extracted during COVID-19 progression . With increased power available through continued data linkage , EpiScore relationships with COVID-19 outcomes may be observed in future work . This study has several limitations . First , we demonstrate that EpiScores carry disease-relevant signals that may be clinically meaningful to delineate early disease risk when comparing relative differences within a cohort . However , projecting a new individual onto a reference set is complicated due to absolute differences in methylation quantification resulting from batch and processing effects . Second , future studies should assess paired protein and EpiScore contributions to traits , as inference from EpiScores alone , while useful for disease risk stratification , is not sufficient to determine mechanisms . This may also highlight EpiScores that outperform the measured protein equivalent in disease . Third , the epitope nature of the protein measurement in the SOMAscan panel may incur probe cross-reactivity and non-specific binding; there may also be differences in how certain proteins are measured across panels ( Pietzner et al . , 2020; Sun et al . , 2018 ) . Comparisons of multiple protein measurement technologies on the same samples would help to explore this in more detail . Fourth , there may be pQTLs with small effect sizes that were not regressed from the proteins prior to generating the EpiScores . Fifth , while training and testing was performed across multiple cohorts , it is likely that further development of EpiScores in larger proteomic samples with diverse ancestries will improve power to generate robust scores . Upper bounds for DNAm prediction of complex traits , such as proteins , can be estimated by variance components analyses ( Hillary et al . , 2020b; Trejo Banos et al . , 2020; Zhang et al . , 2019 ) . Finally , associations present between EpiScore measures and disease incidence may have been influenced by external factors such as prescription medications for comorbid conditions and comorbid disease prevalence . We have shown that EpiScores for circulating protein levels predict the incidence of multiple diseases , up to 14 years prior to diagnosis . Our findings suggest that DNAm phenotyping approaches and data linkage to electronic health records in large , population-based studies have the potential to ( 1 ) capture inter-individual variability in protein levels; ( 2 ) predict incident disease risk many years prior to morbidity onset; and ( 3 ) highlight disease-relevant biological signals for further exploration . The EpiScore weights are publicly available , enabling any cohort with Illumina DNAm data to generate them and to relate them to various outcomes . Given the increasingly widespread assessment of DNAm in cohort studies ( McCartney et al . , 2020; Min et al . , 2021 ) , EpiScores offer an affordable and consistent ( i . e . , array-based ) way to utilise these signatures . This information is likely to be important in risk stratification and prevention of age-related morbidities . The KORA F4 study includes 3080 participants who reside in Southern Germany . Individuals were between 32 and 81 years of age when recruited to the study from 2006 and 2008 . In the current study , there were 944 individuals with methylation , proteomics , and genetic data available . The Infinium HumanMethylation450 BeadChip platform was used to generate DNAm data for these individuals . The Affymetrix Axiom array was used to generate genotyping data and the SOMAscan platform was used to generate proteomic measurements in the sample . Methylation data were generated for 1814 individuals ( Petersen et al . , 2014 ) ; 944 also had protein and genotype measurements available . During preprocessing , 65 SNP probes were excluded and background correction was performed in minfi ( Aryee et al . , 2014 ) . Samples with a detection rate of less than 95% were excluded . Next , the minfi R package was used to perform normalisation on the intensity methylation measures ( Aryee et al . , 2014 ) , with a method consistent with the Lumi:QN + BMIQ pipeline . After excluding non-cg sites and CpGs on sex chromosomes or with fewer than 100 measures available , 470 , 837 CpGs were available for analyses . The SOMAscan platform ( Version 3 . 2 ) ( Gold et al . , 2010 ) was used to quantify protein levels in undepleted plasma for 1129 SOMAmer probes ( Suhre et al . , 2017 ) . Of the 1000 samples provided for analysis , two samples were excluded due to errors in bio-bank sampling and one based on quality control ( QC ) measures . Of the 997 samples available , there were 944 individuals with methylation and genotypic data . Of the 1129 probes available , five failed the QC , leaving a total of 1124 probes for the subsequent analysis . Protein measurements were transformed by rank-based inverse normalisation and regressed onto age , sex , known pQTLs , and 20 genetic principal components of ancestry derived from the Affymetrix Axiom Array to control for population structure . pQTLs for each protein were taken from a previous GWAS in the sample ( Suhre et al . , 2017 ) . The Lothian Birth Cohorts of 1921 ( LBC1921; N = 550 ) and 1936 ( LBC1936; N = 1091 ) are longitudinal studies of aging in individuals who reside in Scotland ( Deary et al . , 2012; Taylor et al . , 2018 ) . Participants completed an intelligence test at age 11 years and were recruited for these cohorts at mean ages of 79 ( LBC1921 ) and 70 ( LBC1936 ) . They have been followed up triennially for a series of cognitive , clinical , physical , and social data , along with blood donations that have been used for genetic , epigenetic , and proteomic measurement . DNAm , proteomic ( Olink platform ) , and genetic data for individuals from Waves 1 ( n=875 at mean age 70 years and sd 0 . 8 ) and 2 ( n=706 at mean age 73 years and sd 0 . 7 ) of the LBC1936 and Wave 3 of the LBC1921 ( n=162 at mean age 87 years and sd 0 . 4 ) were available . DNA from whole blood was assessed using the Illumina 450 K methylation array . Details of QC have been described elsewhere ( Shah et al . , 2014; Zhang et al . , 2018 ) . Raw intensity data were background-corrected and normalised using internal controls . Manual inspection resulted in the removal of low-quality samples that presented issues related to bisulphite conversion , staining signal , inadequate hybridisation , or nucleotide extension . Probes with low detection rate <95% at p < 0 . 01 and samples with low call rates ( <450 , 000 probes detected at p < 0 . 01 ) were removed . Samples were also removed if they had a poor match between genotype and SNP control probes , or incorrect DNAm-predicted sex . Plasma samples were analysed using either the Olink neurology 92-plex or the Olink inflammation 92-plex proximity extension assays ( Olink Bioscience , Uppsala Sweden ) . One inflammatory panel protein ( BDNF ) failed QC and was removed . A further 21 proteins were removed , as over 40% of samples fell below the lowest limit of detection . Two neurology proteins , MAPT and HAGH , were excluded due to >40% of observations being below the lower limit of detection . This resulted in 90 neurology ( LBC1936 Wave 2 ) and 70 inflammatory ( LBC1936 Wave 1 ) proteins in LBC1936 and 92 neurology proteins available in LBC1921 . Protein levels were rank-based inverse normalised and regressed onto age , sex , four genetic components of ancestry derived from multidimensional scaling of the Illumina 610-Quadv1 genotype array and Olink array plate . In LBC1936 , pQTLs were adjusted for , through reference to GWAS in the samples ( Hillary et al . , 2019; Hillary et al . , 2020b ) . Generation Scotland: the Scottish Family Health Study ( GS ) is a large , family-structured , population-based cohort study of >24 , 000 individuals from Scotland ( mean age 48 years ) ( Smith et al . , 2013 ) . Recruitment took place between 2006 and 2011 with a clinical visit where detailed health , cognitive , and lifestyle information was collected along with biological samples ( blood , urine , saliva ) . In GS , there were 9537 individuals with DNAm and phenotypic information available . The STRADL cohort is a subset of 1188 individuals from the GS cohort who undertook additional assessments approximately 5 years after the study baseline ( Navrady et al . , 2018 ) . In the GS cohort , blood-based DNAm was generated in two sets using the Illumina EPIC array . Set 1 comprised 5190 related individuals whereas Set 2 comprised 4583 individuals , unrelated to each other and to those in Set 1 . During QC , probes were removed based on visual outlier inspection , bead count <3 in over 5% of samples , and samples with detection p-value below adequate thresholds ( McCartney et al . , 2018b; Seeboth et al . , 2020 ) . Samples were removed based on sex mismatches , low detection p-values for CpGs and saliva samples and genetic outliers ( Amador et al . , 2015 ) . The quality-controlled dataset comprised 9537 individuals ( nSet1 = 5087 , nSet2 = 4450 ) . The same steps were also applied to process DNAm in STRADL . Measurements for 4235 proteins in 1065 individuals from the STRADL cohort were recorded using the SOMAscan technology; 793 epitopes matched between the KORA and STRADL cohorts and were included for training in KORA and testing in STRADL . There were 778 individuals with proteomics data and DNAm data in STRADL . Protein measurements were transformed by rank-based inverse normalisation and regressed onto age , sex , and 20 genetic principal components ( derived from multidimensional scaling of genotype data from the Illumina 610-Quadv1 array ) . Over 98% of GS participants consented to allow access to electronic health records via data linkage to GP records ( Read 2 codes ) and hospital records ( ICD codes ) . Data are available prospectively from the time of blood draw , yielding up to 14 years of linkage . We considered incident data for 12 morbidities . Ten of the diseases are listed by the World Health Organization ( WHO ) as leading causes of either morbidity or mortality ( Hay et al . , 2017; World Health Organization , 2018 ) . Inflammatory bowel disease ( IBD ) ( Kassam et al . , 2014 ) and RA ( James et al . , 2018 ) are also included as traits as they have been reported as leading causes of disability and morbidity and the global burdens of these diseases are rising ( Alatab et al . , 2020; Safiri et al . , 2019 ) . Prevalent cases ( ascertained via retrospective ICD and Read 2 codes or self-report from a baseline questionnaire ) were excluded . For IBD prevalent cases were excluded based on data linkage alone . Included and excluded terms can be found in Supplementary files 1Q-1B1 . Alzheimer’s dementia was limited to cases/controls with age of event/censoring ≥65 years . Breast cancer was restricted to females only . Recurrent , major and moderate episodes of depression were included in depression . Diabetes was comprised of predominantly type 2 diabetes codes and additional general diabetes codes such as diabetic retinopathy and diabetes mellitus with renal manifestation that often occur in individuals with type 2 diabetes . Type 1 and juvenile diabetes cases were excluded . Penalised regression models were generated for 160 proteins in LBC1936 and 793 proteins in KORA using Glmnet ( Version 4 . 0-2 ) ( Friedman et al . , 2010 ) in R ( Version 4 . 0 ) ( R Development Core Team , 2020 ) . Protein levels were the outcome and there were 428 , 489 CpG features per model in the LBC1936 training and 397 , 630 in the KORA training . An elastic net penalty was specified ( alpha = 0 . 5 ) and cross validation was applied . DNAm and protein measurements were scaled to have a mean of zero and variance of one . In the KORA analyses , 10-fold cross validation was applied and EpiScores were tested in STRADL ( n = 778 ) . Of 480 EpiScores that generated ≥1 CpG features , 84 had Pearson r > 0 . 1 and p < 0 . 05 in STRADL . As test set comparisons were not available for every protein in the LBC1936 analyses , a holdout sample was defined , with two folds set aside as test data and 10-fold cross validation carried out on the remaining data ( ntrain = 576 , ntest = 130 for neurology and ntrain = 725 , ntest = 150 for inflammatory proteins ) . We retained 36 EpiScores with Pearson r > 0 . 1 and p < 0 . 05 . New predictors for these 36 proteins were then generated using 12-fold cross validation and tested externally in STRADL ( n = 778 ) and LBC1921 ( n = 162 , for the neurology panel ) . Twenty-one EpiScores had r > 0 . 1 and p < 0 . 05 in at least one of the external test sets . Four EpiScores did not have external comparisons and were included based on holdout performance . Functional annotations for each of the proteins used to train the finalised set of 109 EpiScores were sourced from the STRING database ( Jensen et al . , 2009 ) . GeneSet enrichment analysis against protein-coding genes was performed using the FUMA database , to quantify which canonical pathways were most commonly implicated across the 109 genes corresponding to the proteins used to train the 109 EpiScores ( Watanabe et al . , 2017 ) . The background gene-set was specified as protein coding genes and a threshold of FDR p < 0 . 05 was applied for enrichment status , with the minimum overlapping genes with gene-sets set to ≥2 . The 109 selected EpiScores were then applied to Generation Scotland ( n = 9537 ) . DNAm at each CpG site was scaled to have a mean of zero and variance of one , with scaling performed separately for GS sets . Mixed effects Cox proportional hazards regression models adjusting for age , sex , and methylation set were used to assess the relationship between 109 EpiScores and 12 morbidities in Generation Scotland . Models were run using coxme ( Therneau , 2020b ) ( Version 2 . 2-16 ) with a kinship matrix accounting for relatedness in Set 1 . Cases included those diagnosed after baseline who had died , in addition to those who received a diagnosis and remained alive . Controls were censored if they were disease free at time of death , or at the end of the follow-up period . EpiScore levels were rank-base inverse normalised . Fully adjusted models included the following additional covariates measured at baseline: alcohol consumption ( units consumed in the previous week ) ; deprivation assessed by the Scottish Index of Multiple Deprivation ( GovScot , 2016 ) ; BMI ( kg/m2 ) ; educational attainment ( an 11-category ordinal variable ) ; and a DNAm-based score for smoking status ( Bollepalli et al . , 2019 ) . A false discovery rate multiple testing correction p < 0 . 05 was applied to the 1306 EpiScore-disease associations ( 109 EpiScores by 12 incident disease traits , with two associations excluded for failing the global proportional hazards assumption ) . Proportional hazards assumptions were checked through Schoenfeld residuals ( global test and a test for the protein-EpiScore variable ) using the coxph and cox . zph functions from the survival package ( Therneau , 2020a ) ( Version 3 . 2-7 ) . For each association failing to meet the assumption ( Schoenfeld residuals p < 0 . 05 ) , a sensitivity analysis was run across yearly follow-up intervals . Fully adjusted Cox proportional hazards models were run with Houseman-estimated white blood cell proportions as covariates ( Houseman et al . , 2012 ) . A further sensitivity analysis added GrimAge acceleration ( Lu et al . , 2019 ) as an additional covariate . Basic and fully adjusted Cox models were also run with estimated monocyte , B-cell , CD4T , CD8T , and natural killer cell proportions as predictors , in addition to models with GrimAge acceleration as the predictor of incident disease . Correlation structures for EpiScores , DNAm-estimated white blood cell proportions , and phenotypic information were assessed using Pearson correlations and pheatmap ( Kolde , 2019 ) ( Version 1 . 0 . 12 ) and ggcorrplot packages ( Version 0 . 1 . 3 ) ( Kassambara , 2019 ) . The psych package ( Version 2 . 0 . 9 ) ( Revelle , 2020 ) was used to perform principal components analysis on EpiScores . Figures 1 and 2 were created with BioRender . com . Associations for EpiScores that were related to a minimum of three morbidities were subset from the fully adjusted Cox proportional hazards results and were visualised using the ggraph package ( Version 2 . 0 . 5 ) ( Pedersen , 2021 ) . This network representation was used ( Figure 5 ) to highlight protein EpiScores that were connected with multiple morbidities . Comparisons were conducted between EpiScore-diabetes associations and type 2 diabetes associations with measured proteins using three previous large-scale proteomic studies ( Elhadad et al . , 2020; Gudmundsdottir et al . , 2020; Ngo et al . , 2021 ) In these studies , six cohorts were included ( Study 1: KORA n = 993 , HUNT n = 940 [Elhadad et al . , 2020] , Study 2: AGES-Reykjavik n = 5438 and QMDiab n = 356 [Gudmundsdottir et al . , 2020] , Study 3: Framingham Heart Study n = 1618 and the Malmo Diet and Cancer Study n = 1221 ) . Study 1 included the KORA dataset , which we use in this study to generate SOMAscan EpiScores . We characterised which SOMAscan-based EpiScore-diabetes associations from our fully adjusted results reflected those observed with measured protein levels . We included basic ( nominal p < 0 . 05 ) and fully adjusted results ( with either FDR or Bonferroni-corrected p < 0 . 05 ) , wherever available , across the lookup cohorts ( Supplementary file 1M ) . Associations between each of the 109 selected protein EpiScores and subsequent long-COVID or COVID-19 hospitalisation were tested in the Generation Scotland population . A binary variable was used for long-COVID based on self-reported COVID-19 duration from the CovidLife study survey 3 questionnaire ( N = 2399 participating individuals ) ( Fawns-Ritchie et al . , 2021 ) . Participants were asked about the total overall time they experienced symptoms in their first/only episode of illness , as well as their COVID-19 illness duration . The dataset is correct as of February 2021 when the survey 3 was administered . Of the 9537 individuals with DNAm that were included in incident disease analyses , 173 indicated that they had COVID-19 and 56 of these individuals reported having long-COVID ( >4 weeks duration of symptoms after infection ) . The mean duration from DNAm measurement to long-COVID for these 56 individuals was 11 . 2 years ( sd 1 . 2 ) . Hospitalisation information , derived from the Scottish Morbidity Records ( SMR01 ) , was used to obtain COVID-19 hospital admissions using ICD-10 codes U07 . 1 ( lab-confirmed COVID-19 diagnosis ) , and U07 . 2 ( clinically diagnosed COVID-19 ) . This data linkage identified 268 of the 9537 individuals that had COVID-19 diagnoses and 29 had been recorded as being hospitalised due to COVID-19 . The mean duration from DNAm measurement to hospitalisation for these 29 individuals was 11 . 9 years ( sd 1 . 4 ) . Logistic regression models with either hospitalisation or long-COVID status as binary outcomes were used , with the 109 scaled protein EpiScores as the independent variables . Sex and age at COVID testing were included as covariates . The latter was defined as the age at positive COVID-19 test or 1 January 2021 if COVID-19 test data were not available .	Although our genetic code does not change throughout our lives , our genes can be turned on and off as a result of epigenetics . Epigenetics can track how the environment and even certain behaviors add or remove small chemical markers to the DNA that makes up the genome . The type and location of these markers may affect whether genes are active or silent , this is , whether the protein coded for by that gene is being produced or not . One common epigenetic marker is known as DNA methylation . DNA methylation has been linked to the levels of a range of proteins in our cells and the risk people have of developing chronic diseases . Blood samples can be used to determine the epigenetic markers a person has on their genome and to study the abundance of many proteins . Gadd , Hillary , McCartney , Zaghlool et al . studied the relationships between DNA methylation and the abundance of 953 different proteins in blood samples from individuals in the German KORA cohort and the Scottish Lothian Birth Cohort 1936 . They then used machine learning to analyze the relationship between epigenetic markers found in people’s blood and the abundance of proteins , obtaining epigenetic scores or ‘EpiScores’ for each protein . They found 109 proteins for which DNA methylation patterns explained between at least 1% and up to 58% of the variation in protein levels . Integrating the ‘EpiScores’ with 14 years of medical records for more than 9000 individuals from the Generation Scotland study revealed 137 connections between EpiScores for proteins and a future diagnosis of common adverse health outcomes . These included diabetes , stroke , depression , Alzheimer’s dementia , various cancers , and inflammatory conditions such as rheumatoid arthritis and inflammatory bowel disease . Age-related chronic diseases are a growing issue worldwide and place pressure on healthcare systems . They also severely reduce quality of life for individuals over many years . This work shows how epigenetic scores based on protein levels in the blood could predict a person’s risk of several of these diseases . In the case of type 2 diabetes , the EpiScore results replicated previous research linking protein levels in the blood to future diagnosis of diabetes . Protein EpiScores could therefore allow researchers to identify people with the highest risk of disease , making it possible to intervene early and prevent these people from developing chronic conditions as they age .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "epidemiology", "and", "global", "health", "genetics", "and", "genomics" ]	2022	Epigenetic scores for the circulating proteome as tools for disease prediction
Neurons communicate through neurotransmitter release at specialized synaptic regions known as active zones ( AZs ) . Using biosensors to visualize single synaptic vesicle fusion events at Drosophila neuromuscular junctions , we analyzed the developmental and molecular determinants of release probability ( Pr ) for a defined connection with ~300 AZs . Pr was heterogeneous but represented a stable feature of each AZ . Pr remained stable during high frequency stimulation and retained heterogeneity in mutants lacking the Ca2+ sensor Synaptotagmin 1 . Pr correlated with both presynaptic Ca2+ channel abundance and Ca2+ influx at individual release sites . Pr heterogeneity also correlated with glutamate receptor abundance , with high Pr connections developing receptor subtype segregation . Intravital imaging throughout development revealed that AZs acquire high Pr during a multi-day maturation period , with Pr heterogeneity largely reflecting AZ age . The rate of synapse maturation was activity-dependent , as both increases and decreases in neuronal activity modulated glutamate receptor field size and segregation . Synaptic vesicle fusion occurs at specialized regions of the presynaptic membrane known as active zones ( AZs ) . Several evolutionarily conserved structural proteins are enriched in this subdomain of the presynaptic terminal , including RIM , RIM binding protein , Syd-1 , Liprin-α , ELKS/CAST/Bruchpilot , Munc13 , and Bassoon/Piccolo/Fife ( Schoch and Gundelfinger , 2006; Südhof , 2012; Van Vactor and Sigrist , 2017; Zhai and Bellen , 2004 ) . These large macromolecular complexes facilitate clustering of synaptic vesicles and voltage-gated Ca2+ channels ( VGCCs ) , allowing action potential-triggered Ca2+ influx to act locally on synaptic vesicles that are docked and primed for release ( Acuna et al . , 2016; Zito et al . , 1999; Bucurenciu et al . , 2008; Eggermann et al . , 2011; Fouquet et al . , 2009; Kawasaki et al . , 2004 ) . Synaptic vesicle fusion occurs through a highly probabilistic process , often with only a small percent of action potentials triggering release from individual AZs ( Körber and Kuner , 2016 ) . Although AZs largely share the same complement of proteins , release probability ( Pr ) for synaptic vesicle fusion is highly variable across different neurons and between AZs formed by the same neuron ( Atwood and Karunanithi , 2002; Branco and Staras , 2009; Melom et al . , 2013; Peled and Isacoff , 2011 ) . Studies have demonstrated that Ca2+ channel abundance and Ca2+ influx are key determinants of Pr ( Borst and Sakmann , 1996; Chen et al . , 2015; Meinrenken et al . , 2002Zito et al . , 1999; Nakamura et al . , 2015; Sheng et al . , 2012; Wang et al . , 2008 ) . In addition , some AZ-specific proteins are non-uniformly distributed , and the molecular composition of AZs can undergo rapid changes ( Glebov et al . , 2017; Graf et al . , 2009; Liu et al . , 2016; Reddy-Alla et al . , 2017; Sugie et al . , 2015; Tang et al . , 2016; Weyhersmüller et al . , 2011; Wojtowicz et al . , 1994 ) . The Drosophila neuromuscular junction ( NMJ ) has emerged as a useful system to study release heterogeneity . At this connection , motor neurons form glutamatergic synapses onto bodywall muscles in a stereotypical fashion , with the axon expanding to form ~10–60 synaptic boutons that each contain many individual AZs ( Harris and Littleton , 2015 ) . Drosophila AZs contain a similar assortment of proteins to those identified at mammalian AZs ( Böhme et al . , 2016; Bruckner et al . , 2017 , 2012; Ehmann et al . , 2014; Feeney et al . , 1998; Fouquet et al . , 2009; Graf et al . , 2012; Jan and Jan , 1976; Kaufmann et al . , 2002; Kittel et al . , 2006; Liu et al . , 2011; Owald et al . , 2010; Wagh et al . , 2006 ) . Each AZ is specifically associated with a postsynaptic glutamate receptor field . Glutamate receptors at the Drosophila NMJ are excitatory inotropic non-NMDA receptors that exist as tetramers , with three obligatory subunits encoded by GluRIII , GluRIID and GluRIIE , and a variable fourth subunit encoded by either GluRIIA ( A-type ) or GluRIIB ( B-type ) ( Featherstone et al . , 2005; Marrus et al . , 2004; Petersen et al . , 1997; Qin et al . , 2005; Schuster et al . , 1991 ) . GluRIIA containing receptors generate a larger quantal size and display slower receptor desensitization than their GluRIIB counterparts ( DiAntonio et al . , 1999 ) . The A- and B-subtypes compete for incorporation into the tetramer at individual postsynaptic densities ( PSDs ) in a developmental and activity-regulated fashion ( Chen and Featherstone , 2005; DiAntonio et al . , 1999; Marrus and DiAntonio , 2004a; Rasse et al . , 2005; Schmid et al . , 2008 ) . The stereotypical alignment of individual AZs to distinct postsynaptic glutamate receptor fields in Drosophila allowed the generation of genetic tools to optically follow quantal fusion events at single release sites by visualizing glutamate receptor activation ( Melom et al . , 2013; Peled and Isacoff , 2011 ) . Classically , studies of synaptic transmission have used electrophysiology to measure the postsynaptic effect of neurotransmitter release over a population of release sites ( Katz and Miledi , 1969 , 1967 ) , precluding an analysis of how individual AZs contribute to the evoked response . By transgenically expressing GCaMP Ca2+ sensors that target to the postsynaptic membrane , single vesicle fusion events at each individual AZ can be imaged by following spatially localized Ca2+ influx induced upon glutamate receptor opening . This allows for the generation of Pr maps for both evoked and spontaneous fusion for all AZs ( Cho et al . , 2015; Melom et al . , 2013; Muhammad et al . , 2015; Newman et al . , 2017; Peled et al . , 2014; Peled and Isacoff , 2011; Reddy-Alla et al . , 2017 ) . One surprising observation using this quantal imaging approach is that AZs formed by a single motor neuron have a heterogeneous distribution of Pr , ranging from 0 . 01 to ~0 . 5 , with neighboring AZs often showing ~50-fold differences in Pr ( Melom et al . , 2013; Peled et al . , 2014; Peled and Isacoff , 2011 ) . These differences in Pr result in AZs with distinct short-term plasticity properties , suggesting release heterogeneity has functional importance for synaptic transmission ( Peled and Isacoff , 2011 ) . Key questions raised by these observations include how Pr is uniquely set for individual AZs and how the heterogeneity in Pr arises during development . Pr variability is likely to be controlled in part by variable Ca2+ channel abundance at release sites , consistent with the heterogeneity in VGCCs and other associated AZ proteins previously documented ( Böhme et al . , 2016; Ehmann et al . , 2014; Fulterer et al . , 2018; Graf et al . , 2012 , 2009; Guerrero et al . , 2005; Peled and Isacoff , 2011 ) . Indeed , Pr has been shown to correlate with BRP abundance at AZs in Drosophila ( Paul et al . , 2015; Peled et al . , 2014; Reddy-Alla et al . , 2017 ) and BRP has a key function in clustering VGCCs ( Kittel et al . , 2006 ) . Furthermore , Pr has previously been shown to correlate with the number of VGCCs at several vertebrate synapses ( Chen et al . , 2015; Nakamura et al . , 2015; Sheng et al . , 2012 ) . Pr could also be regulated by local synaptic vesicle pools and their number and/or state ( i . e . phosphorylation status ) . Beyond the molecular factors that determine AZ Pr , it is unclear how release heterogeneity at Drosophila NMJs arises during development . Intravital imaging at third instar larval NMJs has demonstrated that AZs are born small and gain pre- and postsynaptic components over time in a sequential manner ( Andlauer and Sigrist , 2012; Fouquet et al . , 2009; Füger et al . , 2007; Rasse et al . , 2005; Zhang et al . , 2010 ) . However , this approach has not been used during earlier stages of larval development to determine whether the release heterogeneity observed at third instar NMJs reflects AZ birth order . To characterize factors regulating Pr at individual AZs , as well as the origin of Pr diversity , we employed optical quantal analysis and intravital imaging to examine how Pr heterogeneity arises during development . Recent studies indicate that release sites possess structural and functional heterogeneity ( Éltes et al . , 2017; Holderith et al . , 2012; Maschi and Klyachko , 2017; Melom et al . , 2013; Peled et al . , 2014; Peled and Isacoff , 2011; Reddy-Alla et al . , 2017; Sugie et al . , 2015 ) . Using the Drosophila NMJ , we previously observed that evoked Pr is non-uniform across a population of ~300 AZs formed by motor neuron MN4-Ib onto muscle 4 , ranging from 0 . 01 to ~0 . 5 in HL3 saline containing 1 . 3 mM extracellular Ca2+ and 20 mM Mg2+ ( Melom et al . , 2013 ) . In our original study , each AZ was identified by the location of postsynaptic Ca2+ flashes , but AZs were not directly labeled in the live preparation . To more precisely map AZ Pr heterogeneity , we identified the position of each corresponding PSD by co-expressing the RFP-tagged glutamate receptor subunit GluRIIA under the control of its endogenous promoter ( Rasse et al . , 2005 ) along with a newer version of our previous biosensor , N-terminal myristoylated GCaMP6s , expressed in muscles using Mef2-GAL4 . We monitored postsynaptic Ca2+ influx from activation of glutamate receptors after either spontaneous release or nerve stimulation ( 0 . 3 Hz for 5 min ) in muscle 4 of early stage third instar larvae ( Video 1 ) . Using this approach , we mapped all myrGCaMP6s visualized release events to the position of in vivo GluRIIA-RFP labeled PSDs ( Figure 1A ) . Consistent with previous data , we observed a heterogeneous distribution of AZ Pr , with an average Pr of 0 . 073 ± 0 . 002 ( n = 1933 AZs from 16 NMJs from 16 animals ) . However , there was a ~50-fold difference in Pr between the highest and lowest releasing sites . The AZ Pr dataset did not fit a normal distribution ( D'Agostino K2 test ( p<0 . 0001 ) , Shapiro-Wilk test ( p<0 . 0001 ) , Kolmogorov-Smirnov test ( p<0 . 0001 ) ) and instead was skewed to the right , with a majority of AZs rarely releasing a synaptic vesicle following an action potential ( Pr in the range of 0 . 01 to 0 . 2 ) and a small number of AZs consistently showing high release rates ( 75% percentile of Pr was 0 . 1 , with a maximum Pr of 0 . 73; Figure 1B ) . 9 . 7% of all release sites defined by their apposed GluRIIA receptors displayed only spontaneous fusion events , and another 14 . 6% of the AZ population was silent for both spontaneous and evoked release during the recording period ( Figure 1B ) . We categorized all AZs with a release rate greater than two standard deviations above average as ‘high Pr’ , and the remaining AZs that showed evoked release as ‘low Pr’ . Using these criteria , 65 . 8% of all AZs fell in the low Pr category with an average Pr of 0 . 049 ± 0 . 004 . In contrast , 9 . 9% of AZs were classified as high Pr sites , with an average Pr of 0 . 277 ± 0 . 015 ( Figure 1C ) . High Pr AZs displayed on average a 5 . 7-fold higher chance of vesicle fusion following an action potential compared to low Pr AZs . One potential caveat to the interpretation of Pr heterogeneity is the possibility that multiple closely-positioned release sites could be falsely characterized as single high Pr AZs using conventional light microscopy . Presynaptic AZ position can be precisely identified at the NMJ by labeling the core AZ T-bar component Bruchpilot ( BRP ) , the homolog of mammalian ELKS/CAST ( Fouquet et al . , 2009; Wagh et al . , 2006 ) . To determine if the high Pr sites we observed were actually due to release from closely clustered AZs , we used high-resolution structured illumination microscopy ( SIM ) on fixed tissue stained with anti-BRP ( Figure 1D ) following dual color ( myrGCaMP6s/GluRIIA-RFP ) quantal imaging . SIM provides a lateral resolution of ~110 nm ( Wegel et al . , 2016 ) , providing clear resolution of the BRP ring structure , which has a diameter of ~200 nm ( Owald et al . , 2012 ) , smaller than the resolution limit of conventional light microscopy . The presence of GluRIIA-RFP allowed us to precisely match individual high Pr sites observed during live imaging with their position in fixed and stained tissue during SIM imaging ( Figure 1D ) . Using an automated detection algorithm in the Volocity 3D image analysis software , we were able to identify all AZs labeled with BRP ( Figure 1D , far right panel ) , and to resolve individual AZ clusters that were not separated using conventional spinning disk microscopy ( Figure 1D , white box ) . SIM analysis of distances between neighboring AZs indicated that 2 . 45 ± 0 . 4% ( n = 9 NMJs from nine animals ) of all AZs were located close enough to each other ( within 280 nm ) such that they would not be resolvable during live imaging . In contrast , 9 . 9% ( n = 16 NMJs from 16 animals ) of AZs were functionally classified as high Pr sites , suggesting that the majority of these sites are not likely to be explained by release from multiple closely clustered AZs . SIM visualization of BRP at AZs following Pr mapping revealed that a majority of high Pr sites were represented by a single BRP-positive AZ that was not further resolvable after SIM ( Figure 1D , red circles ) . We identified high Pr sites ( n = 42 AZs from 5 NMJs from five animals ) and then determined what fraction of these sites were truly clusters of multiple neighboring AZs based on their SIM profiles . Only 16 ± 3% of these high Pr sites were resolved into multiple AZs upon SIM analysis , indicating that most high Pr AZs correspond to single release sites . These single BRP clusters at high Pr sites had larger sum fluorescence intensities than most other BRP positive puncta ( Figure 1E ) . The average sum fluorescence of single BRP puncta from high Pr AZs ( 3 . 95 × 106 ± 2 . 67 x 105 , n = 24 AZs from 9 NMJs from nine animals ) was 1 . 7-fold greater than the fluorescence of randomly selected low Pr BRP clusters ( 2 . 33 × 106 ± 0 . 98 x 105 , n = 60 AZs from 9 NMJs from nine animals , p<0 . 0001 ) . To further examine large single BRP clusters that could not be resolved using conventional spinning disk microscopy , all BRP clusters larger than 280 nm were automatically detected and assigned their Pr measured during live imaging . We then determined whether these sites were represented by single or multiple AZs using SIM microscopy . Clusters > 280 nm in diameter that could be resolved to multiple BRP positive AZs after SIM imaging had a lower Pr ( 0 . 10 ± 0 . 02 , n = 35 AZs from 5 NMJs from five animals ) than those comprised of a single large BRP positive AZ ( 0 . 19 ± 0 . 02 , n = 42 AZs from 5 NMJs from five animals; Figure 1F ) . As such , high resolution SIM analysis confirms that most high Pr sites correspond to single AZs with more intense BRP labeling , consistent with previous data supporting the positive role of BRP in regulating Pr ( Paul et al . , 2015; Peled et al . , 2014; Reddy-Alla et al . , 2017 ) . Heterogeneous release rates between AZs could solely reflect stable differences in protein content of the AZs themselves . However , the accumulation of different synaptic vesicle populations with variable levels of Ca2+ sensitivity or fusogenicity might also contribute to release heterogeneity . The synchronous Ca2+ sensor Synaptotagmin 1 ( Syt1 ) resides on synaptic vesicles and plays a major role in Pr determination at Drosophila NMJs ( DiAntonio and Schwarz , 1994; Guan et al . , 2017; Lee et al . , 2013; Littleton et al . , 1994 , 1993; Yoshihara et al . , 2003; Yoshihara and Littleton , 2002 ) . We hypothesized that if differential synaptic vesicle Ca2+ sensitivity is a major determinant of release heterogeneity in addition to its established role in determining overall Pr , then elimination of Syt1 would disrupt Pr heterogeneity . Consistent with electrophysiological findings , quantal imaging in syt1 null mutants expressing GluRIIA-RFP and myrGCaMP6s revealed a dramatic reduction in evoked release , a shift from synchronous to highly asynchronous fusion , and an increase in spontaneous release rates ( Video 2 ) . To estimate AZ release heterogeneity in syt1 nulls , preparations were stimulated at 5 Hz and release events were normalized to the number of stimuli ( Figure 2A ) . The average release rate per AZ per second in syt1 nulls during 5 Hz stimulation was 0 . 03 ± 0 . 001 ( n = 719 AZs from 7 NMJs from six animals; Figure 2B ) . In contrast , spontaneous release rate per AZ in the absence of stimulation was 0 . 018 ± 0 . 001 per second in syt1 nulls ( n = 719 AZs from 7 NMJs from six animals ) compared to 0 . 011 ± 0 . 001 in controls ( n = 559 AZs from 6 NMJs from four animals , p<0 . 0001; Figure 2B ) . Although release rate is dramatically reduced in syt1 nulls , AZs still maintained overall heterogeneity in Pr distribution . Comparing the distribution of AZ release rates for syt1 nulls and controls , release was proportionally decreased across all AZs in syt1 ( Figure 2C ) ; frequency distribution analysis of AZs with normalized release rates ( from 0 to maximum release ) confirmed that there was no significant change in the heterogeneity of release between syt1 mutants and controls ( Figure 2D , E ) . Given that AZ Pr remains highly heterogeneous in the absence of Syt1 , these data indicate that heterogeneity in synaptic vesicle Ca2+ sensitivity between AZs is unlikely to play a major role in Pr distribution . Release heterogeneity in syt1 null animals suggests that synaptic vesicle Ca2+ sensitivity is not a major determinant of Pr heterogeneity at this synapse; however , it is possible that other synaptic vesicle components influence Pr on an AZ-specific level . If unique local synaptic vesicle pools contribute to the distribution of Pr between AZs , we predicted that Pr would be highly dynamic at individual AZs over time . In contrast , stability of Pr would argue that release heterogeneity is more likely associated with stably resident proteins at individual AZs . To assess Pr stability over time , we conducted a 3 min imaging session using 0 . 3 Hz stimulation to generate an initial Pr map , and then allowed the preparation to rest for 5 min without stimulation or imaging before re-mapping Pr in a final 3 min imaging session . We were constrained in our ability to examine Pr continuously over longer time intervals due to bleaching of GCaMP6s from the high frequency capture rate . Pr at individual AZs was very stable between the two sessions ( Figure 3A ) . This was especially evident for high Pr sites , which sustained high levels of activity during both imaging sessions . Plotting release rate for all AZs revealed a strong correlation for Pr across the two imaging sessions ( Pearson r = 0 . 77 , R2 = 0 . 59 , p<0 . 0001 , n = 988 AZs from 8 NMJs from seven animals; Figure 3B ) . We next used a strong stimulation paradigm to drive vesicle cycling to promote intermixing of synaptic vesicles . NMJ preparations were imaged during two low frequency 0 . 3 Hz stimulation periods separated by a 5 min 5 Hz stimulation session ( Figure 3C ) . Release maps were not dramatically altered by 5 Hz stimulation , with the overall correlation of Pr between the two imaging sessions similar with and without stimulation ( Pearson r = 0 . 78 , R2 = 0 . 61 , p<0 . 0001 , n = 613 AZs from 6 NMJs from six animals; Figure 3D ) . Thus , inducing vesicle recycling with 5 Hz stimulation does not dramatically change Pr across the AZ population , arguing that stably resident AZ components , rather than AZ-specific synaptic vesicle populations , are likely to represent the major driver of Pr heterogeneity at this synapse . Given that vesicle fusion is highly sensitive to Ca2+ and most effective in close proximity to VGCCs ( Augustine et al . , 1985; Böhme et al . , 2016; Chen et al . , 2015; Heidelberger et al . , 1994; Katz and Miledi , 1967; Katz , 1969Zito et al . , 1999; Keller et al . , 2015; Meinrenken et al . , 2003 , 2002; Nakamura et al . , 2015; Sheng et al . , 2012; Stanley , 2016; Wang et al . , 2008 ) , Ca2+ channel abundance at individual AZs is predicted to be a key variable for Pr heterogeneity at Drosophila NMJs as well . Cacophony ( cac ) encodes the Drosophila voltage-activated Ca2+ channel α1 subunit required for neurotransmitter release ( Fouquet et al . , 2009; Kawasaki et al . , 2004; 2000; Littleton and Ganetzky , 2000; Liu et al . , 2011; Rieckhof et al . , 2003; Smith et al . , 1996 ) . Transgenic animals expressing fluorescently tagged Cac channels have been previously generated , demonstrating that Cac localizes specifically to AZs and its abundance appears heterogenous across release sites ( Kawasaki et al . , 2004; Matkovic et al . , 2013; Yu et al . , 2011 ) . To examine the heterogeneity of Cac abundance across AZs , we used SIM to measure the distribution of Cac-GFP and BRP at muscle 4 NMJs . Using this approach , we observed a heterogeneous distribution of mean Cac-GFP fluorescence at AZs , similar to the variable levels of BRP intensity described earlier ( Figure 1D , Figure 4—figure supplement 1A , B ) . 5 . 72% of AZs displayed Cac-GFP fluorescence greater than two standard deviations above average ( n = 2011 AZs from 11 NMJs from three animals ) . The mean Cac-GFP fluorescence for these bright AZs was 2 . 1-fold greater than that observed for the remaining sites ( p<0 . 0001; Figure 4—figure supplement 1C ) . Mean AZ intensities of Cac-GFP and BRP were positively correlated ( Pearson r = 0 . 46 , R2 = 0 . 21 , p<0 . 0001 , n = 730 AZs from 6 NMJs from three animals; Figure 4—figure supplement 1D , E ) . To determine whether Cac distribution correlates with Pr heterogeneity , we used dual color imaging experiments where vesicle fusion events were detected by myrGCaMP6s driven in muscles using mef2-GAL4 and Ca2+ channel distribution was visualized using red-labeled Cac-TdTomato expressed pan-neuronally using elav-GAL4 ( Figure 4A ) . We observed a strong positive correlation ( average Pearson r = 0 . 61 , R2 = 0 . 38 , p<0 . 0001 , n = 483 AZs from 7 NMJs from seven animals ) between AZ Cac fluorescence intensity and evoked AZ Pr ( Figure 4B ) . In contrast , single AZ release rates for spontaneous events showed only a mild correlation with Cac intensity ( average Pearson r = 0 . 19 , R2 = 0 . 036 , p<0 . 0001 , n = 483 AZs from 7 NMJs from seven animals; Figure 4C ) . These results are consistent with previous observations that release rates for evoked and spontaneous fusion are not correlated at Drosophila AZs ( Melom et al . , 2013; Peled et al . , 2014 ) , and that spontaneous fusion is largely independent of extracellular Ca2+ at this synapse ( Jorquera et al . , 2012; Lee et al . , 2013 ) . To increase confidence that the observed Cac-TdTomato intensity accurately reflects Cac distribution , we also measured the correlation between Pr and Cac channels transgenically tagged with GFP . We generated transgenic lines expressing myristoylated red Ca2+ indicators previously characterized in the field , including RCaMP1h , R-GECO1 and jRGECO1a . Although RCaMP1h and R-GECO1 were too dim to visualize localized Ca2+ transients at PSDs , transgenic lines expressing the myristoylated Ca2+ sensor jRGECO1a ( Dana et al . , 2016 ) in muscle four allowed detection of Ca2+ influx following vesicle fusion at single AZs ( Video 3 ) . In contrast to the more robust GCaMP6s , jRGECO1a has a shorter fluorescent lifetime and the signal amplitude decays more rapidly . We observed that quantal events imaged with myr-jRGECO1a were dimmer and fully bleached within 7–10 min of imaging . Therefore , preparations were stimulated at 1 Hz for shorter two-minute imaging sessions to generate Pr maps in larvae expressing myr-jRGECO1a ( Figure 4D ) . We observed a strong correlation between AZ Pr detected by myr-jRGECO1a and Cac-GFP intensity ( average Pearson r = 0 . 54 , R2 = 0 . 29 , p<0 . 0001 , n = 651 AZs from 7 NMJs from seven animals , correlation from a representative experiment shown in Figure 4E ) . Again , we found a weaker correlation between spontaneous fusion and Cac-GFP intensity ( Pearson r = 0 . 17 , R2 = 0 . 03 , p<0 . 0001 , n = 651 AZs from 6 NMJs from six animals ) . Hence , Pr for action-potential evoked fusion is strongly correlated with the local abundance of Cac channels at individual Drosophila AZs . We next compared Cac-GFP fluorescence at AZs that were functionally classified as either low or high Pr sites by quantal imaging using myr-jRGECO1a ( Figure 4F ) . The average fluorescence of single Cac-GFP puncta from high Pr AZs ( normalized intensity = 0 . 6 ± 0 . 04 , n = 38 AZs from 7 NMJs from seven animals ) was 2 . 09-fold greater than the average fluorescence of low Pr AZs ( normalized intensity = 0 . 29 ± 0 . 01 , n = 638 AZs from 7 NMJs from seven animals , p<0 . 0001 ) . We also examined the Pr of AZs classified by Cac-GFP fluorescence . The average Pr for AZs displaying high Cac-GFP fluorescence ( >2 standard deviations above average ) was 0 . 2 ± 0 . 016 ( n = 7 NMJs from seven animals ) compared to 0 . 06 ± 0 . 003 ( n = 7 NMJs from seven animals , p<0 . 0001 ) for the remaining AZs with lower levels of Cac-GFP . Although the absolute number of Cac channels at single Drosophila AZs is unknown , these data suggest that a ≥ 2 fold difference in channel number exists between low and high Pr AZs . Given the steep third to fourth order non-linear dependence of synaptic vesicle fusion on Ca2+ ( Dodge and Rahamimoff , 1967; Heidelberger et al . , 1994; Jan and Jan , 1976 ) , a small change in channel number is likely to have a large effect on Pr . VGCCs are extensively modulated by second messenger pathways that can alter channel conductivity ( Catterall and Few , 2008; Dolphin et al . , 1991; Evans and Zamponi , 2006; Reid et al . , 2003; Tedford and Zamponi , 2006; Zamponi and Snutch , 1998 ) . Although Cac channel abundance correlates with AZ Pr , an important readout of channel activity is the local Ca2+ influx occurring at each AZ . Assaying presynaptic Ca2+ influx directly would also be useful to bypass any unknown effects on Pr generated by expressing fluorescently tagged Ca2+ channels . As such , we generated transgenic animals expressing GCaMP6m fused to the N-terminus of BRP , which localizes directly to the base of the AZ where Ca2+ channels cluster ( Fouquet et al . , 2009; Kittel et al . , 2006 ) . These GCaMP6m fusions were made to a BRP fragment ( BRPshort ) corresponding to amino acids 473–1226 of the full 1740 amino acid protein ( Schmid et al . , 2008 ) . At rest , N-terminal GCaMP-BRPshort was dim , consistent with low levels of resting Ca2+ ( Figure 5A ) . Stimulation at 10 Hz resulted in a robust increase in discrete punctated presynaptic fluorescence that remained confined to single AZs during stimulation ( Figure 5A ) . During multiple rounds of 5 s 10 Hz stimulation , GCaMP-BRPshort fluorescence increase ( ΔF ) varied between AZs , but was stable at the same AZ for each independent stimulation ( n = 205 AZs from 6 NMJs from three animals; Figure 5B , C ) . Given that GCaMP-BRPshort abundance at an AZ likely reflects the absolute amount of BRP at that AZ , we assayed if heterogeneity in GCaMP-BRPshort fluorescence during 10 Hz stimulation could be solely explained by differences in sensor distribution across AZs . We applied 200 nM of the Ca2+ ionophore ionomycin to elevate Ca2+ concentrations uniformly throughout the terminal independent of Cac abundance . In the presence of ionomycin , differences in fluorescent signals between AZs should be entirely due to heterogeneity in sensor abundance . We observed a rightward shift in the GCaMP-BRPshort intensity distribution among AZs upon ionomycin application compared to 10 Hz stimulation ( average AZ fluorescence during 10 Hz stimulation was 1924 ± 63 , and following ionomycin addition was 6105 ± 175; Figure 5—figure supplement 1A , B ) , indicating that during 10 Hz stimulation , detection of Ca2+ by GCaMP-BRPshort is not limited by sensor abundance . Furthermore , we observed a significant difference in the shape of the distribution during 10 Hz stimulation compared to both before stimulation and after ionomycin . The distribution of fluorescence intensities is narrower both at rest and upon ionomycin application; these two distributions should primarily reflect sensor distribution . In contrast , the distribution of GCaMP-BRPshort fluorescence during 10 Hz stimulation is wider , indicating that the sensor is reporting local changes in Ca2+ influx and not just sensor distribution ( Figure 5 , Figure 5—figure supplement 1C , D ) . Thus , although GCaMP-BRPshort abundance is likely to contribute to the levels of Ca2+ influx detected , these results are consistent with heterogeneity in Ca2+ influx across individual AZs . We next assayed if Ca2+ influx detected by GCaMP-BRPshort is correlated with Cac channel abundance . Animals expressing both Cac-TdTomato and GCaMP-BRPshort transgenes in the presynaptic compartment displayed a strong correlation between Ca2+ dependent excitation of GCaMP-BRPshort ( ΔF ) and Cac-TdTomato intensity at individual AZs during stimulation ( Pearson r = 0 . 73 , R2 = 0 . 53 , p<0 . 0001 , n = 176 AZs from 7 NMJs from six animals; Figure 5D , E ) . We observed a weaker correlation between Cac intensity and GCaMP-BRPshort ΔF at rest ( Pearson r = 0 . 18 , R2 = 0 . 03 , p<0 . 001 , n = 338 AZs from 8 NMJs from six animals ) . A few instances were noted where specific AZs experienced a disproportionally low GCaMP-BRPshort ΔF signal relative to their Cac-TdTomato intensity ( Figure 5D , arrows ) , suggesting Ca2+ influx may be fine-tuned at certain release sites . We next analyzed the correlation between GCaMP-BRPshort ΔF induced by 10 Hz stimulation and release rate visualized by postsynaptic myr-jRGECO1a during 1 Hz stimulation ( Figure 6A , B ) . AZs that experienced stronger Ca2+ influx displayed higher Pr during stimulation . Overall , there was a strong correlation between Ca2+ influx and AZ Pr ( Pearson r = 0 . 56 , R2 = 0 . 31 , p<0 . 0001 , n = 492 AZs from 6 NMJs from six animals; Figure 6B ) . In contrast , the frequency of spontaneous vesicle fusion per AZ was only mildly correlated with GCaMP-BRPshort ΔF ( Pearson r = 0 . 23 , R2 = 0 . 07 , n = 492 AZs from 6 NMJs from six animals; correlations from a representative experiment shown in Figure 6C ) . It is worth noting that although a strong correlation between Ca2+ influx and evoked Pr was observed at most AZs , a minority population of release sites that displayed robust Ca2+ influx had very low Pr ( Figure 6B ) . To functionally test if the level of Ca2+ influx rather than the structural presence of the Ca2+ channel is responsible for determining Pr , we generated Pr maps in the cacNT27 mutant using dual color quantal imaging with GluRIIA-RFP and myr-GCamP6s ( Figure 6—figure supplement 1A ) . CacNT27 channels have reduced Ca2+ conductance due to a point mutation in the S4 voltage sensor ( Rieckhof et al . , 2003 ) . We observed that cacNT27 results in a global decrease in Pr across AZs; evoked Pr in controls ranged from 0 to 0 . 73 with an average of 0 . 073 ± 0 . 0021 ( n = 1933 AZs from 16 NMJs from 16 animals ) , while cacNT27 Pr ranged from 0 to 0 . 47 with a significantly lower average Pr of 0 . 049 ± 0 . 0045 ( n = 275 AZs from 5 NMJs from five animals; Figure 6—figure supplement 1B ) . While the entire Pr distribution was shifted lower compared to controls ( Figure 6—figure supplement 1C ) , the normalized distribution of Pr was nearly identical to controls ( Figure 6—figure supplement 1D ) . These findings indicate that the levels of Ca2+ influx through Cac channels , rather than the physical presence of the channels , is a primary determinant of Pr . We next examined if postsynaptic glutamate receptor composition varied at low Pr versus high Pr AZs . Drosophila glutamate receptors at the NMJ assemble as heteromeric tetramers containing three essential subunits ( GluRIII , IID and IIE ) and a variable fourth subunit of GluRIIA or GluRIIB , with the GluRIIA subtypes having a higher conductance than GluRIIB ( Featherstone et al . , 2005; Marrus et al . , 2004; Petersen et al . , 1997; Qin et al . , 2005; Schuster et al . , 1991 ) . To determine if the GluR subtypes differentially accumulate at AZs in a manner that correlates with presynaptic Pr , we visualized GluRIIA-RFP and GluRIIB-GFP expressed under the control of their endogenous promoters ( Rasse et al . , 2005 ) . To image myrGCaMP6s activity without obscuring GluRIIB-GFP , myrGCaMP6s was expressed at low levels using the LexA/LexOP system in muscle four with Mef2-LexA . LexA driven myrGCaMP6s signal is dimmer than UAS-myrGCaMP6s at rest and does not obscure the brighter GluRIIB-GFP PSD puncta ( Figure 7A ) . However , upon Ca2+ binding to myrGCaMP6s , the fluorescence dramatically increases compared to the level of endogenous GluRIIB-GFP signal , allowing simultaneous imaging of baseline GluRIIB-GFP levels and synaptic activity detected by myrGCaMP6s ( Video 4 ) . Simultaneous expression of GluRIIA-RFP and GluRIIB-GFP revealed a heterogeneous distribution of each subunit across the AZ population , with GluRIIA levels appearing more variable than GluRIIB ( Figure 7A ) . Similar to the relatively sparse localization of high Pr AZs ( Figure 1 ) , a similar sparse distribution of AZs apposed by very bright GluRIIA fields was observed ( Figure 7A ) . To determine if AZs that preferentially accumulate high levels of GluRIIA corresponded to high Pr release sites , we mapped Pr across the AZ population in GluRIIA-RFP/GluRIIB-GFP expressing animals . Analysis of the Pr map revealed a strong positive correlation between mean GluRIIA-RFP intensity and Pr ( Pearson r = 0 . 56 , R2 = 0 . 32 , p<0 . 0001 , n = 756 AZs from 8 NMJs from four animals; Figure 7B ) . In contrast , correlation with the levels of mean GluRIIB-GFP intensity was weaker ( Pearson r = 0 . 32 , R2 = 0 . 1 , p<0 . 0001 , n = 756 AZs from 8 NMJs from four animals; Figure 7C ) . Heterogeneity in GluRIIA PSD brightness and AZ Pr was also observed in syt1 mutants ( Figure 2A , arrows ) . These findings are consistent with previous observations that glutamate receptors preferentially cluster at sites with high Pr based on electrophysiological studies in a Drosophila GluRIII hypomorphic mutant ( Marrus and DiAntonio , 2004b ) . As observed in controls ( Figure 1 ) , a positive correlation between AZ Pr and BRP levels was also observed in larvae expressing labeled glutamate receptors ( Pearson r = 0 . 44 , R2 = 0 . 2 , p<0 . 0001 , n = 399 AZs from 6 NMJs from four animals; Figure 7D ) . In summary , these data indicate that GluRIIA accumulates more at PSDs apposing high Pr AZs . Beyond the preferential GluRIIA accumulation at high Pr sites , a change in GluRIIA/B distribution within single PSDs was also observed . PSDs associated with the highest Pr AZs showed a segregated distribution of the receptor subtypes , with GluRIIA concentrating in the center of the receptor field immediately apposing the presynaptic BRP cluster ( Figure 7E , arrows ) . At these sites , GluRIIB occupied a more peripheral position around the central GluRIIA cluster . A similar localization pattern with a ring of GluRIIB surrounding a central GluRIIA patch was previously noted with antibody staining for the two receptors at a population of AZs in wildtype late third instar larvae ( Marrus et al . , 2004 ) . To analyze this receptor segregation in greater detail , GluRIIA-RFP and GluRIIB-GFP were examined in the absence of co-expressed myrGCaMP6s . Prior analysis ( Figure 7B ) indicated the brightest GluRIIA PSDs corresponded to high Pr sites . Bright PSDs were selected based on their GluRIIA intensity ( two standard deviations above average ) and line profiles were drawn across each PSD . The intensity of pixels along that line for each fluorophore was then analyzed . Average pixel intensity revealed drastically distinct profiles for GluRIIB distribution between ‘bright’ and ‘dim’ PSDs classified based on their GluRIIA intensity . GluRIIB was more evenly distributed across the entire PSD at dim GluRIIA sites , but was segregated outward , forming a ring around central GluRIIA puncta at bright GluRIIA sites ( Figure 7F ) . In addition , presynaptic BRP intensity was more strongly correlated with postsynaptic GluRIIA levels ( Pearson r = 0 . 53 , R2 = 0 . 28 , p<0 . 0001 , n = 2496 AZs from 19 NMJs from seven animals; Figure 7G ) than with GluRIIB ( Pearson r = 0 . 24 , R2 = 0 . 05 , p<0 . 0001 , n = 2496 AZs from 19 NMJs from seven animals; Figure 7H ) . These findings suggest that the postsynaptic cell accumulates GluRIIA and redistributes GluRIIB to the PSD periphery at high Pr sites . The Drosophila larval NMJ is a highly dynamic structure , with new synaptic boutons and AZs undergoing continuous addition throughout development ( Harris and Littleton , 2015; Rasse et al . , 2005; Schuster et al . , 1996; Zito et al . , 1999 ) . Given the correlation between Ca2+ channel abundance , GluRIIA/B segregation and high Pr , we were interested in determining how AZs acquire a specific Pr during a larval developmental period that lasts 6–7 days . One model is that certain AZs gain a higher Pr status during development through preferential accumulation of key AZ components compared to their neighbors . Alternatively , high Pr AZs might simply be more mature than their low Pr neighbors , having an earlier birthdate and a longer timeframe to accumulate AZ material . To test these models for release heterogeneity , it would be desirable to follow Pr development from the embryonic through larval stages . However , this is not technically feasible due to the small size of AZs and the rapid locomotion that larvae undergo , preventing generation of Pr maps in moving animals . Instead , we employed an alternative intravital approach to repeatedly image the same NMJ at muscle 26 directly through the cuticle of intact larvae during anesthesia ( Andlauer and Sigrist , 2012; Fouquet et al . , 2009; Füger et al . , 2007; Rasse et al . , 2005; Zhang et al . , 2010 ) . During anesthesia , endogenous action potential-induced release and the associated GCaMP signals were eliminated , preventing direct Pr measurements in anesthetized larvae . We instead focused on imaging GluRIIA accumulation and GluRIIA/B segregation , which was strongly correlated with Pr ( Figure 7 ) , as a proxy for the emergence of high Pr sites . Previously described in vivo imaging approaches with anesthesia at Drosophila NMJs employed early third instar larvae as the starting time point , and followed the distribution of fluorescently-labeled synaptic proteins during the final ~36 hr of development prior to pupation ( Andlauer and Sigrist , 2012; Fouquet et al . , 2009; Füger et al . , 2007; Rasse et al . , 2005; Zhang et al . , 2010 ) . To follow AZ Pr development beginning soon after synapse formation , we modified these techniques to allow imaging of glutamate receptors at earlier stages of development ( see Materials and methods ) . This allowed successful birth dating and successive imaging of the same AZ over a 6 day period beginning shortly after synapse formation in the early first instar period through the late third instar stage ( Figure 7—figure supplement 1 ) . In early first instar larvae ( within 12 hr of hatching ) GluRIIA and GluRIIB were largely co-localized at postsynaptic puncta ( Figure 7—figure supplement 1 ) . One exception was the presence of diffuse GluRIIA that accumulated around unusually long axonal extensions that emerged from presynaptic boutons ( Figure 7—figure supplement 2 ) . These structures were devoid of any detectable GluRIIB or the bright GluRIIA puncta that are associated with AZs , and may be remnants of previously described muscle filopodial structures , termed myopodia , that interact with presynaptic filopodia to dynamically regulate early synaptic target recognition ( Kohsaka and Nose , 2009; Ritzenthaler et al . , 2000; Ritzenthaler and Chiba , 2003 ) . GluRIIA was diffusely present on these structures , as has been observed for the leucine-rich repeat cell adhesion protein Capricious ( Kohsaka and Nose , 2009 ) . Repeated imaging of these thinner GluRIIA-positive processes revealed that they were capable of developing into mature synaptic boutons with concentrated GluRIIA and GluRIIB synaptic puncta ( Figure 7—figure supplement 3 ) . By 24 hr of larval growth , GluRIIA rich extensions were no longer observed , indicating these structures are restricted to early developmental stages . Prior studies indicated GluRIIA PSD levels closely track with Cac accumulation at corresponding AZs of third instar larvae ( Fouquet et al . , 2009; Rasse et al . , 2005 ) , indicating the two compartments are likely to mature at similar rates . To examine this directly , we assayed whether GluRIIA and Cac accumulation were correlated during earlier stages of development ( Figure 7—figure supplement 3 ) . Indeed , the intensity of Cac-GFP and GluRIIA-RFP puncta were strongly correlated at individual AZs during both early and late larval development ( first instar; Pearson r = 0 . 82 , R2 = 0 . 6771 , p<0 . 0001 , n = 441 AZs from 8 NMJs from eight animals; third instar; Pearson r = 0 . 63 , R2 = 0 . 395 , n = 874 AZs from 8 NMJs from five animals; Figure 7—figure supplement 3D , E ) . One exception was observed in very early first instar larvae , where a few Cac-GFP puncta accumulated along the previously described GluRIIA positive axonal extensions prior to the specific accumulation of GluRIIA at PSDs ( Figure 7—figure supplement 3A ) . In contrast , postsynaptic appearance of GluRIIA at the PSD could be observed to slightly precede the accumulation of Cac-GFP at later developmental stages ( Figure 7—figure supplement 3B , C ) , similar to previous observations at mature third instar NMJs ( Rasse et al . , 2005 ) . In summary , the intensity of Cac-GFP and GluRIIA-RFP puncta are strongly correlated at AZs during larval development , indicating that GluRIIA provides a robust marker that reflects the corresponding levels of presynaptic Cac at individual AZs . We examined how GluRIIA accumulation and GluRIIA/B segregation emerged during development of the NMJ . Live imaging of GluRIIA and GluRIIB distribution at early PSDs in anesthetized first instar larvae demonstrated that the receptors were co-localized and lacked the segregation where GluRIIB clustered around central GluRIIA puncta that was observed at high Pr sites later in development ( Figure 8A ) . The first emergence of GluRIIA/B segregation was observed after 36 hr of imaging starting from the first instar period . The GluRIIA/B segregation always emerged first at the oldest synapses that existed previously on the first day of imaging ( Figure 8A , B ) . The most mature PSDs also contained more GluRIIA fluorescent signal ( 17430 ± 634 . 0 , n = 86 AZs from 8 NMJs from five animals ) compared to younger synapses that emerged during the 48 hr imaging session ( 8909 ± 289 . 8 , n = 210 AZs from 8 NMJs from five animals ) . During later larval development , the cuticle thickness changed dramatically and prevented reliable comparison of absolute receptor density with earlier stages . However , GluRIIA intensities that were more uniform at the first instar larval stage became more heterogeneous at the third instar stage ( Figure 8—figure supplement 1 ) . Indeed , histograms of normalized fluorescence intensity ( relative intensity scaled from 0 to 1 ) revealed that GluRIIA and GluRIIB were relatively uniformly distributed at first instar larval PSDs , with GluRIIA distribution becoming more skewed at later stages ( Figure 9A , B ) . To determine whether muscle 26 exhibits Pr heterogeneity similar to muscle 4 , we performed Pr mapping using GluRIIA-RFP and myr-GCaMP6s in dissected non-anesthetized third larvae and confirmed that Pr heterogeneity is extremely similar between muscle 26 ( mean Pr = 0 . 068 ± 0 . 0048 , skewness = 2 . 27 , 365 AZs from 5 NMJs from five animals ) and muscle 4 ( mean Pr = 0 . 073 ± 0 . 0021 , skewness = 2 . 23 , 1933 AZs from 16 NMJs from 16 animals; Figure 9—figure supplement 1B–D ) . These results indicate that GluRIIA/B fluorescence and Pr distribution are both highly heterogeneous by the early third instar stage in muscle 26 , with the brightest GluRIIA PSDs , and by extension their corresponding high Pr AZs , representing those that appeared earliest in development . Over what time frame do synapses developmentally acquire markers of high Pr sites ? To estimate the average time required for AZ maturation , we calculated the time interval from the first emergence of a PSD in an imaging session to the time point when segregation of GluRIIB around GluRIIA central puncta occurred . This analysis was restricted to newly formed AZs that appeared during the imaging sessions and excluded AZs that were present in the first imaging session performed in first instar larvae . The average time from the first emergence of a PSD to when it acquired the segregated GluRIIA/B pattern observed at high Pr AZs was 3 . 20 ± 0 . 08 days ( n = 41 AZs from 7 NMJs from three animals; Figure 9C ) . In a small subset of PSDs ( 5% ) , a slightly faster accumulation of GluRIIA and the formation of GluRIIB peripheral rings was observed , but never faster than 2 days . In addition , glutamate receptor fields increased in size throughout the first 72 hr of development ( Figure 9D ) . GluRIIA diameter increased by 1 . 47-fold during the first 24 hr of development but then plateaued , while GluRIIB field diameter continued to grow over 72 hr , increasing in size by 2 . 8-fold over that time . At 72 hr after initial glutamate receptor field formation , the average GluRIIA field diameter was 0 . 59 ± 0 . 058 μm ( range 0 . 45–0 . 80 μm , 12 PSDs from three animals ) and the average GluRIIB field diameter was 1 . 01 ± 0 . 035 μm ( range 0 . 89–1 . 12 μm , 12 PSDs from three animals ) , consistent with the formation of segregated GluRIIA/B fields at mature AZs . To directly assess whether AZ age corresponds to Pr , we followed PSDs in animals expressing GluRIIA-RFP , GluRIIB-GFP and myr-GCaMP for 24 hr using in vivo imaging , and then dissected the animals and mapped Pr directly ( Figure 9F ) . We observed that PSDs that emerged on the second day of imaging ( less than 24 hr old ) were consistently associated with very low Pr AZs . The mean Pr of these newly-formed AZs was 0 . 035 ± 0 . 0047 ( n = 77 AZs from 4 NMJs from four animals ) , with a range of 0–0 . 14 . In contrast , the average Pr for AZs older than 24 hr was 0 . 14 ± 0 . 0093 with a range of 0–0 . 61 ( n = 188 AZs from 4 NMJs from four animals; Figure 9G ) . These findings support a model in which the vast majority of newly formed AZs are very weak , with increased Pr requiring more than 24 hr to develop . Given the developing NMJ is adding AZs at a rapid rate ( Rasse et al . , 2005; Schuster et al . , 1996 ) , we estimated whether AZ maturation time identified over the course of our live imaging experiments could lead to the ~10% of high Pr sites observed at the early third instar stage . We quantified the number of synapses present at the same NMJ from the first instar through the early third instar stage from live imaging experiments ( Figure 9E ) . AZ number roughly doubled each day , such that the average number of AZs found at the first instar stage ( day 1 ) represented 14 . 7 ± 1 . 4% ( n = 8 NMJs from three animals ) of all AZs present by day 4 ( 3 days after initial imaging in first instars ) . Overall , these data are consistent with the hypothesis that AZ maturation is a key factor in regulating Pr , leading to increased accumulation of Ca2+ channels and GluRIIA/B segregation at high Pr sites compared to AZs that are newly formed ( <2 days ) . If heterogeneity in AZ age underlies the majority of diversity in Pr , we would expect to see a different Pr distribution at the earlier second instar stage when there is a smaller range of AZ ages . Since the number of AZs nearly doubles each day and newly formed AZs display low Pr , we hypothesized that third instar NMJs would have a greater proportion of low Pr sites compared to second instar NMJs . To test this hypothesis , we mapped Pr at muscle 4 NMJs in the second instar and compared the distribution to that seen at the third instar stage ( Figure 9—figure supplement 1A , B ) . A rightward shift in the distribution of Pr was observed in the earlier second instar stage , with a greater proportion of AZs in the higher Pr category , a smaller population of low Pr AZs , and a significant increase in mean Pr in second instars compared to third instars ( third instar: 0 . 07 ± 0 . 002 , n = 1933 AZs from 16 NMJs from 16 animals; second instar: 0 . 13 ± 0 . 008 , n = 282 AZs from 6 NMJs from six animals; Figure 9—figure supplement 1C , D ) . We were unable to map Pr at muscle 26 in earlier stages in dissected animals because it is covered by muscles 6 and 7 until the third instar stage . These data support the hypothesis that Pr heterogeneity reflects AZ age and maturation time . To determine whether the developmental time-course of synapse maturation could be modulated by changes in presynaptic activity , we measured the rate of PSD growth ( fold-increase in GluRIIB area over 24 hr ) and the percent of PSDs displaying GluRIIA/B segregation in mutants with altered presynaptic activity ( Figure 10 ) . We first measured postsynaptic maturation in BRP69/def null animals , which have a dramatic reduction in evoked synaptic transmission ( Kittel et al . , 2006 ) . Consistent with previous findings , Cac-GFP intensity in BRP69/def animals was reduced to 25% ( mean fluorescence intensity 3549 ± 23 , 4 NMJs from two animals ) of control levels ( mean intensity 14177 ± 220 , 4 NMJs from four animals; Figure 10—figure supplement 1A , B ) . We expressed GluRIIA-RFP and GluRIIB-GFP in BRP69/def mutants and imaged muscle 26 NMJs in anesthetized larvae intravitally over 24 hr . A significant reduction in postsynaptic maturation rate was observed; newly formed BRP69/def PSDs only increased in GluRIIB area by 1 . 16-fold ( ±0 . 11 , 4 NMJs from three animals ) over the first 24 hr of development , compared to a 1 . 61-fold ( ±0 . 11 , 5 NMJs from five animals ) increase in controls ( Figure 10A , C ) . Furthermore , a significant reduction in the percent of PSDs with GluRIIA/B rings ( defined by a > 10% central dip in the GluRIIB intensity profile ) was observed in the second day of larval development; only 4 . 9% of BRP69/def PSDs ( ±1 . 3% , 10 NMJs from five animals ) showed receptor segregation compared to 22% ( ±5% , 11 NMJs from five animals ) in age-matched controls ( Figure 10D ) . We were unable to observe any GluRIIB rings at the third instar stage in BRP69/def , in contrast to the clear rings seen in controls at this stage . Instead , BRP69/def mutants displayed highly disorganized GluR clusters ( Figure 10B ) . To examine the consequences of reductions in presynaptic activity independent of structural alterations in the AZ that might occur at NMJs in BRP mutants , we measured PSD maturation rate in napTS and syt1null mutants that reduce release through separate mechanisms . Loss of Syt1 causes a dramatic decrease in synchronous fusion ( DiAntonio and Schwarz , 1994; Guan et al . , 2017; Lee et al . , 2013; Littleton et al . , 1994 , 1993; Yoshihara et al . , 2003; Yoshihara and Littleton , 2002 ) , while napTS results in constitutively reduced neuronal excitability due to decreased sodium channel activity ( Kernan et al . , 1991; Wu et al . , 1978 ) . In both syt1 and napTS mutants , significantly reduced PSD growth rate and GluRIIA/B segregation was observed compared to controls; GluRIIB field area in napTS and syt1null mutants only increased by 1 . 24-fold ( ±0 . 05 4 NMJs from three animals ) and 1 . 15-fold ( ±0 . 05 , 4 NMJs from four animals ) respectively during the first 24 hr of PSD development compared to 1 . 61-fold ( ±0 . 11 , 5 NMJs from five animals ) in controls ( Figure 10C ) . At the second instar stage , when 22% of control PSDs displayed GluRIIA/B segregation , only 5 . 4% of syt1 ( ±1 . 5% , 7 NMJs from three animals ) and 6 . 0% of napTS ( ±2 . 0% , 13 NMJs from four animals ) PSDs showed GluR segregation ( Figure 10D ) . Both napTS and syt1null mutants also formed less segregated GluRIIA/B fields compared to controls at the third instar stage ( Figure 10B ) . To determine whether increasing presynaptic release would accelerate maturation , we examined PSD size and GluR segregation in shaker120 , eag1 ( sh , eag ) double mutants that display increased excitability due to loss of several K+ currents ( Ganetzky and Wu , 1983; Salkoff and Wyman , 1981; Wu et al . , 1983 ) . Sh , eag showed a significantly increased rate of GluRIIB field size increase ( 2 . 37-fold ± 0 . 11 , 5 NMJs from four animals ) over 24 hr compared to controls ( 1 . 61 fold ( ±0 . 11 , 5 NMJs from five animals; Figure 10A , C ) . Furthermore , 48 . 7% of sh , eag mutant PSDs ( ±4 . 4% , 13 NMJs from five animals ) displayed GluRIIB rings at the early second instar stage , compared to only 22% of control PSDs ( Figure 10D ) . These observations indicate that increasing ( sh , eag ) or decreasing ( napTS , BRP69/def , syt1 ) presynaptic activity results in a respective increase or decrease in PSD maturation rate . One factor that might contribute to these changes in PSD maturation rate is altered growth rate of the entire NMJ secondary to changes in neuronal activity . Indeed , we observed that the rate of new PSD addition over 24 hr in BRP69/def , napTS , and sh , eag mutants showed a similar trend to PSD maturation , suggesting that overall NMJ growth rate is altered in these mutants ( control animals increase AZ number by 1 . 8 ± 0 . 1 fold over 24 hr compared to sh , eag ( 2 . 2 ± 0 . 1 fold ) , napTS ( 1 . 3 ± 0 . 04 fold ) , syt1 ( 1 . 5 ± 0 . 08 fold ) , and BRP69/def ( 1 . 2 ± 0 . 06 fold ) ; Figure 10E ) . Previous work on activity-dependent NMJ growth in Drosophila focused primarily on bouton number , not AZ addition rate , and it will be interesting to explore the molecular mechanisms mediating activity-dependent seeding of new AZs . To measure PSD maturation rate in the context of an otherwise healthy animal , we took advantage of the rab3 null mutant , where roughly half of the PSDs are apposed by presynaptic areas lacking BRP and Cac , and the remaining AZs contain higher than normal levels of these presynaptic components ( Graf et al . , 2009 ) . Rab3 AZs containing BRP and Cac have been shown to have higher Pr , whereas AZs lacking these presynaptic components are functionally silent ( Peled and Isacoff , 2011 ) . This redistribution of Pr provides an opportunity to compare PSD maturation with less influence of overall animal health and NMJ growth rate . The rab3 phenotype has been previously analyzed at the third instar stage of larval development . To characterize the dynamics of pre- and postsynaptic maturation during earlier development , we followed rab3 animals expressing Cac-GFP and GluRIIA-RFP for 48 hr starting in the early first instar larval stage . AZs populated with Cac-GFP early in development were stable over time; Cac-GFP was never lost at these AZs over 48 hr of imaging ( Figure 10—figure supplement 1E ) . However , a divergence in the rate of PSD versus Cac-GFP addition was observed after the first instar stage; new GluRIIA-containing PSDs were added to the NMJ at a rate similar to control , with total PSD number nearly doubling every 24 hr . However , the rate of addition of new Cac-GFP-containing AZs was much lower than in controls , resulting in a decrease in Cac/GluRIIA apposition in rab3 mutants over development ( Figure 10—figure supplement 1D , F ) . This is consistent with the observation that PSD growth rate was decreased in the rab3 null ( 1 . 23-fold ± 0 . 05 , 6 NMJs from six animals ) when we measured PSDs born in the late first instar or early second instar stage ( Figure 10C ) , while the percent of GluRIIA/B rings seen in the second instar stage was elevated ( 34 . 7 ± 3 . 3% of PSDs had rings , 14 NMJs from six animals ) compared to age matched controls ( 22%; Figure 10D ) . To avoid complications from the atypical dynamics of AZ apposition during rab3 NMJ development , we focused the remainder of our analysis on the early first instar stage where the age distribution of AZs is much narrower . We next examined first instar rab3 mutants to determine whether AZs with higher presynaptic activity had more mature postsynaptic receptor fields than their silent neighbors . We imaged GluRIIA-RFP and GluRIIB-GFP in rab3 1st instars and then dissected and stained for BRP . In muscle 26 of rab3 mutants , roughly 50% of first instar AZs were populated with BRP ( the ratio of GluRIIA-RFP puncta to Cac-GFP puncta was 1 . 9 ± 0 . 15 ) ; these AZs were opposed by large PSDs , many of which already showed GluRIIA/IIB segregation ( Figure 10D , Figure 10—figure supplement 1C ) . In contrast , PSDs lacking BRP were much smaller and lacked receptor segregation . All of the PSDs with GluRIIB rings were BRP-positive , suggesting that PSDs opposite AZs with reduced presynaptic release are less mature at this stage of larval development . Though the mechanism by which rab3 regulates AZ assembly remains unknown , the larger PSD size and enhanced GluRIIA/B segregation opposite BRP-positive AZs and the small size and lack of receptor segregation opposite BRP-negative AZs within the same terminal suggests that PSD maturation depends on presynaptic activity at the level of individual AZs . In the current study we used quantal imaging , super resolution SIM , and intravital imaging to examine the development of heterogeneity in evoked Pr across the AZ population at Drosophila NMJs . We first confirmed that release heterogeneity was not caused by summation of fusion events from multiple unresolvable AZs . Indeed , high Pr sites corresponded to single AZs with enhanced levels of BRP . These findings are consistent with previous observations using conventional light microscopy that indicate Pr correlates with BRP levels ( Muhammad et al . , 2015; Peled et al . , 2014; Reddy-Alla et al . , 2017 ) . By monitoring release over intervals of extensive vesicle fusion during strong stimulation , we also observed that Pr is a stable feature of each AZ . In addition , loss of the synaptic vesicle Ca2+ sensor Syt1 globally reduced Pr without altering the heterogeneous distribution of Pr across AZs , indicating that AZ-local synaptic vesicle pools with differential Ca2+ sensitivity are not likely to account for Pr heterogeneity . Since VGCC abundance , gating , and organization within the AZ are well established regulators of Pr across synapses ( Borst and Sakmann , 1996; Chen et al . , 2015; Meinrenken et al . , 2002Zito et al . , 1999; Nakamura et al . , 2015; Sheng et al . , 2012; Südhof , 2012; Wang et al . , 2008 ) , heterogeneity in presynaptic Ca2+ channel abundance was a clear candidate for the generation of Pr heterogeneity at Drosophila NMJs . Indeed , the Cac Ca2+ channel responsible for neurotransmitter release is heterogeneously distributed across the NMJ ( Fouquet et al . , 2009; Kawasaki et al . , 2004; 2000; Littleton and Ganetzky , 2000; Liu et al . , 2011; Rieckhof et al . , 2003; Smith et al . , 1996 ) . Using transgenically labeled Cac lines , we observed that Cac density at AZs is indeed strongly correlated with Pr . To directly visualize presynaptic Ca2+ influx at single AZs , we generated GCaMP fusions to the core AZ component BRP . Ca2+ influx at single AZs was highly correlated with both Cac density and Pr . The cacNT27 mutant with decreased conductance also resulted in a global reduction in Pr without disrupting heterogeneity , further confirming that Ca2+ influx regulates Pr across the range of release heterogeneity . Postsynaptically , high Pr AZs were enriched in GluRIIA-containing receptors and displayed a distinct pattern of glutamate receptor clustering . While most synapses showed GluRIIA and GluRIIB spread over the entire PSD , high Pr AZs were apposed by PSDs where GluRIIA concentrated at the center of the AZ , with GluRIIB forming a ring at the PSD periphery . Indeed , anti-glutamate receptor antibody staining of wildtype larvae lacking tagged glutamate receptors had previously identified a GluRIIB ring around the GluRIIA core in some mature third instar NMJ AZs ( Marrus et al . , 2004 ) . In addition , activity-dependent segregation of GluRIIA and a GluRIIA gating mutant has been observed at individual AZs in Drosophila ( Petzoldt et al . , 2014 ) . The correlation of Pr with GluRIIA accumulation is especially intriguing considering that this subunit has been implicated in homeostatic and activity-dependent plasticity ( Davis , 2006; Frank , 2014; Petersen et al . , 1997; Sigrist et al . , 2003 ) . By following the developmental acquisition of this postsynaptic property as a proxy for Pr from the first through third instar larval stages via intravital imaging in control and mutant backgrounds , we observed that the earliest formed AZs are the first to acquire this high Pr signature over a time course of ~3 days , and that PSD maturation rate can be modulated by changes in presynaptic activity . Similar to prior observations ( Melom et al . , 2013; Peled and Isacoff , 2011 ) , we found that most AZs at the Drosophila NMJ have a low Pr . For the current study , the AZ pool was artificially segregated into low and high release sites , with high releasing sites defined based on having a release rate greater than two standard deviations above the mean . Given that birthdate is a key predictor of glutamate receptor segregation , and by proxy Pr , we expect the AZ pool to actually reflect a continuum of Pr values based on their developmental history . However , using the two standard deviation criteria , 9 . 9% of AZs fell into the high Pr category , with an average Pr of 0 . 28 . We also observed that 9 . 7% of the AZs analyzed displayed only spontaneous release . We could detect no fusion events for either evoked or spontaneous release for another 14 . 6% of AZs that were defined by a GluRIIA-positive PSD in live imaging . Future investigation will be required to determine whether these cases represent immature AZs with extremely low evoked Pr , or distinct categories reflective of differences in AZ content . The remaining AZs that participated in evoked release had an average Pr of 0 . 05 . Ca2+ channel density and Ca2+ influx at individual AZs was a key determinant of evoked Pr heterogeneity , as Pr and the intensity of Cac channels tagged with either TdTomato or GFP displayed a strong positive correlation . Spontaneous fusion showed a much weaker correlation with both Cac density and Ca2+ influx at individual AZs , consistent with prior studies indicating spontaneous release rates are poorly correlated with external Ca2+ levels at this synapse ( Jorquera et al . , 2012; Lee et al . , 2013 ) . With synaptic vesicle fusion showing a steep non-linear dependence upon external Ca2+ with a slope of ~3–4 ( Dodge and Rahamimoff , 1967; Heidelberger et al . , 1994; Jan and Jan , 1976 ) , a robust change in Pr could occur secondary to a relatively modest increase in Ca2+ channel abundance over development . Although the number of VGCCs at a Drosophila NMJ AZ is unknown , estimates of Cac-GFP fluorescence during quantal imaging indicate a ~ 2 fold increase in channel number would be necessary to move a low Pr AZ into the high Pr category . Similar correlations between evoked Pr and Ca2+ channel abundance have been found at mammalian synapses ( Holderith et al . , 2012; Nakamura et al . , 2015; Sheng et al . , 2012 ) , suggesting this represents a common evolutionarily conserved mechanism for determining release strength at synapses . We did not test the correlation of Pr with other AZ proteins besides Cac and BRP , but it would not be surprising to see a positive correlation with the abundance of many AZ proteins based on the observation that maturation time is a key determinant for Pr . Indeed , recent studies have begun to correlate Pr with specific AZ proteins at Drosophila NMJs ( Reddy-Alla et al . , 2017 ) . We also observed that PSD size was robustly increased by 1 . 6-fold over a 24 hr period of AZ development during the early larval period in control animals . AZ maturation is likely to promote increased synaptic vesicle docking and availability , consistent with observations that correlate AZ size with either Pr or the readily releasable pool ( Han et al . , 2011; Holderith et al . , 2012; Matkovic et al . , 2013; Matz et al . , 2010; Nakamura et al . , 2015; Schikorski and Stevens , 1997; Wadel et al . , 2007 ) . We considered several models for how AZs acquire this heterogeneous nature of Pr distribution during a developmental period lasting several days . One possibility is that unique AZs gain high Pr status through a mechanism that would result in preferential accumulation of key AZ components compared to their neighbors . Given that retrograde signaling from the muscle is known to drive synaptic development at Drosophila NMJs ( Ball et al . , 2010; Berke et al . , 2013; Harris and Littleton , 2015; Keshishian and Kim , 2004; McCabe et al . , 2003; Piccioli and Littleton , 2014; Yoshihara et al . , 2005 ) , certain AZ populations might have preferential access to specific signaling factors that would alter their Pr state . Another model is that AZs compete for key presynaptic Pr regulators through an activity-dependent process . High Pr AZs might also be more mature than their low Pr neighbors , having a longer timeframe to accumulate AZ components . Given the Drosophila NMJ is constantly forming new AZs at a rapid pace during development ( Rasse et al . , 2005; Schuster et al . , 1996 ) , newly formed AZs would be less mature compared to a smaller population of ‘older’ high Pr AZs . To examine if the release heterogeneity observed at the third instar stage reflects AZ birth order over several days of development , we needed to extend the intravital imaging through a longer time period beginning in the first instar larval stage . GCaMP imaging indicated high Pr sites segregate GluRIIA and GluRIIB differently from low Pr sites , with the IIA isoform preferentially localizing at the center of PSDs apposing high Pr AZs . As such , we used developmental acquisition of this property as an indicator of high Pr sites . Although segregation of glutamate receptors may not perfectly replicate the timing of Pr acquisition during development , it is currently the best tool for estimating Pr during sequential live imaging . Based on the acquisition of GluRIIA/B segregation , the data support the hypothesis that increases in Pr reflect a time-dependent maturation process at the NMJ . The continuous addition of new AZs , which double in number during each day of development , ensures that the overall ratio of high to low Pr sites represents a low percentage as the NMJ grows ( Figure 11 ) . We further established confidence in the age-dependency model of Pr by mapping release after following NMJs intravitally for 24 hr; using this approach , we found that newly formed AZs are consistently very low Pr . Finally , we mapped Pr in the second instar stage and observed that the heterogeneity at this stage is shifted towards higher releasing sites , with a reduction in the fraction of low-releasing sites . These data indicate AZs are born with low Pr and gain pre- and postsynaptic material over 3 days on an upward trajectory toward higher Pr status ( Figure 11 ) . Is AZ age a static determinant of Pr , or can growth rate be regulated to allow faster or slower acquisition of high Pr AZs ? To answer this question , we assayed whether mutants that alter presynaptic activity influence the rate of PSD growth and GluRIIA/B segregation . In BRP69/def , syt1null , and napTS mutants with decreased presynaptic release , a significant reduction in postsynaptic maturation rate and in the percentage of PSDs with GluRIIA/IIB rings was observed . Conversely , in the shaker , eag double mutant with increased presynaptic excitability , a significantly increased rate of GluRIIB field size and a significant increase in GluRIIA/IIB rings was found compared to controls . To investigate whether differences in synapse growth and maturation rate could be seen between AZs enriched in BRP and Cac versus neighboring AZs that are deficient in these components , we imaged development in the rab3 null mutant . At the first instar stage when AZs are roughly age matched , release sites enriched in presynaptic components had developed large and mature PSDs in stark contrast with non-enriched AZs , whose PSDs appeared immature . These results indicate that PSD maturation can be influenced by presynaptic activity at the resolution of single AZs . Although how AZs are assembled during development is still being established , our data do not support a model where AZs are fully preassembled during transport and then deposited as a single ‘quantal’ entity onto the presynaptic membrane . Rather , these data support a model of seeding of AZ material that increases developmentally over time as AZs matures , consistent with previous studies of AZ development in Drosophila ( Böhme et al . , 2016; Fouquet et al . , 2009 ) . Although no evidence for rapid changes in Pr were detected in the steady-state conditions used in the current study , homeostatic plasticity is known to alter Pr over a rapid time frame ( ~10 min ) at the NMJ ( Davis and Müller , 2015; Frank , 2014; Frank et al . , 2006 ) . It will be interesting to determine if Cac abundance can change over such a rapid window , or whether the enhanced release is mediated solely through changes in Cac function and Ca2+ influx ( Müller and Davis , 2012 ) . Changes in the temporal order of Pr development could also occur secondary to altered transport or capture of AZ material . For example , the large NMJ on muscle fibers 6 and 7 displays a gradient in synaptic transmission , with terminal branch boutons often showing a larger population of higher Pr AZs ( Guerrero et al . , 2005; Peled and Isacoff , 2011 ) . If AZ material is not captured by earlier synapses along the arbor , it would be predicted to accumulate in terminal boutons , potentially allowing these AZs greater access to key components , and subsequently increasing their rate of Pr acquisition . Alternatively , the gradient of Pr along the axon could be due to terminal boutons being slightly older than the rest of the arbor . In summary , our data indicate that heterogeneity in release correlates highly with Ca2+ channel abundance and Ca2+ influx at AZs . Postsynaptically , PSDs apposed to high-releasing AZs display increased GluRIIA abundance and form segregated receptor fields , with GluRIIB forming a ring around a central core of GluRIIA . Release sites accumulate these high Pr markers during a synapse maturation process in which newly formed AZs are consistently low Pr , with AZs gaining signatures of high releasing sites over several days . Finally , mutations that increase or decrease presynaptic activity result in faster or slower rates of PSD maturation , respectively . These data add to our understanding of the molecular and developmental features associated with high versus low Pr AZs . Flies were cultured at 25°C on standard medium . Actively crawling third instar male and female larvae dwelling on top of the food were used for experiments unless otherwise noted . The following strains were used: UAS-myrGCaMP6s , UAS-GCaMP6m-BRPshort , pBid-lexAop-myrGcaMP6s , UAS-myrjRGECO; Elav–GAL4 , Mef2–GAL4 , UAS-CacGFP ( provided by Richard Ordway ) ; UAS-CacTdTomato ( provided by Richard Ordway ) ; GluRIIA-RFP inserted onto chromosome III under the control of its endogenous promoter ( provided by Stephan Sigrist ) , GluRIIB-GFP inserted onto chromosome III under the control of its endogenous promoter ( provided by Stephan Sigrist ) ; napTS and shaker120b , eag1 ( provided by Barry Ganetzky ) ; BRP69 ( provided by Stephan Sigrist ) ; BRPdef; rab3rup ( provided by Ethan Graf ) and 44H10-LexAp65 ( provided by Gerald Rubin ) . syt1 null mutants were generated by crossing syt1N13 , an intragenic syt1 deficiency ( Littleton et al . , 1994 ) , with syt1AD4 , which truncates Syt1 before the transmembrane domain ( DiAntonio and Schwarz , 1994 ) . brp null mutants were generated by crossing brp69 , a truncation mutant , to a genomic deficiency brpdef . The fluorescent Ca2+ sensor GCaMP6s was tethered to the plasma membrane with an N-terminal myristoylation ( myr ) sequence . The first 90 amino acids of Src64b , containing a myristoylation target sequence , were subcloned into pBID-UASc with EcoRI and BglII ( creating pBID-UASc-myr ) . GCaMP6s cDNA ( Addgene plasmid 40753 ) was cloned into pBID-UASc-myr with BglII and XbaI . To generate the UAS-GCaMP6m-Brp-short line , GCaMP6m ( Addgene plasmid 40754 ) cDNA and Brp-short ( gift from Dr . Tobias Rasse ) were PCR amplified and double digested with EcoRI/BglII and BglII/XbaI , respectively . The two cDNA fragments were ligated and the product was used to PCR amplify the fused GCaMP6m-Brp-short cDNA . The PCR product was inserted into the vector backbone pBID-UASc after digestion with EcoRI and XbaI to generate the final plasmid pBID-UASc-GCaMP6m-Brp-short . To create UAS-myrjRGECO , the vector backbone pBID-UASc-myr was digested with BglII and XbaI . jRGECO sequence was amplified from plasmid pGP-CMV-NES-jRGECO1a ( gift from Dr . Douglas Kim , Addgene plasmid #61563 ) . The digested backbone and insert were fused according to the Gibson assembly protocol using NEBuilder HighFidelity DNA Assembly Cloning Kit ( E5520 ) . To generate pBid-lexAop-myrGcaMP6s , myrGCaMP6s was amplified by PCR and inserted into pBiD-lexAop-DSCP ( gift from Brian McCabe ) between NotI and XbaI sites . All transgenic Drosophila strains were generated by BestGene . Wandering third instar larvae were dissected in Ca2+-free HL3 solution and fixed in 4% paraformaldehyde for 10 min , washed in PBT ( PBS plus 0 . 1% Triton X-100 ) and blocked in 5% normal goat serum ( NGS ) and 5% BSA in PBT for 15 min . Samples were incubated overnight with anti-BRP ( NC82 , 1:200 ) from the Developmental Studies Hybridoma Bank ( DSHB Cat# NC82 , RRID:AB_2314866 ) , washed for 1 hr in PBS and then incubated for 2–3 hr with Alexa Fluor 607-conjugated anti-mouse IgG at 1:1000 ( Invitrogen , #A21237 , RRID:AB_1500743 ) . Confocal images were obtained on a Zeiss Axio Imager two equipped with a spinning-disk confocal head ( CSU-X1; Yokagawa ) and ImagEM X2 EM-CCD camera ( Hammamatsu ) . An Olympus LUMFL N 60X objective with a 1 . 10 NA was used to acquire GCaMP6s imaging data at 7 to 8 Hz . A Zeiss pan-APOCHROMAT 63X objective with 1 . 40 NA was used for imaging stained or live animals . third instar larvae were dissected in Ca2+-free HL3 containing 20 mM MgCl2 . After dissection , preparations were maintained in HL3 with 20 mM MgCl2 and 1 . 3 mM Ca2+ for 5 min . To stimulate the NMJ , motor nerves were cut close to the ventral ganglion and sucked into a pipette . Single pulses of current were delivered every one second for myr-jRGECO mapping or every three seconds for GCaMP6s mapping with an AMPI Master-8 stimulator using a stimulus strength just above the threshold for evoking EJPs . A 3D image stack was taken before the GCaMP imaging session to generate a full map of GluRIIA or Cac channel distribution . Later , single focal planes were imaged continuously for 4–5 min to collect GCaMP signals . Volocity 3D Image Analysis software ( PerkinElmer ) was used to analyze images . All images were Gaussian filtered ( fine ) to reduce noise and a movement-correction algorithm was applied . To enhance identification of myrGCaMP6 flashes , background myrGCaMP fluorescence was subtracted by creating a composite stack of 5–6 images during intervals when no synaptic release occurred . To identify the position of GluRIIA receptors and corresponding Ca2+ events , a 3D stack image of GluRIIA was merged to create a single plane . AZ position was identified using the ‘find spot’ algorithm in Volocity 3 . 2 software that detects fluorescent peaks . ROIs with identical 5-pixel size ( 0 . 138 µm/pixel ) were automatically generated by the software from identified GluRIIA spots . All GCaMP flashes were detected using the intensity threshold tool and assigned to specific ROIs based on proximity of their centroids . The time and location of Ca2+ events were imported into Excel or Matlab for further analysis . The number of observed GCaMP events per AZ was divided by the number of delivered stimuli to calculate AZ Pr . Analysis of Cac , BRP , GluRIIA or GluRIIB intensities was performed similarly , identifying AZ fluorescence peaks and defining three pixel square ROIs around each peak to calculate average fluorescence . Average AZ fluorescence intensities of 3-pixel square ROIs was also used for correlation analysis . SIM microscopy was performed on an Applied Precision DeltaVision-OMX BLAZE-3D-Structural Illumination Microscope equipped with 60X , 1 . 4 NA oil objective and 3 sCMOS cameras . 3D-SIM images were obtained with 125 nm z-steps . Samples were illuminated by three central diffraction orders with 488 , 562 , and 640 nm lasers . For initial identification of specific NMJs , larvae were imaged in conventional confocal mode using a 20X oil objective . The positions of NMJ were marked and recorded to provide transition between objectives . A ZEISS LSM 800 microscope with Airyscan was also used to image anesthetized animals . Fluorescence was detected by a concentrically-arranged hexagonal detector array consisting of 32 single detector elements . Larvae were anesthetized with SUPRANE ( desflurane , USP ) from Amerinet Choice ( Zhang et al . , 2010 ) . Larvae were incubated in a petri dish with a small paper towel containing Suprane for 1–2 min in a fume hood . Anesthetized larvae were positioned ventral side up on a glass slide between spacers made by transparent tape , which prevented extreme compression of the larvae . Different size spacers were required for the various larval stages . Larvae were covered with a thin film of halocarbon oil and then with a cover glass . NMJ synapses on muscle 26 in hemi-segment 2 or three were imaged . After an imaging session , larvae were placed in numbered chambers with food in a 25°C incubator . The same data acquisition settings where used to visualize NMJs at different larval stages . Larvae were imaged with either 6 , 24 and 36 hr intervals for one data set ( Figure 8A–C ) , or for 24 hr intervals for the remaining datasets . To keep the size consistent between different time periods , images of the corresponding NMJ area at younger stages were cut ( dashed areas in figures ) and placed onto a black background . This presentation generated a similar orientation of the different size NMJs for easier comparison for Figure 8 , Figure 9 and Figure 8—figure supplement 1 . Ionomycin ( Sigma Aldrich ) was dissolved in ethanol to make a 10 mM stock solution and was stored at 4°C . Ionomycin was used at a working concentration of 200 nM dissolved in HL3 with 1 . 3 mM Ca2+; this solution was applied to dissected preparations and NMJs were imaged one minute after application . Statistical analysis was performed with GraphPad Prism using one-way ANOVA followed by Dunnett’s Multiple Comparisons test for comparison of samples within an experimental group , or Student’s t-test for comparing two groups . Asterisks denote p values of: p≤0 . 05; p≤0 . 01; and **p≤0 . 001 . All histograms and measurements are shown as mean ± SEM . Pearson coefficient of correlation was calculated in GraphPad Prism using the following parameters: - two-tailed P value and 95% confidence interval .	To send a message to its neighbor , a neuron releases chemicals called neurotransmitters into the gap – or synapse – between them . The neurotransmitter molecules bind to proteins on the receiver neuron called receptors . But what causes the sender neuron to release neurotransmitter in the first place ? The process starts when an electrical impulse called an action potential arrives at the sender cell . Its arrival causes channels in the membrane of the sender neuron to open , so that calcium ions flood into the cell . The calcium ions interact with packages of neurotransmitter molecules , known as synaptic vesicles . This causes some of the vesicles to empty their contents into the synapse . But this process is not particularly reliable . Only a small fraction of action potentials cause vesicles to fuse with the synaptic membrane . How likely this is to occur varies greatly between neurons , and even between synapses formed by the same neuron . Synapses that are likely to release neurotransmitter are said to be strong . They are good at passing messages from the sender neuron to the receiver . Synapses with a low probability of release are said to be weak . But what exactly differs between strong and weak synapses ? Akbergenova et al . studied synapses between motor neurons and muscle cells in the fruit fly Drosophila . Each motor neuron forms several hundred synapses . Some of these synapses are 50 times more likely to release neurotransmitter than others . Using calcium imaging and genetics , Akbergenova et al . showed that sender cells at strong synapses have more calcium channels than sender cells at weak synapses . The subtypes and arrangement of receptor proteins also differ between the receiver neurons of strong versus weak synapses . Finally , studies in larvae revealed that newly formed synapses all start out weak and then gradually become stronger . How fast this strengthening occurs depends on how active the neuron at the synapse is . This study has shown , in unprecedented detail , key molecular factors that make some fruit fly synapses more likely to release neurotransmitter than others . Many proteins at synapses of mammals resemble those at fruit fly synapses . This means that similar factors may also explain differences in synaptic strength in the mammalian brain . Changes in the strength of synapses underlie the ability to learn . Furthermore , many neurological and psychiatric disorders result from disruption of synapses . Understanding the molecular basis of synapses will thus provide clues to the origins of certain brain diseases .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2018	Characterization of developmental and molecular factors underlying release heterogeneity at Drosophila synapses
Fast spiking , parvalbumin ( PV ) expressing hippocampal interneurons are classified into basket , axo-axonic ( chandelier ) , and bistratified cells . These cell classes play key roles in regulating local circuit operations and rhythmogenesis by releasing GABA in precise temporal patterns onto distinct domains of principal cells . In this study , we show that each of the three major PV cell classes further splits into functionally distinct sub-classes during fast network events in vivo . During the slower ( <10 Hz ) theta oscillations , each cell class exhibited its own characteristic , relatively uniform firing behavior . However , during faster ( >90 Hz ) oscillations , within-class differences in PV interneuron discharges emerged , which segregated along specific features of dendritic structure or somatic location . Functional divergence of PV sub-classes during fast but not slow network oscillations effectively doubles the repertoire of spatio-temporal patterns of GABA release available for rapid circuit operations . Brain state-dependent network oscillations occurring at various frequencies provide multiscale temporal windows for the precise timing of neuronal discharges and the temporal binding of spatially distributed cell populations ( Singer , 1993; Buzsaki and Chrobak , 1995 ) . The spatio-temporal control of principal cell assemblies during endogenous brain rhythms is orchestrated by GABAergic interneurons ( Soltesz , 2006; Somogyi et al . , 2014 ) . Versatility and precision in the GABAergic coordination of principal cell excitability arise from the diversity of interneurons that enables the selective timing of GABAergic inputs to specific spatial domains of principal cells . Within the CA1 region of the rodent hippocampus , 21 distinct interneuronal classes are currently recognized , each possessing unique input–output connectivity patterns , developmental profiles , and characteristic molecular and electrophysiological properties ( Klausberger and Somogyi , 2008; Bezaire and Soltesz , 2013 ) . The distinct features of interneuronal classes enable these cells to be selectively recruited to entrain principal cell populations into different oscillatory patterns of activity , including the theta , gamma , epsilon , and ripple waves that are associated with specific behaviors ( O'Keefe and Nadel , 1978; O'Keefe and Recce , 1993; Soltesz and Deschenes , 1993; Buzsaki and Chrobak , 1995; Lubenov and Siapas , 2009; Colgin and Moser , 2010; Maier et al . , 2011 ) . The parvalbumin expressing ( PV ) interneurons , in particular , have been identified to serve major mechanistic roles in a variety of network functions , including local circuit operations , learning and memory , rhythmogenesis , sensory processing , and critical period plasticity ( Pouille and Scanziani , 2004; Buzsaki and Wang , 2012; Kuhlman et al . , 2013; Siegle and Wilson , 2014 ) . Furthermore , PV cells have also been linked to a number of neurological and psychiatric disorders including epilepsy , schizophrenia , and autism ( Lewis et al . , 2005; Ogiwara et al . , 2007; Gibson et al . , 2009; Armstrong and Soltesz , 2012; Verret et al . , 2012; Trouche et al . , 2013 ) . A key property of PV cells is speed; indeed , these interneurons can fire fast action potentials at high frequencies with little accommodation ( Kawaguchi , 1995; Buhl et al . , 1996 ) , possess fast membrane time constants , and release GABA with high temporal precision following the arrival of presynaptic action potentials at the release sites , enabling PV cells to serve as the rapid signaling elements of interneuronal–principal cell networks ( Bartos et al . , 2001; Jonas et al . , 2004; Hu et al . , 2010 ) . Importantly , PV cells are known to sharply segregate into three major classes that are differentiated based on the distinct post-synaptic domains that they innervate ( Klausberger and Somogyi , 2008 ) . Axo-axonic ( chandelier ) cells ( AACs ) make synaptic contacts exclusively on the axon initial segments of principal cells ( Somogyi , 1977; Somogyi et al . , 1983 ) . In contrast , basket cells ( BCs ) innervate the somata and proximal dendrites of the target cells , whereas the bistratified ( Bistrat ) cells provide GABAergic inputs to the basal and apical dendritic domains ( Buhl et al . , 1994 ) . In spite of the key functional importance of these specialized rapid signaling elements within the network , our understanding of the cellular sources of GABA from PV cells during fast network oscillations in vivo without anesthesia remains limited . For example , although AACs have received a lot of attention as playing major roles in the synchronization of principal cell populations ( Howard et al . , 2005; Szabadics et al . , 2006; Woodruff et al . , 2010 ) , to date only a single AAC has been reported from the CA1 region under non-anesthetized in vivo conditions ( Viney et al . , 2013 ) . Similarly , in spite of a series of recent studies examining anatomically and immunocytochemically identified ( Lapray et al . , 2012; Varga et al . , 2012; Viney et al . , 2013; Katona et al . , 2014 ) or optogenetically ‘tagged’ interneurons in vivo without anesthesia ( Royer et al . , 2012; Kvitsiani et al . , 2013 ) , there are no data on the preferential phase-specific discharges of AACs and Bistrat cells during high-frequency ( >90 Hz ) oscillations in vivo in awake animals . In this study , we report the brain state-specific action potential discharge patterns of AACs , BCs , and Bistrat cells during fast network oscillations in the CA1 region of the hippocampus from anesthesia-free head-fixed mice that were running or resting on a spherical treadmill ( Dombeck et al . , 2007; Varga et al . , 2012 ) . The cells were recorded using juxtacellular techniques ( Pinault , 1996 ) that enabled the non-invasive recording ( i . e . , without disturbance of the intracellular milieu of the recorded neuron ) , high-fidelity labeling , and rigorous post-hoc identification of single cells . In addition to providing new insights into the precise patterns of discharges by AACs and Bistrat cells during brain state-dependent rhythms of the hippocampus , we report that during fast , but not slow , network oscillations each of the three major PV interneuron class splits into two functionally distinct sub-classes that segregate according to characteristic morphological properties . These findings increase our understanding of the mechanisms of hippocampal oscillations and describe a unique form of functional fission of previously established PV cell classes that effectively doubles the available repertoire of spatio-temporal patterns of GABA release during rapid circuit operations . Fast spiking PV interneurons ( total: n = 27; Figure 1A ) were post-hoc identified primarily based on their axonal characteristics as BCs ( n = 12; 7 were included in Varga et al . , 2012 ) , Bistrat cells ( n = 8 ) , or AACs ( n = 7 ) . A summary of the n = 27 PV cells , including the immunocytochemical tests and EM verification used for each cell , is presented in Table 1 . Axons of the BCs were largely confined to the stratum pyramidale , where they formed synapses on the somata and proximal dendrites of PCs . Electron microscopic verification of somatic axonal targets of BCs was performed in n = 4 cells . In contrast , the dendritically projecting Bistrat cells had the majority of their axons in the stratum radiatum and some in the stratum oriens . Since the axons of Bistrat cells avoided the stratum pyramidale , the axons provided a feature of Bistrat cells that clearly distinguished them from the other two cell classes . The AACs projected to the stratum pyramidale , where they formed synapses exclusively on the axon initial segments . The targets of n = 5 AACs were identified at the EM level ( number of boutons examined: 11 , 10 , 10 , 8 , 4 , all synapsing on axon initial segments ) . Out of the two AACs not tested at the EM level , one AAC was tested for ankyrinG , a marker for axon initial segments ( Jenkins and Bennett , 2001 ) . The axon terminals of the tested AACs were found to be in close juxtaposition with ankyrinG immunopositive profiles ( not shown ) . The remaining AAC was tested for the expression of special AT-rich sequence binding protein 1 ( SATB1 ) , because SATB1 is expressed in BCs and Bistrat cells but not in AACs ( Viney et al . , 2013 ) . As expected , the AAC was found to be immunonegative for SATB1 ( not shown ) . 10 . 7554/eLife . 04006 . 003Figure 1 . PV cells discharge at similar rates during running-associated theta oscillations but show between-class differences in theta modulation and preferred phases of firing . ( A ) The 3 classes of PV interneuron innervate distinct post-synaptic domains of CA1 PCs ( black ) . AACs exclusively target the axon initial segments ( AAC , blue ) ; BCs innervate the perisomatic region ( somata and proximal dendrites ) ( BC , red ) ; Bistrat cells synapse on the apical and basal dendrites ( Bistrat , green ) . ( B ) Firing frequencies during running are high and not significantly different between the 3 PV cell classes . ( C ) Mean preferred phase vs modulation strength of PV cell discharges . Blue triangles: AACs , red circles: BCs , green squares: Bistrat cells . A thin gray line representing a theta cycle is shown to illustrate the theta phases . ( D–F ) Example traces of LFP , spikes , and filtered theta oscillation of an AAC ( D ) , BC ( E ) , and Bistrat cell ( F ) . Scales: LFPs = 0 . 5 mV; single cell unit firing = 1 mV; theta = 0 . 4 mV . DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 00310 . 7554/eLife . 04006 . 004Table 1 . List of all n = 27 in vivo juxtacellularly filled PV interneurons and their identification based on immunocytochemistry and electron microscopy ( EM ) DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 004Cell ID#Cell typeParvalbuminSomatostatinmGluR1aSATB1Bouton targetsAnkyrinG verificationEM verification123011c3C-BC+n . t . n . t . n . t . n . t . Soma/proximal dendritecvi30C-BC+n . t . n . t . n . t . n . t . Soma/proximal dendritecvi33C-BC+n . t . n . t . n . t . n . t . n . t . cvi35C-BC+n . t . n . t . n . t . n . t . n . t . cvi55C-BCn . t . n . t . n . t . n . t . n . t . Soma/proximal dendritecvi65C-BC+n . t . n . t . n . t . n . t . n . t . cvi75C-BC+n . t . n . t . n . t . n . t . n . t . cvi151C-BCn . t . n . t . n . t . n . t . n . t . n . t . cvi251C-BCn . t . n . t . n . t . n . t . n . t . n . t . cvi240bH-BC+n . t . n . t . n . t . n . t . Soma/proximal dendrite042911c5H-BC+n . t . n . t . n . t . n . t . n . t . gs0920H-BC+n . t . n . t . n . t . n . t . n . t . gs012913O-Bistrat++n . t . n . t . n . t . n . t . gs022713O-Bistrat++n . t . n . t . n . t . n . t . cvi017O-Bistratn . t . +n . t . n . t . n . t . n . t . cvi255O-Bistrat−+n . t . n . t . n . t . n . t . cvi312O-Bistratn . t . +−n . t . n . t . n . t . cvi270C-Bistrat++n . t . n . t . n . t . n . t . imi069C-Bistratn . t . +n . t . n . t . n . t . n . t . cvi190C-Bistrat−+n . t . n . t . n . t . n . t . 090311c3E-AAC+n . t . n . t . n . t . n . t . AIScvi059E-AACn . t . n . t . n . t . n . t . n . t . AIScvi153C-AACn . t . n . t . n . t . n . t . n . t . AIScvi258C-AAC+n . t . n . t . n . t . n . t . AIScvi315C-AACn . t . n . t . n . t . −n . t . n . t . imi075C-AAC+n . t . n . t . n . t . AISn . t07082014cs6C-AACn . t . n . t . n . t . n . t . AISAISAbbreviations: n . t . : not tested; AIS: axon initial segment . In addition , PV or somatostatin ( SOM ) immunoreactivity was also tested . Specifically , PV immunopositivity was confirmed in 8 out of the 8 tested BCs , 3 out of the 5 tested Bistrat cells , and 3 out of the 3 tested AACs . SOM immunopositivity was verified in all 8 out of the 8 tested Bistrat cells , which is important since SOM expression distinguishes Bistrat cells from BCs and AACs ( Klausberger and Somogyi , 2008 ) . Note that Bistrat cells have been observed to exhibit generally weaker PV immunoreactivity compared to BCs ( Ferraguti et al . , 2004 ) , in agreement with recent reports indicating often low PV immunopositivity in cells with concurrently active PV and SOM promoters ( Fenno et al . , 2014 ) . Because the nomenclature of the various fast oscillations is not always consistent across various studies , it is important to clearly state how we used the various terms relating to fast rhythms . Specifically , we examined PV cell discharges during four major oscillations that appeared either when the animal was running or resting , and we used these two distinct behavioral states that could be objectively distinguished in our head-fixed paradigm as an important factor in classifying oscillations . Accordingly , theta ( 5 Hz–10 Hz ) , gamma ( 25 Hz–90 Hz ) , and epsilon ( 90 Hz–130 Hz ) oscillations were studied during running , whereas ripples ( 90 Hz–200 Hz ) occurred during rest . Regarding the running-associated high-frequency oscillations , a unique frequency band has been recently distinguished called epsilon ( Freeman , 2007; Belluscio et al . , 2012; Bieri et al . , 2014; Schomburg et al . , 2014 ) . Epsilon is distinct from other high frequency oscillations in terms of generation mechanisms and location . Specifically , epsilon oscillations reflect network events generated locally in CA1 ( Schomburg et al . , 2014 ) , and most pyramidal cells ( PCs ) show strong phase locking to the trough of the epsilon oscillations , indicating large synchrony of PC discharges during these events ( Belluscio et al . , 2012; Schomburg et al . , 2014 ) . Another important point regarding the nomenclature of fast rhythms is that our running-associated epsilon oscillations are similar , but not identical , to the high gamma oscillations recorded during theta in other studies ( Csicsvari et al . , 1999a; Canolty et al . , 2006; Colgin et al . , 2009 ) , because these latter studies employed a generally wider frequency band for detecting high gamma oscillations compared to our filter setting for epsilon . Regarding the fast rhythms that occur during rest , the 90 Hz–200 Hz fast oscillations are sometimes differentiated into ‘high’ ( or ‘fast’ ) gamma ( 90–140 Hz ) and ripple ( 140–200 Hz ) oscillations ( Csicsvari et al . , 1999b ) . However , because the non-theta-associated 90 Hz–140 Hz and 140 Hz–200 Hz fast oscillations during rest can appear under similar behavioral conditions , share similar generation mechanism , and differ mostly in the strength of incoming CA3 input ( Sullivan et al . , 2011 ) , for the purposes of the current study we considered them together under the collective term ‘ripples’ ( 90 Hz–200 Hz ) . This approach avoids the sorting of events according to a sharp 140 Hz boundary into 90–140 Hz or 140 Hz–200 Hz oscillations and still allows the comparative study of the firing of interneurons during transient events that fall towards the lower- vs higher-end of the continuum of 90 Hz–200 Hz oscillations ( see below ) . Most cells in our sample could be analyzed for running-associated theta , gamma , epsilon , and rest-associated ripple oscillations , with the exception of n = 2 AACs ( cells imi075 and 07082014cs6 in Table 1 ) that yielded data only for ripples , but not for theta , gamma , or epsilon oscillations due to the presence of only brief periods of running and theta waves . We first examined the action potential firing of BCs , Bistrat cells , and AACs during running-associated theta oscillations . The three distinct cell classes discharged at comparable frequencies during theta ( Figure 1B , D–F; AAC: 27 . 1 ± 9 . 1 Hz , n = 5; BC: 28 . 64 ± 9 . 4 Hz , n = 12; Bistrat: 34 . 03 ± 11 . 9 Hz , n = 8; p = 0 . 84 , one way ANOVA ) . While every cell showed theta modulation of its firing ( p < 0 . 001 for all cells , Rayleigh test ) , the strength of theta modulation differed between the three cell classes . Specifically , AACs showed the most prominent theta modulation , followed by the intermediate level of modulation of the BCs , whereas Bistrat cells were only weakly modulated ( Figure 1D–F ) . Statistical analysis of the strength of theta modulation of action potential discharges revealed significant differences between the three cell classes ( modulation strength , from the highest to low: AACs: r = 0 . 44 ± 0 . 1; BCs: r = 0 . 25 ± 0 . 1; Bistrat cells: r = 0 . 17 ± 0 . 05; p < 0 . 01; Kruskal–Wallis test; Figure 1C ) . The three cell classes also displayed preferential firing at distinct phases of the theta waves . The AACs preferred the middle of the descending phase ( 251° ± 17°; 0° is considered to be the trough , 180° the peak of the individual oscillatory cycle ) . Thus , AACs did not preferentially fire close to the theta peak during running , as they were reported to do so under anesthesia ( Klausberger et al . , 2003 ) . The preferential discharges of the AACs during the theta cycle were followed by the BCs that fired on average during the late descending phase ( 310° ± 23° ) , while Bistrat cells discharged near the trough ( 0° ± 17° ) , around the time that the mean preferential discharges of most PCs take place ( Mizuseki et al . , 2011 ) . These data suggest that the different PV interneuron classes deliver GABA in a sequential manner along the long ( axon initial segment–somato-dendritic ) axis of the CA1 PCs , and that the bulk of this spatio-temporally organized inhibition sweeps through the PC populations from the axon initial segment to the dendrites within about 40 ms during running-associated theta ( calculated for an average , 8 Hz theta rhythm ) . The BC , AAC , and Bistrat cell groups occupied different parts of the phase vs modulation strength plot ( Figure 1C ) , with the earliest firing cells ( the AACs ) also being the strongest modulated , while Bistrat cells fired last during the theta cycle and with the least amount of theta modulation . As illustrated in Figure 1C for the AACs , BCs , and Bistrat cells , there was a high degree of within-class uniformity in how members of these three cell groups discharged during theta oscillations . On the other hand , it should also be noted that even though the AACs , BCs , and Bistrat cells comprised statistically different groups in terms of their average strength of theta modulation and phase of preferential firing , the cell clusters were close to each other in the phase vs modulation strength plot in Figure 1C . Therefore , extracellularly recorded fast spiking single units ( e . g . , recorded with tetrodes or silicon probes ) without direct anatomical identification cannot be unambiguously classified as belonging to one of these three cell classes based on the theta-related firing alone , even if the unit is identified as expressing PV ( e . g . , with optogenetic tagging; [Royer et al . , 2012] ) . Gamma oscillations are thought to play important roles in information transfer between brain areas ( Engel and Singer , 2001; Womelsdorf et al . , 2007; Sirota et al . , 2008; Atallah and Scanziani , 2009 ) , with the synchronous gamma-frequency firing in groups of spatially distinct cells hypothesized to contribute to the ‘binding’ of information in downstream neuronal groups ( Engel and Singer , 2001 ) . While the precise roles of gamma rhythm in information processing are not fully understood , there is an overall agreement that perisomatic inhibition is mechanistically important in the generation of gamma-frequency oscillations ( Buzsaki and Wang , 2012 ) . Gamma oscillations appear during the falling phase of the running-associated theta rhythm ( see peak gamma activity centered around 60 Hz in Figure 2A ) . Interestingly , although fast spiking PV cells are important in gamma oscillations ( Buzsaki and Wang , 2012 ) , not all PV cells in our recordings fired in a gamma-modulated manner ( number of cells significantly modulated: BCs: 9/12; Bistrat cells: 6/8; AACs: 4/5 ) . Furthermore , the modulation strength was only moderate ( in agreement with Tukker et al . , 2007; Varga et al . , 2012 ) , and there was no difference between the perisomatically ( BCs , AACs ) vs dendritically ( Bistrat cells ) projecting PV interneurons in terms of the strength of modulation ( BC: r = 0 . 18 ± 0 . 07; Bistrat cells: r = 0 . 18 ± 0 . 05; AAC: r = 0 . 17 ± 0 . 08; p = 0 . 79; Kruskal–Wallis test ) . BCs preferentially fired during the middle ascending phase of the individual gamma waves ( 110 ± 18° ) , significantly earlier than the other two cell classes ( p < 0 . 01; Watson–Williams test ) that discharged on average during the late ascending phase ( Bistrat cells: 148 ± 23° ) or close to the peak of the gamma oscillations ( AACs: 181 ± 15 . 4° ) . These data indicate that the majority of cells in each PV class was moderately modulated by gamma-frequency oscillations , and that the BCs preferentially fired earlier during the individual gamma waves compared to either the Bistrat cells or the AACs . 10 . 7554/eLife . 04006 . 005Figure 2 . PV cell discharges during high-frequency ( >90 Hz ) oscillations associated with running or resting . ( A–C ) Running-associated , theta-nested epsilon oscillations . ( A ) Averaged time–frequency plot triggered by theta troughs during running . A thin gray line representing a theta cycle is overlaid on the plots in panels A and B to illustrate the theta phases . Asterisk indicates the epsilon oscillations occurring during the late descending phase of theta . ( B ) Average firing probabilities of the 3 PV cell classes during theta oscillations . Note that the peak firing probability of BCs and Bistrat cells overlaps with the occurrence of epsilon oscillatory epochs ( asterisk in panel A ) . ( C ) Preferred phase of firing and strength of modulation of discharges during epsilon oscillations ( individual cells ) . A thin gray line representing half an epsilon cycle is overlaid on the plot to illustrate the epsilon phases . ( D–F ) Rest-associated ripples . ( D ) Time–frequency plot of detected ripples ( 90–200 Hz ) during resting . ( E ) The mean firing probabilities of PV cell classes before , during , and after ripples ( error bars: S . D . ) . ( F ) Preferred phase of firing and strength of modulation of discharges during ripple oscillations ( individual cells ) . A thin gray line representing half a ripple cycle is overlaid on the plot to illustrate the ripple phases . Blue triangles: AACs , red circles: BCs , green squares: Bistrat cells . DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 005 Theta oscillations themselves are relatively slow , but high-frequency ( >90 Hz ) oscillations are known to be also present during the theta rhythm ( Canolty et al . , 2006; Colgin et al . , 2009; Belluscio et al . , 2012 ) . The epsilon oscillations occurred during the second half of the falling phase of the running-associated theta waves ( asterisk in Figure 2A ) . From the cell class-specific theta-related firing of PV interneurons described above and the relative timing of epsilon oscillations during theta , we predicted that AACs were unlikely to be strongly modulated by these running-associated high-frequency oscillations . Specifically , as illustrated in Figure 2B ( re-plotted from the data in Figure 1C to emphasize the distribution of firing probabilities during theta ) , the peak firing of AACs took place slightly before the peak of the epsilon oscillations ( asterisk in Figure 2A ) . Indeed , AACs were either not epsilon modulated ( n = 3 ) or they exhibited only weak epsilon modulation ( n = 2; blue triangles in Figure 2C; r = 0 . 19 and 0 . 12; y-axis in Figure 2C ) . In contrast to AACs , the firing probabilities of every BC and Bistrat cell showed significant epsilon modulation ( Figure 2C ) . On average , the modulation strength of the BCs was higher than that of the Bistrat cells ( BC: r = 0 . 38 ± 0 . 09 , Bistrat: r = 0 . 26 ± 0 . 07; p < 0 . 01 , Mann–Whitney test ) . These data indicate a more precise epsilon-related firing activity of BCs compared to Bistrat cells ( and AACs ) , which likely contributes to the regulation of PC discharges during epsilon epochs ( Belluscio et al . , 2012; Schomburg et al . , 2012 ) . PV interneurons were not only modulated by epsilon , but they also showed differential phase preferential firing with respect to the epsilon waves ( x-axis in Figure 2C; as with the theta waves , 0° is the trough , 180° is the peak of the individual epsilon oscillatory cycle ) . The BCs on average fired significantly earlier during the epsilon waves compared to the Bistrat cells ( BCs: 66 ± 17° , n = 12; Bistrat cells: 126 ± 14° , n = 8; p < 0 . 001; Watson–Williams test; 1–1 . 5 ms delay ) ; the 2 significantly epsilon-modulated AACs preferentially fired at phases similar to Bistrat cells ( 137° and 108°; Figure 2C ) . Taken together , these data show that the firing of PV interneurons segregates according to the three established major PV cell classes during epsilon oscillations that take place when the animal is running . High-frequency ( >90 Hz ) oscillations are not confined to running-associated epsilon waves but are well known to be present also during rest . In fact , the rest-associated high frequency oscillations have at least an order of magnitude higher increase in power than the running-associated epsilon oscillations ( compare the heat-colored , Z-scored power above 90 Hz in Figure 2A vs Figure 2D ) . The rest-associated high-frequency oscillations typically appear in the 90 Hz–200 Hz frequency range ( see Figure 2D ) ( Csicsvari et al . , 1999a; Klausberger et al . , 2003 ) and are collectively referred to as ripples in this paper ( see above ) . Ripples , which are known to be associated with sharp-waves in the CA1 area ( O'Keefe and Nadel , 1978; Buzsaki et al . , 1983 ) , represent one of the fastest ( up to 200 Hz , lasting approximately 40 ms–150 ms; for an example , see Figure 3 ) and most synchronized episodic electrographic events in the normal brain ( Buzsaki , 1986 ) and are crucial for memory consolidation by replaying the encoded ensemble firing patterns in a time-compressed manner ( Wilson and McNaughton , 1994; Foster and Wilson , 2006 ) . 10 . 7554/eLife . 04006 . 006Figure 3 . Bistrat cells , in contrast to BCs , differentially fire during faster and slower ripples . ( A and B ) Time–frequency representation ( top ) of example traces ( LFP ) with two high-frequency epochs ( highlighted in red ) . Note that spikes of the recorded Bistrat cell ( A ) occurred only during the first , higher frequency oscillatory event , but remained silent during the second , lower frequency one . However , the BC ( B ) was highly active during both the higher and lower frequency epoch . ( C and D ) Frequency distribution of ripple events with ( gray bars ) or without ( black bars ) Bistrat cell ( C ) or BC ( D ) spiking ( 10 Hz binned ) . Data are pooled from all recorded Bistrat cells ( n = 8 , C ) and BCs ( n = 12 , D ) . Note that Bistrat cells selectively fired during the higher frequency events but BCs discharged during virtually all ripple events . Inset in ( D ) illustrates that the high-frequency events showed a continuum of core frequencies ( n = 16 animals; note that mean and standard error are plotted in the inset to facilitate comparison with Figure 1B in Sullivan et al . , 2011 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 006 How do the three classes of fast spiking interneurons discharge during ripples in non-anesthetized animals ? Although the ripple-related firing rates of BCs ( Lapray et al . , 2012 ) , Bistrat cells ( Katona et al . , 2014 ) , and a single AAC ( Viney et al . , 2013 ) have been reported in recent studies from the CA1 region of non-anesthetized animals , we re-examined the firing of our PV interneurons during ripple events in order to gain a better understanding of the relative firing rates , strength of modulation , and the phase-preferential firing of these three cell classes . In agreement with previous reports ( Lapray et al . , 2012; Katona et al . , 2014 ) , BCs discharged at a high rate ( 77 ± 28 Hz ) during ripple events , dramatically increasing their discharge probability ( Figure 2E ) . Bistrat cells also significantly increased their discharge probabilities during ripples , but their firing frequency was about half of that of the BCs during the ripple events ( 36 ± 16 Hz , p < 0 . 01 with respect to BCs , Mann–Whitney; Figure 2E ) . In contrast to BCs and Bistrat cells , the AACs on average , as a group ( n = 7 ) , showed no significant changes in firing rates during ripple events ( blue line in Figure 2E ) . Further analysis showed that the AACs fired the majority ( 68 . 3 ± 17 . 3% , n = 7 ) of their ripple-related spikes during the beginning of the ripple events ( before the ripple reached its maximal amplitude ) , in broad agreement with the discharge dynamics noted by Klausberger et al . , 2003 . The ripple-related modulation strengths of the firing of BCs and Bistrat cells were high and not different from each other ( BCs: r = 0 . 63 ± 0 . 12; Bistrat: r = 0 . 62 ± 0 . 08 ) . Therefore , Bistrat cells were weakly modulated during the running-associated epsilon waves but strongly modulated during the rest-associated ripples ( Figure 2C , F ) . In terms of phase-preferential firing during ripples , the BCs on average discharged significantly earlier than the Bistrat cells ( 55 ± 20° vs 126 ± 10° , p < 0 . 001 , Watson–Williams test ) , with a sharp division between the two clusters representing the BCs ( red circles ) and the Bistrat cells ( green squares ) in the phase vs modulation strength plot in Figure 2F . Because BCs and Bistrat cells discharged at different rates during ripples ( Figure 2E ) , we further examined what factors may influence the ripple-related firing of these two PV cell classes . Our analysis revealed that BCs discharged on virtually every ripple event ( in agreement with [Varga et al . , 2012] ) , whereas Bistrat cells remained silent ( Figure 3A , B ) on about a third of the ripple episodes ( percentage of ripple events when the interneurons fired: BCs: 95 ± 6%; Bistrat cells: 68 ± 13%; p < 0 . 01 , Mann–Whitney ) . Furthermore , the data showed that whether or not a Bistrat cell fired during a ripple correlated with the oscillation frequency of the individual ripple events ( i . e . , with the intra-ripple or ‘core’ frequency of the LFP oscillations; these are highlighted in red in the example traces in Figure 3A , B ) . Namely , as illustrated in Figure 3C , D , the core frequency of the ripples when Bistrat cells discharged was significantly higher than the core frequency of those ripples during which the Bistrat cells did not discharge ( mean core frequency of ripple events without Bistrat cell spikes: 122 . 5 ± 18 Hz , n = 191 events; with spikes: 140 ± 17 Hz , 347 events; n = 8 cells , p < 0 . 0001 , Mann–Whitney test ) . Therefore , these results indicate that , in sharp contrast to BCs ( Figure 3B , D ) , Bistrat cells ( Figure 3A , C ) differentiate between lower vs higher frequency ripple events ( note that , as illustrated in the inset in Figure 3D , rest-associated high-frequency events in our recordings showed no evidence of a bimodal distribution with a minimum at 140 Hz , the frequency boundary used to separate high gamma and ripple events in sleeping or awake immobile rats [see Figure 1B in Sullivan et al . , 2011] ) . Taken together , these data show that both the epsilon- and the ripple-associated discharges of PV interneurons display characteristic properties that segregate along previously established class lines . Furthermore , these results demonstrate that the phase-specific discharges of BCs characteristically preceded Bistrat cell firing during fast oscillations ( >90 Hz ) , regardless of the behavioral state of the animal ( epsilon during running , ripples during rest ) . The Bistrat cells have been reported to occur in at least two distinct forms . One sub-class has extensive dendrites both in the oriens and the radiatum layers , with the dendrites in the stratum radiatum extending up to the border with the lacunosum-moleculare layer ( Klausberger et al . , 2004; Katona et al . , 2014 ) , which presumably allows these cells to receive inputs both from the local CA1 PCs and the Schaffer collateral/commissural inputs from the CA3 PCs ( note that axons from CA3 PCs are present in the CA1 stratum oriens as well [Sik et al . , 1993; Wittner et al . , 2006] ) . We will refer to these cells as classical or C-Bistrat cells . In contrast , other Bistrat cells have their entire dendritic tree confined to the stratum oriens with no dendrites in the stratum radiatum ( Maccaferri et al . , 2000 ) ; these cells are referred to as oriens or O-Bistrat cells . From our n = 8 Bistrat cells , 3 were C-Bistrat ( Figure 4A , C and 5 ) were O-Bistrat cells ( Figure 4B , C ) . Note that both sub-classes provided dense axonal projections to the radiatum , with minor innervation of the oriens layers ( Figure 4A , B ) . 10 . 7554/eLife . 04006 . 007Figure 4 . Morphologically distinct Bistrat cell sub-classes fire differentially during sharp-wave associated ripple events . ( A and B ) Fully reconstructed ( axons: blue; dendrites: red; soma: black ) representative cells of the two Bistrat cell sub-classes . ( C ) Total dendritic lengths measured in stratum radiatum/pyramidale/oriens of individual cells , with each cell differentially color coded . The O-Bistrat cells ( B ) have no dendrites in the radiatum , the C-Bistrat cells ( A ) have dendrites in stratum radiatum and oriens as well . ( D ) The firing frequencies of the O-bistrat and C-bistrat cells did not show significant differences during running-associated theta oscillations . ( E ) The O-Bistrat cells discharged at significantly higher rates than the C-Bistrat cells during ripples . DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 00710 . 7554/eLife . 04006 . 008Figure 5 . Sub-populations of AACs show distinct action potential discharge probabilities during ripples . ( A and B ) Full reconstruction of dendritic arborizations and cell bodies of two representative AAC cells . Note that the cell body in ( A ) is located within stratum pyramidale ( C-AAC ) , but the cell in ( B ) is in the stratum oriens ( E-AAC; scale: 100 μm ) . ( C and D ) Electron microscopic verification of axon initial segments ( AISs ) as the exclusive post-synaptic targets of the axons of AACs in ( A ) and ( B ) , respectively . The bouton in ( C ) innervates two adjacent AISs . Panel ( D ) also illustrates that targets were verified on longitudinally and transversely ( inset ) sectioned AISs ( arrowheads in C , D: synaptic clefts , scales: 500 nm ) . ( E and F ) Example LFPs ( upper traces ) and the action potentials of cells shown in ( A ) and ( B ) , respectively . Ripples are highlighted with red . Note the different firing activity of cell ( A ) and ( B ) during the events . Voltage scales: 0 . 5 mV; time scale: 100 ms . ( G and H ) Firing probabilities of cells in ( A ) and ( B ) , respectively , as a function of normalized time before , during , and after ripples . Insets in ( G and H ) : average firing frequencies of individual cells outside ( ‘out’; black dots ) and inside ( ‘rip’; red dots ) of ripple events . C-AACs ( G , n = 5 ) did not change significantly their firing frequency during ripples; however , E-AACs ( n = 2 ) showed elevated activity during events ( statistical analysis as in Varga et al . ( 2012 ) , based on randomization of spike times; see ‘Materials and methods’ ) . Significance is indicated with asterisks ( p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 008 In spite of the marked differences in dendritic trees in the radiatum layer , there have been no functional differences reported in the literature between the C- and O-Bistrat cells , possibly because no previous study had both sub-classes sampled in vivo under the same conditions . Indeed , there was no difference in the running-associated theta-related firing of C- and O-Bistrat cells in our sample either ( firing frequency during theta: O-Bistrat: 34 . 6 ± 14 . 7 Hz; C-Bistrat: 27 . 2 ± 6 . 5 Hz; p = 0 . 67 , Mann–Whitney; Figure 4D; theta phase preference: O-Bistrat: 2 . 6 ± 14 . 6°; C-Bistrat: 357 . 6 ± 12 . 3°; strength of modulation of firing during theta , r: O-Bistrat: 0 . 17 ± 0 . 06; C-Bistrat: 0 . 19 ± 0 . 04; p = 0 . 7 , Mann–Whitney ) . In contrast , during rest-associated ripples , O-Bistrat cells fired significantly faster than C-Bistrat cells ( O-Bistrat: 44 . 2 ± 11 . 0 Hz; C-Bistrat: 19 . 8 ± 3 . 7 Hz; p = 0 . 035 , Mann–Whitney; Figure 4E ) , with no differences in other aspects of ripple-related firing ( phase preference: O-Bistrat: 122 . 41 ± 6 . 63°; C-Bistrat: 134 . 07 ± 10 . 35°; strength of modulation during ripples , r: O-Bistrat: 0 . 65 ± 0 . 09; C-Bistrat: 0 . 57 ± 0 . 03; p = 0 . 25 , Mann–Whitney ) . The different firing rates of the O-Bistrat and C-Bistrat cell sub-classes during ripples demonstrate the emergence of intra-class functional differences among Bistrat cells during fast network oscillations . In addition , the fact that O-Bistrat cells had especially high firing frequencies during ripples suggests that dendrites in the radiatum are not required for elevated levels of discharges in these cells during ripple events . Next , we asked the question if there were other functional distinctions beyond the ripple-associated firing of O- vs C-Bistrat cells during fast network oscillations between cells that belong to the same PV cell class . One potential clue was our observation that , as mentioned above , 2 out of the 5 AACs for whom theta , gamma , and epsilon-related firing could be analyzed showed significant modulation of action potential discharges during running-associated epsilon waves ( Figure 2C ) , indicating a certain amount of intra-class heterogeneity among AACs as well . Did the AACs that showed significant epsilon-related modulation of their firing display any distinct anatomical features ? Indeed , we noted that the somata of five of our AACs were situated within the PC layer or at the PC layer and stratum oriens border ( Figure 5A , C; we refer to these cells as Classical or C-AACs ) , whereas the somata of the two AACs that showed epsilon-modulation were located outside of the PC layer ( Figure 5B , D; we refer to these cells as External or E-AACs . Such E-AACs have been also noted before [Ganter et al . , 2004; Forro et al . , 2013] ) . In addition , the C-AACs and E-AACs also differed in their firing during ripples ( number of AACs that could be analyzed for ripples: n = 7 ) . As illustrated in Figure 5E , G , C-AACs either did not alter ( 4 out of 5 ) or decrease ( 1 out of 5 ) their firing rates during ripples ( inset in Figure 5G ) . In sharp contrast to the C-AACs , the two E-AACs significantly increased their discharge rates during ripples ( Figure 5F , H; from 5 . 6 Hz–23 Hz , and from 3 . 8 Hz–7 . 5 Hz; p < 0 . 001 for each cell ) . Furthermore , the strength of modulation of the ripple-related firing by the E-AACs was strong ( r = 0 . 69 and r = 0 . 45; blue triangles in Figure 2F ) , comparable to that of the BCs and Bistrat cells . Apart from in vitro reports ( Hajos et al . , 2013 ) , to our knowledge , this is the first in vivo evidence of a sub-population of AACs enhancing their discharge probabilities during ripple oscillations , with significant implications for GABAergic regulation of PC discharges during fast network oscillations ( see ‘Discussion’ ) . Interestingly , the E-AACs showed preferred phases of firing during ripples comparable to Bistrat cells ( Figure 2F; mean phase values for the E-AACs: 115° and 124°; note that the E-AACs also showed similar phase values to Bistrat cells during the epsilon oscillations as well; Figure 2C ) . Additionally , it is worthwhile to note that while only one of the AACs decreased its firing during the ripples , 5 out of the 5 C-AACs and 1 out of the 2 E-AACs showed significantly decreased probability of firing in the immediate post-ripple period ( see Figure 5G ) . Similar post-ripple decrease in firing was reported previously for C-AACs from anesthetized rats ( Klausberger et al . , 2003 ) . Therefore , these observations demonstrate that there is a functional splitting of the AAC class during fast network events as well , in addition to the differences described above for the ripple-related discharges of the Bistrat cell sub-classes . Finally , we investigated whether intra-class differences in fast network event-related discharges could be identified for BCs as well . As a potential clue , we focused on the morphological differences identified above for the differentially discharging Bistrat cell and AAC sub-classes , namely , the presence of dendrites in the stratum radiatum vs oriens only ( Bistrat cells ) and the soma positions within or outside the PC layer ( AACs ) . Indeed , BC somata have been reported to be occasionally located outside the stratum pyramidale ( Pawelzik et al . , 2002 ) , but no functional differences have been reported between BCs situated inside vs outside the cell layer . Among our n = 12 successfully recorded and post-hoc visualized BCs , there were three cells with cell bodies and dendrites exclusively in the stratum oriens ( Figure 6B ) . All of these three cells were positive for PV ( Figure 6E ) and targeted somata ( Figure 6F; number of boutons examined that synapsed on somata: n = 9; dendrites: n = 3; target unidentified: n = 2 ) . We refer to these BCs with somata and dendrites in the oriens as horizontal or H-BCs ( note that H-BCs had axonal projections into [n = 2] or towards [n = 1] the subiculum ) . The H-BCs were clearly distinct from their more canonical counterparts ( referred to as classical or C-BCs below ) , whose somata were located within the PC layer and the C-BCs had vertically oriented dendrites that spanned across the oriens to the radiatum , with branches extending into the lacunosum-moleculare layer ( Figure 6A ) . These cells also showed PV immunoreactivity ( Figure 6C ) and targeted somata ( Figure 6D ) . 10 . 7554/eLife . 04006 . 009Figure 6 . Horizontal-BCs located outside pyramidal layer fire doublets during sharp-wave associated events . ( A and B ) Full dendritic ( red ) and partial axonal reconstructions of a classical ( A ) and a horizontal ( B ) BC . Axons were partially reconstructed from single 60 μm sections for better visualization . Note that the H-BC had dendrites only in oriens/alveus , whereas the C-BC had dendrites in the stratum radiatum ( and lacunosum-moleculare ) as well . The H-BC sent axons both to the subiculum and CA1 , in contrast to the C-BCs that innervated only the CA1 pyramidal layer . Scales: 100 μm . ( C and E ) Immunohistochemical verification of PV expression of cells in ( A ) and ( B ) , respectively . Arrows point to neurobiotin ( Bio ) filled boutons ( green ) in the stratum pyramidale and to the corresponding PV positive profiles ( scales: 10 µm ) . ( D and F ) Electron micrographs of boutons of the same classical ( A and D ) and horizontal ( B and F ) BCs innervating somata in the CA1 pyramidal layer . Arrows indicate the synaptic specializations . ( G and H ) Example band-pass filtered LFPs ( 90–200 Hz; upper traces ) and the action potentials of the H- and C-BCs ( A and B , respectively ) . Note that the H-BC repeatedly fired two spikes per oscillatory cycle ( doublets: red ) , whereas the C-BC fired one spike per ripple cycle . LFP: 0 . 4 mV; unit traces: 1 mV; time scale: 100 ms . ( I ) The relative number of ripple events with doublets of action potential firing was significantly higher for H-BCs compared to C-BCs . ( J ) Average firing frequencies during ripple events were also significantly higher in H- than in C-BCs . DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 009 Importantly , analysis showed that the morphologically distinct BC sub-classes displayed differential firing behavior during fast network oscillations . Namely , H-BCs often fired two spikes per individual ripple cycle ( i . e . , doublets of action potentials ) ( Figure 6H ) , which occurred only rarely in C-BCs ( Figure 6G ) . In addition , H-BCs also fired more intensely during ripples ( Figure 6H ) . Quantification of the firing patterns of the two BC sub-classes showed statistical differences both in terms of the proportion of ripples when the cells fired doublets of action potentials ( Figure 6I; H-BCs: 77 . 9 ± 11 . 6%; C-BCs: 17 . 3 ± 12%; p = 0 . 009 , Mann–Whitney ) and in terms of the average firing frequency during the ripples ( Figure 6J; H-BCs: 128 . 9 ± 23 . 8 Hz; C-BCs: 62 . 7 ± 12 . 3 Hz; p = 0 . 004 , Mann–Whitney ) . Taken together , these results showed that intra-class differences in firing behavior during fast network oscillations also existed for the most numerically dominant PV interneuronal class , the BCs , similar to what was described above for the Bistrat cells and the AACs . In this study , we examined the firing patterns of anatomically identified , fast-spiking , PV interneurons in the CA1 region of the mouse hippocampus under awake head-fixed conditions in order to provide insights into how the spatio-temporal patterns of GABA release from PV interneurons may regulate the excitability of populations of PCs during brain state-dependent hippocampal rhythms . Given that the canonical property of PV cells is their uniquely rapid signaling capability ( Norenberg et al . , 2010; Armstrong and Soltesz , 2012; Chiovini et al . , 2014 ) , our focus was on the fast ( >90 Hz ) oscillations . Since the epsilon rhythm occurs embedded within the theta oscillations , we first analyzed the discharge patterns of PV interneurons during theta waves . The data showed that the AACs fired with the strongest theta modulation . Unlike AACs from anesthetized rats ( Klausberger et al . , 2003 ) , AACs in our awake head-fixed mice clearly did not fire with the highest probability close to the peak of theta rhythm ( 180° ) , but in fact discharged at 251° ( the 71° difference corresponds to a 25 ms delay for an average , 8 Hz theta with a cycle duration of 125 ms ) . Although inter-species differences cannot be ruled out , the one AAC previously reported from the CA1 of non-anesthetized freely moving rats in the literature also discharged well after the theta peak ( at 225° ) ( Viney et al . , 2013 ) . Furthermore , it is interesting to note that all of the 20 optogenetically tagged PV cells in a recent study conducted in freely moving mice discharged during the descending phase of the theta rhythm and none fired preferentially at the peak ( Royer et al . , 2012 ) . Therefore , it appears that in animals during running , the axon initial segments of PCs receive a barrage of GABA inputs from AACs during the descending phase of the theta cycle ( at 251° on average ) , followed by inputs from BCs onto their cell bodies and proximal dendrites about 20 ms later ( at 310° ) , and then , with another approximately 20 ms delay on average , near the theta trough ( 360° or 0° ) , Bistrat cells release GABA onto the basal and mid-level apical dendrites . Thus , the pattern of discharges by the three cardinal PV interneuron classes is organized to generate a spatio-temporally exquisitely orchestrated axon initial segment to somata/proximal dendrites to mid-level dendrites sweep of fast GABAergic inhibition in a relatively short time window ( Figure 7A , ‘Theta’ ) . 10 . 7554/eLife . 04006 . 010Figure 7 . Summary of the temporal ordering of discharges by the three main PV cell classes ( A ) and the firing properties of the new sub-classes of PV cell during ripples ( B ) . ( A ) Schematic illustration shows a PC with axon terminals from AACs ( blue ) , BCs ( red ) , and Bistrat cells ( green ) ; the same color code is used to illustrate the phase-preferential firing of the three major PV cell classes during theta , gamma , epsilon , and ripple oscillations . Note that the axon initial segment receives the earliest inputs during theta waves , whereas during the faster rhythms ( gamma , epsilon , ripple ) BCs fire first , followed by the Bistrat cells and the AACs . ( B ) Layer-specific locations of somata and dendrites segregate with the ripple-related discharge properties of the 6 PV sub-classes . Note the enhanced firing by the H-BCs , E-AACs , and the O-Bistrat cells compared to their classical counterparts ( C-BCs , C-AACs , and C-Bistrats ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04006 . 010 In terms of fast rhythms , our results revealed clear distinctions between the three major PV cell classes . First , BCs clearly differed from Bistrat cells and AACs in their strong modulation by epsilon oscillations . These results indicate that somatic and proximal dendritic GABAergic inputs from BCs may play primary roles in shaping the excitability of PCs during epsilon-associated behaviors . Second , BCs as a group also differed from the Bistrat cell and AAC classes in exhibiting the highest rates of firing during ripple events . Third , we found marked differences between Bistrat cell and BC discharges during ripples with low core frequencies , indicating that Bistrat cell-derived dendritic inhibition is only recruited during high-frequency sharp-wave ripple events . Fourth , our data revealed that BCs on average fired earlier than Bistrat cells and AACs during all network oscillations faster than the theta range , including the running-associated , theta-embedded gamma and epsilon waves , as well as sharp-wave ripples during rest ( Figure 7A ) . The latter finding indicates a behavioral state-independent leadership position ( referred to as primacy below ) of BCs in interneuronal firing sequences during fast oscillatory cycles , extending our previous observations of a frequency-invariant temporal ordering of BC and oriens-moleculare interneuron ( OLM cells project to the distal dendritic regions of CA1 PCs ) discharges , where the preferred phases of firing of BCs were also significantly earlier than the mean phase of discharges by the OLM cells during both epsilon and ripple events ( Varga et al . , 2012 ) . In addition to these four major inter-class differences during fast network oscillations , our data also showed that AACs in awake mice , when considered together as a single group , exhibited no significant changes in firing rates during ripple events , fired the majority of their spikes at the beginning of the ripples , and decreased their firing in the immediate post-ripple period , in overall agreement with data from anesthetized rats ( Klausberger et al . , 2003 ) . Therefore , AACs in the CA1 do not necessarily stop delivering GABA during ripple events when sub-sets of PCs are known to replay previously experienced firing sequences ( Wilson and McNaughton , 1994; O'Neill et al . , 2008 , 2010 ) , indicating that AACs may actively regulate PC discharges during sharp-wave ripples . The precise nature of the AAC to PC interaction in general ( Howard et al . , 2005; Szabadics et al . , 2006; Glickfeld et al . , 2009 ) , and during ripples in particular , will have to be investigated further in future studies . Importantly , in addition to inter-class differences between PV cells , our data also revealed that robust intra-class distinctions emerged from all three major PV cell classes during fast ( >90 Hz ) , but not slow , network oscillations ( Figure 7B ) . First , Bistrat cells without dendrites in the stratum radiatum , the O-Bistrat cells , fired at significantly higher rates during ripple events than their C-Bistrat counterparts . These closely correlated morphological and physiological differences firmly establish the O-Bistrat and C-Bistrat cell groups as distinct interneuronal sub-classes and highlight the fact that intense action potential discharges during sharp-wave ripples can be generated by interneurons without dendrites in the stratum radiatum . Second , E-AACs were significantly modulated by epsilon and ripple waves , in contrast to their counterparts located in the stratum pyramidale . In addition , E-AACs showed significantly elevated frequency of action potential discharges during ripples . Third , H-BCs fired more intensely during ripples and discharged more doublets of action potentials during individual ripple cycles than the C-BCs . Differences in somatic layer position , in conjunction with significant physiological differences , qualify these distinct AACs and BCs as separate , novel interneuronal sub-classes ( note that layer-specific location together with differential spiking behaviors in vivo was also used as the basis of the differentiation of a previously unified cell class into neurogliaform and ivy cell sub-classes [Fuentealba et al . , 2008] ) . Importantly , these newly identified differentially discharging PV sub-classes are likely to be not negligible in terms of their abundance , as the population of PV interneurons in the stratum oriens has been estimated to be about 30% of the number of PV cells within the stratum pyramidale ( in 1000s: BCs: 1 . 3/3 . 9; Bistrat cells: 0 . 5/1 . 6; AACs: 0 . 3/1 ) ( Bezaire and Soltesz , 2013 ) . Reassuringly , the H-BC/C-BC and E-AAC/C-AAC ratios ( 3/10 and 2/5 , respectively ) in our sample approximated 30% , indicating that we did not under-sample the oriens layer cells . Future studies will be needed to identify the reasons for the differences in firing frequencies of these new PV sub-classes during fast rhythms in the CA1 , but they may include potential differences between intrinsic excitability ( e . g . , as reported for fast spiking BCs and AACs in the neocortex and CA3 ) ( Woodruff et al . , 2009; Papp et al . , 2013 ) and excitatory and inhibitory connectivity ( Hajos et al . , 2013 ) both from within the CA1 as well as from extrinsic sources ( e . g . , AACs have been reported to receive differential inhibitory innervation from the septum [Viney et al . , 2013] ) . Among the circuitry related differences that may contribute to the functional splitting of the PV classes during fast oscillations , it is interesting to consider the potential role of the primacy of the BCs discussed above . What consequences would the consistently early firing of BCs during cycles of fast network oscillations have on the excitability of the other PV cells ? One clue to this question may be related to the fact that BCs are known to synapse not only on PCs but also on the other PV expressing cells as well ( Cobb et al . , 1997 ) . A second potential clue is that BC axons are largely confined to the stratum pyramidale ( Glickfeld and Scanziani , 2006; Foldy et al . , 2007 ) , thus , PV cells whose cell bodies and proximal dendrites are located outside the PC layer will be in a privileged position to escape inhibition from BCs . A third relevant point is that our data showed that the average phase-preferential discharges of Bistrat cells , and the two epsilon and ripple-modulated AACs in our sample , were delayed by about 1 . 4 ms ( calculated for a 140-Hz ripple ) with respect to the mean phase of firing by BCs during fast network events , a period of time that is approximately a monosynaptic delay . Therefore , firing of BCs early during each ripple cycle is expected to deliver synaptic inhibition to other BCs as well as to AACs and Bistrat cells . The inhibition to other BCs would arrive after the BCs fired their first action potentials , thus , it can only prevent the firing of subsequent action potentials by BCs during each individual ripple cycle . Indeed , although BCs are known to be able to fire at 100s of Hz with little or no accommodation ( Kawaguchi , 1995; Buhl et al . , 1996 ) , most BCs fired only a single spike per ripple cycle . The exceptions were precisely those BCs whose cell bodies were outside of the PC layer ( and thus largely outside of the BC axonal cloud ) the H-BCs , which discharged at higher intra-ripple frequencies and fired frequent doublets during ripple cycles ( Figure 7B ) . The early firing of BCs is expected to inhibit the later-discharging AACs and Bistrat cells as well , likely contributing to their relatively lower frequencies of firing during ripples . However , AACs whose cell bodies are outside of the PC body layer would be predicted to be able to largely escape the BC-derived inhibition , which is consistent with our observation that such E-AACs significantly increased their firing rates during ripple events ( Figure 7B ) . While future studies will be needed to identify the precise factors shaping the temporal ordering of interneuronal discharges , the early firing of BCs during fast network events is likely to differentially regulate the excitability of those interneurons whose cell bodies are located inside vs outside the termination zone of BC axons . A particular strength of our study is the rigorous identification of the recorded fast-spiking interneurons . However , the juxtacellular in vivo technique is based on non-targeted ( ‘blind’ ) recordings from actively discharging interneurons , and even after successful post-hoc visualization and cell identification it produces only relatively limited sample sizes . Nevertheless , even though only a single cell was attempted to be juxtacellularly labeled in each animal , the database of PV cells presented in this paper is the largest collection of anatomically and immunocytochemically identified fast-spiking cells that have been recorded from non-anesthetized animals in vivo to date . Interneurons are a minority ( estimated to be 38 , 500/349 , 500 or about 11% ) of the total CA1 neuron population , and PV cells comprise approximately a quarter of the interneurons in the CA1 ( 9200/38 , 500 , 23 . 9% ) ( Bezaire and Soltesz , 2013 ) . Among the PV cell classes , BCs are estimated to be the most numerous ( 14% of all CA1 interneurons ) , followed by the Bistrat cells ( 6% ) and the AACs ( 4% ) ( Baude et al . , 2007; Bezaire and Soltesz , 2013 ) . Reassuringly , these frequencies of occurrence roughly matched the ratio of the cell numbers for the three PV cell classes in our database ( n = 12 BCs; n = 8 Bistrat cells; n = 5 AACs ) , indicating a lack of a major sampling bias for the various PV cell classes in our juxtacellular recordings . Although recordings from head-fixed animals have certain limitations ( e . g . , in studies of eye movements ) ( Wallace et al . , 2013 ) , an advantage of the head-fixed juxtacellular recording approach ( Varga et al . , 2012 ) for the study of network oscillations is that the mice under head-fixed conditions produce relatively long periods of running ( typically tens of seconds ) , which allows robust sampling of continuous theta rhythm and the associated theta-nested epsilon oscillations while maintaining high-quality juxtacellular recordings from single cells . In contrast , theta-related juxtacellular recordings from non-head-fixed freely moving but tethered and spatially confined rats were reportedly limited to short periods of head movements and postural shifts while the animals remained in the same location ( Viney et al . , 2013 ) , highlighting the existence of both advantages and disadvantages of the particular experimental arrangements . An important conclusion from our study is that caution has to be exercised when trying to infer precise interneuronal identity from extracellular unit data , even if combined with optogenetic tagging . Nevertheless , our results also highlight new criteria that may now be employed . For example , although a hypothetical extracellularly recorded unit that increases its firing during sharp-wave ripples may involve any of the three major PV classes , BCs and AACs can be sharply distinguished based on the strength of epsilon modulation of their firing ( note that all BCs were strongly epsilon modulated , whereas AACs were weakly or not significantly modulated ) . Similarly , Bistrat cells can be distinguished from BCs based on their phase-preferential firing during theta and ripples , in combination with their modulation during epsilon waves . Such insights are important in light of the fact that numerous studies have been conducted using extracellular unit recordings from freely moving animals using tetrodes and silicon probes , and post-hoc analysis of existing data may now be attempted to dissect the role of specific PV cell populations in particular behavioral tasks . The existence of three new PV cell sub-classes in the CA1 region beyond the classically identified monolithic BC , Bistrat cell , and AAC groups is likely to have major consequences for the mechanisms by which local GABAergic fast inhibition regulates output from the hippocampus . In particular , our results demonstrating a functional splitting of the three major PV cell classes may serve to increase the available repertoire of inhibitory regulatory processes to facilitate the coordination of PC excitability during fast network rhythms associated with various cognitive and behavioral processes ( Staba et al . , 2004; Mukamel et al . , 2005; Canolty et al . , 2006; Tort et al . , 2008; Colgin et al . , 2009; Kepecs and Fishell , 2014 ) . The preferential ability to generate doublets of spikes by the distinct BC sub-populations identified in this study is particularly interesting from a functional perspective . Indeed , firing of doublets of action potentials by BCs has been suggested to play key roles in synchronization of spatially distinct neuronal sub-populations during gamma rhythm ( Traub et al . , 1996 ) . Therefore , it will be important to determine if the doublet firing that we found preferentially in the H-BCs serves a similar function during ripples . The hippocampus plays key roles in learning and memory processes by transforming input from associative neocortical regions to outputs carried by the long-distance projecting axons of CA1 PCs targeting a variety of brain areas , including the medial entorhinal cortex , amygdala , and the medial prefrontal cortex ( Cenquizca and Swanson , 2007; Lee et al . , 2014 ) . Interestingly , recent results demonstrated that CA1 BCs differentially innervated sub-populations of CA1 PCs , both in terms of their superficial vs deep positions within the stratum pyramidale and with regards to their long-distance projection targets ( Lee et al . , 2014 ) , likely contributing to the sparse and distributed nature of hippocampal network activity . Since a high degree of bias was also noted in the excitatory innervation of BCs by the distinctly projecting CA1 PCs as well ( e . g . , medial prefrontal cortex-projecting PCs were much more likely to innervate BCs than amygdala-projecting ones ) ( Lee et al . , 2014 ) , it will be especially important to determine if such specialized local inhibitory–excitatory circuits also involve AACs and Bistrat cells in general and the newly recognized PV sub-classes in particular . Our study also highlights the importance of future investigations into the developmental origins of the distinct BC , Bistrat cell , and AAC sub-classes . Indeed , functionally distinct , new sub-classes of OLM cells , recently identified based on expression of 5HT ( 3A ) receptor , differentially participated in network oscillations and originated from distinct sub-divisions of the ganglionic eminences during embryonic development ( Chittajallu et al . , 2013 ) . Therefore , developmental mechanisms may drive the formation of specialized microcircuits that can parse the various CA1 PC sub-populations according to the dynamically changing circuit demands during hippocampal memory functions . Minimal duration of episodes for analysis was >20 s for running and >200 s for resting episodes . In the case of two cells ( imi069 and imi075 in Table 1 ) , mice had been injected with saline solution ( 50 nl ) in the amygdala at least 2 weeks before recording as part of a control group in an on-going parallel study; the firing properties of these cells did not differ from their counterparts recorded from non-injected mice , and thus they were included in the database . The analysis was performed with the help of custom-written Matlab scripts as described earlier ( Varga et al . , 2012 ) with some additional features/modifications described below . The low-pass filtered ( Bessel , 5 kHz ) signals of the two adjacent channels were sampled at 12 or 20 kHz and stored for offline analysis . Running-associated theta was filtered ( 5–10 Hz ) on downsampled ( 20 times ) signal of the unit channel , high-frequency oscillations were analyzed on the LFP channel . Running-associated LFP signals were filtered for epsilon oscillations between 90–130 Hz . For high-frequency oscillatory epochs during resting ( ripples ) , LFP was filtered between 90–200 Hz . First , envelope-detection was performed on the absolute values of the signals , and instances where the envelope crossed the threshold of 5 standard deviations were considered as ripples . The start and end of ripples were specified as the time points where the signal reached a second threshold of 2 standard deviations . For display purposes ( Figures 2E and 5G , H ) , before and after ripple periods ( lasting two times the duration of the ripples ) were also included , and spiking probabilities were binned into 40 , 20 , 40 bins corresponding to before , during , and after ripple periods , respectively . Time–frequency plots were generated on the LFP channel signal . First , the signals were narrow-band filtered ( 4 Hz bandwidths from 10–200 Hz , 2 Hz steps [Canolty et al . , 2006] ) , then envelope-detection was performed on the absolute values of the signals and the power of the envelope was normalized in every frequency band . Theta trough or ripple peak power triggered averaging of the resulting data was performed and displayed for cross-frequency coupling during running-associated theta and for resting-associated ripple activity , respectively . Spike occurrences during the various oscillations were determined and uniformity was tested with Rayleigh-test ( Varga et al . , 2012 ) . Preferred firing phase and strength of modulation during oscillations of individual cells were calculated as described earlier ( Varga et al . , 2012 ) and expressed as orientation and length of vectors calculated by summing the spike phases normalized by the number of spikes . Depending on the orientations of the different phase angles , the length of the normalized vector ( r ) ranged from 0 ( no phase preference ) to 1 ( all phase angles identical ) . This length of the vector was used to measure the magnitude of phase modulation , and the direction of r indicated the mean phase angle of the cell ( Varga et al . , 2012 ) . Differences between cell classes ( BCs , Bistrat cells , AACs ) and sub-classes ( Figure 7B ) in terms of phase locking were evaluated using Watson–Williams circular test ( Senior et al . , 2008 ) . In order to determine whether the firing probability of a given cell was significantly lower or higher during ripples than outside ripples , simulations were performed either by randomizing the spike location ( Varga et al . , 2012 ) or by shuffling the ripple location ( Lasztoczi et al . , 2011 ) . The core frequency of ripple epochs was calculated with three different methods . The first method used the inverse of the averaged times between zero crossings of the ripple signal within the ripple boundaries and divided by two . The second method computed the inverse of the time between neighboring peaks of ripples within the ripple boundaries ( Tukker et al . , 2013 ) . The third method generated wavelet transforms on the 90–200 Hz filtered LFP signal using ‘cmor5-1’ function in the Matlab Wavelet Toolbox . The peak values of the wavelet transforms were measured and counted as the core frequency ( Patel et al . , 2013 ) . All three methods were tested on artificial ripple events and found to show similar core frequency detection . We report our data throughout the manuscript with the wavelet method . For the analysis associated with Figure 3 , events detected on the 90–200 Hz band-pass filtered LFP signals during rest were sorted based on the presence or lack of action potentials of the recorded Bistrat and BC cells . Core frequencies of events with or without spikes were statistically compared with Wilcoxon signed rank test . With the exception of the inset in Figure 3D , data are presented as mean ± standard deviation .	The brain continuously processes information from outside and inside the body to cope with the challenges of everyday life . As the brain carries out these processes , networks of neurons produce patterns of electrical activities called oscillations . Fast-spiking PV cells are neurons that orchestrate the precise timing of these oscillations in a region of the brain called the hippocampus , which is important for the formation of memories . PV cells perform this role by releasing a chemical called GABA that suppresses electrical activity . The hippocampus contains three distinct sub-classes of fast-spiking PV cells , but it is not clear how these different sub-classes collaborate to control the network oscillations in the hippocampus . Varga et al . have now explored this question by recording the electrical activity of PV cells in mice , while they were resting and also while they were running . PV cells are involved in both fast and slow network oscillations . As had been found in previous experiments , Varga et al . found that the three different sub-classes of PV cells behaved similarly during slow network oscillations . During fast oscillations , however , the neurons within each sub-class displayed two distinct types of behavior , depending on their shape and location . PV cells release GABA in patterns that depend on both space and time: the work of Varga et al . shows that the repertoire of patterns that can be employed by PV cells is about twice as big as was previously thought . Future studies are needed to explore the influence of this behavior on memory .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Functional fission of parvalbumin interneuron classes during fast network events
Recent evidence suggests that autophagy facilitates the unconventional secretion of the pro-inflammatory cytokine interleukin 1β ( IL-1β ) . Here , we reconstituted an autophagy-regulated secretion of mature IL-1β ( m-IL-1β ) in non-macrophage cells . We found that cytoplasmic IL-1β associates with the autophagosome and m-IL-1β enters into the lumen of a vesicle intermediate but not into the cytoplasmic interior formed by engulfment of the autophagic membrane . In advance of secretion , m-IL-1β appears to be translocated across a membrane in an event that may require m-IL-1β to be unfolded or remain conformationally flexible and is dependent on two KFERQ-like motifs essential for the association of IL-1β with HSP90 . A vesicle , possibly a precursor of the phagophore , contains translocated m-IL-1β and later turns into an autophagosome in which m-IL-1β resides within the intermembrane space of the double-membrane structure . Completion of IL-1β secretion requires Golgi reassembly and stacking proteins ( GRASPs ) and multi-vesicular body ( MVB ) formation . Most eukaryotic secretory proteins with an N-terminal signal peptide are delivered through the classical secretion pathway involving an endoplasmic reticulum ( ER ) -to-Golgi apparatus itinerary ( Lee et al . , 2004; Schatz and Dobberstein , 1996 ) . However , a substantial number of secretory proteins lack a classical signal peptide , called leaderless cargoes , and are released by unconventional means of secretion ( Nickel and Rabouille , 2009; Nickel and Seedorf , 2008 ) . The range of unconventional secretory cargoes encompasses angiogenic growth factors , inflammatory cytokines and extracellular matrix components etc . most of which play essential roles for development , immune surveillance and tissue organization ( Nickel , 2003; Rabouille et al . , 2012 ) . Unlike a unified route for classical protein secretion , leaderless cargoes undergoing unconventional secretion employ multiple means of protein delivery , the details of which are largely unknown ( Ding et al . , 2012; Nickel , 2010; Rabouille et al . , 2012; Zhang and Schekman , 2013 ) . IL-1β is one of the most intensely investigated cargoes of unconventional secretion . A biologically inactive 31 kDa precursor , pro-IL-1β , is made following initiation of the NF-κB signaling cascade . Pro-IL-1β is subsequently converted into the active form , the 17 kDa mature IL-1β , by the pro-inflammatory protease caspase-1 which is activated , in response to extracellular stimuli , after its recruitment to a multi-protein complex called the inflammasome ( Burns et al . , 2003; Cerretti et al . , 1992; Rathinam et al . , 2012; Thornberry et al . , 1992 ) . Interpretation of the mechanism of unconventional secretion of IL-1β is complicated by the fact that one of the physiologic reservoirs of this cytokine , macrophages , undergoes pyroptotic death and cell lysis under conditions of inflammasome activation of caspase-1 . Indeed , many reports including two recent publications make the case for cell lysis as a means of release of mature IL-1β ( Liu et al . , 2014; Shirasaki et al . , 2014 ) . In contrast , other reports demonstrate proper secretion of mature IL-1β without cell lysis in , for example , neutrophils , which are nonetheless dependent on the inflammasome response to activate caspase-1 and secrete mature IL-1β ( Chen et al . , 2014 ) . Quite aside from the possible complication of cell lysis , another body of work has suggested an unconventional pathway for the proper secretion of IL-1β . Pro-IL-1β lacks a typical signal peptide and the propeptide is processed in the cytosol rather than the ER ( Rubartelli et al . , 1990; Singer et al . , 1988 ) . Although mature IL-1β appears to be incorporated into a vesicular transport system , secretion is not blocked by Brefeldin A , a drug that blocks the traffic of standard secretory proteins form the Golgi apparatus ( Rubartelli et al . , 1990 ) . Multiple mechanisms have been implicated in the unconventional secretion of IL-1β , including autophagy , secretory lysosomes , multi-vesicular body ( MVB ) formation and micro-vesicle shedding ( Andrei et al . , 1999; Andrei et al . , 2004; Brough et al . , 2003; Lopez-Castejon and Brough , 2011; MacKenzie et al . , 2001; Qu et al . , 2007; Verhoef et al . , 2003 ) . However , a clear demonstration of the mechanism for the entry of IL-1β into a vesicular carrier , e . g . the autophagosome , is lacking . Macroautophagy ( hereafter autophagy ) is a fundamental mechanism for bulk turnover of intracellular components in response to stresses such as starvation , oxidative stress and pathogen invasion ( Mizushima and Levine , 2010; Yang and Klionsky , 2010 ) . The process is characterized by the formation of a double-membrane vesicle , called the autophagosome , through the elongation and closure of a cup-shaped membrane precursor , termed the phagophore , to engulf cytoplasmic cargoes ( Hamasaki et al . , 2013; Lamb et al . , 2013 ) . Completion of autophagosome formation requires a sophisticated protein-vesicle network organized by autophagic factors , such as autophagy-related ( ATG ) proteins , and target membranes ( Feng et al . , 2014; Mizushima et al . , 2011 ) . Besides the degradative function , autophagy or ATG proteins have recently been implicated in multiple secretory pathways including the delivery of leaderless cargoes undergoing unconventional secretion , such as the mammalian pro-inflammatory cytokines IL-1β and IL-18 , the nuclear factor HMGB1 , and the yeast acyl coenzyme A-binding protein Acb1 , to the extracellular space ( Bruns et al . , 2011; Dupont et al . , 2011; Duran et al . , 2010; Manjithaya and Subramani , 2011; Pfeffer , 2010; Subramani and Malhotra , 2013 ) . The Golgi reassembly and stacking protein ( s ) GRASP ( s ) ( GRASP55 and GRASP65 in mammals , dGRASP in Drosophila , GrpA in Dictyostelium and Grh1 in yeast ) are required for autophagy-regulated unconventional secretion ( Giuliani et al . , 2011; Kinseth et al . , 2007; Levi and Glick , 2007; Manjithaya et al . , 2010 ) . Dupont et al . , 2011 documented a role for autophagy in the secretion of mature IL-1β ( Dupont et al . , 2011 ) , but how a protein sequestered within an autophagosome could be exported as a soluble protein was unexplained . Here , we sought to understand how conditions of starvation-induced autophagy could localize IL-1β into an autophagosomal membrane . We reconstituted the autophagy-regulated secretion of IL-1β in cultured cell lines and detected a vesicle intermediate , possibly an autophagosome precursor , containing mature IL-1β . Three-dimensional ( 3D ) Stochastic Optical Reconstruction Microscopy ( STORM ) demonstrated that , after entering into the autophagosome , IL-1β colocalizes with LC3 on the autophagosomal membrane , which , together with an antibody accessibility assay and observations from biochemical assays , implies a topological distribution in the intermembrane space of the autophagosome . This distribution of IL-1β explains the mechanism accounting for its secretion as a soluble protein through either a direct fusion of autophagosome with the plasma membrane or via the MVB pathway . A dual effect of autophagy has been proposed on the secretion of IL-1β in macrophages ( Deretic et al . , 2012; Jiang et al . , 2013 ) . On one hand , induction of autophagy directly promotes IL-1β secretion after inflammasome activation by incorporating it into the autophagosomal carrier ( Dupont et al . , 2011 ) . On the other hand , autophagy indirectly dampens IL-1β secretion by degrading components of the inflammasome as well as reducing endogenous triggers for inflammasome assembly , including reactive oxygen species ( ROS ) and damaged components , which are required for the activation of caspase-1 and the production of active IL-1β ( Harris et al . , 2011; Nakahira et al . , 2011; Shi et al . , 2012; Zhou et al . , 2011 ) . To focus our study specifically on the role of autophagy in IL-1β secretion , we reconstituted a stage of IL-1β secretion downstream of inflammasome activation by co-expressing pro-IL-1β ( p-IL-1β ) and pro-caspase-1 ( p-caspase-1 ) in non-macrophage cells . As shown in Figure 1A , the generation and secretion ( ~5% ) of mature IL-1β ( m-IL-1β ) was achieved by co-expression of p-IL-1β and p-caspase-1 in HEK293T cells . Mature IL-1β was not produced or secreted without p-caspase-1 , whereas a low level of secreted p-IL-1β ( ~0 . 2% ) was detected with or without the expression of p-caspase-1 . Furthermore , little cell lysis occurred during the treatment we used to induce IL-1β secretion: Much less precursor than mature IL-1β and little cytoplasmic tubulin was detected released into the cell supernatant during the 2 hr incubation in starvation medium ( Figure 1A ) . Starvation , a condition that stimulates autophagy , enhanced IL-1β secretion ( ~3 fold ) and reduced the level of IL-1β in the cell lysates ( Figure 1A , B ) . Inhibition of autophagy by the phosphatidylinositol 3-kinase ( PI3K ) inhibitors 3-methyladenine ( 3-MA ) or wortmannin ( Wtm ) blocked IL-1β secretion activated by starvation and caused the accumulation of mature IL-1β in the cell ( Figure 1B ) . Likewise , in an autophagy-deficient cell line , Atg5 knockout ( KO ) mouse embryo fibroblasts ( MEFs ) ( Mizushima et al . , 2001 ) , IL-1β secretion was reduced and failed to respond to starvation ( Figure 1C ) . Moreover , IL-1β secretion was also inhibited in a dose-dependent manner in the presence of an ATG4B mutant ( C74A ) ( Fujita et al . , 2008 ) , or after the depletion of ATG2A and B ( Velikkakath et al . , 2012 ) , or FIP200 ( Hara et al . , 2008 ) , which block autophagosome biogenesis at different stages ( Figure 1D–F ) . Therefore , the reconstituted system recapitulates the autophagy-regulated secretion of IL-1β . 10 . 7554/eLife . 11205 . 003Figure 1 . Reconstitution of autophagy-regulated IL-1β secretion in cultured cells . ( A ) Reconstitution of starvation-induced IL-1β secretion in HEK293T cells . HEK293T cells were transfected with a single plasmid encoding p-IL-1β or together with the p-caspase-1 plasmid . After transfection ( 24 h ) , the cells were either treated in regular ( DMEM ) or starvation ( EBSS ) medium for 2 hr . The medium and cells were collected separately and immunoblot was performed to determine the level of indicated proteins . ( B ) PI3K inhibitors 3-methyladenine ( 3-MA ) or wortmannin ( Wtm ) inhibit IL-1β secretion . HEK293T cells transfected with p-IL-1β and p-caspase-1 plasmids were cultured in DMEM , EBSS , or EBSS containing 10 mM 3-MA or 20 nM wortmannin for 2 hr . The medium and cells were collected separately and immunoblot was performed as shown in ( A ) . ( C ) IL-1β secretion is blocked in Atg5 KO MEFs . Control WT or Atg5 KO MEFs were transfected with p-IL-1β and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were either cultured in DMEM or EBSS for 2 hr followed by immunoblot as shown in ( A ) . ( D ) IL-1β secretion is inhibited by the ATG4B mutant ( C74A ) . HEK293T cells were transfected with plasmids encoding p-IL-1β , p-caspase-1 and different amounts of ATG4B ( C74A ) plasmid DNA as indicated . After transfection ( 24 hr ) , cells were starved in EBSS for 2 hr followed by immunoblot as shown in ( A ) . ( E ) Knockdown of Atg2 reduces IL-1β secretion . HEK293T cells were transfected with control siRNA or siRNAs against Atg2A , Atg2B alone or both . After transfection ( 48 hr ) , the cells were transfected with p-IL-1β and p-caspase-1 plasmids . After another 24 hr , the cells were starved in EBSS for 2 h followed by immunoblot as shown in ( A ) . ( F ) Knockdown of FIP200 reduces IL-1β secretion . HEK293T cells were transfected with control siRNA or FIP200 siRNA . IL-1β secretion under starvation conditions was determined as shown in ( E ) . Quantification of IL-1β secretion was calculated as the ratio between the amount of IL-1β in the medium and the total amount ( the sum of IL-1β in both medium and lysate ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 00310 . 7554/eLife . 11205 . 004Figure 1—figure supplement 1 . Depletion of ESCRT or GRASPs affects IL-1β secretion . HEK293T cells were transfected with indicated siRNAs ( Hrs ( ESCRT-0 ) ( A ) , Tsg101 ( ESCRT-I ) ( A ) , GRASP55 ( B ) or GRASP65 ( B ) ) . After transfection ( 48 hr ) , the cells were transfected with p-IL-1β and p-caspase-1 plasmids . After another 24 hr , the cells were starved in EBSS for 2 hr followed by immunoblot as shown in Figure 1A . Quantification of IL-1β secretion was calculated as the ratio between the amount of IL-1β in the medium and the total amount ( the sum of IL-1β in both medium and lysate ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 004 In macrophages , MVB formation and GRASP proteins are required for IL-1β secretion ( Dupont et al . , 2011; Qu et al . , 2007 ) . Inhibiting MVB formation by depletion of the ESCRT components , hepatocyte growth factor receptor substrate ( Hrs ) or TSG101 , compromised secretion of IL-1β and CD63 , an exosome marker ( Figure 1—figure supplement 1A ) . Knockdown of the GRASP55 or GRASP65 also led to the reduction of IL-1β secretion ( Figure 1—figure supplement 1B ) . Therefore , in addition to functions required for autophagy , the secretion of IL-1β in HEK293T cells depends on GRASP proteins and at least two proteins implicated in MVB formation , as reported previously ( Dupont et al . , 2011; Qu et al . , 2007 ) . To study if autophagy directly regulates IL-1β secretion , we employed a three-step membrane fractionation procedure as described previously ( Figure 2A ) ( Ge et al . , 2013 ) . We first performed a differential centrifugation to obtain 3k , 25k and 100k membrane pellet fractions . Both IL-1β and the lipidated form of LC3 ( LC3-II ) , a protein marker of autophagosome , were mainly enriched in the 25k membrane fraction ( Figure 2B ) . We then separated the 25k membrane through a sucrose step gradient ultracentrifugation where both IL-1β and LC3-II co-distributed in the L fraction at the boundary between 0 . 25 M and 1 . 1 M layer of sucrose ( Figure 2B ) . Further fractionation of the L fraction using an OptiPrep gradient showed co-fractionation of IL-1β with LC3-II ( Figure 2C ) . To confirm the presence of IL-1β in the autophagosome , we performed immunoisolation of LC3-positive autophagosomes from the 25k fraction and found that IL-1β , especially the mature form , co-sedimented with autophagosomes ( Figure 2D ) . Consistent with our observations , a recent study also showed a colocalization of IL-1β and LC3 in the form of puncta in macrophages ( Dupont et al . , 2011 ) . These data demonstrate that at least a fraction of intracellular mature IL-1β associates with the autophagosome , possibly related to its role in IL-1β secretion . 10 . 7554/eLife . 11205 . 005Figure 2 . IL-1β vesicles co-fractionate with LC3 vesicles . ( A ) Membrane fractionation scheme . Briefly , HEK293T cells transfected with p-IL-1β and p-caspase-1 plasmids were starved in EBSS for 2 hr , collected and homogenized . Cell lysates were subjected to differential centrifugations at 3000×g ( 3k ) , 25 , 000×g ( 25k ) and 100 , 000×g ( 100k ) . The level of IL-1β in each membrane fraction was determined by immunoblot . The 25k pellet , in which IL-1β was mainly enriched , was selected and a sucrose gradient ultracentrifugation was performed to separate membranes in the 25k pellet to the L ( light ) and P ( pellet ) fractions . The L fraction , which contained the majority of IL-1β , was further resolved on an OptiPrep gradient after which ten fractions from the top were collected . ( B , C ) Immunoblot was performed to examine the distribution of IL-1β , LC3 as well as the indicated membrane markers in the indicated membrane fractions . T , top; B , bottom ( D ) HEK293T cells transfected with p-IL-1β , p-caspase-1 and FLAG-tagged LC3-I plasmids were starved in EBSS for 2 hr . LC3 positive membranes were immunoisolated with anti-FLAG agarose from the 25 k pellet and the presence of IL-1β was determined by immunoblot analysis . FT , flowthroughDOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 005 To determine if IL-1β is localized to the phagophore in the absence of autophagosome completion , we fractionated membranes from ATG2-depleted cells , which are deficient in phagophore elongation and therefore fail to form mature autophagosomes ( Velikkakath et al . , 2012 ) , and examined the distribution of LC3-II , which remains attached to immature phagophore membranes , and mature and precursor IL-1β . We performed the three-step fractionation described above . In control cells , IL-1β co-distributed with LC3-II in all three steps ( Figure 3 ) . Depletion of ATG2 did not affect the co-fractionation of IL-1β and LC3-II ( Figure 3 ) , indicating that IL-1β enters into the phagophore membrane before the completion of the autophagosome . 10 . 7554/eLife . 11205 . 006Figure 3 . IL-1β co-distributes with LC3 in Atg2-depleted cells . ( A ) HEK293T cells were transfected with siRNAs against Atg2A and Atg2B followed with p-IL-1β and p-caspase-1 plasmids as shown in Figure 1E . The cells were starved in EBSS for 2 hr . Membrane fractions ( 3k , 25k , 100k ( ×g ) , L and P ) were separated from the post-nuclear supernatant as depicted in Figure 2B . ( B ) Ten membrane fractions were collected from the OptiPrep gradient ultracentrifugation as depicted in Figure 2C . Immunoblot was performed to examine the distribution of IL-1β , LC3 as well as the indicated membrane markers . T , top; B , bottom . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 006 We asked how IL-1β enters into the autophagosome . One possibility is engulfment through the closure of the phagophore membrane during autophagosome maturation as in the capture of autophagic cargo . In this scenario , closure of the phagophore to complete autophagosome formation would be required to sequester IL-1β away from the cytoplasm . Alternatively , we considered the possibility that IL-1β may be translocated through a membrane into the lumen of the phagophore envelope and be sequestered from the cytoplasm even before the mature autophagosome is sealed . To test this possibility , we performed proteinase K protection experiments with the membranes from ATG2-depleted cells ( Figure 4A ) . In control cells , p62 ( an autophagic cargo ) and a fraction of LC3-II ( which was encapsulated after autophagosome completion ) , as well as mature IL-1β , were largely resistant to proteinase K digestion similar to the ER luminal protein , protein disulfide isomerase ( PDI ) . In contrast , SEC22B , a membrane anchored SNARE protein exposed to the cytoplasm , was sensitive to proteinase K digestion ( Figure 4A ) . Triton X-100 treatment permeabilized the membrane and rendered all proteins tested sensitive to proteinase K digestion ( Figure 4A ) . This demonstrated that the majority of membrane localized IL-1β was sequestered within an organelle , likely the autophagosome , as demonstrated by the fractionation results of Figures 2 and 3 . However , the result did not pinpoint where within the autophagosome IL-1β was housed . In ATG2-depleted cells , p62 and LC3-II remained sensitive to proteinase K digestion , consistent with the hypothesis that ATG2 is essential for maturation and closure of the autophagosome ( Figure 4A ) . However , in the same samples the majority of IL-1β resisted degradation by proteinase K treatment ( Figure 4A ) , except on addition of Triton X-100 to permeabilize membranes . Although the precursor form of IL-1β remained associated with isolated autophagosome and phagophore membranes ( Figure 3 ) , the protein was degraded when membranes from normal and ATG2-depleted cells were treated with protease in the presence or absence of Triton X-100 ( data not shown ) . Thus , the mature but not the precursor IL-1β appears to be transported into the phagophore . 10 . 7554/eLife . 11205 . 007Figure 4 . Closure of the autophagosome is not required for the entry of IL-1β into vesicles . ( A ) HEK293T cells were transfected with siRNAs against Atg2A and Atg2B followed by transfection with p-IL-1β and p-caspase-1 plasmids as shown in Figure 1E . The cells were starved in EBSS for 2 hr and proteinase K digestion was performed with the 25k membrane fractions . ( B ) Atg5 WT , KO MEFs were transfected with p-IL-1β and p-caspase-1 plasmids as shown in Figure 1B . The cells were starved in EBSS for 2 hr followed by proteinase K digestion as shown in ( A ) . ( C ) HEK293T cells were transfected with siRNA against FIP200 followed by analysis of membrane entry of IL-1β as shown in ( A ) . The level of proteinase K protection was calculated as the percentage of the total protein . Error bars represent standard deviations of at least three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 007 A most recent study showed that small , closed double-membrane structures could be observed in ATG2-depleted cells ( Kishi-Itakura et al . , 2014 ) . To rule out the possibility that IL-1β was engulfed by the small closed autophagosomes , we employed Atg5 KO MEFs in which the phagophore could not be closed ( Kishi-Itakura et al . , 2014; Mizushima et al . , 2001 ) . Similar to what we observed in ATG2-depleted cells , IL-1β was protected from proteinase K digestion in membranes from Atg5 KO MEFs ( Figure 4B ) . In addition , IL-1β was sequestered within vesicles in FIP200 ( another early factor in phagophore development ( Hara et al . , 2008 ) ) knockdown cells ( Figure 4C ) . These data indicate that the entry of IL-1β into the vesicle carrier is not dependent on the formation of the autophagosome . These results are inconsistent with a role for engulfment of IL-1β by the maturing phagophore and suggest instead that IL-1β may be translocated across a membrane into a vesicle precursor of the phagophore , possibly at a very early stage in the development of the organelle . We then sought to test if IL-1β could directly translocate across the membrane of a vesicle carrier . As protein unfolding is usually required for protein translocation , we adopted an approach used in many other circumstances wherein a targeted protein is fused to dihydrofolate reductase ( DHFR ) , an enzyme whose three-dimensional structure is stabilized by the folate derivative aminopterin , hence providing a chemical ligand to impede the unfolding process ( Backhaus et al . , 2004; Eilers and Schatz , 1986; Wienhues et al . , 1991 ) . We first determined the secretion of the DHFR-fused IL-1β . As shown in Figure 5A , secretion of a mature IL-1β-DHFR fusion protein was enhanced by starvation similar to the untagged counterpart . Importantly , IL-1β-DHFR secretion was reduced in a dose-dependent manner in the presence of aminopterin ( Figure 5B ) . Of notice , treatment of aminopterin did not completely abolish IL-1β secretion perhaps due to a cell death-induced release of IL-1β at high concentrations of aminopterin , as indicated by the release of a low level of tubulin into the medium fraction ( Figure 5B ) . As a control , aminopterin did not reduce the secretion of untagged IL-1β , confirming its specific effect on DHFR ( Figure 5—figure supplement 1 ) . Fractionation of cells incubated with aminopterin showed a reduced level of IL-1β in the membrane fraction with a corresponding increase in the cytosol fraction ( Figure 5C ) . The residual DHFR-tagged IL-1β associated with membranes from aminopterin-treated cells was sensitive to proteinase K digestion ( Figure 5D ) , indicating that this pool of membrane-associated IL-1β did not translocate into the lumen of the vesicle . The data suggest that entry of IL-1β into a vesicle carrier involves a process of protein unfolding and translocation . 10 . 7554/eLife . 11205 . 008Figure 5 . Protein unfolding is required for the entry of IL-1β into vesicles . ( A ) Secretion of DHFR-tagged IL-1β . HEK293T cells were transfected with p-IL-1β-DHFR and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were treated with DMEM or EBSS for 2 hr . Release of IL-1β was determined as shown in Figure 1 . ( B ) Secretion of IL-1β-DHFR was inhibited by aminopterin . HEK293T cells were transfected with p-IL-1β-DHFR and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were treated with EBSS , or EBSS containing different concentrations of aminopterin as indicated for 15 min followed by determination of IL-1β secretion as shown in ( A ) . Quantification of IL-1β secretion was calculated as the ratio between the amount of IL-1β in the medium and the total amount ( the sum of IL-1β in both medium and lysate ) . ( C ) Less IL-1β enters into membrane in the presence of aminopterin . HEK293T cells were transfected with p-IL-1β-DHFR and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were either untreated or treated with 5 μM aminopterin in EBSS for 2 hr . The membrane fraction was collected from the top fractions of a Nycodenz density gradient resolved from membranes in a 25k pellet as described in Material and Methods . The cytosolic fraction was collected as the supernatant after 100k×g centrifugation . All fractions were analyzed by immunoblotting using indicated antibodies . ( D ) IL-1β-DHFR is not protected from proteinase K in the presence of aminopterin . Nycodenz -floated membrane fraction collected as shown in ( C ) was subjected to proteinase K digestion and then analyzed by immunoblotting using indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 00810 . 7554/eLife . 11205 . 009Figure 5—figure supplement 1 . Secretion of IL-1β is not affected by aminopterin . HEK293T cells were transfected with p-IL-1β and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were treated with EBSS , or EBSS containing different concentrations of aminopterin as indicated for 15 min followed by determination of IL-1β secretion as shown in Figure 1 ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 009 If IL-1β is directly translocated across the membrane of a vesicle intermediate , fusion of these vesicles to form a double-membrane autophagosome would deposit IL-1β in the lumen between the two membranes of the autophagosome . To visualize the subcellular localization of IL-1β , we employed U2OS cells , which formed large and distinct autophagosomes after starvation . U2OS cells co-expressing p-IL-1β and p-caspase-1 secreted IL-1β in a starvation-enhanced and PI3K-dependent manner similar to HEK293T cells ( Figure 6—figure supplement 1 ) . To prepare for the subsequent fluorescence imaging , we also employed a FLAG-tagged m-IL-1β , which allowed us to directly determine the topological localization of the m-IL-1β . Secretion of m-IL-1β-FLAG from U2OS cells was stimulated by starvation and dependent on PI3K ( Figure 6—figure supplement 1 ) . To determine the topological distribution of IL-1β , we first performed confocal immunofluorescence labeling experiments . After starvation , cells were exposed to 40 μg/ml of digitonin to permeabilize the plasma membrane , harvested and washed with cold PBS to remove the excess cytosolic m-IL-1β-FLAG . In cells expressing either p-IL-1β and p-caspase-1 , or m-IL-1β alone , LC3 and IL-1β were observed by confocal microscopy to localize together or adjacent to one another on the edge of ring-shaped autophagosomes ( Figure 6—figure supplement 2 ) . To further resolve these ring structures , we employed 3D STORM ( Huang et al . , 2008; Rust et al . , 2006 ) super-resolution microscopy ( Hell , 2007; Huang et al . , 2010 ) ( Figure 6 and Figure 6—figure supplements 3 , 4 and Videos 1 and 2 ) . Ring-shaped autophagosomes positive for LC3 ( cyan ) formed after starvation . Some IL-1β ( magenta ) also organized in ring-shaped structures that co-localized with LC3 ( Figure 6 and Figure 6—figure supplement 3 ) . Around 18 ring structures of IL-1β accounting for ~5% of the total IL-1β signal were observed in each cell . A 3D virtual Z-stack analysis confirmed the spatial co-distribution of LC3 and IL-1β on a ball-shaped vesicle ( Video 1 and 2 ) . The diameter of the structures double-labeled with LC3 and IL-1β are ~700 nm ( larger structures up to 2 μm in diameter were also found ) which is comparable to the size of the autophagosome . Occasionally , we also found IL-1β localized in the center of the ring structure , where cytoplasmic autophagic cargoes fill , surrounded by LC3 ( Figure 6—figure supplement 4 ) . This portion of IL-1β was possibly being engulfed by the autophagosome . 10 . 7554/eLife . 11205 . 010Figure 6 . Topological localization of IL-1β in the autophagosomal carrier determined by STORM . U2OS cells were transfected with a plasmid containing the expression cassette of FLAG-tagged mature IL-1β ( m-IL-1β-FLAG ) . After transfection ( 24 hr ) , the cells were starved in EBSS for 1 hr followed by immunofluorescence labeling with mouse monoclonal anti-LC3 and rabbit polyclonal anti-FLAG antibodies . STORM analysis imaging and data analysis were performed as described in Materials and methods . Cyan , LC3; Magenta , IL-1β; Bars: 2 μm ( original image ) and 500 nm ( magnified inset ) DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01010 . 7554/eLife . 11205 . 011Figure 6—figure supplement 1 . Secretion of IL-1β in U2OS cells . U2OS cells were transfected with plasmids encoding the p-IL-1β and p-caspase-1 ( first 4 lanes ) or m-IL-1β-FLAG ( last 4 lanes ) . After transfection ( 24 hr ) , the cells were untreated or starved in the absence or presence of indicated PI3K inhibitors ( 3-MA or wortmannin [Wtm] ) followed by measurement of secretion as indicated in Figure 1 ( A ) and ( B ) . α-m-IL-1β , IL-1b antibody; α-FLAG , FLAG antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01110 . 7554/eLife . 11205 . 012Figure 6—figure supplement 2 . Localization of IL-1β determined by confocal microscopy . U2OS cells were transfected with plasmids encoding the p-IL-1β and p-caspase-1 ( A ) or m-IL-1β-FLAG ( B ) . After transfection ( 24 hr ) , the cells were starved for 1 hr followed by immunofluorescence labeling and confocal microscopy analysis . Bar: 10 μmDOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01210 . 7554/eLife . 11205 . 013Figure 6—figure supplement 3 . Extra images for Figure 6 . Bars: 2 μm ( original image ) and 500 nm ( magnified inset ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01310 . 7554/eLife . 11205 . 014Figure 6—figure supplement 4 . A minority of IL-1β engulfed by autophagosome . U2OS cells were transfected and treated followed by STORM analysis as shown in Figure 6 . Arrow head points to the autophagosome with engulfed IL-1β . Bar: 2 μmDOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01410 . 7554/eLife . 11205 . 015Figure 6—figure supplement 5 . Determination of the topological localization of IL-1β in the autophagosome and phagophore . ( A , C ) Diagrams of autophagosome ( A ) /phagophore ( B ) and omegasome , antibody accessibility for each possible situation of IL-1β localization , and summaries of the antibody accessibility of m-IL-1β-FLAG ( red ) and EGFP-DFCP1 ( green ) are illustrated . ( B , D ) U2OS cells ( B ) and Atg5 KO MEFs ( D ) were transfected with plasmids encoding the m-IL-1β-FLAG and EGFP-DFCP1 . After transfection ( 24 hr ) , the cells were starved in EBSS for 1 hr followed by digitonin treatment and fixation ( see Materials and methods ) . The cells were either labeled with anti-FLAG ( to label IL-1β ) and anti-EGFP ( to label EGFP-DFCP1 ) antibodies ( Digitonin ) or further treated with Saponin followed by antibody labeling ( Digitonin + Saponin ) . Images were acquired by confocal microscopy . Bar: 10 μm ( E ) Quantification of the percentage of EGFP-DFCP1 labeled by EGFP antibody . Percentage was counted by the ratio of puncta numbers of antibody labeled EGFP-DFCP1 and EGFP-DFCP1 according to the EGFP signal . Error bars are standard deviations of more than 50 cells in two independent experiments . ( F ) Quantification of the puncta number for m-IL-1β-FLAG puncta ( red ) and those colocalized with DFCP1 ( yellow ) . Error bars are standard deviations of more than 50 cells in two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01510 . 7554/eLife . 11205 . 016Video 1 . 3D section of the magnified structure in Figure 6 ( upper one ) . The virtual Z-section thickness is 150 nm , and the step size is 50 nm . Cyan , LC3; Magenta , IL-1β; Bar 500 nmDOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01610 . 7554/eLife . 11205 . 017Video 2 . 3D section of the magnified structure in Figure 6 ( lower one ) . The virtual Z-section thickness is 150 nm , and the step size is 50 nm . Cyan , LC3; Magenta , IL-1β; Bar 500 nmDOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 017 The visual detection of IL-1β localized to ring-shaped autophagosomes is consistent with our biochemical assays that place IL-1β in the intermembrane space between the outer and inner membrane of the autophagosome . We devised a further visual test of this conclusion using selective permeabilization of cell surface and intracellular membranes with digitonin and saponin , respectively ( Figure 6—figure supplement 5 ) . We compared antibody accessibility to IL-1β and DFCP1 , a marker located on the cytosolic surface of the omegasome ( a harbor for the phagophore ) in both WT and Atg5 KO cells . Consistent with a cytosolic surface localization , DFCP1 was readily labeled in cells treated with digitonin alone ( selectively permeabilizes the plasma membrane ) in both WT and Atg5 KO cells ( Figure 6—figure supplement 5A–E ) . In contrast , IL-1β was accessible to the antibody only after treatment with digitonin and saponin ( gently permeabilizes the endomembrane ) ( Figure 6—figure supplement 5A , B and F ) in WT cells . This by itself would not distinguish localization of IL-1β to the intermembrane space vs the cytoplasmic enclosed space of a mature autophagosome . However , in Atg5 KO cells where the phagophore precursor envelope remains open and exposed to the cytosol , saponin treatment was necessary to expose IL-1β to antibody and roughly half of the labeled structures coincided with the phagophore marker DFCP1 ( Figure 6—figure supplement 5C , D and F ) . This visual assay further confirms the intermembrane localization of IL-1β in the phagophore and autophagosome . In chaperone-mediated autophagy ( CMA ) , cargoes are recognized by a KFERQ sequence motif for transport into the lysosome ( Dice et al . , 1986; Kaushik and Cuervo , 2012 ) . We analyzed the primary sequence of IL-1β and found three KFERQ-like motifs on IL-1β including 127LRDEQ131 , 132QKSLV136 and 198QLESV202 ( Figure 7A ) . We mutated the glutamine , which has been shown to be essential for the function of the motif , as well as an adjacent amino acid in each motif ( E130Q131 , Q132K133 and Q198L199 ) to alanines and examined the secretion efficiency of these mutants . The 130-131AA mutant did not affect secretion of IL-1β ( Figure 7B ) . However , the Q132K133 and Q198L199 mutations were both defective in secretion of mature IL-1β which instead accumulated in the cytoplasmic fraction ( Figure 7B ) . A low level of release of the pro-forms persisted as seen with WT and mutant protein ( Figure 7B ) . The cytoplasmic mature forms of the mutant proteins were less abundant in the membrane fraction compared with the WT mature IL-1β ( Figure 7C , compare the lanes without proteinase K treatment ) . In addition , the membrane associated mutant IL-1β remained proteinase K accessible ( less than 10% of protection compared with ~45% of WT IL-1β ) , demonstrating that these two KFERQ-like motifs are required for the membrane translocation of IL-1β ( Figure 7C ) . Equal amounts of WT and mutant p-IL-1β associated with the membrane but both remained largely proteinase K accessible ( Figure 7C ) . 10 . 7554/eLife . 11205 . 018Figure 7 . Mutation of the KFERQ-like motif affects IL-1β secretion and entry into vesicles . ( A ) Protein sequence of IL-1β . The yellow region indicates mature IL-1β . Three KFERQ-like motifs ( aa127-131 , aa132-136 and aa198-202 ) are highlighted in red underlined bold . ( B ) Secretion of IL-1β mutants . HEK293T cells were transfected with p-IL-1β-DHFR and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were either treated with DMEM or EBSS for 2 hr . Secretion of IL-1β mutant proteins was detected by immunoblot . ( C ) IL-1β mutant 132-133AA or 198-199AA is accessible to proteinase K digestion . HEK293T cells were transfected with plasmids encoding p-caspase-1 and IL-1β mutant 132-133AA or 198-199AA . After transfection ( 24 hr ) , the cells were treated with EBSS for 2 hr . The 25k membrane fraction was collected and subjected to proteinase K digestion assay and then analyzed by immunoblot using indicated antibodies . The level of proteinase K protection was calculated as the percentage of the total protein . Error bars represent standard deviations of at least three experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 018 The chaperone proteins HSC70 and HSP90 have been reported to function in chaperone-mediated autophagy ( CMA ) ( Kaushik and Cuervo , 2012; Majeski and Dice , 2004 ) . HSP70 has also been implicated in autophagy and stress responses ( Murphy , 2013 ) . We performed shRNA-mediated knockdown of the three chaperone proteins to assess their potential role in the membrane translocation of IL-1β . Knockdown of Hsp90 , but not of Hsp70 or Hsc70 substantially reduced IL-1β secretion ( Figure 8A ) . As a control , knockdown of Hsc70 compromised CMA as indicated by the stabilization of a CMA cargo , GAPDH ( Figure 8—figure supplement 1A ) . Moreover , secretion of mature IL-1β was inhibited in a dose-dependent manner by an HSP90 inhibitor geldanamycin ( Figure 8B ) . In both experiments , mature IL-1β accumulated in the cytosol fraction at the expense of secretion . Knockdown of Hsp90 also rendered IL-1β accessible to proteinase K digestion ( Figure 8C ) , consistent with a role for HSP90 in the translocation of IL-1β as opposed to some later secretion event . Furthermore , in a co-immunoprecipitation assay , HSP90 associated with m-IL-1β but not the translocation-deficient mutants Q132K133 and Q198L199 ( Figure 8D ) . Although p-IL-1β also formed a complex with HSP90 , the efficiency appeared lower than for m-IL-1β . These results suggest that HSP90 binds to a region of the mature IL-1β , including the essential residues Q132K133 and Q198L199 , to promote the translocation event . Cleavage of p-IL-1β by caspase-1 may potentiate the recruitment of HSP90 to the mature form of IL-1β however chaperone binding is not required for this proteolytic event ( Figures 8D and 7B ) . 10 . 7554/eLife . 11205 . 019Figure 8 . HSP90 is involved in the entry of IL-1β into vesicles . ( A ) Knockdown of Hsp90 inhibits IL-1β secretion . HEK293T cells were transduced with lentivirus carrying control ( Ctrl ) shRNA or shRNA against Hsc70 , Hsp90 or Hsp70 . Then the cells were transfected with p-IL-1β and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were cultured in EBSS for 2 hr followed by determination of IL-1β secretion by immunoblot . ( B ) IL-1β secretion is reduced in the presence of HSP90 inhibitor geldanamycin . HEK293T cells were transfected with p-IL-1β and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were treated with EBSS containing different concentrations of geldanamycin as indicated . Immunoblot was performed as shown in Figure 1 . Quantification of IL-1β secretion was calculated as the ratio between the amount of IL-1β in the medium and the total amount ( the sum of IL-1β in both medium and lysate ) . ( C ) IL-1β remains accessible to proteinase K in Hsp90 knockdown cells . HEK293T cells were transduced with lentivirus carrying control ( Ctrl ) shRNA or shRNA against Hsp90 . Then the cells were transfected with p-IL-1β and p-caspase-1 plasmids . After transfection ( 24 hr ) , the cells were cultured in EBSS for 2 hr . The 25k membrane fraction was collected and digested with proteinase K and then analyzed by immunoblotting using indicated antibodies . ( D ) Association of HSP90 with IL-1β WT and mutants . HEK293T cells transfected with p-caspase-1 and IL-1β mutant 132-133AA or 198-199AA were starved in EBSS for 2 hr . Immunoprecipitation ( IP ) with anti-HSP90 antibody coupled to protein G-agarose was performed , followed by an immunoblot with anti-IL-1β and anti-HSP90 antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 01910 . 7554/eLife . 11205 . 020Figure 8—figure supplement 1 . Translocation of IL-1β is mechanistically different from CMA . ( A ) Knockdown of Hsc70 reduces CMA . HEK293T cells transduced with lentivirus carrying control ( Ctrl ) shRNA or shRNA against Hsc70 were incubated with regular medium ( -CMA ) or DMEM ( +CMA ) in the presence of 20 μg/ml cycloheximide for 24 hr . The cells were lysed and analyzed by immunoblotting using indicated antibodies . For quantification , the ratio of GAPDH and tubulin was calculated and normalized by that in control ( -CMA ) treatment which was set as one . ( B ) Co-immunoprecipitation of HSC70 or HSP90 with IL-1β . HEK293T cells transfected with m-IL-1β-FLAG were starved in EBSS for 2 hr . Immunoprecipitation ( IP ) with anti-HSC70 or anti-HSP90 antibody coupled to protein A/G-agarose was performed , followed by an immunoblot with indicated antibodies . ( C ) Knockdown of Lamp2 blocks CMA . HEK293T cells were transfected with control or LAMP2 siRNA . After transfection ( 48 hr ) , the cells were trypsinized and plated . After 24 hr , siRNA transfection was repeated . After another 48 hr , the cells were trypsinized and plated . After 24 hr , the cells were incubated with regular medium ( -CMA ) or DMEM ( +CMA ) in the presence of 20 μg/ml cycloheximide for 24 hr . The cells were lysed and analyzed by immunoblotting using indicated antibodies . For quantification , the ratio of GAPDH and Tubulin was calculated and normalized by that in control ( -CMA ) treatment which was set as one . ( D ) Knockdown of LAMP2 does not affect IL-1β secretion . HEK293T cells were transfected with control or LAMP2 siRNA as show in ( C ) . After the second siRNA transfection ( 24 hr ) , the cells were transfected with m-IL-1β-FLAG plasmid . After transfection ( 24 h ) , the cells were either cultured in DMEM or EBSS for 2 hr followed by determination of IL-1β secretion by immunoblot as shown in Figure 1A . Quantification of IL-1β secretion was calculated as the ratio between the amount of IL-1β in the medium and the total amount ( the sum of IL-1β in both medium and lysate ) . ( E ) Level of IL-1β in the membrane fraction was not affected by lysosome disruption . HEK293T cells transfected with m-IL-1β were cultured in EBSS for 2 hr and then treated with DMSO or 0 . 5 mM glycyl-L-phenylalanine-2-naphthylamide ( GPN ) for 10 min . The membrane fraction was collected from the top fractions of a Nycodenz density gradient resolved from membranes in a 25k pellet as described in Material and Methods . Both membrane fraction and cell lysis were analyzed by immunoblotting using indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 020 In the CMA pathway , HSC70 and HSP90 play different roles . HSC70 binds to cargoes and delivers them into the lysosome as well as disassembling LAMP2A oligomers , whereas HSP90 is required for the oligomerization and stability of LAMP2A ( Bandyopadhyay et al . , 2008; Chiang et al . , 1989 ) . Co-immunoprecipitation indicated that IL-1β associates with HSP90 but not HSC70 ( Figure 8—figure supplement 1B ) . In addition , knockdown of Lamp2A compromised CMA but did not affect the secretion of IL-1β , and disruption of the lysosome did not result in the release of IL-1β from the membrane carrier ( Figure 8—figure supplement 1C–E ) . These data suggest that the translocation of IL-1β into the vesicle carrier is mechanistically distinct from CMA . We next asked if starvation regulated the association between HSP90 and IL-1β . We performed an HSP90 co-immunoprecipitation experiment with cytosol prepared from cells grown in nutrient-rich or starvation conditions ( Figure 9A ) . Starvation led to a ~2 . 5 fold increase of the association of HSP90 and IL-1β ( Figure 9A ) . This increase was likely not due to starvation-stimulated processing of p-IL-1β because starvation had no effect on the cleavage of mutant forms of IL-1β unable to bind HSP90 ( Figure 7B ) . Starvation led to a ~ 2 fold increase in the membrane localization and cytosolic depletion of mature IL-1β ( Figure 9B ) . Starvation may stimulate the recruitment of a complex of m-IL-1β/HSP90 to the membrane responsible for IL-1β translocation ( Figure 9B ) . 10 . 7554/eLife . 11205 . 021Figure 9 . Induction of autophagy enhances the membrane incorporation of IL-1β . ( A ) Starvation enhances the association of IL-1β with HSP90 . HEK293T cells transfected with p-IL-1β and p-caspase-1 were cultured in DMEM or EBSS for 2 hr . Immunoprecipitation with anti-HSP90 antibody was performed followed by an immunoblot with anti-IL-1β and anti-HSP90 antibodies . ( B ) Starvation promotes the entry of IL-1β into the membrane fraction . HEK293T cells transfected with p-IL-1β and p-caspase-1 were cultured in DMEM or EBSS for 2 hr . The membrane fraction was collected from the top fractions of a Nycodenz density gradient resolved from membranes in a 25k pellet as described in Material and Methods . The cytosolic fraction was collected as the supernatant after 100k×g centrifugation . Immunoblot was performed to determine the levels of IL-1β in both fractions . ( C ) A proposed model for autophagy-mediated IL-1β secretion . Cytosolic IL-1β associates with HSP90 which facilitates the translocation of IL-1β into the lumen of a vesicle carrier which later either turns into a phagophore and an autophagosome or fuses with them . IL-1β localizes between the outer and inner membrane after the double membrane autophagosome forms . The topological distribution ensures the secretion of IL-1β in a soluble form . The IL-1β-containing autophagosome may directly fuse with the plasma membrane or further fuse with a MVB followed by fusion with the plasma membrane . DOI: http://dx . doi . org/10 . 7554/eLife . 11205 . 021 Genetic and cell biological studies have implicated autophagy in the transport of several leaderless cargoes to the extracellular space ( Bruns et al . , 2011; Dupont et al . , 2011; Duran et al . , 2010; Manjithaya et al . , 2010 ) . Unconventional secretory cargoes , such as IL-1β and Acb1 , have been shown to have overlapping requirements with formation of the autophagosome or its precursor suggesting that the autophagosome may physically convey these cargo proteins to the cell surface . A key question is if and how these cargoes engage the autophagosome and how this structure exports soluble cargo molecules . In this study , we probed the organelle association and molecular requirements for the secretion of one such unconventional cargo protein , IL-1β . Using surrogate cell lines rather than macrophages to reconstitute autophagy-mediated secretion of IL-1β ( Figure 1 ) , we find mature IL-1β localized to the lumen of the membrane in early intermediates and mature autophagosomes ( Figures 2–4 , 6 ) . This surprising location may help to explain how mature IL-1β is secreted in a soluble form to the cell surface ( Figure 9C ) . However , localization to the lumen between the two membranes of the autophagosome would require that IL-1β is translocated from the cytoplasm across the membrane precursor of a phagophore , rather than being engulfed as the phagophore membrane matures by closure into an autophagosome . Our evidence suggests that IL-1β must unfold or be held in an unfolded state to promote membrane translocation ( Figure 5 ) and that a complex sorting signal in the mature portion of IL-1β interacts with HSP90 to deliver the chaperone and its cargo to a site on a phagophore precursor membrane where the mature species is translocated ( Figures 7–9 ) . The unconventional secretory cargo fibroblast growth factor 2 ( FGF2 ) has been shown to directly translocate across the plasma membrane as a folded protein without the apparent aid of chaperones ( Backhaus et al . , 2004; Steringer et al . , 2015 ) . Unlike FGF2 , the entry of IL-1β into the autophagosomal carrier appears to be dependent on protein unfolding in a conformational state that may be fostered by the association of HSP90 with two KFERQ-like sequences within the mature portion of IL-1β ( Figure 5 and 8 ) . This translocation mechanism appears superficially similar to another delivery process termed HSC70-dependent CMA in which autophagic cargoes bearing KFERQ targeting motifs are directed into the lysosome for degradation . Indeed , using a cell-free approach to study the import of CMA cargo into isolated lysosomes , Salvador et al . ( 2000 ) reported that DHFR fused to a CMA cargo is blocked in translocation by addition of methotrexate , a drug that stabilizes DHFR to unfolding , just as we have shown that IL-1β fused to DHFR is blocked in cells treated with a cell permeable folate analog , aminopterin ( Wei et al . , 2013 ) . In our fractionation study , IL-1β distributed in LC3-positive autophagosomal carriers that were separated from the lysosome marker LAMP2 , the proposed receptor or channel for uptake of CMA cargo ( Kaushik and Cuervo , 2012 ) ( Figure 2B ) . This observation , together with the involvement of a different chaperone i . e . HSP90 , suggests distinct routes for IL-1β and cargoes of the CMA pathway . The target membrane for IL-1β translocation may be a vesicle that could fuse with or form the autophagosome . We find that mature IL-1β can be detected within protease inaccessible membranes in cells blocked early in the autophagic pathway ( e . g . ATG5 null cells and cells depleted of FIP200 , both of which block at a stage prior to the lipidation of LC3 ) . The identity of the vesicle carrier is unknown and could be any one of those reported to supply membrane to the formation of the autophagosome ( Ge et al . , 2014a; Lamb et al . , 2013 ) . Although we have ruled out the involvement of LAMP2A IL-1β translocation , it is likely that a membrane receptor locating on the membrane of the vesicle carrier , perhaps a functional equivalent of LAMP2A , recruits the protein complex of HSP90 and IL-1β , therefore designating the correct membrane targeting of IL-1β . In addition , a protein conducting channel may be involved in the translocation of IL-1β into the membrane . It seems unlikely that a standard translocation channel , such as SEC61 , is involved in this import process , but no current evidence bears on this point . The exact route by which the autophagosome delivers mature IL-1β to the cell surface as well as how it avoids fusion with degradative lysosome remains obscure , possibly involving interaction with the multi-vesicular body or some form of lysosome as a prelude to fusion at the cell surface ( Figure 9C ) , and this process may require selective recruitment of membrane sorting and targeting factors such as Rabs and SNAREs . Fusion of the autophagosome directly with the plasma membrane would lead to the release of soluble IL-1β available to trigger an inflammatory response in the surrounding tissue . If mature IL-1β were engulfed within the cytoplasmic interior of the autophagosome , fusion of this organelle at the cell surface might release an intact vesicle corresponding to the inner membrane-enclosed cytoplasmic compartment of the autophagosome . We found mature IL-1β secreted by macrophages or in our surrogate cell system to be completely soluble , thus inconsistent with the engulfment model ( data not shown ) . An alternative possibility may be that the autophagosome fuses with another intracellular organelle such as the MVB or the lysosome under conditions where the inner membrane of the autophagosome is degraded . If so , mature IL-1β would be available for secretion if the combined organelle ( amphisome , Figure 9C ) fused with the plasma membrane . However , for this model to be viable , the mature IL-1β released on dissolution of the autophagosome inner membrane would have to withstand proteolytic attack such as may be encountered in an amphisome or lysosome . Because mature IL-1β is clearly sensitive to proteolysis ( Figure 4 ) , thus any compartment engaged in presenting autophagosomal content to the cell surface must be depleted of proteases . The nature of the organelle that delivers autophagosome content to the plasma membrane may be probed by selective ablation of different Rab proteins , e . g . Rab11 , Rab27 and Rab35 , which appear to be required for fusion of the MVB with the cell surface ( Hsu et al . , 2010; Ostrowski et al . , 2010; Savina et al . , 2002 ) , or Rab27a and Rab38 , implicated in the fusion of lysosomes at the cell surface ( Blott and Griffiths , 2002; Hume et al . , 2001; Jager et al . , 2000 ) . The plasmid encoding p-IL-1β was kindly provided by Russell Vance lab ( University of California , Berkeley ) . The plasmids encoding FLAG-tagged p-caspase-1 and ATG4B ( C74A ) were from Addgene ( Cambridge , MA ) . The 3×FLAG-LC3 plasmid was generated by PCR insertion of a 3×FLAG peptide into MYC-LC3 ( provided by the Zhong lab , UT Southwestern , Dallas ) . P-IL-1β mutants 130-131AA , 132-133AA and 198-199AA were generated by PCR-based site-directed mutagenesis . The p-IL-1β-DHFR plasmid was constructed by subcloning the DHFR peptide from MTS-GFP-DHFR ( provided by the Nickel lab , Heidelberg , Germany ) into pro-IL-1β . The plasmid encoding the FLAG-tagged m-IL-1β was constructed by deleting the sequence encoding AA2-117 of pro-IL-1β and inserting a DNA sequence encoding a single FLAG before the stop codon by PCR-based mutagenesis . The resulting protein is the m-IL-1β with a FLAG at the C-terminus . Small interference siRNAs against Hrs or TSG101 were purchased from Qiagen ( Valencia , CA ) . The target sequence against Hrs was CCGGAACGAGCCCAAGTACAA . The target sequence against TSG101 was CAGTTTATCATTCAAGTGTAA . A pool of four siRNAs against Atg2A , Atg2B , FIP200 , LAMP2 , GRASP55 or GRASP65 was purchased from Qiagen ( Valencia , CA; GeneSolution siRNAs ) , Dharmacon ( Lafayette , CO; siGENOME SMART pool siRNAs ) or Thermo Scientific ( Rockford , IL; siGENOME SMART pool siRNAs ) . HEK293T and U2OS cells were grown in a tissue culture facility . Atg5 KO and WT MEFs were generously provided by Noboru Mizushima ( University of Tokyo , Japan ) . Cells were grown at 37°C in 5% CO2 and maintained in Dulbecco’s modified Eagle’s medium ( DMEM ) supplemented with 10% FBS . For starvation , the cells were incubated in Earle’s Balanced Salt Solution ( EBSS ) for the indicated time durations in the absence or presence of the drugs indicated in the manuscript . Transfection of DNA constructs into cells was performed using X-tremeGENE HP ( Roche , Indianapolis , IN ) according to the manufacture's protocols . SiRNA transfection was performed on HEK293T cells with lipofectamine RNAiMAX ( Invitrogen , Carlsbad , CA ) according to the manufacture's protocols . ShRNA constructs targeting Hsc70 , Hsp90 and Hsp70 were inserted into pLKO . 1-puro vector ( Addgene ) . As previously described ( Hubbi et al . , 2013; Zhong et al . , 2011; Zuo et al . , 2012 ) , the following sequences were used: Ctrl , CAACAAGATGAAGAGCACCAA; Hsc70-A , GCCCGATTTGAAGAACTGAAT; Hsc70-B , GCAACTGTTGAAGATGAGAAA; Hsp90 , CCTGTGGATGAATACTGTATT; Hsp70 , GGCCAACAAGATCACCATC . For lentiviral transduction , plasmid pLKO . 1 carrying Ctrl , Hsc70 , Hsp90 or Hsp70 was transfected into HEK293T cells along with lentiviral packaging plasmids pMD2 . G and psPAX2 ( Addgene ) using X-tremeGENE HP to produce lentiviral particles to infect HEK293T cells . Transduced cells were selected using 2 μg/ml Puromycin ( Invitrogen ) . We obtained EBSS from Invitrogen ( Grand Island , NY ) ; 3-methyladenine ( 3-MA ) , wortmannin ( Wtm ) , aminopterin , geldanamycin , proteinase K , 4- ( 2-Aminoethyl ) -benzenesulfonyl fluoride hydrochloride ( AEBSF ) , cycloheximide , anti-FLAG M2 agarose and 3×FLAG tag peptide from Sigma ( St . Louis , MO ) ; Protein G-Sepharose beads from GE Healthcare ( Piscataway , NJ ) ; glycyl-L-phenylalanine-2-naphthylamide ( GPN ) from Santa Cruz Biotechnology ( Dallas , TX ) . Mouse anti-GM130 , anti-transferrin receptor ( TFR ) , anti-PMP70 , anti-FLAG , anti-PDI , anti-tubulin and rabbit anti-Prohibitin-1 , anti-RPN1 , anti-SEC22B , anti-LAMP2 , anti-LC3 and anti-ERGIC53 antibodies were described before ( Ge et al . , 2013 , 2014b ) . We purchased goat anti-IL-1β antibody from R&D Systems ( Minneapolis , MN ) ; rabbit anti-IL-1β , anti-LAMP2A and mouse anti-HSC70 antibodies from Abcam ( Cambridge , MA ) ; rabbit anti-Caspase-1 , rabbit anti-CD63 , mouse anti-GAPDH and goat anti-GRASP65 antibodies from Santa Cruz ( Dallas , TX ) ; rabbit anti-ATG2A and anti-p62 antibodies from MBL ( Woburn , MA ) ; rabbit anti-FLAG and anti-ATG2B antibodies from Sigma ( St . Louis , MO ) ; mouse anti-FIP200 antibody from Millipore ( Billerica , MA ) ; mouse anti-Hrs antibody from Enzo Life Sciences ( Lörrach , Germany ) ; mouse anti-TSG101 antibody from GeneTex ( San Antonio , TX ) ; mouse anti-HSP90 antibody from EMD ( Billerica , MA ) ; mouse anti-HSP70 antibody from Enzo Life Sciences ( Plymouth Meeting , PA ) ; mouse anti-DHFR antibody from BD Biosciences Pharmingen ( San Diego , CA ) ; rabbit anti-GRASP55 antibody from ProteinTech Group ( Chicago , IL ) . HEK293T or MEF cells were transfected with plasmids encoding different forms of p-IL-1β , p-caspase-1 as well as other plasmids as indicated in figure legends . After transfection ( 24 hr ) , cell culture media were replaced with DMEM , EBSS or EBSS containing indicated drugs for indicated time durations . Media were collected and concentrated ( 20-fold ) by a 10 kD Amicon filter ( Millipore , Billerica , MA ) . Cells were lysed in SDS-PAGE loading buffer and analyzed by immunoblot analysis . For RNAi experiments , cells were transfected with indicated siRNAs . After 6 hr , a similar plasmid transfection indicated above was performed . After 60 hr , IL-1β secretion was determined . The procedure is modified from our previous work ( Ge et al . , 2013 ) . Cells ( ten 10-cm dishes ) were cultured to confluency , harvested and homogenized in a 2 . 7× cell pellet volume of B1 buffer ( 20 mM HEPES-KOH , pH 7 . 2 , 400 mM Sucrose , 1 mM EDTA ) plus a cocktail of protease and phosphatase inhibitors ( Roche , Indianapolis , IN ) and 0 . 3 mM DTT by passing through a 22 G needle until ~85% lysis analyzed by Trypan Blue staining . Homogenates were subjected to sequential differential centrifugation at 3 , 000×g ( 10 min ) , 25 , 000×g ( 20 min ) and 100 , 000×g ( 30 min , TLA100 . 3 rotor , Beckman ) to collect the membranes sedimented at each speed . The 25 , 000×g membrane pellet , which contained the highest level of IL-1β , was suspended in 0 . 75 ml 1 . 25 M sucrose buffer and overlaid with 0 . 5 ml 1 . 1 M and 0 . 5 ml 0 . 25 M sucrose buffer ( Golgi isolation kit; Sigma ) . Centrifugation was performed at 120 , 000×g for 2 hr ( TLS 55 rotor , Beckman ) , after which two fractions , one at the interface between 0 . 25 M and 1 . 1 M sucrose ( L fraction ) and the pellet on the bottom ( P fraction ) , were separated . IL-1β protein levels of the two fractions were then tested and the L fraction was selected and suspended in 1 ml 19% OptiPrep for a step gradient containing 0 . 5 ml 22 . 5% , 1 ml 19% ( sample ) , 0 . 9 ml 16% , 0 . 9 ml 12% , 1 ml 8% , 0 . 5 ml 5% and 0 . 2 ml 0% OptiPrep each . Each density of OptiPrep was prepared by diluting 50% OptiPrep ( 20 mM Tricine-KOH , pH 7 . 4 , 42 mM sucrose and 1mM EDTA ) with a buffer containing 20 mM Tricine-KOH , pH 7 . 4 , 250 mM sucrose and 1mM EDTA . The OptiPrep gradient was centrifuged at 150 , 000×g for 3 hr ( SW 55 Ti rotor , Beckman ) and subsequently ten fractions , 0 . 5 ml each , were collected from the top . Fractions were diluted with B88 buffer ( 20 mM HEPES-KOH , pH 7 . 2 , 250 mM sorbitol , 150 mM potassium acetate and 5 mM magnesium acetate ) and membranes were collected by centrifugation at 100 , 000×g for 1 hr . Samples were normalized using a measured level of phosphatidylcholine ( Ge et al . , 2013 ) and subjected to SDS-PAGE followed by immunoblot analysis with the indicated antibodies . Cells ( five 10-cm dishes ) transfected with indicated plasmids were starved in EBSS for 2 hr and harvested as indicated above . Membranes from a 25 , 000×g membrane pellet were resuspended in 300 μl 60% ( wt/vol ) Nycodenz ( Accurate Chemical , Westbury , NY ) in B88 buffer and transferred to a Beckman tube ( Polycarbonate , 11 × 34 mm ) . Aliquots were overlaid with 600 μl of 40% Nycodenz in B88 buffer and 100 μl B88 buffer , and then centrifuged for 2 hr at 100 , 000×g ( TLS 55 rotor , Beckman ) . Ten fractions were collected from top to bottom and analyzed by immunoblot . For determining the level of IL-1β in the membrane fraction , top fractions were combined and diluted with B88 buffer and membranes were collected by centrifugation at 100 , 000×g for 40 min followed by immunoblot analysis . HEK293T cells ( ten 10-cm dishes ) transfected with p-IL-1β , p-caspase-1 and 3×FLAG-LC3 were starved in EBSS for 2 hr and harvested as indicated above . Membranes from a 25 , 000×g pellet were collected , resuspended in immunoisolation buffer containing 25 mM HEPES , pH 7 . 4 , 140 mM potassium chloride , 5 mM sodium chloride , 2 . 5 mM magnesium acetate , 50 mM sucrose and 2 mM EGTA . Anti-FLAG M2 agarose was added to a 1 ml membrane suspension with or without 0 . 2 mg/ml 3×FLAG tag blocking peptides and mixed by rotation at 4°C overnight . Beads with the associated membranes were washed with 1 ml immunoisolation buffer three times and membranes bound to the beads were eluted by incubating with 0 . 5 mg/ml of 3×FLAG peptides for 0 . 5 hr at room temperature . The eluted membranes were collected by centrifuging at 100 , 000×g for 40 min and analyzed by immunoblot . The 25 , 000×g membrane pellet separated from cell homogenates were aliquoted into several fractions and resuspended in 30 μl of B88 or B88 containing indicated concentrations of proteinase K with or without 0 . 5% Triton X-100 , and stored on ice for 30 min . The reactions were stopped by sequentially adding AEBSF and 3×SDS loading buffer , which were then heated in boiling water for 10 min and analyzed by immunoblot . Co-immunoprecipitation was performed as previously described ( Zhang et al . , 2010 ) . Briefly , 24 hr after transfection , the cells were lysed on ice for 30 min in lysis buffer ( 50 mM Tris/HCl pH 7 . 4 , 150 mM NaCl , 1 mM EDTA , 1% Triton X-100 , 10% glycerol ) with protease inhibitor mixture , and the lysates were cleared by centrifugation . The resulting supernatants were incubated with mouse anti-HSP90 or a mouse IgG control antibody overnight at 4°C and then precipitated with Protein G-Sepharose beads for 2 hr at 4°C . After washing , 3×SDS loading buffer was added to the beads and immunoblot was performed . U2OS cells were starved in EBSS for 1 hr and permeabilized with 40 μg/ml of digitonin diluted in PBS on ice for 5 min . The cells were then washed once with cold PBS and immediately incubated with 4% cold paraformaldehyde for 20 min at room temperature . The cells were further permeabilized with 0 . 1% saponin diluted in PBS at room temperature for 10 min followed by blocking with 10% FBS diluted with PBS for 1 hr and primary antibody incubation for 1 hr . For confocal microscopy , procedures were as described previously ( Ge et al . , 2014b ) . For STORM , cells were washed three times with 0 . 2% BSA in PBS , followed by incubation with CF568 anti-mouse ( Biotium ) and Alexa Fluor 647 anti-rabbit ( Invitrogen ) secondary antibodies in 3% BSA in PBS for 1 hr at room temperature . Cells were washed three times before mounting for STORM imaging . Dye-labeled cell samples were mounted on glass slides with a standard STORM imaging buffer consisting of 5% ( w/v ) glucose , 100 mM cysteamine , 0 . 8 mg/mL glucose oxidase , and 40 µg/mL catalase in 1M Tris-HCI ( pH 7 . 5 ) ( Huang et al . , 2008; Rust et al . , 2006 ) . Coverslips were sealed using Cytoseal 60 . STORM imaging was performed on a homebuilt setup based on a modified Nikon Eclipse Ti-U inverted fluorescence microscope using a Nikon CFI Plan Apo λ 100x oil immersion objective ( NA 1 . 45 ) . Dye molecules were photoswitched to the dark state and imaged using either 647- or 560-nm lasers ( MPB Communications ) ; these lasers were passed through an acousto-optic tunable filter and introduced through an optical fiber into the back focal plane of the microscope and onto the sample at intensities of ~2 kW cm-2 . A translation stage was used to shift the laser beams towards the edge of the objective so that light reached the sample at incident angles slightly smaller than the critical angle of the glass-water interface . A 405-nm laser was used concurrently with either the 647- or 560-nm lasers to reactivate fluorophores into the emitting state . The power of the 405-nm laser ( typical range 0–1 W cm-2 ) was adjusted during image acquisition so that at any given instant , only a small , optically resolvable fraction of the fluorophores in the sample were in the emitting state . Emission was recorded with an Andor iXon Ultra 897 EM-CCD camera at a framerate of 110 Hz , for a total of ~80 , 000 frames per image . Multicolor imaging was performed by imaging each color channel separately and sequentially using separate emission filters and excitation lasers . For 3D STORM imaging , a cylindrical lens was inserted into the imaging path so that images of single molecules were elongated in opposite directions for molecules on the proximal and distal sides of the focal plane ( Huang et al . , 2008 ) . The raw STORM data was analyzed according to previously described methods ( Huang et al . , 2008; Rust et al . , 2006 ) .	Cells release a large number of proteins to the extracellular space . The majority of these ‘secreted’ proteins first pass through two structures inside cells called the endoplasmic reticulum and Golgi . However , a growing number of proteins have been identified that are released by an unconventional mechanism that bypasses the endoplasmic reticulum and Golgi . Autophagy is a process that destroys damaged proteins and other unwanted material in cells . It gets triggered when cells are starved of nutrients , leading them to digest their own materials and recycle the resources into new molecules . During autophagy , a cup-like structure with a double layer of membrane forms around the material that is to be digested . This structure then elongates and eventually engulfs the material to form a bubble-like compartment called the autophagosome . Recent evidence has suggested that autophagosomes are involved in the unconventional secretion of a protein called interleukin-1β; this protein is crucial for the body’s immune response against infection . However , it was not clear how these proteins entered the autophagosomes . Zhang et al . have now explored the link between interleukin-1β and autophagy in more detail . The experiments showed that when autophagy was triggered by starvation , the secretion of interleukin-1β was enhanced . Conversely , when autophagy was inhibited , interleukin-1β accumulated inside the cells and could not be secreted . Further experiments then revealed unexpectedly that interleukin-1β was not engulfed by the cup-like structure ( as is the case for material that is destined to be removed ) . Instead , interleukin-1β was found to enter into smaller bubble-like packages ( called vesicles ) that turn into the autophagosome . Zhang et al . also found that a protein called HSP90 binds to interleukin-1β and enables it to cross the membrane ( or translocate ) into the vesicles , and that this means that interleukin-1β actually resides in the space between the outer and inner membranes of the autophagosome . How many other proteins share this unusual route out of the cell and what membrane channel is used for this translocation event remain open questions for the future .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "cell", "biology" ]	2015	Translocation of interleukin-1β into a vesicle intermediate in autophagy-mediated secretion
Binary fission has been well studied in rod-shaped bacteria , but the mechanisms underlying cell division in spherical bacteria are poorly understood . Rod-shaped bacteria harbor regulatory proteins that place and remodel the division machinery during cytokinesis . In the spherical human pathogen Staphylococcus aureus , we found that the essential protein GpsB localizes to mid-cell during cell division and co-constricts with the division machinery . Depletion of GpsB arrested cell division and led to cell lysis , whereas overproduction of GpsB inhibited cell division and led to the formation of enlarged cells . We report that S . aureus GpsB , unlike other Firmicutes GpsB orthologs , directly interacts with the core divisome component FtsZ . GpsB bundles and organizes FtsZ filaments and also stimulates the GTPase activity of FtsZ . We propose that GpsB orchestrates the initial stabilization of the Z-ring at the onset of cell division and participates in the subsequent remodeling of the divisome during cytokinesis . Bacterial cell division has been extensively studied in rod-shaped organisms such as Escherichia coli and Bacillus subtilis ( Adams and Errington , 2009; Lutkenhaus et al . , 2012; Rowlett and Margolin , 2015; Tsang and Bernhardt , 2015 ) . However , spherical bacteria lack several key components found in these well-studied model organisms ( Pinho et al . , 2013 ) , so fundamental features of how they divide are poorly understood . The Gram-positive human pathogen Staphylococcus aureus is a spherical bacterium that is commensal in ~30% of the U . S . population ( Kuehnert et al . , 2006 ) , but in immunocompromised individuals , it is a leading cause of bacteremia and nosocomial infections in industrialized nations ( Klevens et al . , 2007 ) . The emergence of several antibiotic resistant strains of S . aureus has necessitated the identification of novel antibiotic targets ( Pendleton et al . , 2013 ) . In recent years , components of the bacterial cell division machinery have been proposed as such targets ( Lock and Harry , 2008; Sass and Brötz-Oesterhelt , 2013 ) . GpsB is a small coiled-coil cell division protein ( Claessen et al . , 2008; Rismondo et al . , 2016; Tavares et al . , 2008 ) that is widely conserved in the Firmicutes phylum and is conditionally required for growth in certain species , depending on growth media and temperatures ( Claessen et al . , 2008; Fleurie et al . , 2014; Land et al . , 2013; Rismondo et al . , 2016; Tavares et al . , 2008 ) . GpsB is highly co-conserved ( Pinho et al . , 2013 ) with the cell division protein DivIVA . Like DivIVA , GpsB is relatively small and harbors a highly homologous N-terminal α-helical domain . However , the C-terminus differs from that of DivIVA: whereas DivIVA assembles into an anti-parallel tetramer , the GpsB structure was reported to hexamerize with a parallel alignment of helices ( Rismondo et al . , 2016 ) . Similar to DivIVA ( Kaval and Halbedel , 2012 ) , GpsB orthologs perform slightly different functions in different species . In the rod-shaped Bacillus subtilis and Listeria monocytogenes , GpsB participates in shuttling a cell wall assembly protein ( PBP1 or PBP A1 , respectively ) to help maintain the characteristic rod shape of the bacterium ( Claessen et al . , 2008; Rismondo et al . , 2016 ) . In the ovoid-shaped Streptococcus pneumoniae , GpsB additionally has been reported to interact with PBP2a , PBP2b , and MreC , and has been implicated in recruiting a Ser/Thr kinase to mid-cell that activates cell wall assembly machinery specifically at the division septum , thereby modulating a switch between peripheral and medial cell wall assembly to again maintain the proper shape of the cell ( Fleurie et al . , 2014; Rued et al . , 2017 ) . In all reported cases , GpsB interacts with a peripheral divisome component , EzrA , but not necessarily core components of the division machinery , to mediate its role in cell shape maintenance ( Claessen et al . , 2008; Fleurie et al . , 2014; Rued et al . , 2017; Steele et al . , 2011 ) . In S . aureus , GpsB is an essential protein ( Santiago et al . , 2015 ) ( M . Santiago , personal communication ) , but its cellular function is poorly understood . Herein , we report that GpsB interacts directly with bacterial tubulin homolog FtsZ , the core component of the division machinery , and orchestrates the dynamics of its assembly . In vivo , we show that GpsB localizes to mid-cell at the onset of cell division and co-constricts with the divisome during cytokinesis . Depletion of GpsB in vivo arrested cell division and prevented the robust assembly of the divisome at mid-cell . In vitro , we show that purified GpsB promotes lateral interactions between FtsZ polymers in a manner reminiscent of bundling , thereby increasing the local concentration of FtsZ , and organizes the polymers . Unlike other proteins that exhibit FtsZ bundling activity , GpsB stimulated FtsZ GTPase activity . Consistent with this activity , overproduction of GpsB in vivo inhibited cell division and resulted in the production of large S . aureus cells . Our data suggest that , compared to GpsB orthologs in other Gram-positive bacteria , S . aureus GpsB plays a significantly different role by directly interacting with central component of the division machinery to regulate the remodeling of the divisome during cytokinesis: first , by bundling and stabilizing FtsZ polymers at mid-cell by promoting lateral interactions between FtsZ filaments , which increases the local concentration and triggers the GTPase activity of FtsZ and allows cytokinesis to proceed . To initially investigate if Staphylococcal GpsB ( GpsBSa ) performs a similar function as the B . subtilis GpsB ortholog ( GpsBBs ) , we expressed gpsBSa under the control of an inducible promoter in B . subtilis . In the presence of inducer , otherwise WT B . subtilis harboring either gpsBSa or gpsBSa-GFP exhibited a severe growth defect ( Figure 1 ) . In contrast , cells similarly expressing gpsBBs or gpsBBs-GFP did not exhibit a growth defect ( Figure 1A ) , suggesting that cell toxicity was specifically due to expression of the S . aureus ortholog of gpsB ( Figure 1—figure supplement 1A ) . Immunoblotting with antisera specific to GpsBSa revealed a ~ 3 . 2 fold overproduction of GpsBSa at a population level in the presence of inducer ( Figure 1—figure supplement 1B; note that the anti-GpsBSa antiserum did not recognize GpsBBs ) . In the absence of inducer , B . subtilis cells harboring gpsBSa examined by epifluorescence microscopy were of uniform length and displayed division septa at mid-cell ( Figure 1B ) , but in the presence of inducer , these cells were filamentous with segregated chromosomes that rarely elaborated division septa ( Figure 1C ) . GpsB interacts with several cell division proteins in different Gram-positive bacteria ( Claessen et al . , 2008; Cleverley et al . , 2016; Pompeo et al . , 2015 ) . Deletion of ezrA , ponA , prkC , or gpsB resulted in minor morphological defects in B . subtilis , but overproduction of GpsBSa in these strain backgrounds nonetheless resulted in filamentation ( Figure 1D–K ) . Additionally , while deletion of divIVA resulted in cell elongation ( Edwards and Errington , 1997 ) , overproduction of GpsBSa in the absence of DivIVA resulted in further filamentation ( Figure 1L–M ) . Thus , the B . subtilis filamentation phenotype resulting from GpsBSa overproduction does not require these peripheral cell division factors . We next examined if GpsBSa affects FtsZ localization . In the absence of inducer , FtsZ-GFP localized properly to mid-cell at incipient and active sites of cell division ( Figure 1N ) . However , upon overproduction of GpsBSa , filamentous cells displayed diffuse localization of FtsZ-GFP in the cytosol ( Figure 1O ) , suggesting that GpsBSa overproduction interferes with the localization of the central component of the B . subtlis cell division machinery . In B . subtilis , gpsB is not essential for growth , but deletion of ezrA ( a peripheral component of the divisome ) together with gpsB is synthetically lethal ( Claessen et al . , 2008 ) . In B . subtilis cells harboring a gpsB deletion and expressing gpsBBs , we obtained 865 ± 157 transformants harboring a deletion in ezrA ( n = 3 , per ~400 ng of transformed DNA containing ezrA deletion ) , whereas we did not recover any transformants when we attempted to delete ezrA in ∆gpsB cells that expressed gpsBSa , indicating that gpsBSa was unable to complement the gpsBBs deletion phenotype . Together with the different phenotypes observed upon overexpression of either gpsBBs or gpsBSa in B . subtilis , the data suggested that S . aureus GpsB may exhibit a different function or activity . To test the effect of GpsBSa overproduction in Staphylococci , we cloned gpsBSa in a high copy plasmid under control of an inducible promoter , introduced the construct into S . aureus strain SH1000 , stained the cells with a fluorescent membrane dye , and examined cell size using epifluorescence microscopy . Immunoblotting revealed an ~5 . 4 fold overproduction of GpsB at a population level relative to endogenous levels of GpsB ( Figure 1—figure supplement 1C ) . 100% of WT cells we observed ( n = 676 ) were less than 1 . 2 μm in diameter , as were WT cells harboring the empty vector ( n = 100 ) ( Figure 2A–D ) . In the absence of inducer , 6 . 4% ( n = 971 ) of cells harboring the inducible copy of gpsBSa were larger than 1 . 2 μm in diameter; in the presence of inducer , 56 . 9% ( n = 770 ) of cells were larger than 1 . 2 μm ( Figure 2E–F ) . Interestingly , overproduction of gpsBBs did not result in a similar enlargement of S . aureus cells ( Figure 2G–H ) , suggesting that the cell division inhibition phenotype in B . subtilis and S . aureus was unique to the overproduction of the S . aureus ortholog of GpsB . Quantification of cell diameters of 200 individual cells overproducing GpsBSa revealed a range of cell diameters higher than 1 . 2 µm in over half of the cells ( Figure 2K ) . The variation in cell diameters was likely due to unequal expression of gpsBSa in every cell , since control experiments in which gfp was placed under control of the inducible promoter revealed that only ~34% of cells ( n = 263 ) produced GFP in the presence of inducer . The gpsB gene is essential for viability in S . aureus ( Santiago et al . , 2015 ) . Consistent with this observation , we were unable to knockout the gene , even in the presence of a complementing multicopy plasmid , presumably due to the disruptive overproduction phenotype described above . We therefore sought to deplete GpsB by overexpressing gpsB antisense RNA under the control of an inducible promoter from a multicopy plasmid and examined the morphology of cells using fluorescence microscopy ( antisense resulted in ~2 . 5-fold reduction in GpsB; Figure 1—figure supplement 1F ) . Immediately after addition of the inducer , cells harboring this construct were morphologically similar to cells harboring the empty vector ( Figure 2L–M ) . At later time points , we observed that cells harboring the depletion construct that had already elaborated a division septum ( Figure 2L , arrow ) did not complete cytokinesis . Instead , the division septa became deformed and membrane aberrantly accumulated as foci . Cells that had not yet initiated cell division at the time of induction ( Figure 2L , arrowhead ) did not elaborate division septa and also accumulated aberrant membrane foci . In contrast , cells harboring only the empty vector ( Figure 2M ) elaborated division septa and completed cytokinesis during the observation period . The severe growth defect imposed by gpsBSa overexpression in B . subtilis permitted us to isolate suppressor mutations that could correct this defect . One such mutation , an intragenic single nucleotide change in gpsBSa , altered the specificity of a highly conserved codon at position 35 from Leu to Ser ( Figure 1—figure supplement 1A , boxed residue ) , and allowed B . subtilis cells overexpressing gpsBSa-L35S to grow normally . To check if the L35S substitution caused in a major structural change in the protein , we purified WT GpsBSa and the L35S variant and examined the α-helical content of both proteins using circular dichroism ( CD ) spectroscopy ( Figure 1—figure supplement 1D ) . The CD spectrum revealed similar profiles for each protein , suggesting that the L35S substitution did not grossly affect the secondary structure of GpsBSa . In the presence of inducer , S . aureus cells harboring inducible gpsBSa-L35S did not exhibit a cell enlargement defect ( Figure 2I–J ) . Taken together , we conclude that overproduction of GpsBSa , but not GpsBBs , inhibits cell division in both S . aureus and B . subtilis , resulting in cell filamentation ( in B . subtilis ) or cell enlargement ( in S . aureus ) , like the depletion phenotype of FtsZ ( Pinho and Errington , 2003 ) . Depletion of GpsB in S . aureus , however , arrested cell division without a coincident enlargement of cells and ultimately caused aberrant membrane accumulation . Furthermore , substitution of Leu35 to Ser abolished the toxicity resulting from GpsB overproduction , suggesting that this residue is critical for GpsBSa function . We next examined the subcellular localization of GpsBSa-GFP in S . aureus . In non-dividing cells GpsBSa-GFP ( produced at lower levels that did not result in cell division inhibition ) localized near the cell periphery ( Figure 3A , arrowhead ) . In dividing cells , GpsBSa-GFP localized to mid-cell , between the segregated chromosomes , and co-localized with the constricting membrane ( Figure 3A , arrow ) . In contrast , GpsBSa-L35S-GFP localized primarily in the cytosol in both dividing and non-dividing cells ( Figure 3A ) . Likewise , when produced at lower levels in B . subtilis , GpsBSa-GFP accumulated at division septa , whereas the L35S variant localized primarily in the cytosol ( Figure 1—figure supplement 2 ) . Since the L . monocytogenes GpsB ortholog is membrane-associated ( Rismondo et al . , 2016 ) , we next tested if the L35S substitution could have disrupted any intrinsic membrane affinity of GpsBSa by fractionating S . aureus cell extracts and examining the distribution of GpsBSa-GFP and GpsBSa-L35S-GFP by immunoblotting ( Figure 1—figure supplement 1E ) . Unlike L . monocytogenes GpsB , we detected S . aureus GpsB-GFP exclusively in the soluble fraction , suggesting that it does not directly associate with the Staphylococcal membrane . GpsBSa-L35S-GFP was similarly detected in the cytosolic fraction . Association of L . monocytogenes GpsB with the membrane is reportedly mediated by Leu24 , since substitution of Leu24 with Ala disrupted membrane association ( Rismondo et al . , 2016 ) . Interestingly , the corresponding residue in S . aureus GpsB is Ala ( Figure 1—figure supplement 1A , asterisk ) , consistent with the apparent lack of intrinsic affinity of S . aureus GpsB for the membrane . We conclude that , unlike L . monocytogenes GpsB , S . aureus GpsB ( hereafter , simply ‘GpsB’ ) is likely not directly membrane-associated and that cell peripheral localization of S . aureus GpsB may be mediated by another factor . To discern if GpsB-GFP co-localized with , or at sites adjacent to , the site of membrane constriction , we employed structured illumination microscopy ( SIM ) ( Gustafsson , 2005 ) , a super-resolution technique that previously provided enough resolution to discern the localization of DivIVA-GFP on either side of the ~80 nm division septum ( Eswaramoorthy et al . , 2011 ) . At the onset of cell division , mid-plane images of S . aureus cells displayed only two GpsB-GFP foci that co-localized with sites of membrane invagination at mid-cell ( Figure 3Bi ) . Reconstruction of deconvolved Z-stacks and rotation of the reconstructed image around the axis of cell division revealed that GpsB-GFP formed an irregular ring-shaped structure , reminiscent of the structure of an assembling divisome ( Figure 3Bi ) ( Lund et al . , 2018 ) . In cells that were further advanced in cell division , the two foci of GpsB-GFP followed the leading edges of the constricting membrane ( Figure 3Bii ) and formed a ring structure that was smaller than the diameter of the cell , ( Figure 3Bii ) ( Buss et al . , 2015; Ebersbach et al . , 2008 ) , suggesting that the GpsB ring structure co-constricts with the division machinery . In a cell approaching completion of cytokinesis , GpsB-GFP collapsed into a single focus at the center of the invaginating membrane ( Figure 3Biii ) , and immediately after the completion of cell division , we observed that GpsB-GFP localized largely to the cell periphery in the adjacent daughter cells ( Figure 3Biv ) , suggesting that GpsB may dynamically localize during the cell cycle . Phototoxicity induced by SIM precluded us from performing super-resolution time lapse experiments of actively dividing cells using this method . To test the dynamic nature of GpsB-GFP localization , we followed the fate of GpsB-GFP in individual cells through three rounds of cell division using diffraction-limited epifluorescence microscopy . At the onset of our measurements , GpsB-GFP localized primarily at mid-cell in a cell that had completed cytokinesis and was poised to separate into two daughter cells ( Figure 3; [Steele et al . , 2011] ) . After cell separation , GpsB-GFP redistributed to the periphery of each daughter cell ( Figure 3 ) . Beginning the next round of cell division , GpsB-GFP re-localized to the mid-cell of each daughter cell as two foci that coincided with the invaginating membrane ( Figure 3C , t ) . It again localized with the invaginating membrane , followed by redistribution of fluorescence to the cell peripheries of the daughter cells ( Figure 3 ) . The redistribution of peripherally-localized GpsB to the division septum is reminiscent of the FtsZ-dependent late localization of GpsB reported in S . pneumoniae ( Land et al . , 2013 ) . We therefore conclude that GpsB localizes as a single ring-shaped structure at mid-cell at the onset of cell division , constricts with the invaginating membrane during cytokinesis , and ultimately , after daughter cell separation , uniformly redistributes to the periphery of each daughter cell . Although the S . aureus ortholog of GpsB was non-functional in B . subtilis , its ability to localize at mid-cell suggested that it is capable of recognizing an intrinsic feature of the divisome shared between B . subtilis and S . aureus ( Figure 1—figure supplement 2 ) . The bacterial divisome is composed of approximately 10 core proteins ( Lutkenhaus et al . , 2012; Margolin , 2005 ) that direct the cell wall assembly machinery to mid-cell and mediate cell membrane constriction during cell division . The core divisome component is the bacterial tubulin homolog , FtsZ ( Coltharp et al . , 2016; Osawa and Erickson , 2013 ) , which is a target of cell division regulators in different systems ( Ortiz et al . , 2016 ) . To investigate a potential interaction between the divisome and GpsB , we examined the localization of GpsB-GFP in S . aureus cells grown in the presence and absence of the PC190723 , a small ligand that inhibits GTPase activity of FtsZ and inhibits cell division ( Andreu et al . , 2010; Haydon et al . , 2008 ) . In the presence of the drug , 92 . 5% of cells ( n = 200 ) harboring empty vector exhibited a diameter larger than 1 . 5 µm , compared to just 30 . 5% of cells in the absence of inhibitor , consistent with a block in cell division ( Figure 4A–B’ ) . To confirm that the drug was inhibiting divisome assembly , we examined the localization of ZapA-GFP , a known early stage cell division protein that assembles concomitantly with FtsZ ( Gamba et al . , 2009; Reichmann et al . , 2014 ) and is used as a proxy for localization of FtsZ . In untreated cells , ZapA localized at mid-cell at the onset of cell division ( Figure 4C–C’ ) , but in the presence of the inhibitor , 96 . 5% of cells ( n = 200 ) displayed diffuse and/or punctate localization in the cytosol that was not located at mid-cell ( Figure 4D–D’ ) , indicating that the divisome was not assembling correctly due to inhibition of FtsZ . In the absence of inhibitor , GpsB-GFP localized at mid-cell or the periphery in 55% or 30% of cells , respectively , and mis-localized in the cytosol in 15% of cells ( n = 200; Figure 4E–E’ ) . In contrast , in the presence of inhibitor , GpsB did not localize to mid-cell in any cell observed ( n = 200 ) and instead displayed a combination of diffuse cytosolic localization and aggregation along the cell periphery ( Figure 4F–F’ ) . The data therefore indicated that that GpsB localization to mid-cell depends directly or indirectly on functional FtsZ . We next tested how GpsB influences divisome assembly . In otherwise wild type cells producing ZapA-GFP , no cell enlargement was detected; among them , 53% of cells displayed ZapA-GFP localized to mid-cell ( these were cells that were actively undergoing cell division ( Figure 4G–G’ , arrow ) and 37% of cells displayed ZapA-GFP as a ring that corresponded to the subsequent plane of cell division in daughter cells that had recently completed cytokinesis ( Figure 4G–G’ , arrowhead; n = 200 ) . In contrast , in cells harboring inducible gpsB , addition of inducer resulted in the enlargement of cells and ZapA-GFP was mis-localized in 86% of the enlarged cells ( n = 100; Figure 4H–I’ ) . Assuming that the FtsZ bundling activity of ZapA is not synergistically participating with GpsB overexpression , this suggests that the cell enlargement phenotype caused by overproduction of GpsB was due to the mis-assembly of the divisome . To determine the behavior of the divisome in GpsB-depleted cells , we monitored the localization of ZapA-GFP . In cells harboring empty vector , ZapA-GFP localized to mid-cell in 82 . 5% of the cells ( n = 200; Figure 4J–J’ ) . Quantification of fluorescence intensity in individual cells revealed that the fluorescence of mid-cell-localized ZapA-GFP was 2429 ± 1346 units/cell ( n = 75 ) . At earlier time points after induction to deplete GpsB , before cell lysis , we observed faint ZapA-GFP signals at mid-cell in 41% of the cells and diffuse localization in the remaining cells ( n = 200; Figure 4K–K’ ) , but the mean fluorescence intensity of the ZapA-GFP ring structure ( 614 ± 450 units/cell; n = 75 ) was nearly four-fold lower than that observed for ZapA-GFP intensity in the absence of gpsB depletion . Together , the observations suggest that divisome assembly and GpsB localization are reciprocally influenced: GpsB requires FtsZ for localization to mid-cell; overproduction of GpsB disrupts divisome assembly; and depletion of GpsB prevents robust divisome assembly at mid-cell that precedes membrane deformities that ultimately lead to cell lysis . To test if GpsB directly influences FtsZ behavior , we purified recombinant S . aureus FtsZ , GpsB , and GpsBL35S ( Figure 5A ) and examined the GpsB variants by size exclusion chromatography ( Figure 5B ) . GpsB eluted in two peaks by size exclusion chromatography , which approximately corresponds to the predicted sizes of hexameric ( Rismondo et al . , 2016 ) and dodecameric GpsB ( Figure 5B , top ) , indicating that S . aureus GpsB could potentially exist in two forms . In contrast , GpsBL35S eluted exclusively as a dodecamer ( Figure 5B , bottom ) , suggesting that its inability to form hexamers could underlie its loss of function in vivo . We next measured the GTP hydrolysis activity of purified S . aureus FtsZ with time at increasing protein concentrations . Unlike the well-characterized E . coli FtsZ , which robustly hydrolyzes GTP ( Arjes et al . , 2015; Buske and Levin , 2012; Mukherjee and Lutkenhaus , 1998; Romberg and Levin , 2003 ) , S . aureus FtsZ hydrolyzed GTP poorly below ~30 μM ( Figure 5C ) ( Anderson et al . , 2012 ) . The rate of hydrolysis continued to increase with FtsZ concentration ( Figure 5D ) , displaying a behavior more similar to FtsZ from the Gram-positive Streptococcus pneumoniae , which has a critical concentration above 10 µM , than to E . coli FtsZ ( although a lag observed for S . pneumonia FtsZ was not detected for S . aureus FtsZ ) ( Salvarelli et al . , 2015 ) . It should be noted that this result contrasts with that of Elsen et al . , which reported a low critical concentration for S . aureus FtsZ ( ~5 µM ) ( Elsen et al . , 2012 ) . However , a recent report by Wagstaff et al . showed GTP hydrolysis rates at high S . aureus FtsZ concentration ( 10 and 20 µM ) , and similar to the rates reported here ( Wagstaff et al . , 2017 ) . These differences could be due to varying populations of conformationally active FtsZ in different preparations ( Elsen et al . , 2012 ) . We next measured the effect of GpsB on the GTP hydrolysis rate of FtsZ . Incubation of 30 µM FtsZ with increasing amounts of GpsB resulted in a non-linear stimulation of GTP hydrolysis activity , wherein appreciable stimulation of GTP hydrolysis was only seen above 8 μM GpsB ( Figure 5E ) . At 10 µM GpsB ( 1:3 ratio of monomeric GpsB:FtsZ; 1:18 ratio of hexameric GpsB:FtsZ; 1:36 ratio of dodecameric GpsB:FtsZ ) , GTP hydrolysis was stimulated ~3 fold . As a control , GpsB alone did not exhibit appreciable GTPase activity ( Figure 5E ) . In contrast , incubation of FtsZ with GpsBL35S did not appreciably stimulate GTPase activity of FtsZ ( Figure 5F ) , nor did the addition of GpsB , even at equimolar amounts , to lower concentrations ( 10 µM ) of FtsZ . Thus , wild type GpsB , which purifies as a hexamer and dodecamer , stimulates the GTPase activity of FtsZ at substoichiometric levels at sufficiently high enough concentrations of FtsZ ( above 30 μM ) , whereas GpsBL35S , which is locked in the dodecameric form , fails to do so . We next investigated if GpsB directly interacts with polymerized FtsZ using a high-speed sedimentation assay performed with a non-hydrolyzable GTP analog ( GMPCPP ) , which promotes the assembly of stable FtsZ polymers . In the absence of nucleotide , FtsZ was largely detected in the supernatant after ultracentrifugation , but in the presence of GMPCPP , more than 50% of FtsZ was detected in the pellet fraction , indicating that it had polymerized ( Figure 6A ) . When GpsB was incubated with the reaction , 94% of GpsB co-sedimented with FtsZ , whereas only 20% of the nonfunctional GpsBL35S co-sedimented with FtsZ . Finally , in the absence of FtsZ , GpsB and GpsBL35S were largely soluble , suggesting that GpsB , but not GpsBL35S , interacts with FtsZ polymers . To test if GpsB altered the ultrastructure of assembled FtsZ , we repeated the centrifugation at a slower speed to distinguish between individual or short FtsZ polymers and larger supramolecular assemblies of FtsZ . At a slower centrifugation speed , we detected 43% of FtsZ in the pellet fraction in the presence of GMPCPP ( Figure 6—figure supplement 1A ) . Addition of GpsB increased the fraction of FtsZ in the pellet to 55% , whereas addition of GpsBL35S did not significantly alter the pelleted fraction of FtsZ . In the presence of GTP , 29% of FtsZ was detected in the pellet fraction ( Figure 6—figure supplement 1B ) , and this fraction increased to 39% in the presence of GpsB , but not GpsBL35S . The differential centrifugation patterns suggested that direct interaction with GpsB could alter the assembly of FtsZ polymers . To visualize the morphology of purified FtsZ polymers with and without GpsB , we examined purified proteins in the presence and absence of GTP using negative stain transmission electron microscopy ( TEM ) . Purified FtsZ or GpsB alone did not show any distinguishable structures by TEM ( Figure 6B–C , F–G ) . In the presence of GTP , FtsZ formed linear polymers , ~100 nm in length , that were abundant and scattered in different directions on the grid , indicating that it had polymerized successfully in a GTP-dependent manner ( Figure 6D , H ) . In the presence GpsB and GTP , however , FtsZ polymers formed long filaments , closer to 1 µm in length , which were oriented in a similar direction ( Figure 6E , I ) . This pattern of orientation on an EM grid was reminiscent of the bundling behavior reported previously for proteins in E . coli that could promote lateral interactions between FtsZ filaments in vitro ( Hale et al . , 2000; Small et al . , 2007 ) . With GMPCPP , FtsZ polymers were very long and in the presence of GpsB also exhibited extensive lateral interactions between FtsZ filaments , indicating that GpsB-mediated bundling of FtsZ did not require GTP hydrolysis ( Figure 6—figure supplement 1C–E ) . We next monitored the kinetics of FtsZ assembly in vitro using 90° angle light scattering ( Mukherjee and Lutkenhaus , 1999 ) as a function of FtsZ concentration . Incubation of either 10 µM or 20 µM purified S . aureus FtsZ with GTP did not result in an appreciable increase in light scattering ( Figure 6J ) , consistent with the apparent high critical concentration for FtsZ assembly suggested in GTP turnover experiments ( Figure 5C–D ) . Incubation of 30 µM or 40 µM FtsZ with GTP resulted in a rapid increase in light scattering that likely corresponds to the assembly of FtsZ polymers . The increase was followed by a brief plateau , likely reflecting that the reaction was at steady state , then a decrease in light scattering , corresponding to disassembly of FtsZ polymers coincident with depletion of GTP in the reaction . Such kinetics were not detected when 30 µM FtsZ was incubated with GDP , ATP , or ADP ( Figure 6—figure supplement 2A ) , suggesting that the light scattering assay specifically reflects GTP-dependent dynamics of FtsZ assembly . To confirm that the decrease in light scattering corresponded to the depletion of GTP and accumulation of GDP in the reaction , we repeated the assay in the presence of a regeneration system to replenish GTP . As expected , addition of a GTP regeneration system prevented the rapid loss of scatter following the plateau ( Figure 6—figure supplement 2B ) , suggesting that the decrease in Figure 6J represents the disassembly of FtsZ polymers as GTP becomes limiting . Next , we tested the effect of GpsB on FtsZ assembly . Addition of 10 µM GpsB to the reaction with GTP and 30 µM FtsZ resulted in an initial increase in light scattering that was much more rapid and larger in amplitude than that of FtsZ alone in the presence of GTP ( Figure 6K ) , whereas incubation of GpsB alone with GTP did not result in an increase in light scattering . To determine if the increase in light scattering due to GpsB was reversible , we followed the assembly reaction for a longer period ( Figure 6L ) . Upon addition of GTP , the reaction containing FtsZ and GpsB displayed a rapid increase in light scattering , which was not observed when FtsZ was incubated with GpsBL35S . Note that , due to saturation of the detector in the presence of GpsB , the slit width in Figure 6L was adjusted , precluding a direct comparison between the signal amplitudes shown in Figure 6J K . After reaching a plateau , the reaction containing WT GpsB displayed a steady decrease in light scattering , suggesting that the assembly of the higher order FtsZ structures generated in the presence of GpsB was reversible , in contrast to the behavior of other FtsZ bundling proteins reported in other systems . We therefore conclude that GpsB directly interacts with polymerized FtsZ and bundles FtsZ filaments . Taken together with the observation that GpsB also triggers GTP hydrolysis by FtsZ , we propose that FtsZ bundling by GpsB increases FtsZ local concentration and triggers GTP hydrolysis which , in the light scattering assay , is linked to the disassembly of FtsZ polymers as GTP is depleted . Since binary fission has been traditionally studied in rod-shaped model organisms , the roles of factors that participate in cell division of spherical bacteria have been less well characterized ( Eswara and Ramamurthi , 2017 ) . In this report , we investigated the role of a coiled-coil protein , GpsB , during cell division in the spherical bacterium S . aureus . Unlike the orthologs of GpsB in other systems , we report that GpsB directly interacts with FtsZ , the core component of the bacterial cell division machinery and increases the GTPase activity of FtsZ . We also demonstrate that GpsB promotes bundling of FtsZ filaments in vitro . We propose a model in which the bundling of S . aureus FtsZ by GpsB raises the local concentration of FtsZ transiently so that it may robustly hydrolyze GTP , and thereby participates in remodeling the constricting divisome during cytokinesis . A recent report suggested that cell division in S . aureus proceeds in two principal steps: an initial FtsZ treadmilling-dependent step in which membrane invagination initiates , followed by recruitment of peptidoglycan remodeling enzymes by later arriving divisome components that mediates the progression and completion of cell division ( Monteiro et al . , 2018 ) . We propose that GpsB may participate in the initial step that stabilizes FtsZ at mid-cell and activates GTP hydrolysis ( by increasing the local concentration of FtsZ via a bundling-like mechanism ) to trigger FtsZ treadmilling , which is linked to constriction of the membrane and concurrent peptidoglycan synthesis ( Bisson-Filho et al . , 2017; Yang et al . , 2017 ) . In our model , FtsZ requires a high concentration to polymerize and to hydrolyze GTP efficiently . In the presence of GpsB , though , FtsZ filaments are laterally bridged to form higher order supramolecular structures ( Figure 7A ) . Intracellular levels of FtsZ and GpsB are reported to be approximately 4452 and 1659 molecules per cell , respectively ( Zühlke et al . , 2016 ) ( S . Fuchs , personal communication ) . Considering a cell diameter of 0 . 8 µm , this equates to intracellular concentrations of 28 uM FtsZ and 10 uM GpsB and corresponds closely with our in vitro reaction conditions . We show that GpsB is a multimer and propose that it may harbor 6–12 binding sites per multimer to recruit and bridge multiple FtsZ proteins . We propose that the bridging of FtsZ filaments also serves to increase the local FtsZ concentration and enhances GTP hydrolysis . Unlike other proteins that bundle FtsZ irreversibly in vitro by inhibiting GTP hydrolysis , incubation with GpsB ultimately allows for the subsequent disassembly of FtsZ polymers once GTP is depleted ( Figure 7A ) . To our knowledge , this is the first report of an FtsZ regulatory protein that promotes both lateral interactions between FtsZ filaments while also stimulating GTP hydrolysis . Furthermore , considering the intracellular concentrations of FtsZ and GpsB and what we have observed biochemically , FtsZ polymers and regulators appear poised at the threshold between assembly and disassembly , enabling tight control over the division process . Our view is supported by multiple lines of evidence . First , overexpression of gpsB resulted in the enlargement of S . aureus cells , reminiscent of the phenotype caused by depletion of FtsZ ( Pinho and Errington , 2003 ) , likely due to increased FtsZ GTPase activity leading to the inability of FtsZ to polymerize and treadmill in a concerted fashion . Curiously , overexpression of S . aureus gpsB in B . subtilis , but not the B . subtilis gpsB ortholog , resulted in filamentation , suggesting that S . aureus GpsB harbors a unique cell division-modulating activity that is not exhibited by the B . subtilis version . Second , depletion of GpsB in S . aureus resulted in the arrest of cytokinesis and abrogation of initiation of cell division . Third , we observed that GpsB co-localized with the cell division machinery at the onset of cytokinesis , and co-constricted with the invaginating membrane during cell division , consistent with its role in modulating the activity of FtsZ . Fourth , we found that purified GpsB directly interacted with FtsZ in vitro and stimulated the GTPase activity of FtsZ , consistent with the ability of GpsB to inhibit cell division in vivo when overproduced . Finally , when incubated with FtsZ in vitro , GpsB promoted lateral interactions between FtsZ polymers , but also allowed for the ultimate disassembly of FtsZ in vitro once GTP had been depleted . Our genetic , cytological , and biochemical data in sum suggest a model in which Staphylococcal FtsZ begins assembling at mid-cell and recruits GpsB to that location ( Figure 7B ) where GpsB stabilizes the Z-ring via a bundling-like mechanism that concentrates and organizes FtsZ polymers . We propose that this reinforces the faithful and robust assembly of FtsZ at mid-cell at the onset of cell division and drives an increase in the local concentration of FtsZ , which stimulates its GTPase activity , which is linked to treadmilling- an activity that is likely coincident with the initial membrane constriction that initiates cytokinesis . After completion of cytokinesis , GpsB redistributes to the cell periphery and awaits the next round of cell division . It is tempting to speculate that this dynamic redistribution of GpsB , presumably coincident with a dynamic ability to modulate FtsZ activity , is dependent on the multimerization state of GpsB . In this way , the hexameric and dodecameric populations of purified GpsB could represent the active and inactive forms , respectively , of the protein that may mediate its interaction with FtsZ . Consistent with this model , the inactive GpsBL35S was locked in the dodecameric form and did not exhibit the dynamic redistribution from the cell periphery to the cell division site in vivo . Identifying the factors that regulate the multimerization state of GpsB could therefore provide an understanding into the temporal regulation of cell division in S . aureus . Interestingly , depletion of a known interaction partner of GpsB , EzrA , also leads to cell enlargement , hinting at a possible collaboration between these two proteins in regulating cell division ( Jorge et al . , 2011; Steele et al . , 2011 ) . Several divisome proteins in E . coli and B . subtilis that positively regulate cell division by bundling FtsZ polymers do so via inhibition of FtsZ GTPase activity ( Durand-Heredia et al . , 2011; Hale et al . , 2011; Lutkenhaus et al . , 2012; Mohammadi et al . , 2009; Pacheco-Gómez et al . , 2013; Singh et al . , 2008; Small et al . , 2007; Tsang and Bernhardt , 2015 ) . In our model , the seemingly contradictory observation that GpsB stimulates GTP hydrolysis , even though it promotes FtsZ bundling may be explained by the proposition that FtsZ bundling is not the ultimate activity of GpsB . Rather , we envision that FtsZ bundling is an intermediate step that increases local FtsZ concentration to stimulate GTP hydrolysis ( Figure 7A ) . This set of opposing activities exhibited in S . aureus in a single protein is reminiscent of the model in E . coli , where FtsZ polymers bundled by other proteins require a separate protein , FtsA , that can disrupt the bundles and destabilize FtsZ polymers ( Conti et al . , 2018; Krupka et al . , 2017 ) . In a slight variation of this model , since FtsZ bundling in B . subtilis requires C-terminal positively charged residues ( Buske and Levin , 2012 ) , it is conceivable that GpsB modulates exposure of the C-terminus of FtsZ to promote FtsZ self-interactions or remodels FtsZ to stabilize a conformation associated with enhanced GTP hydrolysis . In this way , GpsB , an essential S . aureus protein , may orchestrate the organization , stabilization , and activity of FtsZ to remodel the divisome during cell division . B . subtilis strains used in this study are derivatives of PY79 ( Youngman et al . , 1984 ) , and S . aureus strains are derivatives of SH1000 ( Horsburgh et al . , 2002 ) . To overproduce GpsB or GpsB-GFP orthologs in B . subtilis , gpsB ( HindIII/SphI; primers oP36/oP38 , please see Key Resources Table for primers ) or gpsB-gfp ( HindIII/NheI; oP36/37 for gpsB without stop codon; and NheI/SphI; oP46/24 for gfp with stop codon ) were PCR amplified and cloned into the 5’ HindIII and 3’ NheI restriction sites in pDR111 ( D . Rudner ) to place it under control of the isopropyl β-D-1-thiogalactopyranoside ( IPTG ) -inducible Phyperspank promoter . The resulting plasmids ( pGG3 , gpsB; pGG4 , gpsB-gfp ) were integrated into the amyE locus in the B . subtilis chromosome by double recombination . Similarly , B . subtilis gpsB was constructed using primers oP100/102 ( gpsB; SalI/NheI ) and gpsB-gfp was constructed by ligating the products of oP100/101 ( gpsB no stop codon; SalI/NheI ) and oP46/24 ( gfp; NheI/SphI ) . To produce S . aureus GpsB or GpsB-GFP in S . aureus , gpsB or gpsB-gfp were PCR amplified and cloned into the 5’ HindIII and 3’ SphI restriction sites in the pCL15 plasmid ( Luong and Lee , 2006 ) , downstream of the IPTG-inducible Pspac promoter , to create pPE45 and pPE46 , respectively . The L35S substitution was introduced using the QuikChange Site-Directed Mutagenesis kit ( Agilent ) using pPE45 , pPE46 , pGG3 , or pGG4 as a template . To express B . subtilis gpsB in S . aureus , a pCL15-based vector pPE83 was constructed by amplifying and inserting the B . subtilis gpsB fragment with the help of primer pairs oP100/195 ( SalI/BamHI ) . To express the antisense RNA of the gpsB open reading frame and ribosome binding site under control of a xylose-inducible promoter , using primers oP187/188 abutted by EcoRI and BamHI restriction sites and cloned into plasmid pEPSA5 ( Forsyth et al . , 2002 ) to create plasmid pGG59 . Plasmid pRB42 ( zapA-gfp ) was constructed using primers oP236/237 ( zapA no stop codon; SalI/NheI ) and oP46/47 ( gfp; NheI/BamHI ) and inserted into cadmium-inducible plasmid pJB67 ( Windham et al . , 2016 ) . Plasmids were first introduced into S . aureus RN4220 by electroporation , then transduced into strain SH1000 . Expression was induced by addition of 1 mM IPTG or 1% xylose or 1 . 25 µM CdCl2 , as required , in the growth medium . For immunoblot analysis of cell extracts , overnight cultures of S . aureus were diluted 1:50 into 10 ml tryptic soy broth ( TSB ) and were grown to mid-logarithmic phase , harvested by centrifugation , and resuspended in 1 ml buffer A ( see below ) containing 200 mM KCl , 1 mM dithiothreitol , and 10 mg/ml lysostaphin and incubated for 15 min at room temperature . Suspensions were then sonicated ( 3 intervals at 10 s each at 20% power level ) , then cleared by centrifugation at 14 , 000 × g for 10 min . Supernatants were isolated and centrifuged at 100 , 000 × g for 1 hr to separate soluble ( supernatant ) fraction from insoluble ( pellet ) fraction . Supernatants were removed for analysis . Pellets were resuspended in 1 ml buffer ( no lysostaphin ) containing 0 . 01% SDS . Samples were separated using 8–16% SDS-PAGE ( BioRad ) , transferred to nitrocellulose membrane , and probed with rabbit antisera raised against purified S . aureus GpsB or B . subtilis SigA antibody . Overnight B . subtilis cultures grown at 22°C in Luria-Bertani ( LB ) medium were diluted 1:20 into fresh LB medium and grown for 2 . 5 hr at 37°C . Overnight cultures of S . aureus in TSB , containing 15 μg/ml chloramphenicol and/or 5 μg/ml erythromycin for plasmid maintenance if necessary , were diluted into fresh medium and grown to mid-logarithmic phase . 1 mM IPTG was added as required for 3 hr . 1 ml cultures were washed with PBS and resuspended in ~100 μl PBS containing 1 μg/ml fluorescent dye FM4-64 and/or 2 μg/ml DAPI to visualize membranes and DNA , respectively . 5 μl was spotted on a glass bottom culture dish ( Mattek ) and covered with a 1% agarose pad made with distilled water and imaged at 25°C . For time lapse , a 5 μl aliquot of SH1000 pGG59 cells grown in TSB/chloramphenicol until mid-log phase was spotted on a glass bottom culture dish and covered with an agarose pad made with TSB/chloramphenicol containing 1% xylose to induce expression of the gpsB antisense RNA . After 20 min of equilibration in the microscopy environmental chamber , images were obtained at 15 min intervals for 4 hr at 25°C . For FtsZ inhibition experiments , mid-log phase cells were incubated with 2 μg/ml PC190723 and samples for imaging were collected after 3 hr . Cells were viewed with a DeltaVision Core microscope system ( Applied Precision/GE Healthcare ) equipped with a Photometrics CoolSnap HQ2 camera and an environmental chamber . Seventeen planes for standard microscopy and four planes for time-lapse microscopy were acquired every 200 nm , and the data were deconvolved using SoftWorx software as described previously ( Tan et al . , 2015 ) . For structured illumination microscopy , cells were viewed using a DeltaVision OMX ( Applied Precision/GE Healthcare ) comprising an OMX optical microscope ( version 4 ) , equipped with a sCMOS camera . To purify FtsZ , S . aureus ftsZ was PCR amplified and cloned into the pET28a ( + ) vector ( EMD Millipore ) using 5’ NdeI and 3’ XhoI restriction sites , resulting in the addition of an N-terminal histidine tag followed by a thrombin cleavage site . Expression was induced in BL21 ( λDE3 ) ::ΔclpP cells grown in LB broth supplemented with 50 μg/ml Kanamycin for plasmid maintenance , at 30°C by adding 1 mM IPTG after cells reached an optical density ( 600 nm ) of 1 . 0 . Cells were harvested by centrifugation , resuspended in buffer A [20 mM HEPES ( pH 7 . 5 ) , 50 mM KCl , 5 mM MgCl2 and 10% glycerol] , and lysed by French press . Soluble extract was collected by centrifugation at 30 , 000 × g for 30 min at 4°C and applied to an IMAC column ( TALON Superflow , GE Healthcare ) , and washed with Buffer A containing 10 mM imidazole . Untagged FtsZ was eluted with thrombin ( 4 U; Novagen ) and then 0 . 5 mM phenylmethylsulphonyl fluoride was added to inactivate thrombin . To purify GpsB-His6 , S . aureus gpsB was PCR amplified and cloned into pET28a ( + ) using 5’ XbaI and 3’ BamHI restriction sites , using primers to append a His6 tag to the C-terminus . Overproduction of GpsBSa-His6 in B . subtilis resulted in cell filamentation similar to untagged GpsBSa , suggesting that the His6-tagged protein was functional . The L35S substitution was introduced using the QuikChange Site-Directed Mutagenesis kit ( Agilent ) . Expression was induced in BL21 ( λDE3 ) ::ΔclpP cells grown in LB broth supplemented with 25 μg/ml Kanamycin for plasmid maintenance , at 37°C by adding 0 . 5 mM IPTG for 2 hr after cells reached an optical density ( 600 nm ) of 0 . 6 . Cells were harvested by centrifugation and resuspended in 30 ml cold buffer B [50 mM sodium phosphate ( pH 8 . 0 ) , 500 mM NaCl , 20 mM imidazole , 1 mM EDTA , 10% Glycerol , 3 mM DTT] and lysed by sonication ( 5 s on/10 s off cycle for 5 min ) . Lysate was cleared by centrifugation for 30 min at 40 , 000 × g; cleared lysate was passed through a Ni2+-NTA column equilibrated with buffer B , washed with 20 column volumes buffer B , and eluted with buffer B containing 200 mM imidazole . Imidazole was removed with a PD10 desalting column and eluted with buffer A containing 250 mM KCl and 1 mM DTT . To ensure the final buffer composition the protein was dialyzed over night at 4°C against buffer A . FtsZ assembly was monitored by 90° angle light scattering using an Agilent Eclipse fluorescence spectrophotometer with excitation and emission wavelengths set to 450 nm and slit widths of 5/5 or 2 . 5/5 , where indicated . FtsZ ( 30 μM ) was added to reactions ( 80 μl ) containing assembly buffer ( 20 mM HEPES pH 7 . 5 , 140 mM KCl , 5 mM MgCl2 ) with and without GpsB or GpsBL35S ( 1 or 10 μM ) , where indicated . Baseline readings were collected for 3 min , 2 mM GTP was added and light scattering was measured for up to 300 min . GMPCPP-stabilized FtsZ polymers were assembled by incubating FtsZ ( 30 μM ) with 0 . 5 mM GMPCPP in the absence and presence of GpsB or GpsBL35S ( 10 μM ) for 10 min and collected by centrifugation either for 30 min at 129 , 000 x g ( Figure 6A ) , 20 min at 20 , 000 x g , or 20 min at 90 , 000 x g ( Figure 2—figure supplement 1A–B ) , as indicated . Where indicated , polymerization was stimulated with GTP ( 2 mM ) and a nucleotide regenerating system containing acetate kinase ( 25 μg ml−1 ) and acetyl phosphate ( 15 mM ) was included to prevent GDP accumulation . Supernatants and pellets were resuspended in equivalent volumes of LDS sample buffer ( Life Technologies ) and analyzed by SDS-PAGE and Coomassie staining . The relative amounts of FtsZ , GpsB and GpsBL35S in supernatant and pellet fractions were quantified by densitometry using ImageJ ( NIH ) . FtsZ GTP hydrolysis activity was monitored by detection of free phosphate using Biomol Green ( Enzo Life Sciences ) . Reactions containing FtsZ ( 0–40 μM ) in the absence and presence of GpsB and GpsB ( L35S ) ( 0–10 μM ) were incubated with 2 mM GTP in assembly buffer at room temperature . Phosphate was measured at 0 and 15 min by comparison to a phosphate standard curve . Rates were calculated by measuring the amount of free phosphate released during the incubation period . At low FtsZ concentrations , reactions were incubated for 60 min . FtsZ ( 30 μM ) polymers were assembled in buffer ( 20 mM HEPES pH 7 . 5 , 140 mM KCl , 5 mM MgCl2 ) in the presence or absence of GpsB ( 10 μM ) by addition of 2 mM GTP . After 10 min , reactions were applied to formvar/carbon coated 300 mesh grids , fixed with 2 . 5% glutaraldehyde in 0 . 15M sodium cacodylate buffer ( pH 7 . 4 ) and stained with 2% aqueous uranyl acetate . Samples were imaged by transmission electron microscopy using a FEI Tecnai G2 Spirit BioTWIN 80Kv instrument equipped with a SIS Morada 11 Megapixel camera .	A bacterium called Staphylococcus aureus causes many infections in humans , especially in hospital patients with weakened immune systems . These infections are generally treated with drugs known as antibiotics that interact with specific proteins in the bacteria to kill the cells , or stop them from growing . However , some S . aureus infections are resistant to the antibiotics currently available so there is a need to develop new drugs that target different bacterial proteins . Bacteria multiply by dividing to make identical copies of themselves . When a bacterium is preparing to divide , filaments made of a protein called FtsZ form a ring at the site where the cell will split . Many other proteins are involved in controlling how and when a cell divides . For example , several species of bacteria harbor a dispensable cell division protein called GpsB . In at least one organism , it helps to maintain the proper shape of the cell during cell division . In S . aureus , though , GpsB is essential for cells to survive and could therefore be a potential target for new antibiotics . However , its role in S . aureus has not been studied . Eswara et al . have now used genetic and biochemical approaches to study the S . aureus form of the GpsB protein . The experiments show that GpsB moves to the middle of S . aureus cells just before they begin to divide and binds directly to FtsZ . This helps to secure the position of FtsZ across the middle of the cell and activates the protein so that the cell can begin to divide into two . In cells that produce too much GpsB , the FtsZ proteins become active too early , leading to the cells growing larger and larger until they burst . The findings of Eswara et al . reveal that GpsB plays a different role in S . aureus cells than in some other species of bacteria . Further studies into such differences could help researchers to develop new antibiotics , as well as improving our understanding of why bacteria are so diverse .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2018	An essential Staphylococcus aureus cell division protein directly regulates FtsZ dynamics
Tissue mechanics is important for development; however , the spatio-temporal dynamics of in vivo tissue stiffness is still poorly understood . We here developed tiv-AFM , combining time-lapse in vivo atomic force microscopy with upright fluorescence imaging of embryonic tissue , to show that during development local tissue stiffness changes significantly within tens of minutes . Within this time frame , a stiffness gradient arose in the developing Xenopus brain , and retinal ganglion cell axons turned to follow this gradient . Changes in local tissue stiffness were largely governed by cell proliferation , as perturbation of mitosis diminished both the stiffness gradient and the caudal turn of axons found in control brains . Hence , we identified a close relationship between the dynamics of tissue mechanics and developmental processes , underpinning the importance of time-resolved stiffness measurements . All animal experiments were approved by the Ethical Review Committee of the University of Cambridge and complied with guidelines set by the UK Home Office . Single-cell-stage , wild-type Xenopus laevis embryos of both sexes were obtained via in vitro fertilisation . Embryos were reared in 0 . 1× Modified Barth’s Saline ( MBS ) at 14–18°C to reach the desired developmental stage , as described by Nieuwkoop and Faber , 1958 . All embryos used in this study were below stage 45 . Data were collected from at least three independent experiments ( N ≥ 3 ) . The order of data collection was randomized with no blinding and no data were excluded from the analysis . Non-parametric tests as well as linear regression analysis were used for statistical analyses of the data as described in the figure captions . R2 values provide an estimate of the quality of the fits used in the plots . Pearson’s correlation coefficient , on the other hand , provides a measure of the magnitude of correlation between the cell body density gradient and the stiffness gradient .	Neurons in the brain form an intricate network that follows a precise template . For example , in a young frog embryo , the neurons from the eyes send out thin structures , called axons , which navigate along a well-defined path and eventually connect with the visual centres of the brain . This journey requires the axons to take a sharp turn so they can wire with the right brain structures . Axons find their paths not only by following chemical signals but also by reacting to the stiffness of their environment . In an older frog embryo for instance , the brain is stiffer at the front , and softer at the back . As neurons from the eyes make their way through the brain , they turn to follow this gradient , moving away from stiffer areas towards the softer regions . Here , Thompson , Pillai et al . investigate when and how this stiffness gradient is established in frogs . To do so , a new technique was developed . Called time-lapse in vivo atomic force microscopy , the method measures how brain stiffness changes over time in a live embryo , while also taking images of the growing axons . The experiments show that the stiffness gradient arose within tens of minutes , just as the first ‘pioneering’ axons from the eyes began to grow across the brain . These axons then responded to the gradient , turning towards the softer tissue . Changes in the number of cells in the underlying brain tissue governed the formation of the gradient , with rapidly stiffening areas containing more cells than those that remained soft . In fact , using drugs that stop cells from dividing reduced both the mechanical gradient and the turning response of the axons . The technique developed by Thompson , Pillai et al . is a useful tool that can help elucidate how variations in stiffness control the brain wiring process . It could also be used to look into how other developmental or regenerative processes , such as the way neurons reconnect after injuries to the brain or spinal cord , may be regulated by mechanical tissue properties .	[ "Abstract", "Materials", "and", "methods" ]	[ "developmental", "biology", "short", "report", "physics", "of", "living", "systems" ]	2019	Rapid changes in tissue mechanics regulate cell behaviour in the developing embryonic brain
Lamina-associated polypeptide 1 ( LAP1 ) resides at the nuclear envelope and interacts with Torsins , poorly understood endoplasmic reticulum ( ER ) -localized AAA+ ATPases , through a conserved , perinuclear domain . We determined the crystal structure of the perinuclear domain of human LAP1 . LAP1 possesses an atypical AAA+ fold . While LAP1 lacks canonical nucleotide binding motifs , its strictly conserved arginine 563 is positioned exactly where the arginine finger of canonical AAA+ ATPases is found . Based on modeling and electron microscopic analysis , we propose that LAP1 targets Torsin to the nuclear envelope by forming an alternating , heterohexameric ( LAP1-Torsin ) 3 ring , in which LAP1 acts as the Torsin activator . The experimental data show that mutation of arginine 563 in LAP1 reduces its ability to stimulate TorsinA ATPase hydrolysis . This knowledge may help scientists understand the etiology of DYT1 primary dystonia , a movement disorder caused by a single glutamate deletion in TorsinA . Lamina-Associated Polypeptide 1 ( LAP1 ) and Luminal domain-Like LAP1 ( LULL1 ) are type II integral membrane proteins with ∼30 kDa luminal domains that share 62% identity . LAP1 localizes to the nuclear envelope ( NE ) via its lamin-interacting domain , whereas LULL1 is found throughout the endoplasmic reticulum ( ER ) ( Senior and Gerace , 1988; Goodchild and Dauer , 2005 ) . LAP1 and LULL1 associate with Torsins ( Goodchild and Dauer , 2005; Naismith et al . , 2009; Zhao et al . , 2013 ) , which are ER-resident members of the ATPases Associated with a variety of cellular Activities ( AAA+ ATPases ) superfamily ( Iyer et al . , 2004; Hanson and Whiteheart , 2005; Erzberger and Berger , 2006 ) . AAA+ ATPases use ATP hydrolysis to undergo conformational changes and to exert mechanical force on a substrate . AAA+ ATPases are involved in many processes , including vesicle fusion and scission , protein folding/unfolding , complex assembly/disassembly , protein transport , and nucleic acid remodeling . Detailed mechanistic knowledge exists for a number of AAA+ ATPases , but not for Torsins . Torsins are found in all animals , but not in single-cell organisms and plants . Deletion of a single glutamate in TorsinA ( humans have four Torsin orthologs: TorsinA , TorsinB , Torsin2A , and Torsin3A ) at position 302/303 causes the movement disorder DYT1 primary dystonia ( Ozelius et al . , 1997; Breakefield et al . , 2008 ) . Mammalian TorsinA has been assigned to a variety of possible functions including nuclear envelope ( NE ) organization , synaptic vesicle transport and turnover , operation of the secretory pathway , protein degradation , cytoskeletal organization and transport via NE budding ( Vander Heyden et al . , 2009; Granata and Warner , 2010; Jokhi et al . , 2013 ) . While LAP1 and LULL1 were first considered as Torsin substrates ( Naismith et al . , 2009 ) , they are now being proposed as possible activators of Torsin ( Zhao et al . , 2013 ) . To better understand the function of LAP1 in relation to Torsin , we set out to determine its structure . Our data suggest that LAP1 and LULL1 can both form heterohexameric ring assemblies with Torsin , whereby Torsin is activated through an arginine finger at amino acid ( aa ) R563 of LAP1 ( R449 of LULL1 ) . Thus LAP1 and LULL1 each have dual roles , namely the targeting and activation of Torsins . The conserved luminal domain of LAP1 ( aa356–583 ) was recombinantly expressed , purified , and set up for crystallization . Because the protein failed to yield crystals alone , we produced a camelid single domain ( VHH ) antibody fragment , VHH-BS1 , against LAP1 as a crystallization chaperone . The complex of VHH-BS1 and LAP1 yielded well-diffracting crystals ( Table 1 ) . The structure was solved using molecular replacement with VHH as the search model ( for details , see ‘Materials and methods’ ) . In the asymmetric unit we find two LAP1 molecules related by a non-crystallographic symmetry axis , each identically bound by one VHH . The final structure is completely built , except for four N-terminal residues in LAP1 , which seem disordered in the crystal ( six N-terminal residues in the second LAP1 copy ) . 10 . 7554/eLife . 03239 . 003Table 1 . X-ray data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 003ProteinHuman LAP1356–583-VHH-BS1 complexPDB ID4TVSData collection Space groupP21 a , b , c ( Å ) 69 . 79 , 74 . 02 , 85 . 43 α , ß , γ ( ° ) 90 , 108 . 8 , 90 Wavelength ( Å ) 1 . 2548 Resolution range ( Å ) 66 . 1–1 . 60 ( 1 . 66–1 . 60 ) Total reflections630 , 312 ( 46 , 516 ) Unique reflections105 , 976 ( 10 , 195 ) Completeness ( % ) 97 . 6 ( 94 . 3 ) Redundancy5 . 9 ( 4 . 6 ) Rsym ( % ) 9 . 1 ( 141 . 2 ) Rp . i . m . ( % ) 4 . 0 ( 68 . 6 ) I/σ12 . 5 ( 1 . 1 ) CC1/2 ( % ) 99 . 8 ( 54 . 7 ) Refinement Resolution range ( Å ) 66 . 1–1 . 60 Rwork ( % ) 18 . 1 Rfree ( % ) 22 . 9 Coordinate error ( Å ) †0 . 24 Number of reflections Total105 , 565 Rfree reflections2810 Number of non-hydrogen atoms Protein5520 Ligands29 Water612 R . m . s . deviations Bond lengths ( Å ) 0 . 019 Bond angles ( ° ) 1 . 66 B-factors ( Å2 ) Protein39 . 3 Ligands56 . 4 Water47 . 8 Ramachandran ( % ) ‡ Favored ( % ) 98 . 3 Outlier ( % ) 0 . 0 Clashscore5 . 65 MolProbity score‡1 . 29 MolProbity percentile‡96thNumbers in brackets refer to the highest resolution shell ( 10% of all reflections ) . †Maximum likelihood based ( as determined by PHENIX; Adams et al . , 2010 ) . ‡As determined by MolProbity ( Chen et al . , 2010 ) . The LAP1 structure clearly has an AAA+ like fold ( Figure 1 ) . AAA+ proteins are a functionally diverse group within the vast family of ‘P-loop’-type NTP-binding proteins ( Iyer et al . , 2004 ) . Typically , AAA+ proteins are characterized by strongly conserved sequence motifs involved in nucleotide recognition . Subcategories within AAA+ proteins are based on distinct secondary structure topologies . The D2 domain of the two-ring AAA+ ATPase ClpB superimposes on LAP1 with an rmsd of 3 . 28 Å over 111 Cα positions ( Figure 1; Zeymer et al . , 2014 ) . ClpB belongs to the additional strand conserved E ( ASCE ) family ( Iyer et al . , 2004 ) . Like ASCE-type AAA+ proteins , such as ClpA , ClpB , and p97 , LAP1 also has a central , five-stranded parallel β-sheet , surrounded on both sides by 10 helices in total ( Figure 1A , B ) . The secondary structure topology is α ( −1 ) α ( 0 ) β1α ( 1 ) β2-α ( 2a ) α ( 2b ) β3α ( 3a ) α ( 3b ) β4α ( 4 ) β5α ( 5 ) α ( 6 ) , using the AAA+ nomenclature ( Erzberger and Berger , 2006 ) . In contrast to canonical ASCE-type ATPases ( Figure 1C ) , LAP1 lacks a C-terminal domain . Instead the C terminus is attached to helix α1 via a disulfide bond between the highly conserved residues C424 and C582 ( Figure 1—figure supplement 1 ) . In a AAA+ ATPase , the nucleotide binding pocket is the most sequence-conserved region , defined by the essential Walker A and B motifs . LAP1 lacks a recognizable Walker A region , typically located between β1 and α1 ( Figure 2A , B , Figure 1—figure supplement 1 ) . Instead , helix α1 is N-terminally extended compared to AAA+ ATPases , thereby sterically blocking the nucleotide-binding pocket . The Walker B motif with the canonical sequence hhhhDE ( h , hydrophobic ) , is instead hhhhHR in human LAP1 , and therefore cannot interact with nucleotide . Finally , the conserved disulfide bridge would also interfere with nucleotide binding ( Figure 2B , Figure 1—figure supplement 1 ) . 10 . 7554/eLife . 03239 . 004Figure 1 . Crystal structure of human LAP1 . ( A ) Crystal structure of the luminal domain of LAP1 from Homo sapiens . The helices are numbered according to the nomenclature used for AAA+ proteins ( Erzberger and Berger , 2006 ) . A disulfide bridge ( yellow ) attaches the C terminus to helix α1 . ( B ) Same as ( A ) , but rotated around the x-axis by 50° . ( C ) Crystal structure of the AMPPCP-bound ClpB-D2 domain from Thermus thermophilus ( 4LJ6; Zeymer et al . , 2014 ) in the same orientation as LAP1 in ( A ) revealing the striking topological similarity . In comparison to ClpB , LAP1 does not bind a nucleotide and has no C-terminal domain . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 00410 . 7554/eLife . 03239 . 005Figure 1—figure supplement 1 . Phylogenetic analysis of LAP1 homologs . Sequence alignment of phylogenetically diverse sequences of the luminal domain of LAP1 and LULL1 . Residues are color-coded from dark blue ( most conserved ) to white ( not conserved ) . Secondary structure elements of human LAP1 are shown above the sequences . The conserved disulfide bond is indicated with a yellow line . The strictly conserved arginine ( ‘R-finger’ ) is marked by a green circle . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 00510 . 7554/eLife . 03239 . 006Figure 1—figure supplement 2 . Representative 2Fo-Fc electron density of the final model . Stereo view of the binding site of VHH-BS1 ( gray ) with LAP1 ( blue ) . Contacting residues between LAP1 and the variable loops of VHH-BS1 ( red ) are shown as sticks and are labeled . Waters , with direct contact to both binding partners , are in green . The 2Fo-Fc map is contoured at 1σ . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 00610 . 7554/eLife . 03239 . 007Figure 1—figure supplement 3 . VHH-BS1 competes with TorsinA for LAP1 binding . ( A ) Overall structure of the LAP1-VHH-BS1 complex . LAP1 is in light blue , VHH-BS1 in gray . The variable loops , responsible for high affinity interaction with LAP1 , are colored in red . The binding site on LAP1 is centered around arginine 563 ( blue sticks ) . ( B ) Model of LAP1 ( light blue ) interacting with TorsinA ( green ) in a putative heterohexameric ring assembly in the same orientations as in ( A ) . TorsinA-bound ATP molecule is in red . Note the overlap between the VHH-BS1 and the putative TorsinA binding site on LAP1 . ( C and D ) In vitro precipitation assay shows that VHH-BS1 competes with TorsinA ( E171Q ) for LAP1 binding . Preformed TorsinA ( E171Q ) -LAP1 complex was incubated for several hours with VHH-BS1 ( C ) or a control VHH ( D ) . The mixtures were centrifuged at the indicated times to separate soluble from insoluble protein . The samples were analyzed by SDS-PAGE . Due to the competition with VHH-BS1 , TorsinA ( E171Q ) is released from LAP1 . Since free TorsinA ( E171Q ) is insoluble , the process of the reaction can be monitored . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 00710 . 7554/eLife . 03239 . 008Figure 2 . Nucleotide binding site . ( A and B ) Comparison of the nucleotide binding pocket in ClpB-D2 ( A ) with the equivalent region in LAP1 ( B ) . The nucleotide sensing elements of canonical ATPases are indicated . The nucleotide in ( B ) is modeled to illustrate how LAP1 blocks binding of it . ( C and D ) To activate ATP hydrolysis , an arginine residue ( R-finger ) from the neighboring protomer in the typical hexameric ring assembly is necessary . For ClpB-D2 , this is R747 ( light blue ) ( C ) . Modeled as a heterohexameric LAP1-TorsinA assembly , the strictly conserved R563 in LAP1 is positioned as an R-finger to point into the nucleotide-binding pocket of a neighboring TorsinA protomer ( D ) . Residues important for nucleotide interaction are labeled in the TorsinA model . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 00810 . 7554/eLife . 03239 . 009Figure 2—figure supplement 1 . Phylogenetic analysis of Torsin . Sequence alignment of phylogenetically diverse sequences of Torsin . Residues are color-coded from dark blue ( most conserved ) to white ( not conserved ) . The canonical nucleotide interaction motifs , and other Torsin-specific regions , are labeled . Conserved cysteines are marked with yellow circles . Note that within the R-finger region , Torsin has no conserved arginine . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 009 Having established that LAP1 represents a nucleotide-free AAA+ domain , we asked how this might inform us about LAP1's interaction with Torsins , which can be modeled with great confidence as AAA+ ATPases ( Zhu et al . , 2010 ) . The four canonical nucleotide-sensing elements , Walker A , Walker B , Sensor 1 , and Sensor 2 ( Erzberger and Berger , 2006 ) , are well conserved and immediately recognizable in TorsinA when compared to ClpB-D2 ( Figure 2C , D , Figure 2—figure supplement 1; Kock et al . , 2006; Zhu et al . , 2010 ) . AAA+ ATPases are often activated by an arginine residue ( ‘arginine finger’ ) in the neighboring protomer in the hexameric ring assembly , such that this arginine is positioned to reach the phosphate binding site ( Wendler et al . , 2012 ) . While we do not find a conserved arginine in TorsinA ( Figure 2—figure supplement 1 ) at the expected position at the end of helix α5 , in LAP1 this residue is a strictly conserved arginine ( R563 ) . Consequently , modeling and phylogenetic analysis strongly suggest that LAP1 and TorsinA might form a heterohexameric ring ( Figure 3A , B ) with three active sites in which TorsinA and its partner subunits alternate to form the ring . This hypothesis is further strengthened by the recent observation that LAP1 and LULL1 activate TorsinA in vitro ( Zhao et al . , 2013 ) . 10 . 7554/eLife . 03239 . 010Figure 3 . Heterohexameric ring assembly . ( A ) Heterohexameric model of alternating LAP1 ( light blue ) and TorsinA ( green hues ) , based on the hexameric ring of ClpC-D2 domains ( 3PXI; Wang et al . , 2011 ) . The C-terminal domain of Torsin is colored bright green . An ATP molecule ( red ) is modeled into the TorsinA nucleotide binding pocket . The conserved arginine finger in LAP1 is in blue . The disulfide bridge within LAP1 is in yellow . ( B ) Schematic diagram of ( A ) , same coloring scheme . ( C ) A montage of representative particle classes of the TorsinA ( E171Q ) :LULL1 complex obtained from negatively stained particles are shown . The particles are consistent with a toroidal hexameric conformation as shown in ( A and B ) . The particle diameter is approximately 120 Å ( double arrow ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 01010 . 7554/eLife . 03239 . 011Figure 3—figure supplement 1 . Analytical gel filtration of TorsinA ( E171Q ) :LAP1 and TorsinA ( E171Q ) :LULL1 complexes . Elution profiles of TorsinA ( E171Q ) :LAP1 ( red ) and TorsinA ( E171Q ) :LULL1 ( red dashes ) are compared with uncomplexed LAP1 ( blue ) and LULL1 ( blue dashes ) , respectively . The experiment was performed on a Superdex S200 HR10/300 column . Elution volumes of standard globular protein markers are displayed . SDS-PAGE analysis of peak fractions is shown . The fractions are indicated below the chromatogram . The TorsinA:LAP1 and TorsinA:LULL1 protein complexes elute at positions consistent with a heterodimer . SDS-PAGE analysis indicates a 1:1 stoichiometry for the complexes . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 01110 . 7554/eLife . 03239 . 012Figure 3—figure supplement 2 . Negative-stain micrographs . Protein samples were negatively stained with uranyl acetate and analyzed by electron microscopy . All samples were measured at closely comparable concentrations . Scale bars represent 100 nm . While ring-like structures are visible in the TorsinA ( E171Q ) :LULL1 and TorsinA ( E171Q ) :LAP1 samples ( A and B ) , no discernible objects are seen on the micrographs of uncomplexed LAP1 and LULL1 ( C and D ) . This indicates that LAP1 and LULL1 do not form homohexameric rings in the absence of TorsinA . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 012 To experimentally confirm assembly of LAP1 and TorsinA into a heterohexameric ring , we co-expressed and co-purified TorsinA ( E171Q ) in complex with either LAP1 and LULL1 via nickel-affinity and size-exclusion chromatography . We obtained complexes with TorsinA ( E171Q ) :LAP1 and TorsinA ( E171Q ) :LULL1 1:1 stoichiometry ( Figure 3—figure supplement 1 ) . The E171Q mutation traps TorsinA in the ATP-bound form , which helps to stabilize the interaction with LAP1 and LULL1 , respectively ( Goodchild and Dauer , 2005 ) . For negative-stain electron microscopic analysis , the nickel eluate was used directly . On individual micrographs , rings of expected size were observed for the two-complex preparation , but not for individually purified LAP1 ( 356–583 ) or LULL1 ( 322–570 ) ( Figure 3—figure supplement 2 ) . TorsinA ( E171Q ) is insoluble without binding partner , therefore the observed rings should not be homomeric ( Figure 1—figure supplement 3 ) . The best micrographs , obtained with the TorsinA ( E171Q ) :LULL1 preparation , were further processed . Class averaging of 808 TorsinA ( E171Q ) :LULL1 particles yielded multiple classes showing ring assembly ( Figure 3C ) . The rings have a diameter of ∼120 Å , in good agreement with typical hexameric AAA+ ATPase assemblies . We then tested the ATPase activity of TorsinA in the context of LAP1 and LULL1 . In comparison to wild-type TorsinA:LAP1 , the mutants TorsinA:LAP1 ( R563A ) and TorsinA:LULL1 ( R449A ) showed substantial reduction in ATPase activity , while both TorsinA ( E171Q ) :LAP1 and TorsinA ( E171Q ) :LULL1 are essentially inactive ( Figure 4 ) . We note that the ATP hydrolysis rates are very slow and that the effect of the arginine finger mutation is not as drastic as seen with other AAA+ ATPases . We speculate that the ATP hydrolysis rate in the presence of substrate is likely higher , and that the effect of the arginine finger is likely more pronounced in a more physiological context . 10 . 7554/eLife . 03239 . 013Figure 4 . ATPase assay . ( A ) TorsinA:LAP1 and TorsinA:LULL1 complexes were tested for ATPase activity in a coupled ADP/NADH assay , where the oxidation of NADH is monitored spectrometrically ( Norby , 1988 ) . Shown are the oxidation rate plots , from which the ATP hydrolysis rates are calculated using linear regression . Reactions were performed in triplicate . ( B ) Bar graph representation of ATPase assay results shown in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 013 In summary , we conclude that LAP1 and LULL1 both activate Torsins by providing the arginine finger in a heterohexameric ring assembly in which LAP1 and LULL1 alternate with Torsin . Since LAP1 and LULL1 contain transmembrane helices immediately N-terminal to their luminal , nucleotide-free AAA+ Activator domain , they bring Torsin into close proximity to the membrane ( Figure 5 ) . 10 . 7554/eLife . 03239 . 014Figure 5 . Model for Torsin activation and localization . LAP1 ( light blue ) is localized to the nuclear envelope due to its interaction with the nuclear lamina , while LULL1 ( dark blue ) is found throughout the endoplasmic reticulum . Both proteins can bind inactive TorsinA ( red ) and target it to their respective locations . Both LAP1 and LULL1 activate TorsinA ( green ) when assembled into heterohexameric rings . DOI: http://dx . doi . org/10 . 7554/eLife . 03239 . 014 Here we provide structural evidence to explain how LAP1 and LULL1 function as activating proteins for Torsin . This confirms recent work which biochemically showed that LAP1 and LULL1 activate TorsinA ( Zhao et al . , 2013 ) . A heterohexameric ring assembly for AAA+ ATPases is unusual , but not unprecedented ( Gribun et al . , 2008; Saffian et al . , 2012 ) . However , to our knowledge this is the first description of a heterohexameric assembly where a canonical AAA+ domain alternates with a nucleotide-free AAA+ like domain , which we now call AAA+ Activator domain . It is striking that LAP1 and LULL1 lack all canonical nucleotide-binding elements . This feature has undoubtedly complicated their detection by sequence-based methods . Using the LAP1 structure , we performed a reverse hidden Markov model search to possibly identify structurally similar proteins in various organisms using Backphyre ( Kelley and Sternberg , 2009 ) . Interestingly , the top hits that we find besides LAP1 homologs are the Torsins . This indicates that both proteins are possibly derived from a common ancestor . In Drosophila as well as in Caenorhabditis , we find LAP1 homologs that have so far escaped detection ( Figure 1—figure supplement 1 ) . Gratifyingly , for all species where so far an orphaned Torsin has been found , we now also find a LAP1/LULL1 homolog ( for many species , it is difficult to distinguish between LAP1 and LULL1 based on sequence alone ) . This lends further support to our model suggesting that a heterohexameric LAP1-Torsin ring is the functionally relevant protein assembly . Our structure was obtained using a VHH raised against LAP1 . We can only speculate as to why VHH-BS1 helped in crystallizing the protein . Perhaps VHH-BS1 fortuitously generates packing contacts important for crystal formation . We note that VHH-BS1 binds to the area immediately surrounding the arginine finger of LAP1 , thereby interfering with TorsinA binding ( Figure 1—figure supplement 3 ) . Inclusion of VHH-BS1 might prevent cryptic and poorly ordered hexamerization of LAP1 at high protein concentrations and therefore favor a more ordered conformation conducive to crystallization . Interestingly , VHH-BS1 can compete with LAP1 for TorsinA binding in vitro ( Figure 1—figure supplement 3 ) . When expressed intracellularly , it might be a useful tool to further characterize TorsinA-LAP1 function in vivo . What are the substrate ( s ) of Torsin ? Torsin acts in close proximity to the NE or the ER membrane , depending on whether it interacts with membrane-bound LAP1 or LULL1 , respectively . Membrane proximity is further ensured by the N-terminal hydrophobic region within TorsinA ( Vander Heyden et al . , 2011 ) . Using an ATP-trapped mutant , it was shown that TorsinA is involved in the exit of large ribonucleoprotein ( mRNP ) granules from the nucleus through a pathway akin to the nuclear egress of herpes-type viruses ( Rose and Schlieker , 2012; Speese et al . , 2012; Jokhi et al . , 2013 ) . A substrate-trap mutant of TorsinA localized to the neck of mRNP-filled vesicles that bud from the inner nuclear membrane ( INM ) into the perinuclear space ( Jokhi et al . , 2013 ) , consistent with our model of LAP1-mediated activation of TorsinA close to the INM . We therefore speculate that the so far elusive Torsin substrates include proteins involved in membrane scission . This is akin to the action of the AAA+ ATPase Vps4 involved in the scission of narrow membrane necks mediated by ESCRT-III ( endosomal sorting complexes required for transport ) components ( Hill and Babst , 2012; McCullough et al . , 2013 ) , but with an important difference: while Vps4 acts on the inside of the necks , Torsins presumably act on the outside . We further speculate that under more physiological conditions , that is , properly complexed with LAP1 or LULL1 and engaged with substrate , the ATPase activity of Torsin might be substantially higher than that reported under the in vitro conditions tested so far . Apart from the identification of Torsin substrate ( s ) , understanding the catalytic mechanism of the Torsin-LAP1/LULL1 machinery is equally important . Because of the unusual heterohexameric architecture with only three active sites , we expect substantial differences between this system and those that are better understood . One important aspect will be to unravel the role of the highly conserved cysteines , both in LAP1/LULL1 and in Torsin . While both proteins have conserved disulfides , these do not account for all cysteines and suggest a possible redox mechanism , as has been discussed previously ( Zhu et al . , 2010 , 2008 ) . We note that in our model the most conserved cysteine C496 in LAP1 cannot engage in an intramolecular disulfide bridge , but is close to the nucleotide-binding site of the neighboring TorsinA . This remarkable conservation might indicate a role in a redox mechanism that is coupled to the catalytic cycle of Torsin-LAP1/LULL1 . The lumen of the endoplasmic reticulum is a site rich in oxidoreductases that assist in protein folding and assembly . The catalytic function ( s ) of the Torsin-LAP1/LULL1 assemblies may have co-opted part of this machinery . Recombinant proteins were expressed in Escherichia coli . Human TorsinA ( aa51–322 ) , the luminal domain of human LAP1 ( aa356–583 ) , and the luminal domain of human LULL1 ( aa233–470 ) were expressed from a modified ampicillin resistant pETDuet-1 ( EMD Millipore , Billerica , MA ) vector as N-terminally 6×His-7×Arg-tagged fusion proteins . A cleavage site for human rhinovirus 3C protease was inserted after the 7×Arg tag . For co-expression of TorsinA with LAP1 or LULL1 , cells were co-transformed with the TorsinA vector described above and a second , modified kanamycin resistant pETDuet-1 vector containing untagged LAP1 or LULL1 . VHH-BS1 was expressed as a C-terminally 6×His-tagged fusion protein from a kanamycin resistant pET-28b ( + ) ( EMD Millipore ) vector . Mutations were introduced by site-directed mutagenesis . All proteins were expressed in LOBSTR ( DE3 ) RIL strains ( Kerafast , Boston , MA ) ( Andersen et al . , 2013 ) . Bacterial cultures were grown at 30°C to an optical density ( OD600 ) of 0 . 6 , shifted to 18°C for 30 min , and induced overnight at 18°C with 0 . 2 mM IPTG . LAP1 and VHH-BS1 expressing cells were resuspended in lysis buffer A ( 50 mM potassium phosphate pH 8 . 0 , 400 mM NaCl , 40 mM imidazole ) and lysed . The lysate was supplemented with 1 U/ml Benzonase ( Sigma-Aldrich , St . Louis , MO ) and 1 mM PMSF , cleared by centrifugation , and loaded onto a Ni-affinity resin . After washing with lysis buffer , bound protein was eluted with elution buffer ( 10 mM potassium phosphate pH 8 . 0 , 150 mM NaCl , 250 mM imidazole ) . For LAP1 , the eluted protein was purified by cation-exchange chromatography against a gradient of 0 . 150–2 M NaCl with 10 mM potassium phosphate pH 8 . 0 , followed by dialysis into cleavage buffer ( 10 mM potassium phosphate pH 8 . 0 , 150 mM NaCl ) and tag removal with 3C protease followed by another round of cation-exchange chromatography to remove tag and protease . The flow-through from the cation-exchange chromatography was concentrated and purified via size exclusion chromatography on a Superdex S200 column ( GE Healthcare ) equilibrated in buffer ( 10 mM Tris/HCl pH 7 . 4 , 150 mM NaCl ) . For VHH-BS1 , the eluted protein was concentrated and purified via size exclusion chromatography on a Superdex S75 column ( GE Healthcare ) equilibrated in buffer ( 10 mM Tris/HCl pH 7 . 4 , 150 mM NaCl ) . TorsinA-containing preparations were performed according to the same procedure with the following modifications: ( 1 ) the lysis buffer contained 10 mM MgCl2 , 1 mM ATP , and 50 mM HEPES/NaOH pH 8 . 0 instead of potassium phosphate; ( 2 ) the elution buffer contained 10 mM MgCl2 , 1 mM ATP , and 10 mM HEPES/NaOH pH 8 . 0 instead of potassium phosphate; ( 3 ) no ion-exchange step was performed; and ( 4 ) size exclusion buffer contained 10 mM MgCl2 and 0 . 5 mM ATP . The LAP1-VHH-BS1 complex was formed by mixing the individual components using a twofold molar excess of VHH-BS1 and incubating for 30 min on ice , followed by size exclusion chromatography on a Superdex S75 column ( GE Healthcare ) equilibrated in buffer ( 10 mM Tris/HCl pH 7 . 4 , 150 mM NaCl ) . Purified LAP1-VHH-BS1 was concentrated to 10 mg/ml prior to crystallization . The complex crystallized in 20% ( wt/vol ) polyethylene glycol ( PEG ) 3350 , 100 mM Bis-Tris pH 5 . 5 , and 200 mM ammonium sulfate by the hanging drop vapor diffusion method in 2 µl drops at 18°C . Crystals grew within 4–12 days with dimensions of 35 μm × 35 μm × 35 μm . Prior to X-ray data collection , crystals were cryoprotected in the reservoir solution supplemented with 30% ( wt/vol ) PEG 3350 and 15% ( vol/vol ) glycerol . Data were collected at beamlines 24ID-C/-E at Argonne National Laboratory . Data reduction was carried out using HKL2000 software ( Otwinowski and Minor , 1997 ) ; all other software was collectively used through SBGrid ( Morin et al . , 2013 ) . The LAP1-VHH-BS1 structure was solved using molecular replacement . The MRage pipeline procedure from the PHENIX suite ( Adams et al . , 2010 ) was used for initial search model determination . The Phaser-MR tool was subsequently used for phasing and initial refinement . In the initial map , with phases only provided by the VHH search model , the outlines of the major secondary structure elements of LAP1 were visible . With iterative model building and refinement , the model phases gradually improved and the electron density maps became better defined . The final model was refined against native data extending to 1 . 6 Å . The data were cut judged by the CC1/2 value for the highest resolution shell , and visual inspection of the 2Fo-Fc map ( Figure 1—figure supplement 2 ) . Model building was carried out with Coot ( Emsley et al . , 2010 ) and refinement was done with phenix refine from the PHENIX suite . Ni-affinity purified TorsinA ( E171Q ) -LAP1 ( 356–583 ) , TorsinA ( E171Q ) -LULL1 ( 233–470 ) , LAP1 ( 356–583 ) , and LULL1 ( 233–470 ) ( ∼0 . 1 mg/ml ) were used for negative stain electron microscopy grids . Continuous carbon-film grids were glow-discharged ( PELCO easiGlow , Redding , CA ) at 15 mA for 30 s . Next , 5 µl specimen solution was loaded immediately onto the grid and blotted after 15 s . The specimen was then stained by uranyl acetate ( 1% wt/vol ) for 10 s , blotted , and dried . Thirty single-particle electron micrographs were recorded using a 2K × 2K CCD camera on an FEI Tecnai Spirit electron microscope at 80 keV , 98 , 000× nominal magnification ( 3 . 63 Å/pixel ) . For TorsinA ( E171Q ) -LULL1 ( 233–470 ) , 808 particles were boxed out and subjected to ‘direct classification’ , an unbiased , reference-free , alignment-free classification function in the PARTICLE software package ( www . sbgrid . org/software/title/PARTICLE ) . ATP hydrolysis rates of TorsinA:LAP1 and TorsinA:LULL1 complexes were measured by an NADH-coupled assay ( Norby , 1988 ) . In this assay , each ATP hydrolysis event allows conversion of one molecule of phosphoenolpyruvate into pyruvate by pyruvate kinase . Thereafter , pyruvate is converted to lactate by L-lactate dehydrogenase , which results in oxidation of a single NADH molecule . Loss of NADH over time , which is quantifiably proportional to ATP hydrolysis rates , is monitored by a decrease in absorbance at 340 nm . All of the assays were conducted at room temperature in a buffer containing 10 mM HEPES/NaOH pH 8 . 0 , 150 mM NaCl , and 10 mM MgCl2 , and in the presence of 2 mM ATP . Measurements were performed in triplicates using 5 μM of protein complex . Absorbance was measured in 30 µl reaction volume using a 384-well plate reader . Data analysis was performed using Prism . TorsinA ( E171Q ) :LAP1 complex was used in precipitation experiments . First , 100 pmol of TorsinA ( E171Q ) :LAP1 complex was mixed with 200 pmol of VHH molecules and incubated at 4°C in the presence of 10 mM HEPES/NaOH pH 8 . 0 , 150 mM NaCl , 10 mM MgCl2 , and 0 . 25 mM ATP . Reactions were centrifuged at 13 , 000×g for 5 min at different time points to separate soluble and insoluble proteins , and the pellets were solubilized in 8 M urea . Equal volumes of samples from soluble and insoluble proteins were analyzed by SDS-PAGE gel electrophoresis . Size exclusion chromatography was performed on a 10/300 Superdex 200 column in 10 mM HEPES/NaOH pH 8 . 0 , 150 mM NaCl , 10 mM MgCl2 , and 0 . 5 mM ATP . Samples ( 7 . 5 nmol ) of LAP1 and TorsinA ( E171Q ) :LAP1 were loaded in 200 μl injections and peak fractions were analyzed by SDS-PAGE gel electrophoresis . For comparison , 2 . 5 nmol of LULL1 and TorsinA ( E171Q ) :LULL1 were loaded in 200 μl injections . Multiple sequence alignments of Torsin and LAP1 were performed using MUSCLE ( Edgar , 2004 ) and visualized with Jalview ( Waterhouse et al . , 2009 ) . Since differentiating between LAP1 and LULL1 is difficult based on primary sequence , we refer to the sequences as LAP1 for simplicity . The species nomenclature used for the alignments is as follows: Homo sapiens ( hs ) , Ornithorhynchus anatinus ( oa ) , Gallus gallus ( gg ) , Takifugu rubripes ( tr ) , Danio rerio ( dr ) , Branchiostoma floridae ( bf ) , Strongylocentrotus purpuratus ( stp ) , Ciona savignyi ( cs ) , Ciona intestinalis ( ci ) , Nematostella vectensis ( nv ) , Caenorhabditis elegans ( ce ) , and Drosophila melanogaster ( dm ) . The structure of human TorsinA was modeled using HHpred in combination with Modeler through the Bioinformatics Toolkit platform ( Biegert et al . , 2006 ) . The D2 domain of ClpA ( PDB entry 1R6B; Xia et al . , 2004 ) was picked as the closest template structure . Models were generated with another four closely related structures . While there are a few uncertain loops and some discrepancies in modeling the C-terminal domain , the model of the nucleotide binding site is nearly identically in all cases . The structure of the heterohexameric TorsinA-LAP1 was modeled based on the hexameric D2 ring within the MecA-ClpC assembly ( 3PXI; Wang et al . , 2011 ) , by alternately superimposing the LAP1 structure and the TorsinA model onto neighboring ClpC-D2 domains within the ring . An adult male alpaca ( Lama pacos ) was purchased locally , maintained in pasture , and immunized following a protocol authorized by the Tufts University Cummings Veterinary School Institutional Animal Care and Use Committee . Recombinantly expressed LAP1 was used for immunization following a standard protocol ( Maass et al . , 2007 ) . Following the final boost , peripheral blood lymphocytes ( PBLs ) were harvested from blood as described ( Maass et al . , 2007 ) . Total RNA was isolated from 106 freshly isolated PBLs using an RNeasy Plus Mini Kit ( Qiagen , Hilden , Germany ) , following the manufacturer's guidelines . First strand cDNA synthesis was performed using SuperScript III reverse transcriptase ( ThermoFisher Scientific , Waltham MA ) and a combination of poly ( A ) oligo dT , random hexamer or primers specific to alpaca immunoglobulin , AlCH2 and AlCH2 . 2 . Subsequent PCR amplification of VHH sequences and phage library generation followed the procedure described in Maass et al . ( 2007 ) , including the use of alpaca-specific primers for VHH gene amplification and a phagemid vector adapted for VHH expression as a pIII fusion . Following transformation into TGI cells ( Agilent , Santa Clara , CA ) , the total number of independent clones was estimated to be 4 × 106/ml . Ninety six clones were selected at random and sequenced to assess library diversity . The resulting phagemid library was stored at −80°C . A 1 µl sample of the 4 × 106 library was used to inoculate 100 ml SOC with 10 µg/ml ampicillin . The culture was grown to mid-log phase , and infected with 100 µl 1014 pfu/ml VCSM13 helper phage . The culture was then incubated for 2 hr at 37°C , and the cells harvested by centrifugation , and resuspended in 100 ml 2YT , 0 . 1% glucose , 50 µg/ml kanamycin , and 10 µg/ml ampicillin . Cultures were incubated overnight at 30°C , then centrifuged for 20 min at 8000 rpm , and phage precipitated from the resulting supernatant with 1% PEG 6000/500 mM NaCl at 4°C , and resuspended in PBS . A 100 μg sample of LAP1 was labeled via coupling to primary amines with a fivefold molar excess of Chromalink NHS biotin reagent ( Solulink , San Diego , CA ) for 90 min in 100 mM phosphate buffer pH 7 . 4 , 150 mM NaCl . The reaction was then run through an Amicon 10 kDa MWCO concentrator ( EMD Millipore ) to remove unreacted biotin . Incorporation of biotin was monitored spectrophotometrically . A 100 μl sample of MyOne Streptavidin T1 Dynabeads ( ThermoFisher Scientific ) was blocked in PBS/2% BSA for 2 hr at 37°C . Following blocking , 20 µg biotinylated antigen in PBS was added to the beads , and incubated for 30 min at room temperature ( RT ) , with rotation . The beads were then washed 3× in PBS , and 200 µl of 1014 pfu/ml M13 phage displaying the VHH library was added in PBS/2% BSA for 1 hr at RT . The beads were then washed 15× in PBS/0 . 1% Tween 20 . Phage was eluted by the addition of ER2738 E . coli ( NEB , Ipswich , MA ) for 15 min at 37°C , followed by elution with 200 mM glycine pH 2 . 2 for 10 min at RT . The glycine elution was neutralized and pooled with ER2738 culture , plated onto 2YT agar plates supplemented with 2% glucose , 5 µg/ml tetracycline , and 10 µg/ml ampicillin , and grown overnight at 37°C . This enriched library was then used for a second round of panning as described above with the following exceptions: 2 µg of biotinylated antigen was used as bait , and incubated with 2 µl 1014 pfu/ml M13 phage displaying the VHH library for 15 min at 37°C , followed by longer washes in PBS/0 . 1% Tween 20 . Following two rounds of phage display panning , 96 colonies were isolated in 96-well round bottom plates and grown to mid-log phase at 37°C in 200 µl 2YT , 10 µg/ml ampicillin , and 5 µg/ml tetracycline , then induced with 3 mM IPTG , and grown overnight at 30°C . Plates were centrifuged at 2500 rpm for 10 min , and 100 µl of supernatant mixed 1:1 with PBS plus 5% non-fat dry milk was then added to an ELISA plate coated with 1 µg/ml antigen . Following multiple washes in PBS plus 1% Tween 20 , anti-Etag-HRP antibody ( Bethyl ) was added at a 1:10 , 000 dilution in PBS plus 5% non-fat dry milk for 1 hr at RT . The plate was developed with fast kinetic TMB ( Sigma-Aldrich ) , quenched with 1 M HCl and read out at 450 nm ( Spectramax , Molecular Devices ) . ELISA positive clones were sequenced and subcloned into a bacterial expression vector ( see above ) .	Cells are filled with activity—proteins must be folded into intricate shapes and unfolded , complex molecules must be built and taken apart—and all this activity requires energy . Cells use a molecule called ATP as a source of energy , with the energy being released when enzymes called ATPases remove a phosphate group from the ATP molecule . There are many different ATPases , and when one of them does not work properly , the consequences can be severe . A single mutation in the gene for an ATPase called TorsinA , for example , can lead to a painful and severely disabling disorder called primary dystonia . However , scientists have yet to discover what the TorsinA enzyme does in cells and how mutated TorsinA causes primary dystonia . It is known that the TorsinA enzyme is found in the endoplasmic reticulum—a region of the cell where , among other things , proteins are folded and unfolded—and that it interacts with two membrane proteins , LAP1 and LULL1 . Also , TorsinA is a member of the AAA+ family of ATPases ( AAA+ is short for ATPases Associated with diverse cellular Activities ) , and these enzymes tend to combine with each other to form functional ring-shaped arrays . Now Sosa et al . have used a combination of X-ray crystallography , electron microscopy , computer modeling , and biochemistry to study the interactions between TorsinA and the two membrane proteins . This work revealed that , somewhat surprisingly , certain parts of the membrane proteins have structures that are similar to those of AAA+ ATPases . This allows the TorsinA enzymes and the membrane proteins to combine to form rings that contain three enzymes and three LAP1 proteins or three LULL1 proteins . Within these rings , the LAP1 and LULL1 proteins activate the TorsinA enzyme by supplying an amino acid called arginine to the neighboring TorsinA molecule . The amino acid is supplied to a site on the enzyme that can bind nucleotides ( which are the building blocks of DNA and RNA ) ; the LAP1 and LULL1 proteins themselves cannot bind nucleotides . The work of Sosa et al . may help scientists to better understand how TorsinA causes primary dystonia . One of the important next steps is to work out what molecules TorsinA acts on .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2014	How lamina-associated polypeptide 1 (LAP1) activates Torsin
Transcription factors organize gene expression profiles by regulating promoter activity . However , the role of transcription factors after transcription initiation is poorly understood . Here , we show that the homeoprotein Nkx2-5 and the 5’-3’ exonuclease Xrn2 are involved in the regulation of alternative polyadenylation ( APA ) during mouse heart development . Nkx2-5 occupied not only the transcription start sites ( TSSs ) but also the downstream regions of genes , serving to connect these regions in primary embryonic cardiomyocytes ( eCMs ) . Nkx2-5 deficiency affected Xrn2 binding to target loci and resulted in increases in RNA polymerase II ( RNAPII ) occupancy and in the expression of mRNAs with long 3’untranslated regions ( 3’ UTRs ) from genes related to heart development . siRNA-mediated suppression of Nkx2-5 and Xrn2 led to heart looping anomaly . Moreover , Nkx2-5 genetically interacts with Xrn2 because Nkx2-5+/-Xrn2+/- , but neither Nkx2-5+/-nor Xrn2+/- , newborns exhibited a defect in ventricular septum formation , suggesting that the association between Nkx2-5 and Xrn2 is essential for heart development . Our results indicate that Nkx2-5 regulates not only the initiation but also the usage of poly ( A ) sites during heart development . Our findings suggest that tissue-specific transcription factors is involved in the regulation of APA . Transcription factors and chromatin regulators orchestrate the processes of heart development by positively and negatively regulating thousands of genes ( Bruneau , 2010; Nimura et al . , 2009; Prall et al . , 2007; Srivastava , 2006; Takeuchi and Bruneau , 2009; Takeuchi et al . , 2011 ) . Mutations in transcription factors and chromatin regulators cause congenital heart disease ( CHD ) by disrupting gene expression profiles that are tightly regulated by transcription factor networks ( Bruneau , 2008; Srivastava , 2006 ) . However , the mechanisms by which transcription factor deficiencies cause CHD are not fully understood . Nkx2-5 , Gata4 , and Tbx5 are key cardiac transcription factors that coordinate transcription networks during heart development ( Akazawa and Komuro , 2005; Stennard and Harvey , 2005 ) . Haploinsufficiency of these genes can cause CHD in humans , and mice lacking any of these transcription factors exhibit severe defects in heart development ( Akazawa and Komuro , 2005; Stennard and Harvey , 2005 ) . Recent genome-wide analyses of transcription factors , including Nkx2-5 , Gata4 , and Tbx5 , have revealed that these transcription factors assist in the formation of active enhancers in the HL1 cardiomyocyte cell line and in murine adult hearts ( He et al . , 2011; Schlesinger et al . , 2011; van den Boogaard et al . , 2012 ) . He et al . demonstrated that multiple transcription factors activate certain cardiac enhancers without p300 ( He et al . , 2011 ) . In Drosophila , cardiac transcription factors converge on heart enhancer regions; however , the collective binding of these transcription factors to the enhancer regions does not require a conserved DNA motif ( Junion et al . , 2012 ) . Although the regulation of enhancer activity by cardiac transcription factors has been extensively studied , the other roles of these transcription factors during heart development remain poorly understood . Transcription factors require several chromatin regulators to precisely regulate gene expression . Indeed , the results from several histone methyltransferase-knockout mouse studies and genetic studies of human CHD patients have revealed the importance of histone modifications during heart development ( Delgado-Olguín et al . , 2012; Nimura et al . , 2009 ) . In particular , Nkx2-5 regulates gene expression in conjunction with Whsc1 ( Wolf-Hirschhorn Syndrome 1 , also known as NSD2 or MMSET ) , a histone H3 lysine 36 ( H3K36 ) methyltransferase . H3K36 methylation is associated with transcribed genomic regions and has several roles in transcription , including transcriptional repression , alternative RNA splicing , and DNA mismatch repair ( Li et al . , 2013; Luco et al . , 2010 ) . Therefore , Nkx2-5 may be involved in both transcriptional activation and the subsequent events . However , it is not clear whether particular transcription factors regulate events that occur after transcription initiation during heart development . In this study , we determined the genome-wide occupancy of transcription factors that are critical for heart development and the factors associated with these transcription factors as well as the histone modification signatures in embryonic hearts . The results indicate a role for transcription factors in alternative polyadenylation ( APA ) . Furthermore , we discovered that Nkx2-5 is associated with both transcription start sites ( TSSs ) and downstream regions of genes and that Nkx2-5 controls APA in conjunction with the 5’-3’ exonuclease Xrn2 . Simultaneously suppressing Nkx2-5 and Xrn2 caused heart-looping abnormalities . Moreover , Nkx2-5 genetically interacted with Xrn2 during heart development . Our findings suggest that Nkx2-5 is involved in the regulation of the length of the 3’ UTR and may help to elucidate the mechanisms by which transcription factor deficiencies can cause diseases such as CHD . To elucidate how Nkx2-5 , Gata4 , and Tbx5 regulate transcription , we examined the genomic target regions of these transcription factors and the transcription factor-associated factors as well as the chromatin status in mouse E12 . 5 hearts using chromatin immunoprecipitation-sequencing ( ChIP-seq ) ( Figures 1A , B , Figure 1—figure supplement 1 , Figure 1—figure supplement 2 , Figure 1—figure supplement 3 , Figure 1—source data 1 , and Figure 1—figure supplement 4 ) ( Li et al . , 2007; Lickert et al . , 2004; Nimura et al . , 2009 ) . Embryonic hearts at this stage are primarily ( >90% ) composed of cardiomyocytes that do not express Thy1 ( fibroblasts , T-lymphocytes , and neuronal markers ) ( Ieda et al . , 2009 ) . Defects in the ventricular septum and atrial septum , which are most frequently found in CHDs ( Feng et al . , 2002 ) , form during this stage ( Henderson and Copp , 1998 ) . Nkx2-5 and Tbx5 co-occupied the same global genomic regions as RNA polymerase II ( RNAPII ) and the heterogeneous nuclear ribonucleoprotein Raver1 ( Figure 1A ) . Nkx2-5 and Tbx5 were found to associate with the poised serine 5-phosphorylated form of RNAPII ( RNAPII-S5P ) ( Kuehner et al . , 2011 ) , although a physical association between Nkx2-5 and Tbx5 was not detected ( Figure 1B ) . In contrast , Gata4 was found in a different cluster ( Figure 1A ) . We next examined whether the occupancy of Nkx2-5 and Tbx5 is correlated with gene expression levels because these two transcription factors are associated with RNAPII . Nkx2-5 and Tbx5 were significantly enriched around the TSSs and downstream regions of genes that were highly expressed ( Figures 1C , Figure 1—figure supplement 5 , Figure 1—figure supplement 6 , and Figure 1—figure supplement 7 ) . These results suggest the possibility that Nkx2-5 and Tbx5 may be involved not only in enhancer activity regulation ( He et al . , 2011; Schlesinger et al . , 2011; van den Boogaard et al . , 2012 ) but also in post-transcriptional mRNA processing . To elucidate whether Nkx2-5 and Tbx5 play a role in regulating 3’-end processing , we examined changes in poly ( A ) -tailed mRNA in transcription factor-knockdown embryonic cardiomyocytes ( eCMs ) using mRNA-seq . Nkx2-5 knockdown increased the expression of long 3’ UTRs in Tnnt2 ( Troponin T2 , cardiac ) and Atp2a2 ( ATPase , Ca++ transporting , cardiac muscle , slow twitch 2 ) transcripts ( Figures 1D , Figure 1—figure supplement 5 , and Figure 1—figure supplement 8 ) . We also detected increased expression of long 3’ UTRs in Nkx2-5-knockout E9 . 5 hearts ( Figures 1E and Figure 1—figure supplement 9 ) . These results suggest a role for Nkx2-5 in the regulation of APA . 10 . 7554/eLife . 16030 . 003Figure 1 . Nkx2-5 deficiency increases transcription from regions downstream of transcription termination sites . ( A ) Co-occupancies of each pair of factors and histone modifications are shown . White indicates a high correlation , and red indicates a low correlation . ( B ) Nkx2-5 , Tbx5 , and Gata4 were immunoprecipitated from nuclear extracts of E12 . 5 hearts with the indicated antibodies . Co-immunoprecipitates and aliquots ( 6% ) of the input proteins were analyzed by Western blotting with the indicated antibodies . ( C ) Average ChIP-seq signal profiles over a 3-kb meta-gene , including 3 kb upstream and 3 kb downstream . The lines correspond to genes with High , Middle , Low , and No expression and all RefSeq genes . ( D and E ) Genome browser representation of strand-specific RNA-seq tag counts from eCMs transfected with the indicated siRNAs ( D ) and E9 . 5 Nkx2-5-/- hearts ( E ) . The red boxes indicate read-through RNAs . neg . , negative strand; pos . , positive strand . The arrow heads show polyadenylation sites . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 00310 . 7554/eLife . 16030 . 004Figure 1—source data 1 . Overlap of peaks between transcription factors and between the results from this study and those from previousely published studies . ( A ) Overlap of peaks between Nkx2-5 , Tbx5 , and Gata4 ChIPseq data in this study . ( B ) Overlap of Nkx2-5 peaks among E12 . 5 hearts ( this study ) , HL1 cells with BirA-fused Nkx2-5 ( HL1_BirA ) ( He et al . , 2011 ) , and Adult hearts ( van den Boogaard et al . , 2012 ) . ( C ) Overlap of Tbx5 between E12 . 5 hearts ( this study ) and HL1 cells with BirA-fused Tbx5 ( HL1_BirA ) ( He et al . , 2011 ) . ( D ) Overlap of Gata4 peaks between native ChIPseq in this study and previousely published crosslink ChIPseq ( He et al . , 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 00410 . 7554/eLife . 16030 . 005Figure 1—figure supplement 1 . Transcription factors-associated proteins in E12 . 5 hearts . This figure is related to Figure 1b . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 00510 . 7554/eLife . 16030 . 006Figure 1—figure supplement 2 . ChIPseq replicate correlations . Scatterplots of pair-wise ChIPseq replicates and Pearson correlation are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 00610 . 7554/eLife . 16030 . 007Figure 1—figure supplement 3 . In vivo transcription factor binding motif by native ChIPseq . De novo motif analysis by Homer using all peaks in ChIPseq data . Obtained motifs are compared with the most matched known motif , respectively . The motif ( T/C ) GATTGG found in Gata4 peaks is similar to the motif TGATTG that Gata proteins strongly bind ( Merika and Orkin , 1993 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 00710 . 7554/eLife . 16030 . 008Figure 1—figure supplement 4 . Validation of the antibodies used for ChIP-seq . The indicated proteins were immunoprecipitated from nuclear extracts of E12 . 5 hearts using the corresponding antibodies . Arrowheads indicate immunoprecipitated proteins . Asterisks indicate the IgG heavy or light chains . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 00810 . 7554/eLife . 16030 . 009Figure 1—figure supplement 5 . Genome browser representation at Tnnt2 and Atps2a2 loci . Genome browser representation of the indicated histone modifications , transcription factors , and transcription factor-associated protein enrichment profiles in E12 . 5 hearts is shown for the Tnnt2 and Atp2a2 loci , with strand-specific RNA-seq tag counts from eCMs transfected with the indicated siRNA . The arrow heads show polyadenylation sites . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 00910 . 7554/eLife . 16030 . 010Figure 1—figure supplement 6 . Heatmap of factor occupancy and histone modification enrichment for 8 kb regions centred on TSSs ( left panel ) and TTSs ( middle panel ) are shown with reference to the RefSeq gene expression level ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01010 . 7554/eLife . 16030 . 011Figure 1—figure supplement 7 . Average signal profiles over a 3 kb meta-gene including 3 kb upstream and 3 kb downstream . The ChIP-seq data were analysed using CEAS . The lines correspond to genes with High , Middle , Low , and No expression and all RefSeq genes . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01110 . 7554/eLife . 16030 . 012Figure 1—figure supplement 8 . Genome browser representation of strand-specific RNA-seq tag counts from eCMs transfected with the indicated siRNAs . Red boxes indicate read-through RNAs . neg . , negative strand . , pos . , positive strand . The arrow heads show polyadenylation sites . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01210 . 7554/eLife . 16030 . 013Figure 1—figure supplement 9 . mRNA with long 3’UTR in Nkx2-5-knockout embryonic hearts . ( A ) Genome browser representation of the read-through RNAs in Nkx2-5-knockout embryonic hearts . Red boxes indicate long 3’UTR . The arrow heads show polyadenylation sites . ( B ) Average profiles of read-through RNAs that are increased in Nkx2-5-knockout embryonic hearts . Significance was assessed using the two-sample Kolmogorov-Smirnov test . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 013 Recent studies have revealed that enhancer-promoter looping mediated by transcription factors is important for gene regulation ( Kagey et al . , 2010; Wang et al . , 2011 ) . As shown in Figure 1 , Nkx2-5 is located at the TSSs and downstream regions of highly expressed genes and is involved in regulating 3’-end processing , which implies that Nkx2-5 may organize chromatin conformations between TSSs and downstream regions of genes and that this Nkx2-5-mediated chromatin conformation may be related to 3’-end processing . A chromatin conformation capture ( 3C ) assay demonstrated that TSSs interact with the downstream regions of two genes , Tnnt2 and Atp2a2 , which are highly expressed in eCMs . The looping between the TSSs and downstream regions of the genes ( TSS-downstream looping ) was dependent on Nkx2-5 but not on Gata4 or Tbx5 ( Figure 2A and Figure 2—figure supplement 1 ) . However , Nkx2-5 knockdown did not affect the TSS-downstream looping of Tnni1 and did not increase the expression of long 3’ UTRs ( Figure 2—figure supplement 2 ) . Furthermore , the binding of Nkx2-5 to the TSSs and downstream regions of Tnnt2 and Atp2a2 in eCMs was also confirmed using ChIP-qPCR at the Nkx2-5-bound downstream regions ( Figure 2B ) . This finding suggests that the loss of TSS-downstream looping is related to the increased expression of long 3’ UTRs . 10 . 7554/eLife . 16030 . 014Figure 2 . A link between Nkx2-5-dependent chromatin conformation and RNAPII . ( A ) Chromatin conformation capture ( 3C ) analysis of the TSSs and downstream regions of Tnnt2 and Atp2a2 in the indicated siRNA-treated eCMs . The corresponding BACs for the regions were used as controls . Undigested regions at Tnnt2 and Atp2a2 were used as Controls . Red arrow , direction of transcription; blue arrow , anchoring primer; black arrow , primer; red line , restriction enzyme site . The arrow heads show polyadenylation sites . ( B–D ) , Relative Nkx2-5 ( B ) and RNAPII-S2P occupancy ( C and D ) at the TSSs ( -TSS ) and downstream regions ( -3 ) of Tnnt2 ( 18 . 4 kb ) , Atp2a2 ( 47 . 2 kb ) , and the silent histone H1foo gene , which served as a negative control , was analyzed by ChIP . ( E and F ) qRT-PCR analysis of mRNA expression from the coding region ( E ) and expression of long 3’ UTRs ( F ) of Tnnt2 and Atp2a2 , normalized to Rplp2 . Error bars indicate the mean ± s . e . m . ( n = 3 ) . * , p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01410 . 7554/eLife . 16030 . 015Figure 2—source data 1 . Source data for Figure 2 and Figure 2—figure supplement 1 and 3 . Numeric data for Figure 2B , C , D , E , F , Figure 2—figure supplement 1A , B , Figure 2—figure supplement 3A , B . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01510 . 7554/eLife . 16030 . 016Figure 2—figure supplement 1 . Quantification of 3C and western blotting data . ( A ) Quantification of 3C data by qPCR at Figure 2A . Tnnt2 , n = 5 . Atp2a2 , n = 4 . * , p < 0 . 05 . ( B ) Quantification of western blotting data at Figure 4H . n = 3 . * , p < 0 . 05 , compared to siControl . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01610 . 7554/eLife . 16030 . 017Figure 2—figure supplement 2 . The chromatin conformation of Tnni1 is independent of Nkx2-5 . ( A ) Genome browser representation of strand-specific RNA-seq tag counts from eCMs transfected with the indicated siRNAs . neg . , negative strand; pos . , positive strand . ( B ) Chromatin conformation capture analysis of the TSS and the downstream region of Tnni1 in the indicated siRNA-treated eCMs . BACs corresponging to the region were used as a control . Red arrow , direction of gene; blue arrow , anchoring primer; black arrow , primer; red line , restriction enzyme site . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01710 . 7554/eLife . 16030 . 018Figure 2—figure supplement 3 . siRNA knockdown efficiencies of three different siRNAs for each gene . ( A ) eCMs were transfected with three different siRNAs for each gene , and the expression level of Nkx2-5 , Gata4 , and Tbx5 was measured by real-time PCR . Expression values were normalised against Rplp2 relative to control siRNA-treated eCMs . ( B ) Quantitative RT-PCR analysis of read-through RNA expression in eCMs transfected with the indicated siRNAs . Error bars indicate the mean ± s . e . m . ( n = 3 ) . siNkx2-5-8958 , siGata4-5094 , and siTbx5-9160 were used in this study because these siRNAs most efficiently reduced the expression of each target gene and minimally affect the expression of other genes . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 01810 . 7554/eLife . 16030 . 019Figure 2—figure supplement 4 . Functional annotations of genes with increased and decreased mRNA with long 3’UTR . ( A ) The knockdown of Nkx2-5 , Gata4 , and Tbx5 in eCMs was analysed by Western blotting . siNkx2-5 , Nkx2-5 siRNA; siTbx5 , Tbx5 siRNA; siGata4 , Gata4 siRNA; WT , wild type . ( B , C ) Enriched gene ontologies in eCMs ( for up-regulated genes: siNkx2-5 , 89 genes; siTbx5 , 58 genes; siGata4 , 182 genes . for down-regulated genes: siNkx2-5 , 12 genes; siTbx5 , 9 genes; siGata4 , 13 genes ) and E9 . 5 Nkx2-5-/- hearts ( up-regulated genes , 200 genes; down-regulated genes , 90 genes ) are shown . N . D . , not detected . The p value is plotted on the x-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 019 To investigate whether TSS-downstream looping is involved in transcription regulation , we examined whether the occupancy of the elongating serine 2-phosphorylated form of RNAPII ( RNAPII-S2P ) ( Kuehner et al . , 2011 ) was increased at these downstream regions in accordance with the increased expression of long 3’ UTRs . RNAPII-S2P binding downstream of Tnnt2 and Atp2a2 was increased in Nkx2-5-knockdown eCMs ( Figures 2C and D ) . Consistent with the increase in RNAPII at downstream regions , Nkx2-5 knockdown increased the expression of long 3’ UTRs; however , the mRNA expression of the coding regions was not substantially increased ( Figures 2E , F , and Figure 2—figure supplement 3 ) . Although Gata4 knockdown may affect the regulation of APA , we could not detect a significant difference in the APA of Tnnt2 and Atp2a2 ( Figure 2F ) . These findings further support the possibility that Nkx2-5-mediated TSS-downstream looping is associated with RNAPII termination . To investigate whether mRNAs with long 3’ UTRs in Nkx2-5-knockdown eCMs are enriched for specific functional annotations , we performed Gene Ontology ( GO ) analysis ( Huang et al . , 2009a; 2009b ) . We observed that the mRNAs with long 3’ UTRs that were increased by the knockdown and knockout of Nkx2-5 were enriched for GO terms related to heart development ( Figure 2—figure supplement 4 ) , whereas in mRNAs with long 3’ UTRs that were increased by the knockdown of Tbx5 and Gata4 , none of GO terms were linked to heart development ( Figure 2—figure supplement 4 ) . These results suggest that Nkx2-5 plays essential roles in regulating the APA of genes related to heart development . To elucidate the molecular mechanisms by which Nkx2-5 regulates APA , we examined the transcription termination factors associated with Nkx2-5 in the embryonic heart . We could detect an association between Nkx2-5 and the transcription termination factor Xrn2 , which has 5’-3’ exonuclease activity ( Kuehner et al . , 2011 ) ( Figure 3A ) . However , no association was detected between Nkx2-5 and the RNA helicases , Senataxin , Ddx5 , and Dhx9 ( Figure 3A ) , although it has been reported that these RNA helicases play an essential role in processing the ends of transcripts and that this process is regulated by the circadian clock PERIOD complex ( Padmanabhan et al . , 2012 ) . To confirm the association between Nkx2-5 and Xrn2 , we examined regions that may be important for their association . We found that the C-terminus of Xrn2 was important due to its association with Nkx2-5 ( Figure 3B ) . The N- and C-termini of Nkx2-5 were crucial due to their association with Xrn2 , whereas the Nkx2-5 homeodomain ( HD ) was not required ( Figure 3C ) . Next , we examined whether the association between Nkx2-5 and Xrn2 is dependent on DNA or RNA . Xrn2 co-immunoprecipitated with Nkx2-5 from eCM nuclear extracts that were treated with ethidium bromide ( EtBr ) to release DNA from proteins ( Heale et al . , 2006 ) ( Figure 3D ) . Although this concentration of EtBr ( 20 µg/ml ) reduced the association of Whsc1 with histone ( Nimura et al . , 2009 ) , the association of Nkx2-5 with Xrn2 was resistant to EtBr treatment . Furthermore , the association between these factors was also resistant to RNase A ( 50 µg/ml ) treatment ( Calvo and Manley , 2001 ) . These data suggest that Nkx2-5 may be associated with Xrn2 in a DNA- or RNA-independent manner and that Nkx2-5 could recruit Xrn2 to target regions of the genome based on its ability to recognize these target regions with its HD ( Figure 3E ) . 10 . 7554/eLife . 16030 . 020Figure 3 . Nkx2-5 associates with the 5’-3’ exonuclease Xrn2 . ( A ) Co-immunoprecipitates derived using the indicated antibodies from nuclear extracts of E12 . 5 hearts and aliquots ( 6% ) of the input proteins were analyzed by Western blotting . ( B ) Xrn2 and Nkx2-5 deletion mutants were transfected into C3H10T1/2 cells . Co-immunoprecipitates derived using the M2 antibody and aliquots ( 7% ) of the input proteins were analyzed by Western blotting . Schematic presentation of Xrn2 and its deletion mutants is shown at the right panel . ( C ) Nkx2-5 and Xrn2 deleting mutants were transfected into C3H10T1/2 cells . Co-immunoprecipitates derived using the HA antibody and aliquots ( 7% ) of the input proteins were analyzed by Western blotting . Schematic presentation of Nkx2-5 and its deletion mutants is shown at the right panel . ( D ) Co-immunoprecipitates derived using the indicated antibodies from nuclear extracts of E12 . 5 hearts exposed to 20 µg/ml EtBr or 50 µg/ml RNaseA as well as aliquots ( 6% ) of the input proteins were analyzed by Western blotting . ( E ) Summary of interacting regions between Nkx2-5 and Xrn2 . HD , homeodomain; Xrn , Xrn domain; ZF , zinc finger . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 020 To determine whether Xrn2 regulates the APA of Tnnt2 and Atp2a2 , Xrn2 was repressed by transfection with a sequence-specific siRNA in eCMs ( Figures 4A , B , and C ) . Xrn2 knockdown increased the expression of Tnnt2 and Atp2a2 transcripts with long 3’UTRs but did not alter transcription of the coding sequence of these genes ( Figures 4B and C ) . The increase in expression of long 3’UTR of Tnnt2 and Atp2a2 was also detected in chromatin-fractioned RNA ( Figure 4—figure supplement 1 ) . Furthermore , Xrn2 expression was not affected by the knockdown of Nkx2-5 , Gata4 , or Tbx5 ( Figure 4D ) . Thus , these results suggest that Nkx2-5 functions together with Xrn2 to regulate APA . 10 . 7554/eLife . 16030 . 021Figure 4 . Nkx2-5 functions together with Xrn2 to regulate APA . ( A ) Xrn2 knockdown was analyzed by Western blotting . ( B and C ) qRT-PCR analysis of mRNAs expression of the long 3’ UTRs ( B ) and gene bodies ( C ) of Tnnt2 and Atp2a2 in Xrn2-knockdown eCMs , normalized to Rplp2 . ( D ) Xrn2 expression levels were measured by qRT-PCR and normalized to Rplp2 . ( E ) Xrn2 binding in eCMs transfected with the indicated siRNAs was analyzed by ChIP-qPCR . The control values were set to 1 . 0 . ( F ) Long 3’ UTRs in eCMs transfected with the indicated siRNA were analyzed by Northern blotting using probes against Tnnt2 and Atp2a2 mRNA . Red brackets indicate mRNAs with long 3’ UTRs . ( G ) The original lengths of the Tnnt2 and Atp2a2 mRNAs and the lengths of the Tnnt2 and Atp2a2 mRNAs with long 3’ UTRs that were used in the Northern blot analysis were measured by BAS5000 . The ratio of the siControl was set to 1 . 0 . ( H ) Tnnt2 and Atp2a2 proteins in Nkx2-5 and Xrn2 knockdown eCMs were analyzed by western blotting . ( I ) The average profiles of the mRNAs with long 3’ UTRs that were increased in Nkx2-5-knockdown eCMs are shown in eCMs transfected with the indicated siRNAs . The gray area indicates the coding region . Significance was assessed using the two-sample Kolmogorov-Smirnov test . For B , C , D , E , and G , error bars indicate the mean ± s . e . m . ( n = 3 ) . * , p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 02110 . 7554/eLife . 16030 . 022Figure 4—source data 1 . Source data for Figure 4 and Figure 4-figure supplement 1 and 2 . Numeric data for Figure 4B , C , D , E , G , Figure 4—figure supplement 1B , Figure 4—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 02210 . 7554/eLife . 16030 . 023Figure 4—figure supplement 1 . Knockdowns of Nkx2-5 or Xrn2 affect the expression of the long 3’UTR regions in chromatin-fractioned RNA . ( A ) Genome browser representation of strand-specific chromatin-fractioned RNA-seq tag counts from eCMs transfected with the indicated siRNAs . ( B ) qRT-PCR analysis of mRNA expression of long 3’ UTRs of Tnnt2 and Atp2a2 , normalized to Rplp2 on chromatin-fractioned RNA . Error bars indicate the mean ± s . e . m . ( n = 3 ) . * , p < 0 . 05 . The arrow heads show polyadenylation sites . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 02310 . 7554/eLife . 16030 . 024Figure 4—figure supplement 2 . Knockdown of Nkx2-5 affects Xrn2-binding to Myl7 . Xrn2 binding in eCMs transfected with the indicated siRNAs was analyzed by ChIP-qPCR . The control values were set to 1 . 0 . Error bars indicate the mean ± s . e . m . ( n = 3 ) . * , p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 02410 . 7554/eLife . 16030 . 025Figure 4—figure supplement 3 . Nkx2-5 functions together with Xrn2 to regulate APA . The EtBr staining gels and the blottings of b-actin as an internal control are shown , related to Figure 4F . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 02510 . 7554/eLife . 16030 . 026Figure 4—figure supplement 4 . Correlation analysis of Nkx2-5 binding and long 3’UTR expression . ( A ) The average profiles of the mRNAs with long 3’ UTRs in genes that Nkx2-5 binds to TTS ( 3997 genes ) in eCMs transfected with the indicated siRNAs . The gray area indicates the coding region . Significance was assessed using the two-sample Kolmogorov-Smirnov test . ( B ) The average profiles of Nkx2-5 , Gata4 , and Tbx5 binding in genes that their long 3’ UTRs were increased in Nkx2-5-knockdown eCMs . ( C ) Correlation plot of long 3’UTR expression between indicated knockdowns and Pearson correlation , related to Figure 4I . ( D ) Violin plots of fold change in long 3’UTR expression with box plots . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 026 Xrn2 was recently reported to be recruited to TSSs and the downstream regions of genes ( Brannan et al . , 2012 ) . Because Nkx2-5 was also localized at both regions , we next investigated whether Nkx2-5 recruited Xrn2 to genes for which the 3’ UTR length was found to be regulated by Nkx2-5 . Nkx2-5 knockdown reduced Xrn2 binding at the TSSs and downstream regions of genes but not at 3’ UTR regions ( Figure 4E and Figure 4—figure supplement 2 ) . This result indicates a role for Nkx2-5 in Xrn2 binding to the TSSs and downstream regions of genes . Next , we examined whether the knockdown of Nkx2-5 and Xrn2 generated extended poly ( A ) -tailed mRNAs . Northern blot analysis of Tnnt2 and Atp2a2 using poly ( A ) -tailed mRNA showed that the knockdown of Nkx2-5 and Xrn2 significantly increased the expression of extended poly ( A ) -tailed mRNAs ( Figures 4F , G , and Figure 4—figure supplement 3 ) . Long 3’UTRs have been reported to be related to decrease translation efficiency ( Mayr and Bartel , 2009 ) . Consistent with this , the amounts of Tnnt2 and Atp2a2 protein were slightly decreased in Nkx2-5- and Xrn2- knockdown eCMs ( Figure 4H and Figure 2—figure supplement 1B ) . Next , we analyzed the increase in extended poly ( A ) -tailed mRNAs transcribed within 3 kb of the end of the 3’ UTR of a reference sequence in Nkx2-5- and Xrn2-knockdown eCMs . Xrn2 knockdown also increased the expression of long 3’ UTRs of genes for which the expression of long 3’ UTRs was increased in Nkx2-5 , but not Tbx5 nor Gata4 , -knockdown eCMs ( Figure 4I and Figure 4—figure supplement 4 ) . Furthermore , Xrn2 knockdown strongly increased the expression of mRNAs at the transcription termination sites ( TTSs ) , suggesting a role for Xrn2 in processing the ends of mRNAs . These results suggest that Xrn2 is involved in Nkx2-5-dependent regulation of APA . To demonstrate the coordinated functions of Nkx2-5 and Xrn2 in heart development , we introduced siRNAs directed against Nkx2-5 and Xrn2 along with GFP expression plasmids into early-head-fold-stage embryos ( E7 . 5 , before the formation of the linear heart tube ) ( Figures 5A and B ) . These embryos were divided into three categories according to their heart looping morphologies ( Figure 5C ) . The knockdown of both genes together significantly increased the number of abnormal hearts that remained in an essentially linear conformation compared with the knockdown of each gene alone ( Figure 5D ) . We next examined whether the expression of genes related to Left-Right signaling pathway are affected by the injection of siRNAs against Nkx2-5 and Xrn2 ( Figures 5E and F ) , since heart looping is known to be regulated by genes related to Left-Right signaling pathway ( Hamada et al . , 2002 ) . Knockdown of Nkx2-5 and Xrn2 did not significantly change the expression of these genes including Pitx2 , Nodal , Cryptic , and Lefty ( Figure 5E ) . Moreover , Pitx2 expression profile was not affected by knockdown of Nkx2-5 and Xrn2 ( Figure 5F ) . These results suggest that the coordinated functions of Nkx2-5 and Xrn2 are essential for heart formation . 10 . 7554/eLife . 16030 . 027Figure 5 . Knockdown of both Nkx2-5 and Xrn2 perturbs heart looping . ( A ) Transfection of siRNA into embryonic hearts . GFP was used to detect transfected fields . ( B ) We discarded the embryos with low transfection efficiency . ( C ) Representative morphologies of heart looping . D-Loop , the normal rightward loop; L-Loop , situs inversus; abnormal , hearts remained in an essentially linear conformation . ( D ) Knockdown of Nkx2–5 and Xrn2 in embryonic hearts . Graph bars indicate the% morphologies of heart looping . Significance was examined with Fisher’s exact test . * , p < 0 . 05 . ( E ) Looping-related genes expression level in siRNA-transfected embryonic hearts ( n = 3 ) , normalized to Rplp2 expression level . n . s . , not significant . ( F ) In situ hybridization of Pitx2 . The numbers indicate Pitx2 expression pattern among the right side , the left side , and the both sides . White arrows indicate lateral plate mesoderm . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 027 Finally , we generated Xrn2-deleted mice by disrupting exons1 and 2 using the CRISPR/Cas9 system ( Figure 6 and Figure 6—figure supplement 1 ) ( Wang et al . , 2013 ) . We deleted 11 . 6 kb from Xrn2 , including exons1 and 2 , which encode a part of the domain with enzymatic activity ( Figure 6—figure supplement 1 ) . To examine the genetic interaction between Nkx2-5 and Xrn2 , we obtained Nkx2-5+/-Xrn2+/- newborns by intercrossing Nkx2-5+/- mice with two Xrn2+/- mice ( #2 and #3 ) . Although neither Nkx2-5+/-nor Xrn2+/- hearts showed a muscular ventricular septal defect ( VSD ) at postnatal day 0 ( P0 ) , we found a VSD in Nkx2-5+/-Xrn2+/- newborn hearts ( n = 3 of 7 ) ( Figure 6 ) . An atrial septal defect was observed in both Xrn2+/- and Nkx2-5+/-Xrn2+/- mice , suggesting that Xrn2 contributes to atrial septum formation ( Figure 6 ) . These results indicate a genetic interaction between Nkx2-5 and Xrn2 . 10 . 7554/eLife . 16030 . 028Figure 6 . Nkx2-5 genetically interacts with Xrn2 . Histological analysis of Nkx2-5+/-and Xrn2+/- newborn hearts . Frontal sections from newborn hearts were stained with hematoxylin and eosin . ASD was observed in Xrn2+/- ( n = 6 of 9 ) and Nkx2-5+/-Xrn2+/- ( n = 6 of 7 ) newborns . VSD was observed in Nkx2-5+/-Xrn2+/- newborns ( n = 3 of 7 ) . lv , left ventricle; rv , right ventricle; la , left atrium; ra , right atrium; p , septum primum; s , septum secundum . * , p < 0 . 05 . ** , p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 02810 . 7554/eLife . 16030 . 029Figure 6—figure supplement 1 . Nkx2-5 genetically interacts with Xrn2 . ( A ) Scheme illustrating the targeting of exon 1 and 2 in Xrn2 . Two gRNAs and the Cas9 RNA were injected into fertilized eggs . F , Forward primer; WT-R , reverse primer for the wild-type allele; MT-R , Reverse primer for the deleted allele . The genomic sequences around the gRNAs are shown . ( B ) Genotyping of Xrn2+/- mice obtained from crossing Xrn2 F0 and wild-type mice . Sequences of two deleted alleles ( #2 and #3 ) are shown . WT , wild-type allele , 582 bp; Del , deleted allele #2 is 283 bp and #3 is 290 bp . DOI: http://dx . doi . org/10 . 7554/eLife . 16030 . 029 One of the critical roles of transcription factors is to regulate the expression level of target genes by recruiting transcription factor-associated factors such as chromatin remodeling factors , histone modifiers , and RNAPII to the promoter and enhancer regions of these genes ( Graf and Enver , 2009 ) . During heart development , the importance of cardiac transcription factors had been shown in many reports; for example , mutations in transcription factors , such as Nkx2-5 , caused defects in heart development ( Bruneau , 2008; Srivastava , 2006 ) . However , the molecular mechanisms underlying the functions of cardiac transcription factors have remained unclear . Our results suggest that Nkx2–5 is involved in the 3’-end processing of target genes in conjunction with Xrn2; furthermore , Nkx2–5 and Xrn2 deficiency caused abnormalities in heart development . We identified the genome-wide occupancies of transcription factors and transcription factor-associated proteins and determined the histone modification signatures in E12 . 5 hearts . Active histone modifications , including H3K4me1/3 , H3K36me3 , and H3K27ac , were enriched in actively transcribed genomic regions , and repressive histone modifications , including H3K9me3 and H3K27me3 , were enriched in silenced genes ( Figure 1 and Figure 1—figure supplement 6 ) , which agrees with previous reports ( Barski et al . , 2007; Wang et al . , 2008 ) . While RNAPII-S5P has been reported to be associated with promoters ( Rahl et al . , 2010 ) , we found enrichment of RNAPII-S5P at promoters and downstream regions of highly expressed genes in eCMs . This difference of RNAPII-S5P binding profile may be caused by differences of ChIP protocol and cell type . The transcription factor binding regions were different from those of exogenously tagged transcription factors in the murine adult cardiomyocyte cell line HL−1 ( He et al . , 2011 ) . This difference may be caused by differences in gene expression during distinct developmental stages of cardiomyocytes . Metagene analysis revealed that Nkx2-5 is localized at both the TSSs and downstream regions of highly expressed genes in E12 . 5 hearts ( Figure 1 , Figure 1—figure supplement 5 , and Figure 1—figure supplement 6 ) . We compared ChIP and input data for obtaining peaks . Moreover , active histone modification ChIP-seq data including H3Keme1 , H3K4me3 , and H3K27ac showed the enrichment only around TSS regions ( Figure 1—figure supplement 7 ) . These indicate the enrichment of Nkx2–5 at the downstream regions is not bias in native ChIPseq experiment . Based on these data , Nkx2-5 plays unique roles in regulating the APA of highly expressed genes , including Tnnt2 and Atp2a2 , during heart development . The regulation of APA has been reported as one of the mechanisms for determining the length of the 3’ UTR ( Elkon et al . , 2013 ) . Long 3’ UTRs are likely to decrease translation efficiency because they are recognized by miRNAs ( Elkon et al . , 2013 ) . One function of Nkx2-5 may be to regulate the lengths of the 3’ UTRs of genes involved in heart development through the regulation of APA , although there is a possibility that Nkx2-5 is also involved in the regulation of splicing . This process allows the production of necessary quantities of muscle proteins , such as cardiac troponin T ( cTnT ) , a component of troponin that is essential for sarcomere assembly and is encoded by Tnnt2 ( Nishii et al . , 2008 ) , and Atp2a2 ( also known as SERCA2 ) , a cardiac sarcoplasmic reticulum Ca2+ pump that is essential for normal cardiac performance ( Periasamy et al . , 1999 ) . Moreover , Nkx2–5 has been reported to regulate the expression of miRNAs such as miR-1 , which recognizes the 3’ UTR of Cdc42 and decreases the Cdc42 protein level during heart development ( Qian et al . , 2011; Zhao et al . , 2007 ) . In addition , miRNAs might be involved in the regulation of translation efficiency by recognizing long 3’ UTRs . Thus , APA dysregulation in Nkx2–5- and Xrn2-depleted hearts might cause abnormalities in heart development by decreasing the level of heart-related proteins such as Tnnt2 and Atp2a2 . Our results suggest that in addition to the regulation of gene expression by Nkx2–5 , the regulation of APA by Nkx2-5 and Xrn2 is an important mechanism for the precise progression of heart development . Recent studies have reported significant interactions between chromatin regions in the nucleus and have shown the importance of interactions between promoters and enhancers for the activation of gene expression ( Kagey et al . , 2010; Li et al . , 2012; Wang et al . , 2011 ) . Although the function of enhancer-promoter DNA looping mediated by transcription factors and mediators has previously been demonstrated ( Kagey et al . , 2010; Wang et al . , 2011 ) , genome-wide chromatin conformational analyses have revealed that ( 1 ) several chromatin regions form DNA loops between the enhancer and promoter and between the promoter and downstream regions of the gene ( promoter-downstream looping ) and ( 2 ) these interactions are associated with RNAPII ( Li et al . , 2012 ) . Promoter-downstream looping has been shown to enhance RNAPII recycling from the TTS to the promoter ( Cavalli and Misteli , 2013 ) and to determine the direction of transcription ( Tan-Wong et al . , 2012 ) . In this study , we showed that Nkx2-5 deficiency increased the expression of long 3’ UTRs beyond the TTS of the studied genes , which occurred concomitantly with changes in the stability of promoter-downstream looping and the amount of RNAPII localized downstream of the genes ( Figure 2 ) . Thus , promoter-downstream looping is likely to be involved not only in recycling RNAPII ( Cavalli and Misteli , 2013 ) but also in regulating APA . Xrn2 is known to have 5’-3’ exonuclease activity and to contribute to transcription termination at the ends of genes ( Kim et al . , 2004; West et al . , 2004 ) . Recently , it was reported that Xrn2 along with mRNA decapping factors bound to TSSs and other downstream regions and was involved in the regulation of RNAPII elongation ( Brannan et al . , 2012 ) . Our results suggest that Nkx2-5 is involved in the regulation of Xrn2 binding to both TSSs and the downstream regions of target genes during heart development and that promoter-downstream looping support the function of Xrn2 at these regions ( Figures 2 and 4 ) . The association of Nkx2–5 with Xrn2 was EtBr- and RNaseA-resistant ( Figure 3 ) , and the Nkx2–5 HD was not required for the association of Nkx2−5 with Xrn2 ( Figure 3 ) . These findings suggest that Nkx2–5 may be able to recruit Xrn2 to target genomic regions that Nkx2-5 recognized via its HD . Although Xrn2 is a 5’-3’ exonuclease and is believed to eliminate 3’ cleavage products that remain associated with RNAPII after mRNA cleavage ( Kuehner et al . , 2011; Richard and Manley , 2009 ) , Xrn2 knockdown increased the expression of mRNAs with long 3’ UTRs , suggesting that Xrn2 is also critical for regulating APA and promotes the usage of proximal poly ( A ) sites . Nkx2–5 functions together with Xrn2 but not with RNA helicases , which are also transcription termination factors; therefore , both Xrn2 and transcription factors might be involved in determining the usage of poly ( A ) sites . Recent deep-sequencing analyses have revealed that over 70% of human genes are estimated to have 3’ UTRs of variable lengths and that these differences are dependent on cell type and are regulated by APA ( Elkon et al . , 2013 ) . APA is modulated by the processing of mRNA at proximal or distal poly ( A ) sites . mRNAs with short 3’ UTRs are generated in highly proliferative cells including cancer cells ( Mayr and Bartel , 2009 ) , and these mRNAs lack miRNA recognition regions; thus , their translation efficiency can be increased by avoiding miRNA ( Mayr and Bartel , 2009 ) . Nerve cells are known to generate mRNAs with long 3’ UTRs by utilizing distal poly ( A ) sites ( Miura et al . , 2013 ) . The termination of transcription by RNAPII requires RNA 3’-end processing and termination factors ( Kuehner et al . , 2011 ) , and defects in these factors increase the level of RNA with long 3’UTR due to the dysregulation of RNAPII termination ( Liu et al . , 2012; Padmanabhan et al . , 2012 ) . Although cleavage and polyadenylation factors including cleavage and polyadenylation specificity factor ( CPSF ) , cleavage-stimulating factor ( CSTF ) , and cleavage factor Im ( CFIm ) , regulate APA ( Elkon et al . , 2013; Masamha et al . , 2014 ) , the mechanisms that regulate 3’ UTR length remain undefined . Our findings suggest that Nkx2–5 , an essential cardiac transcription factor , regulates APA through the recruitment of 5’-3’ exonuclease Xrn2 during heart development ( Figure 4 ) and that tissue-specific transcription factors play an important role in tissue-specific 3’ UTR length . Although knockdown of Nkx2-5 and Xrn2 showed the defect of heart looping ( Figure 5 ) , Nkx2-5 and Xrn2 double heterozygotes have ASD and VSD ( Figure 6 ) . This difference may be caused by the remaining amount of protein expression and / or acute decrease of expression by gene knockdowns or stable decrease of expression by gene knockout , which is supported by the recent study reporting the difference of phenotypes caused by gene knockdowns or gene knockouts ( Rossi et al . , 2015 ) . Consistent with our results , Nkx2-5 heterozygotes do not display defective heart looping , although Nkx2-5 homozygotes show this defect ( Biben et al . , 2000 ) . In summary , our findings suggest that Nkx2–5 regulates APA by recruiting Xrn2 to targeted genomic regions . Deficiencies in Nkx2–5 and Xrn2 disrupted the regulation of 3’ UTR length and resulted in abnormalities in heart formation . Although we cannot exclude the possibility of other defects in addition to the dysregulation of APA , the data suggest that APA dysregulation could be one of the mechanisms that cause CHD in patients with mutated Nkx2-5 , nevertheless , it is still unclear whether long 3’UTR per se is related to the pathogenesis of cardiac abnormality . Although further studies are required to elucidate the molecular role of Nkx2-5 in the regulation of APA , our findings provide a conceptual framework for understanding how transcription factors regulate 3’ UTR length . Embryonic hearts were obtained from C57BL/6 wild-type and Nkx2-5-deficient mice ( Moses et al . , 2001 ) in accordance with protocols 3422–1 and 24-084-012 approved by the Ethics Committee for Animal Experiments of the Osaka University Graduate School of Medicine . Cas9 RNA and gRNA were generated by in vitro transcription with the SP6 mMessage mMachine kit ( Ambion , Foster City , CA ) as previously described , with minor modifications ( Wang et al . , 2013 ) . Cas9 RNA and two gRNAs were injected into fertilized eggs obtained from C57BL/6 mice to delete exon 1 and 2 of Xrn2 . Deletion of these exons was confirmed by sequencing . Xrn2 heterozygous mice were obtained by intercrossing Xrn2 F0 with C57BL/6 mice . Nuclear extracts from four E12 . 5 embryonic hearts were used in each native ChIP experiment following a previously described protocol with minor modifications ( Nimura et al . , 2006; 2009 ) . Isolated nuclei from embryonic hearts were treated at 25°C for 30 min with 4 . 8 U ml−1 micrococcal nuclease in 250 µl of a nuclear isolation buffer containing 400 mM NaCl , which was then diluted to 200 mM NaCl . The digested chromatin was immunoprecipitated with 15–50 µg of antibody . Only mono- and di-nucleosomes size DNA was used for construction of sequencing libraries . Sequencing libraries were prepared from two or more biological-replicate ChIP samples and from an input according to the instructions provided with the SOLiD Fragment Library Barcoding Kit ( Life Technologies , Carlsbad , CA ) . The libraries were sequenced with SOLiD 4 . The resulting reads were mapped using BioScope software ( Life Technologies ) with the default configuration , combined biological replicates , and analyzed using Homer ( Heinz et al . , 2010 ) , CEAS ( Shin et al . , 2009 ) and R software programs . The mapping results are shown in Supplementary file 1A and were generated using IGV software ( Robinson et al . , 2011 ) . The heatmap was generated using Java TreeView ( http://jtreeview . sourceforge . net/ ) . RNA was extracted from primary embryonic cardiomyocytes ( from E12 . 5 hearts ) and E9 . 5 hearts from Nkx2-5-deficient/wild-type littermates using the TRIzol reagent ( Invitrogen , Carlsbad , CA ) according to the manufacturer’s instructions . Strand-specific sequencing libraries from two biological replicate RNA samples were prepared according to the Life Technologies protocol as previously described ( Mori et al . , 2012 ) . Briefly , digested poly ( A ) -tailed mRNA was ligated to the SOLiD Adaptor Mix and then reverse-transcribed using the SOLiD Total RNA-Seq Kit ( Life Technologies ) . Size-selected first-strand cDNA was amplified by using SOLiD 5’ PCR primers and barcoded SOLiD 3’ PCR primers ( Life Technologies ) . RNA-seq libraries were sequenced with SOLiD 4 . The resulting reads were mapped using BioScope software ( Life Technologies ) and analyzed using BioScope and Cufflinks ( Trapnell et al . , 2010; 2012 ) . The mapping results are shown in Supplementary file 1B . RefSeq genes were divided into three categories ( High , Middle , and Low ) according to the FPKM ( fragments per kilobase of exon per million fragments mapped ) of each gene in wild-type eCMs: High: FPKM > 500; Middle: FPKM 100–500; and Low: FPKM 10–100 . The genes identified for analysis were either upregulated to more than a ln ( fold change ) of 0 . 2 or downregulated to less than a ln ( fold change ) of 0 . 2 compared with control siRNA-treated eCMs . The High , Middle , and Low categories included 90 , 717 , and 5516 genes , respectively . The expression of read-through RNA ( FPKM > 0 . 5 ) was measured within 5 kb in eCMs and in Nkx2-5-deficient embryonic hearts using BioScope ( Life Technologies ) . Genes with an average change in read-through RNA expression of more than a 1 . 4-fold compared with control siRNA-treated eCMs or wild-type littermate hearts were analyzed using DAVID . Genes for which the read-through RNA was expressed at less than 0 . 5-fold compared with controls were considered to have downregulated read-through RNA . Genes with expression within regions of a read-through RNA were removed from the analysis . The normalized read-through RNA tags were counted in 50-bp bins using CEAS , and the obtained values were then used to calculate the average expression in each bin . The polyadenylation sites were obtained from APADB ( http://tools . genxpro . net/apadb/ ) . Chromatin-fractioned RNA was purified from 2~5 X 106 cells of eCMs as previously described ( Nojima et al . , 2015 ) . rRNA was depleted using Ribominus Eukaryote kit ver2 ( Life technologies ) from 3 . 5~4 . 0 µg chromatin RNA . The antibodies used in this study are shown in Supplementary file 1D . All the antibodies used for ChIP-seq have been previously reported as suitable or were verified to be suitable for the immunoprecipitation of target proteins ( Figure 1—figure supplement 4 ) . Primary eCMs were obtained from E12 . 5 hearts by overnight digestion with trypsin-EDTA at 4°C and separation from cardiac fibroblasts by pre-plating on collagen type I-coated dishes for 1 hr , as previously described ( Springhorn and Claycomb , 1989 ) with some modifications . The cardiomyocytes were transfected with Nkx2-5 ( Mm_Nkx2-5_8958 , SIGMA ) , Tb x 5 ( Mm_Tb x 5_9160 , SIGMA ) , Gata4 ( Mm_Gata4_5094 , SIGMA ) , Xrn2 ( Mm_Xrn2_3520 and Mm_Xrn2_3522 , SIGMA ) , or control ( SIC-001 , SIGMA ) siRNAs using RNAiMAX ( Invitrogen ) and cultured for 48 hr in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum . The siRNAs directed against Nkx2-5 , Tbx5 , and Gata4 were chosen from three different siRNAs after determining the knockdown efficiency and specificity of each candidate using quantitative RT-PCR and Western blotting . Total RNA was extracted with the TRIzol reagent ( Invitrogen ) . Reverse transcription was performed with SuperScript III ( Invitrogen ) as previously described ( Nimura et al . , 2009 ) and analyzed using the CFX384 Real-Time System ( BIO-RAD , Hercules , CA ) . Genomic DNA contamination was evaluated by examining reverse transcription reaction samples lacking reverse transcriptase . The values were normalized to Rplp2 ( ribosomal protein , large , P2 ) and expressed relative to the values obtained with control siRNA-treated eCMs . Nuclei isolated from 106 siRNA-transfected E12 . 5 primary cardiomyocytes were digested overnight at 37°C with EcoRI , BglII , or HindIII ( 100 units per 106 cells , TOYOBO , JAPAN ) in 100 µl of buffer with 0 . 4% NP-40 and complete EDTA-free protease inhibitors ( Roche , Indianapolis , IN ) . The nuclei were ligated overnight with Ligation High ( TOYOBO ) and then extracted using phenol-chloroform . The primers were designed using the 3C Primer ( http://dostielab . biochem . mcgill . ca/index . php ) and Primer3 programs ( http://frodo . wi . mit . edu/ ) . BAC plasmids RP23-2E23 for Tnnt2 and RP23-128I8 for Atp2a2 were used as controls . PCR was performed using the THUNDERBIRD SYBR qPCR mix ( TOYOBO ) and a CFX384 thermocycler ( BIO-RAD ) . The amplicons were separated on 2% agarose gels stained with ethidium bromide for visualization . Poly A-tailed mRNA was purified from total RNA that was extracted using TRIzol ( Invitrogen ) and the Ambion MicroPoly ( A ) Purist Kit ( Ambion ) . The poly ( A ) -tailed mRNA ( 360 ng ) was separated by 1% agarose gel electrophoresis and transferred to a Hybond-N+ nylon transfer membrane ( Amersham Biosciences , Pittsburgh , PA ) . Tnnt2 and Atp2a2 mRNA were detected with 32P-labeled cDNA that contained a portion of the gene-body regions of Tnnt2 and Atp2a2 . The intensities of the bands were quantitatively measured using a BAS5000 ( GE Healthcare , Pittsburgh , PA ) . Expression constructs containing HA-Nkx2-5 , Flag-Xrn2 , and their deletion mutants were cotransfected into C3H10T1/2 cells , purchased from ATCC , in 10 cm dishes as previously described ( Nimura et al . , 2009 ) . This cell was neither authenticated nor tested for mycoplasma contamination . HA-Nkx2-5 , Flag-Xrn2 , and their respective deletion mutants were immunoprecipitated from nuclear extracts that were prepared as previously described ( Nimura et al . , 2009 ) . Co-immunoprecipitated proteins were analyzed by Western blotting . E7 . 5 embryos were dissected from the uterus , and embryos at the early-head-fold stage were carefully selected . Liposomes were created by mixing 3 . 33 µM siRNAs directed against Nkx2-5 , Xrn2 , or control together with 53 ng/µl EGFP expression vector in 15 µl of OPTI-MEM with 3 µl of RNAiMAX ( Invitrogen ) diluted in 15 µl of OPTI-MEM . Liposomes were injected into heart fields as previously described ( Yamamoto et al . , 2004 ) . The embryos were rotationally cultured for 48 hr in Dulbecco’s modified Eagle’s medium supplemented with 75% rat serum . GFP signals and heart morphology were examined with a Leica M165FC microscope , and the embryos with weak GFP signals were discarded ( Figure 5 ) . Whole-mount in situ hybridization was performed according to standard procedures ( Wilkinson , 1992 ) and probe specific for Pitx2 mRNA ( Yoshioka et al . , 1998 ) . The data are presented as the mean ± s . e . m . P values were calculated using the two-tailed t-test and the Tukey-Kramer HSD test using JMP , Fisher’s exact test was performed using http://aoki2 . si . gunma-u . ac . jp/exact/fisher/getpar . html , and the two-sample Kolmogorov-Smirnov test was performed using R . p < 0 . 05 was considered statistically significant . The number of n shows biological replication .	About one in every hundred babies is born with problems that either affect the structure of the heart or how it works . These problems are known as congenital heart disease , and result when the development of the heart is disrupted . How the heart develops is determined by thousands of genes whose activity or “expression” must be precisely regulated . Proteins called transcription factors can control gene expression; therefore , researchers may discover new ways of treating congenital heart disease if they can understand how transcription factors work during normal heart development . To produce a protein , the information in a gene must first be “transcribed” to form a molecule of messenger RNA ( mRNA ) . Not all of the mRNA sequence is subsequently “translated” to form the protein; this includes a stretch at the end of the mRNA called the 3’ untranslated region . The length of the 3’ untranslated region for a particular mRNA may vary depending on the type of cell it has been produced in , and this length can influence how efficiently the mRNA is translated to form a protein . However , it was not clear what changes the length of the 3’ untranslated region . Nimura et al . have now studied mice to investigate the role of a transcription factor called Nkx2-5 , which was known to be important for heart development . This revealed that in addition to its expected role in starting the transcription of genes that are important for heart development , Nkx2-5 also controls the length of 3’ untranslated regions of certain mRNAs . To do so , Nkx2-5 binds to a protein called Xrn2 that stops transcription when the end of the gene is reached . Mouse embryos that lacked Nkx2-5 produced mRNAs containing long 3’ untranslated regions from genes related to the development of the heart . Furthermore , suppressing the activity of both Nkx2-5 and Xrn2 resulted in the embryos developing heart defects . The findings of Nimura et al . suggest that transcription factors found in specific tissues are responsible for the different lengths of 3’ untranslated regions in mRNAs in different tissues . Furthermore , incorrectly regulating the length of these regions appears to be linked to the development of congenital heart disease . The next step is to understand exactly how the failure to correctly regulate the length of 3’ untranslated regions contributes to congenital heart disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression" ]	2016	Regulation of alternative polyadenylation by Nkx2-5 and Xrn2 during mouse heart development
Folding of mammalian genomes into spatial domains is critical for gene regulation . The insulator protein CTCF and cohesin control domain location by folding domains into loop structures , which are widely thought to be stable . Combining genomic and biochemical approaches we show that CTCF and cohesin co-occupy the same sites and physically interact as a biochemically stable complex . However , using single-molecule imaging we find that CTCF binds chromatin much more dynamically than cohesin ( ~1–2 min vs . ~22 min residence time ) . Moreover , after unbinding , CTCF quickly rebinds another cognate site unlike cohesin for which the search process is long ( ~1 min vs . ~33 min ) . Thus , CTCF and cohesin form a rapidly exchanging 'dynamic complex' rather than a typical stable complex . Since CTCF and cohesin are required for loop domain formation , our results suggest that chromatin loops are dynamic and frequently break and reform throughout the cell cycle . Mammalian interphase genomes are functionally compartmentalized into topologically associating domains ( TADs ) spanning hundreds of kilobases . TADs are defined by frequent chromatin interactions within themselves and they are insulated from adjacent TADs ( Dekker and Mirny , 2016; Dixon et al . , 2012; Hu et al . , 2015; Merkenschlager and Nora , 2016; Nora et al . , 2012; Wang et al . , 2016 ) . Most TAD or domain boundaries are strongly enriched for CTCF ( Figure 1A ) , an 11-zinc finger DNA-binding protein ( Ghirlando and Felsenfeld , 2016 ) , and cohesin ( Figure 1B ) , a ring-shaped multi-protein complex composed of Smc1 , Smc3 , Rad21 and SA1/2 that is thought to topologically entrap DNA ( Ivanov and Nasmyth , 2005; Skibbens , 2016 ) . The subset of TADs which are folded into loops are referred to as loop domains and tend to be demarcated by convergent CTCF-binding sites ( Rao et al . , 2014 ) . Targeted deletions of CTCF-binding sites demonstrate that CTCF causally determines loop domain boundaries ( Guo et al . , 2015; Sanborn et al . , 2015; de Wit et al . , 2015 ) . Moreover , disruption of loop domain boundaries by deletion or silencing of CTCF-binding sites allows abnormal contact between previously separated enhancers and promoters , which can induce aberrant gene activation leading to cancer ( Flavahan et al . , 2016; Hnisz et al . , 2016a ) or developmental defects ( Lupiáñez et al . , 2015 ) . Finally , genetically engineered depletion of both CTCF ( Nora et al . , 2017 ) and cohesin ( Schwarzer et al . , 2016 ) causes most loops to disappear . Yet , despite much progress in characterizing TADs and loop domains , how they are formed and maintained remains unclear . Since CTCF and cohesin causally control domain organization , here we investigated their dynamics and nuclear organization using single-molecule imaging in live cells . 10 . 7554/eLife . 25776 . 003Figure 1 . CTCF and cohesin can be endogenously tagged and form a complex . ( A ) Sketch of CTCF and its consensus DNA-binding sequence . ( B ) Sketch of cohesin , with subunits labeled , topologically entrapping DNA . ( C ) Western blot of mESC and U2OS wild-type ( wt ) and knock-in cell lines demonstrating homozygous insertions . ( D ) Sketch of covalent dye-conjugation for Halo or SNAPf-Tag . ( E ) CTCF ChIP-Seq read count ( Reads Per Genomic Content ) for wild-type and C59 plotted at MAC2-called wt-CTCF peak regions centered around the peak . ( F ) Rad21 ChIP-Seq read count ( Reads Per Genomic Content ) for wild-type and C59 plotted at MACS2-called wt-Rad21 peak regions . ( G ) Co-IP . CTCF was immunoprecipitated and we immunoblotted for cohesin subunits Rad21 , Smc1 and Smc3 . ( H ) Sketch of a loop maintenance complex ( LMC ) composed of CTCF and cohesin holding together a spatial domain as a loop . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 00310 . 7554/eLife . 25776 . 004Figure 1—figure supplement 1 . Specific labeling of HaloTagged and SNAPf-Tagged proteins in live cells . Wild-type ( wt ) mouse embryonic stem cells ( mESCs ) and C59 mESCs expressing endogenously tagged FLAG-Halo-mCTCF and mRad21-SNAPf-V5 were labeled with the indicated dye by 30-min incubation in a 37°C incubator followed by washing with PBS and fresh medium , preparation for FACS ( dissociation with trypsin , cell collection by centrifugation and filtering through 40 μm mesh ) and their fluorescence was then measured using analytical flow cytometry . Around 30 , 000 cells were measured in each case and live cells were gated using the area of forward and side scattering . JF549/TMR fluorescence was measured using a 561 nm excitation laser and a 610/20 nm emission filter . JF646 fluorescence was measured using a 640 nm excitation laser and a 670/30 nm emission filter . Each panel shows fluorescence in the JF549/TMR channel and in the JF646 channel as well as a JF549/TMR vs . JF646 scatterplot . First panel: no dye control . Second panel: 500 nM SNAP-TMR ( NEB #S9105S ) labeling only . Third panel: 500 nM cp-JF549 labeling only ( the cp handle also labels SNAPf-Tag proteins , but more specifically than the SNAP-handle ) . Fourth panel: 500 nM Halo-JF549 labeling only . Fifth panel: 500 nM Halo-JF646 labeling only . Bottom panel: double 500 nM cp-JF549 and 500 nM Halo-JF646 labeling . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 00410 . 7554/eLife . 25776 . 005Figure 1—figure supplement 2 . Teratoma assay demonstrates that tagging CTCF and Rad21 does not affect pluripotency in mESCs . 350 , 000 wild-type and C59 ( FLAG-Halo-mCTCF; mRad21-SNAPf-V5 ) JM8 . N4 mouse embryonic stem cells were injected into the testis and kidney of Fox Chase SCID-beige male 8-week-old mice ( Charles River ) and tumors were harvested 27–33 days after injection ( top row ) . Tumors were fixed with 10% formalin overnight , embedded in paraffin and cut into 5 μm serial sections and then H and E stained . Representative sections from each of the three germ layers ( endoderm , mesoderm and ectoderm; highlighted by black arrows ) are shown for both wild-type and C59 mES cells . Black scale bar: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 00510 . 7554/eLife . 25776 . 006Figure 1—figure supplement 3 . Tagging CTCF and Rad21 does not affect expression of key pluripotency genes or CTCF and Rad21 protein levels . ( A ) Expression of key mouse embryonic stem cell genes measured by qPCR was similar in wild-type ( blue ) and C59 ( FLAG-Halo-mCTCF; mRad21-SNAPf-V5 ) ( red ) JM8 . N4 mouse embryonic stem cells . ( B ) CTCF ( red ) and Rad21 ( green ) protein levels as measured by western blot and normalized to either H3 levels ( solid bar ) or TBP ( hashed bar ) was similar between wild-type and tagged mouse embryonic stem cells ( WT , C87 , C45 , C59 ) and similar between wild-type and tagged human U2OS cells ( WT , C32 ) . Error bars show standard deviation among three replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 00610 . 7554/eLife . 25776 . 007Figure 1—figure supplement 4 . CTCF and Rad21 ChIP-Seq results in wt and C59 mESCs . ( A ) Venn diagram showing overlap of called peaks ( MACS2 ) of CTCF and Rad21 in wild-type mESCs . ( B ) Representative genome-browser view of IgG , CTCF and Rad21 ChIP-Seq for wild type ( wt ) and the double knock-in C59 mES cells at the Nanog locus on chromosome 6 . As can been seen , the binding pattern of both CTCF and Rad21 is unaltered after tagging . Scale is ( number of reads ) / ( 106/ ( total read count ) ) . ( C ) Representative genome-browser view of IgG , CTCF and Rad21 ChIP-Seq for wild type ( wt ) and the double knock-in C59 mES cells at the Oct4 ( Pou5f1 ) locus on chromosome 17 . As can been seen , the binding pattern of both CTCF and Rad21 is unaltered after tagging . Scale is ( number of reads ) / ( 106/ ( total read count ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 00710 . 7554/eLife . 25776 . 008Figure 1—figure supplement 5 . Tagging CTCF and Rad21 does not affect the ChIP-Seq genomic binding pattern . ( A ) ChIP-Seq enrichment ( RPGC: Reads Per Genomic Content ) shown for all called CTCF peaks in wild-type mESCs , sorted by wild-type CTCF enrichment . The six columns show IgG ChIP-Seq , CTCF ChIP-Seq and mRad21 ChIP-Seq for wild-type ( left ) and C59 ( FLAG-Halo-mCTCF; mRad21-SNAPf-V5 ) genome-edited ( right ) cells . On the right , the average in each case for the same conditions . ( B ) ChIP-Seq enrichment ( RPGC: Reads Per Genomic Content ) shown for all called Rad21 peaks in wild-type mESCs , sorted by wild-type Rad21 enrichment . The six columns show IgG ChIP-Seq , CTCF ChIP-Seq and mRad21 ChIP-Seq for wild-type ( left ) and C59 ( FLAG-Halo-mCTCF; mRad21-SNAPf-V5 ) genome-edited ( right ) cells . On the right , the average in each case for the same conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 008 In order to image CTCF and cohesin without altering their endogenous expression levels , we used CRISPR/Cas9-mediated genome editing to homozygously tag Ctcf and Rad21 with HaloTag in mouse embryonic stem ( mES ) cells ( Figure 1C , clones C87 and C45 ) . We also generated a double Halo-mCTCF/mRad21-SNAPf knock-in mESC line ( Figure 1C , C59 ) as well as a Halo-hCTCF knock-in human U2OS cell line ( Figure 1C , C32 ) . Halo- and SNAPf-Tags can be covalently conjugated with bright cell-permeable small molecule dyes suitable for single-molecule imaging ( Figure 1D; Figure 1—figure supplement 1; Grimm et al . , 2015 ) . To examine the effect of tagging CTCF and Rad21 , which are both essential proteins , we performed control experiments in the doubly tagged mESC line ( C59 ) , and observed no effect on mESC pluripotency in a teratoma assay ( Figure 1—figure supplement 2 ) , expression of key stem cell genes ( Figure 1—figure supplement 3A ) or tagged protein abundance ( Figure 1—figure supplement 3B ) . Next , to further validate our endogenous tagging approach , we performed chromatin immunoprecipitation followed by DNA sequencing ( ChIP-Seq ) using antibodies against CTCF and Rad21 in both wild-type ( wt ) and the double knock-in C59 line . We compared ChIP-Seq enrichment for both wt and C59 at called wt peaks and observed similar enrichment ( Figure 1E–F ) . Notably , 97% of the 33 , 434 called Rad21 peaks co-localize with one of the 68 , 077 called CTCF peaks ( Figure 1—figure supplements 4–5; Supplementary file 1 ) , suggesting an intrinsic link between CTCF and cohesin and largely confirming previous reports of ~70–90% overlap ( Parelho et al . , 2008; Wendt et al . , 2008 ) . However , chromatin co-occupancy by ChIP-seq at the same sites does not necessarily mean that CTCF and Rad21 bind simultaneously . Thus , to determine whether CTCF and cohesin physically interact , we performed co-immunoprecipitation ( co-IP ) studies . CTCF IP pulled down cohesin subunits Rad21 , Smc1 and Smc3 in both wt and C59 mES cells ( Figure 1G ) , demonstrating a physical interaction between CTCF and cohesin , which is not affected by endogenous tagging . Together , our ChIP-Seq co-localization ( 97% of Rad21 peaks overlap with a CTCF peak ) and co-IP interaction studies suggest that CTCF and cohesin form a complex on chromatin . The Hi-C study with the highest resolution found ~10 , 000 loops in human GM12878 cells using very conservative and stringent loop calling and found these loops to be largely conserved between cell types and between mouse and human ( Rao et al . , 2014 ) . Since each loop is anchored by at least two CTCF/cohesin ChIP-Seq-called sites , but often by clusters of CTCF/cohesin sites , we estimate ( see Appendix 1 for a full discussion ) that at least one-third of cognate-bound CTCF molecules and the majority of chromatin-bound G1 cohesin molecules are involved in chromatin looping . Integrating these results with the recent demonstrations ( Nora et al . , 2017; Schwarzer et al . , 2016 ) that CTCF and cohesin are causally required for chromatin looping , we refer to the subpopulation of CTCF and cohesin involved in looping as a ‘loop maintenance complex’ ( LMC; Figure 1H ) . To investigate the dynamics of the LMC , we measured the residence time of CTCF and cohesin on chromatin . First , we used highly inclined and laminated optical sheet illumination ( Tokunaga et al . , 2008 ) ( Figure 2A ) and single-molecule tracking ( SMT ) to follow single Halo-CTCF molecules in live cells . By using long exposure times ( 500 ms ) , to ‘motion-blur’ fast moving molecules into the background ( Chen et al . , 2014 ) , we could visualize and track individual stable CTCF-binding events ( Figure 2B; Video 1 ) . We recorded thousands of binding event trajectories and calculated their survival probability . A double-exponential function , corresponding to specific and non-specific DNA binding ( Chen et al . , 2014 ) , was necessary to fit the Halo-CTCF survival curve ( Figure 2C ) . After correcting for photo-bleaching ( Figure 2—figure supplement 1A ) , we estimated an average residence time ( RT ) of ~1 min for CTCF in mES cells and a slightly longer RT in U2OS cells ( Figure 2D ) . DNA-binding defective CTCF mutants or Halo-3xNLS alone interacted very transiently with chromatin ( RT ~1 s; Figure 2D ) . The measured RT did not depend on the dye or exposure time ( Figure 2—figure supplement 1B ) . We note that a CTCF RT of ~1 min is a genomic average and that some binding sites likely exhibit a slightly longer or shorter mean residence time . We also note that there is likely an oversampling of binding events at CTCF-binding sites showing the strongest ChIP-Seq enrichment ( Figure 1E ) , which tend to be the sites involved in looping ( Merkenschlager and Nora , 2016 ) . To cross-validate these results using an orthogonal technique , we performed fluorescence recovery after photo-bleaching ( FRAP ) on Halo-CTCF and quantified the dynamics of recovery ( Figure 2—figure supplement 2A–B ) . Both Halo-CTCF in mES cells ( Figure 2E ) and Halo-hCTCF in U2OS cells ( Figure 2—figure supplement 2C ) exhibited FRAP recoveries consistent with a RT ~1 min , but fitting the FRAP curves with a reaction-dominant model suggested a RT of 3–4 min ( Figure 2—figure supplement 2D ) . Whereas our SMT measurements are limited by photobleaching , estimating RTs from FRAP modeling is more indirect and tends to significantly overestimate the RT of transcription factors ( Mazza et al . , 2012 ) and is also affected by anomalous diffusion . Therefore , we interpret 1 min as a lower bound and 4 min as an upper bound for CTCF’s RT in mESCs , but expect the true RT to be closer to 1 min than 4 min . 10 . 7554/eLife . 25776 . 009Figure 2 . CTCF and cohesin have very different residence times on chromatin . ( A ) Sketch illustrating HiLo ( highly inclined and laminated optical sheet illumination ) ( Tokunaga et al . , 2008 ) . ( B ) Example images showing single Halo-mCTCF molecules labeled with JF549 binding chromatin in a live mES cell . ( C ) A plot of the uncorrected survival probability of single Halo-mCTCF molecules and one- and two-exponential fits . Right inset: a log-log survival curve . ( D ) Photobleaching-corrected residence times for Halo-CTCF , Halo-3xNLS and a zinc-finger ( 11 His→Arg point-mutations ) mutant or entire deletion of the zinc-finger domain . Error bars show standard deviation between replicates . For each replicate , we recorded movies from ~6 cells and calculated the average residence time using H2B-Halo for photobleaching correction . Each movie lasted 20 min with continuous low-intensity 561 nm excitation and 500 ms camera integration time . Cells were labeled with 1–100 pM JF549 . ( E ) FRAP recovery curves for Halo-mCTCF , H2B-Halo and Halo-3xNLS in mES cells labeled with 1 μM Halo-TMR . ( F ) FRAP recovery curves for mRad21-Halo and H2B-Halo in mES cells labeled with 1 μM Halo-TMR . Right: sketch of Fucci cell-cycle phase reporter ( Sakaue-Sawano et al . , 2008; Sladitschek and Neveu , 2015 ) . We modified the system to contain mCitrine-hGem ( aa1-110 ) and SCFP3A-hCdt ( aa30-120 ) to avoid overlap in the red region of the electromagnetic spectrum . Each FRAP curve shows mean recovery from >15 cells from ≥3 replicates and error bars show the standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 00910 . 7554/eLife . 25776 . 010Figure 2—figure supplement 1 . Illustration of how residence times are inferred from SMT and control experiments . ( A ) Illustration of how residence times were inferred from single-molecule tracking experiments . Left , a double-exponential fit to Halo-CTCF and H2B-Halo survival curves measured from the two cell lines on the same day . H2B-Halo shows negligible turnover within the observation time window and thus the measured off-rate for H2B-Halo was taken as the rate of photo-bleaching . Right , a sketch showing the inferred residence time ( true CTCF off rate = measured CTCF off rate - measured H2B off rate ) as a function of the frame threshold used for the two-exponential fitting . The residence time was taken as the inferred residence time once convergence was reached ( threshold = 2 . 5 s ) . ( B ) Bar graph showing that the CTCF residence time does not depend on the exposure time or dye used . Left , C87 Halo-mCTCF was labeled with JF549 and residence time was measured as in ( A ) using exposure times of either 300 ms , 500 ms or 800 ms . Right , residence time was measured as in ( A ) using an exposure time of 500 ms with C87 Halo-mCTCF and HaloTag dyes TMR , JF549 and JF646 . Error bars show the standard deviation between replicates – each replicate consisted of 20 min movies of ~6 cells corrected for photobleaching using a similar number of H2B-Halo cells labeled with the same dye and using the same exposure time . At least three replicates were performed . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 01010 . 7554/eLife . 25776 . 011Figure 2—figure supplement 2 . Supplementary and control CTCF FRAP experiments . ( A ) Representative raw confocal microscopy images of H2B-Halo , C59 Halo-mCTCF and C32 Halo-hCTCF each labeled with 1 μM TMR just before , 1 s after , 10 s after and 4 min after bleaching a 1 μm circular spot ( green circle ) . mES cells especially show significant movement on the minute time-scale and all movies were drift-corrected . ( B ) Photo-bleaching corrected FRAP curves measured at one frame per second in mESCs . C59 and C87 Halo-mCTCF show the behavior of endogenously tagged CTCF , and transiently transfected Halo-mCTCF expressed from a CMV-promoter is shown in black ( OE: over-expressed ) . As can be clearly seen , the CTCF FRAP recovery is much faster when over-expressed . Error bars show standard error . ( C ) Photobleaching-corrected FRAP curves measured at one frame per second in human U2OS cells . Halo-3xNLS and H2B-Halo are controls for rapid and negligible recovery , respectively , demonstrating the validity our photobleaching- and drift-correction approaches . CTCF recovery in human U2OS cells is similar , albeit slightly slower than in mESCs . ( D ) Model-fitting of C87 and C59 Halo-mCTCF FRAP curves using a reaction dominant model ( Materials and methods ) for FRAP experiments at either 11 min and 0 . 5 Hz or 5 min and 1 Hz . The model-inferred residence time and the 95% confidence interval ( CI ) from the fitting is shown . ( E ) Dynamics of Halo-mCTCF ( C59 mESCs ) does not change between G1- and S/G2-phase of the cell cycle . G1 and S/G2 were distinguished using the Fucci system ( Figure 2—figure supplement 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 01110 . 7554/eLife . 25776 . 012Figure 2—figure supplement 3 . Supplementary and control cohesin FRAP experiments . ( A ) Top , sketch of the Fucci-system , which distinguishes cells in G1-phase from S/G2-phase of the cell cycle . Bottom , representative raw confocal microscopy images of mESC C45 mRad21-Halo in G1 or S/G2 labeled with 1 μM TMR just before , 2 s after , 2 min after and 10 min after bleaching a 1 μm circular spot ( green circle ) . We corrected for cell drift . ( B ) FRAP curves for SNAPf-tagged mRad21 ( C59 mRad21-SNAPf-V5 ) similar to FRAP curves for C45 presented in main Figure 2F . C59 mRad21-SNAPf co-expressing Fucci reporters or H2B-SNAPf mESCs were labeled with cp-JF549 , a small molecule dye that specifically labels the SNAPf- but not the Halo-Tag , and FRAP recovery measured , photobleaching and drift-corrected . Overall , the FRAP recovery of mRad21-Halo and mRad21-SNAPf were identical within error . Error bars show standard error . ( C ) show exponential-model fits ( black line ) to C45 mRad21-Halo and C59 mRad21-SNAPf FRAP curves ( blue data ) in G1-phase of the cell cycle . We stress that while direct modeling of FRAP curves suffers from fitting models with several fitted parameters to relatively simply data curves , a rough residence time estimate can , nevertheless , be obtained . The average inferred residence time and 95% confidence intervals ( CI ) are shown . ( D ) FRAP of endogenous and transiently transfected mRad21-Halo in interphase cells . For C45 mRad21-Halo we ‘simulated’ interphase cells by scaling the G1 to S/G2 FRAP by 1:7 ( roughly the ratio of G1 to S/G2 cells for mESCs ) . For transient transfections , wild-type mESCs were transfected with a plasmid encoding either pCMV-mRad21-Halo ( high expression ) or pL30-mRad21-Halo ( lower expression ) . Cells were transfected with Lipofectamine 3000 using the ThermoFisher Scientific protocol the day before the FRAP imaging . On the imaging day , cells were labeled with 1 μM TMR . Error bars show standard error . ( E ) Radial bleach spot profile . The radial bleach spot profile was determined by calculating the FRAP recovery at the perimeter of circles of increasing radii in increments of 100 nm . This was averaged over multiple cells ( ~20–30 cells ) . Curves show the mean radial profile along with the standard deviation ( error bars ) for C59 Halo-mCTCF , C45 mRad21-Halo in G1 and mESC H2B-Halo . As can be seen , diffusion does not significantly affect the recovery at a radius of 0 . 6 μm , which is what we used for calculating the FRAP recovery . ( F ) FRAP recovery for H2B-SNAPf labeled with 500 nM cp-JF549 , extensively washed and then with or without blocking with 500 nM SNAP-block ( a ‘dark’ ligand ) . As can be seen , there was no effect on the FRAP recovery within error ( error bars show standard error ) . Thus , re-binding of insufficiently washed out dye after the bleach pulse does not significantly affect our FRAP recovery at the time-scales we are working at . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 01210 . 7554/eLife . 25776 . 013Figure 2—figure supplement 4 . Validation of Fucci reporters . ( A ) FACS gating of C59 mESCs co-expressing the Fucci reporters mCherry-hCdt ( aa30-120 ) and mCitrine-hGem ( aa1-110 ) with the four polygon gates shown ( Sakaue-Sawano et al . , 2008; Sladitschek and Neveu , 2015 ) . Cells in G1 express mCherry-hCdt ( aa30-120 ) but not mCitrine-hGem ( aa1-110 ) , whereas cells in S/G2 express mCitrine-hGem ( aa1-110 ) but not mCherry-hCdt ( aa30-120 ) . For FRAP and SMT experiments where small-molecule dyes in the red part of the electromagnetic spectrum were used ( TMR , JF549 or JF6464 ) , we used SCFP3A-hCdt instead of mCherry-hCdt . We also note that strictly speaking , only the mCitrine-hGem ( aa1-110 ) reporter is necessary since cells negative for this reporter must be in G1 . ( B–E ) histograms showing DAPI ( which stains DNA ) signal for the gated populations in ( A ) . Cells in G1 have not replicated their DNA and thus have 2n chromosomes ( D ) , whereas cells in G2 have finished replicating their DNA and have 4n chromosomes . Cells in S-phase are actively replicating DNA and have an intermediate number of chromosomes ( E ) . As can be seen , the Fucci system allows us to enrich for a relatively pure G1 population ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 01310 . 7554/eLife . 25776 . 014Video 1 . Single-molecule tracking of Halo-mCTCF in mESCs at 2 Hz . Related to Figure 2 . Using long 500 ms camera integration causes most diffusing molecules to ‘motion-blur’ into the background . Laser: 561 nm . Dye: JF549 . One pixel: 160 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 014 Our results differ considerably from a previous CTCF FRAP study using over-expressed transgenes , which reported rapid 80% recovery in 20 s ( Nakahashi et al . , 2013 ) . However , when we used similar transiently over-expressed Halo-CTCF instead of endogenous knock-in cells , we also observed similarly rapid recovery ( Figure 2—figure supplement 2B ) , suggesting that over-expression of target proteins can result in artefactual measurements . This finding underscores the importance of studying endogenously tagged and functional proteins . Thus , although CTCF ( RT ~1–2 min ) binds chromatin much more stably than most sequence-specific transcription factors ( RT ~2–15 s ) ( Chen et al . , 2014; Mazza et al . , 2012 ) , its binding is still highly dynamic . We next investigated the cell-cycle dependent cohesin binding dynamics ( Gerlich et al . , 2006 ) . In addition to its role in holding together chromatin loops , cohesin mediates sister chromatid cohesion from replication in S-phase to mitosis . Thus , since TAD demarcation is strongest in G1 before S-phase ( Naumova et al . , 2013 ) , we reasoned that cohesin dynamics in G1 should predominantly reflect the chromatin looping function of cohesin . To control for the cell-cycle , we deployed the Fucci system ( Sakaue-Sawano et al . , 2008 ) to distinguish G1 from S/G2-phase using fluorescent reporters in the C45 and C59 mESC lines ( Figure 2—figure supplements 3A and 4 ) . We then performed FRAP on mRad21-Halo ( Figure 2F ) and mRad21-SNAPf ( Figure 2—figure supplement 3B ) . We observed significantly faster mRad21 recovery in G1 than in S/G2-phase consistent with Gerlich et al . ( 2006 ) , but nevertheless much slower recovery than CTCF and CTCF showed the same recovery in G1 and S/G2 ( Figure 2—figure supplement 2E ) . The slow mRad21 turnover precluded SMT experiments . Model-fitting of the G1 mRad21 FRAP curves ( Figure 2—figure supplement 3C ) revealed an RT ~22 min . Previous cohesin FRAP studies have reported differing RTs ( Gerlich et al . , 2006; Huis in 't Veld et al . , 2014 ) and as was seen for CTCF , over-expressed mRad21-Halo also showed much faster recovery than endogenous mRad21-Halo ( Figure 2—figure supplement 3D ) . Although we cannot completely exclude a very small population ( <5% ) of CTCF or cohesin molecules with a somewhat shorter or longer RT , these RTs reflect chromatin-bound CTCF/cohesin . Since at least one-third of CTCF and the majority of G1 cohesin molecules bound to chromatin mediate looping ( see Appendix 1 for estimate ) , we are confident that these RTs hold for most CTCF/cohesin molecules involved in looping . Overall , while kinetic modeling of FRAP curves should be interpreted with some caution ( Mazza et al . , 2012 ) , these results , nevertheless , demonstrate a surprisingly large ( ~10–20x ) difference in RTs between CTCF and cohesin , which is difficult to reconcile with the notion of a biochemically stable LMC assembled on chromatin . However , although CTCF and cohesin do not form a stable complex on chromatin , it is still possible that CTCF and cohesin form a stable complex in solution when not bound to DNA . To investigate this possibility , we analyzed how CTCF and cohesin each explore the nucleus . Tracking fast-diffusing molecules has been a major challenge . To overcome this issue , we took advantage of bright new dyes ( Grimm et al . , 2016 ) and developed stroboscopic ( Elf et al . , 2007 ) photo-activation ( Manley et al . , 2008 ) single-molecule tracking ( paSMT; Figure 3—figure supplement 1A ) , which makes tracking unambiguous ( Materials and methods ) . We tracked individual Halo-mCTCF molecules at ~225 Hz and plotted the displacements between frames ( Figure 3A ) . Most Halo-mCTCF molecules exhibited displacements similar to our localization error ( ~35 nm; Materials and methods ) indicating chromatin association , whereas a DNA-binding defective CTCF mutant exhibited primarily long displacements consistent with free diffusion ( Figure 3B; Videos 2–3 ) . To characterize the nuclear search mechanism , we performed kinetic modeling of the measured displacements ( Figure 3—figure supplement 1B; Materials and methods; Mazza et al . , 2012 ) and found that in mES cells , ~49% of CTCF is bound to cognate sites , ~19% is non-specifically associated with chromatin ( e . g . 1D sliding or hopping ) and ~32% is in free 3D diffusion ( Table 1 ) . Thus , after dissociation from a cognate site , CTCF searches for ~66 s on average before binding the next cognate site: ~65% of the total nuclear search is random 3D diffusion ( ~41 s on average ) , whereas ~35% ( ~25 s on average ) consists of intermittent non-specific chromatin association ( e . g . 1D sliding; Table 1; note this search time is based on a CTCF RT of ~1 min ) . The nuclear search mechanism of CTCF in human U2OS cells was similar albeit slightly less efficient ( Table 1; Figure 3—figure supplement 1F ) . We note that CTCF’s search mechanism , with similar amounts of 3D diffusion and 1D sliding , is close to optimal according to the theory of facilitated diffusion ( Mirny et al . , 2009 ) . 10 . 7554/eLife . 25776 . 015Table 1 . Nuclear search mechanism parameters . Table 1 lists key parameters for the nuclear search mechanism inferred from model fitting of the displacements in Figure 3 and the residence times in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 015Fraction bound ( specific ) Fraction bound ( nonspecific ) Free 3D diffusion fractionApparent DFREE ( μm2/s ) τSEARCH ( total ) Fraction of τSEARCH in free 3D diffusionFraction of τSEARCH in non-specific chromatin associationmESC C59 Halo-mCTCF48 . 9%19 . 1%32 . 0%2 . 565 . 9 s41 . 3 s24 . 6 smESC C87 Halo-mCTCF49 . 3%19 . 1%31 . 6%2 . 362 . 6 s39 . 0 s23 . 6 sU2OS C32 Halo-hCTCF39 . 8%17 . 7%42 . 5%2 . 5102 . 8 s71 . 9 s30 . 9 smESC C45 mRad21-Halo: G139 . 8%13 . 7%46 . 5%1 . 533 . 0 min25 . 5 min7 . 5 minmESC C45 mRad21-Halo: S/G249 . 8%13 . 7%36 . 5%1 . 5n/an/an/a10 . 7554/eLife . 25776 . 016Figure 3 . Dynamics of CTCF and cohesin’s nuclear search mechanism . Single-molecule displacements from ~225 Hz stroboscopic ( single 1 ms 633 nm laser pulse per camera integration event ) paSMT experiments over multiple time scales for ( A ) C59 Halo-mCTCF , ( B ) a Halo-mCTCF mutant with the zinc-finger domain deleted , C45 mRad21-Halo in S/G2 phase ( C ) and G1 phase ( D ) and ( E ) a Rad21 mutant that cannot form cohesin complexes . Kinetic model fits ( three fitted parameters ) to raw displacement histograms are shown as black lines . All calculated and fitted parameters are listed in Table 1 . Displacement histograms were obtained by merging data from at least 24 cells from at least three replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 01610 . 7554/eLife . 25776 . 017Figure 3—figure supplement 1 . Supplementary stroboscopic paSMT experiments and controls . ( A ) Sketch illustrating stroboscopic paSMT . Sketch illustrating labeling Halo-tagged proteins , e . g . CTCF or Rad21 , with PA-JF646 . This dye remains dark until 405 nm activation , which converts it to regular fluorescent JF646 . The advantage is that thousands of single-molecule trajectories can be recorded from a single cell at a density of ~0 . 5 fluorescent molecules per nucleus per frame , which makes tracking unambiguous , by using very low intensity 405 nm excitation . Since high 633 nm laser powers are used , most molecules bleach within 3–8 frames . We use PA-JF646 instead of PA-JF549 since the red-shifted 633 nm laser induces less photo-toxicity , although the displacement histograms were identical between PA-JF549 and PA-JF646 . Moreover , we never record for more than 2 min per cell . We observed no visible signs of photo-toxicity after 2 min of paSMT . Below , sketch illustrating stroboscopic illumination . To minimize ‘motion-blurring’ of fast-diffusing molecules , we used pulsed 633 nm excitation with 1 ms pulses . The camera integration time was 4 ms + ~0 . 447 ms ( frame-transfer mode ) resulting in a frame rate of roughly 225 Hz . Below , raw microscopy images demonstrating that even fast-diffusing molecules can be imaged and tracked ( red lines ) at high-signal-to-background . ( B ) Overview of two-state dynamic displacement model . Full details are provided in the Materials and methods . Briefly , the model assumes molecules can exist in either a chromatin associating ( specific and non-specific ) state called ‘bound’ or in a free 3D diffusion state called ‘free’ . A mathematical model describing how the distribution of displacements , r , depends on the time delay , fraction bound , diffusion constants , localization error and axial detection slice is shown below . Overall , the model contains three fitted parameters , which were estimated using least squares fitting to the raw displacements considering the first seven displacements ( Δτ ~4 . 5 ms to 31 . 5 ms ) . For ease of visualization , we show displacement histograms in ( C–F ) , but the fitting was performed on cumulative distribution functions ( CDFs ) to minimize binning artifacts . ( C–F ) displacement histograms for various cell lines all measured using the approach in ( A–B ) . For ease of visualization , the displacement histograms are cut off at 1050 nm , but longer trajectories were included in the model fitting . ( C ) shows Halo-only and Halo-3xNLS in mESCs , which show negligible binding . Note , that most fast-diffusing molecules eventually move out of the focal plane . ( D ) shows various Halo-mCTCF constructs . C59 and C87 are endogenous Halo-mCTCF knock-ins . pL30-wt-Halo-mCTCF was transiently expressed using a weak promoter ( L30 ) . Compared to Halo-mCTCF overexpressed by the strong CMV promoter , overexpressing Halo-mCTCF using a weak promoter ( L30 ) causes only a minor ( 10 percentage points; likely due to saturation of binding sites ) underestimation of the fraction bound . Right , two transiently transfected Halo-mCTCF mutants: 11ZF-mut-Halo-mCTCF is a CTCF mutant with an essential His amino acid in all 11 zinc-fingers mutated to Arg , which should abolish specific DNA-binding . We used this mutant to estimate the non-specifically bound fraction . ΔZF-Halo-mCTCF has the entire 11-zinc-finger domain deleted . We used this mutant to verify that the zinc-finger domain solely is responsible for chromatin association . ( E ) H2B-Halo and Rad21 experiments in mESCs . We used H2B-halo as a control for a protein that is almost exclusively bound . Note that since we use an EF1a promoter to express H2B , which is not cell-cycle regulated , some H2B-molecules do show free diffusion . mRad21-Halo in S/G2 and G1 are also shown , as is transiently transfected wt-mRad21-Halo expressed using the low-expression promoter , L30 . Even though mRad21-Halo is only weakly overexpressed , most molecules show free 3D diffusion . The overexpression artifact may be caused by the fact that without similar overexpression of Smc1 , Smc3 and SA1/2 , most Rad21 cannot form cohesin complexes . Finally , a Rad21-mutant ( F601R , L605R , Q617K ) , which cannot form cohesin complexes was used to estimate the non-specifically bound Rad21 fractions . ( F ) Stroboscopic paSMT experiments in human U2OS cells . H2B-Halo and Halo-3xNLS were used as controls for mostly bound and free molecules and the same zinc-finger mutants as in mESCs were transiently transfected as control for non-specific chromatin association . Note that C32 Halo-hCTCF shows slightly more free diffusion than C59 and C87 in mESCs . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 01710 . 7554/eLife . 25776 . 018Video 2 . Single-molecule tracking of Halo-mCTCF in mESCs at 225 Hz . Related to Figure 3 . Stroboscopic ( 1 ms of 633 nm ) paSMT allows tracking of fast-diffusing molecules . Lasers: 405 and 633 nm . Dye: PA-JF646 . One pixel: 160 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 01810 . 7554/eLife . 25776 . 019Video 3 . Single-molecule tracking of ΔZF-Halo-mCTCF in transiently transfected mESCs at 225 Hz . Related to Figure 3 . Stroboscopic ( 1 ms of 633 nm ) paSMT allows tracking of fast-diffusing molecules . Lasers: 405 and 633 nm . Dye: PA-JF646 . One pixel: 160 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 019 Similar analysis of mRad21-Halo in G1 and S/G2 ( Figure 3C–D ) revealed that cohesin complexes diffuse rather slowly compared to CTCF ( Table 1 ) and that roughly half of cohesins are topologically engaged with chromatin ( G1: ~40%; S/G2: ~50% ) compared to ~13% in non-specific , non-topological chromatin association and the remainder in 3D diffusion ( G1: ~47%; S/G2: ~37% ) . Conversely , a Rad21 mutant ( Haering et al . , 2004 ) unable to form cohesin complexes displayed rapid diffusion and little chromatin association ( Figure 3E ) . Like this Rad21 mutant , overexpressed wild-type mRad21-Halo also showed negligible chromatin association ( Figure 3—figure supplement 1E ) again underscoring the importance of studying endogenously tagged proteins at physiological concentrations . Importantly , this also shows that essentially all endogenously expressed mRad21-Halo proteins are incorporated into cohesin complexes . Topological association and dissociation of cohesin is regulated by a complex interplay of co-factors such as Nipbl , Sororin and Wapl ( Skibbens , 2016 ) . If we , nevertheless , apply a simple two-state model to analyze cohesin dynamics ( Materials and methods ) , we estimate an average search time of ~33 min in between topological engagements of chromatin in G1 , with ~77% of the total search time spent in 3D diffusion ( ~26 min ) compared to ~23% in non-specific chromatin association ( 7 min ) . Thus , for each specific topological cohesin chromatin binding-unbinding cycle in G1 , CTCF binds and unbinds its cognate sites ~20–30 times . These results are certainly not consistent with a model wherein CTCF and cohesin form a stable LMC . Moreover , since CTCF diffuses much faster than cohesin ( Table 1 ) , it also seems unlikely that CTCF and cohesin form stable complexes in solution . To resolve these apparently paradoxical findings , we investigated the nuclear organization of CTCF and cohesin simultaneously in the same nucleus . We labeled Halo-mCTCF and mRad21-SNAPf in C59 mES cells with the spectrally distinct dyes JF646 and JF549 ( Grimm et al . , 2015 ) , respectively , and performed two-color direct stochastic optical reconstruction microscopy ( dSTORM ) super-resolution imaging in formaldehyde-fixed cells ( Figure 4A ) . We localized individual CTCF and Rad21 molecules with a precision of ~20 nm , less than half the size of the cohesin ring . We observe significant clustering of both CTCF and Rad21 and a large fraction of these clusters overlap ( Figure 4A and Figure 4—figure supplement 1A–C ) . We next confirmed clustering using photo-activation localization microscopy ( PALM ) and found that both CTCF and Rad21 predominantly form small clusters ( Figure 4B and Figure 4—figure supplement 1; mean cluster radius ~30–40 nm ) . To determine whether individual CTCF and cohesin molecules co-localize , we calculated the pair cross correlation , C ( r ) ( Stone and Veatch , 2015 ) . C ( r ) quantifies spatial co-dependence as a function of length , r , and C ( r ) =1 for all r under complete spatial randomness ( CSR ) . CTCF and cohesin exhibited significant co-localization ( C ( r ) >1 ) at very short distances in mES cells ( Figure 4C ) . Conversely , CTCF and cohesin were nearly independent at length scales beyond the diffraction limit , emphasizing the importance of super-resolution approaches . A mES cell line co-expressing histone H2B-SNAPf and Halo proteins imaged under the same dSTORM conditions showed no pair cross-correlation ( Figure 4C ) , thereby ruling out technical artifacts . Thus , our two-color dSTORM results provide compelling evidence that a large fraction of CTCF and cohesin molecules indeed co-localize at the single-molecule level inside the nucleus consistent with the LMC model and reveals a clustered nuclear organization . 10 . 7554/eLife . 25776 . 020Figure 4 . Models of CTCF/cohesin mediated chromatin loop dynamics . ( A ) Two-color dSTORM of C59 mESCs with mRad21-SNAPf labeled with 500 nM JF549 ( green ) and Halo-mCTCF labeled with 500 nM JF646 ( magenta ) . High-intensity co-localization is shown as white . Low-intensity co-localization is not visible . Zoom-in on red 3 μm square . Note , the SNAP dye cp-JF549 shows slight artefactual labeling of the nuclear envelope , which was removed during image rendering . ( B ) Cluster radii distributions for CTCF ( C87 and C59 ) and Rad21 ( C45 ) from single-color PALM experiments using PA-JF549 dyes . ( C ) Pair cross correlation of C59 and mESC H2B-SNAPf co-expressing Halo-only . Error bars are standard error from 12 to 18 dSTORM-imaged cells over three replicates . ( D ) Sketch illustrating the concept of a dynamic loop maintenance complex ( LMC ) composed of CTCF and cohesin with frequent CTCF exchange and slow , rare cohesin dissociation , which causes loop deformation and topological re-orientation of chromatin . ( E ) Sketch illustrating how dynamic CTCF exchange during loop extrusion of cohesin may explain alternative loop formations when two competing convergent sites ( B and C ) for another site A ) exist . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 02010 . 7554/eLife . 25776 . 021Figure 4—figure supplement 1 . Overview of super-resolution PALM approach and control experiments . ( A ) Representative super-resolution PALM reconstruction of Halo-mCTCF in C59 mouse embryonic stem cells . Left: full nucleus . Right: zoom-in on a 3 μm square before ( top ) and after ( bottom ) cluster assignment using a Bayesian clustering algorithm . ( B ) Representative super-resolution PALM reconstruction of mRad21-Halo in C45 mouse embryonic stem cells . Left: full nucleus . Right: zoom-in on a 3 μm square before ( top ) and after ( bottom ) cluster assignment using a Bayesian clustering algorithm . ( C ) Bar graphs showing fraction of molecules in clusters for different mES cell lines as inferred from Bayesian cluster assignments . Bar graphs show mean per 3 μm square and the error bars show the standard error . ( D ) Bar graphs showing cluster radii in clusters for different mES cell lines as inferred from Bayesian cluster assignments . Bar graphs show mean per 3 μm square and the error bars show the standard error . ( E ) Representative control for photo-blinking in apparent PALM clustering . U2OS C32 Halo-hCTCF was labeled with ~50:50 PA-JF549:PA-JF646 and imaged using two-color PALM . Clusters were assigned as in ( C–D ) and the fraction of CTCF molecules in each cluster labeled with JF549 and JF646 plotted . If PALM clustering was solely a photo-blinking artifact , clusters should be exclusively composed of either JF549 or JF646 , whereas under ideal conditions the distribution should follow a binomial distribution . The observed distributions resemble the expected binomial distributions suggesting that most called clusters are not a photo-blinking artifact ( Kullback-Leibler divergence ~0 . 3 bits ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25776 . 021 Chromatin loop domains are widely believed to be very stable structures ( Andrey et al . , 2017; Ghirlando and Felsenfeld , 2016; Hnisz et al . , 2016b ) held together by a LMC composed of two CTCFs and cohesin ( whether cohesin acts as a single ring or as a pair of rings remains a matter of debate [Skibbens , 2016] ) . While our in vitro biochemical ( Figure 1G ) and co-localization ( Figure 4A–C ) experiments do demonstrate complex formation between CTCF and cohesin , our SMT experiments paradoxically reveal this complex to be highly transient and dynamic ( Figures 2–3 ) . To reconcile these observations , we therefore propose a ‘dynamic LMC’ model . Consistent with previous studies , CTCF mainly functions to position cohesin at loop boundaries , whereas cohesin physically holds together the two chromatin strands . However , in the ‘dynamic LMC’ model , while cohesin holds together a given chromatin loop , different CTCF molecules are frequently alighting and departing in a dynamic exchange thus giving rise to a ‘transient protein complex’ with a molecular stoichiometry that cycles over time ( Figure 4D ) . Since topological chromatin association of cohesin is infrequent ( ~33 min in G1 ) , dissociation of cohesin ( ~22 min ) likely causes the loop to fall apart ( Figure 4D ) . Even if the CTCF and cohesin co-clusters that we observe ( Figure 4A–C; Figure 4—figure supplement 1 ) are LMC clusters that hold together loop domains , their lifetimes are unlikely to be more than 1–2 hr . Thus , our results suggest that chromatin loops are continuously formed and dissolved throughout a typical 14–24 hr mammalian cell cycle . Our results suggesting that loops are dynamic also provide experimental support for theoretical polymer simulation studies , which found that only dynamic , but not static , loop structures can reproduce experimentally observed chromatin interaction frequencies ( Benedetti et al . , 2014; Fudenberg et al . , 2016; Giorgetti et al . , 2014; Sanborn et al . , 2015 ) . We note that our quantitative characterization of CTCF and cohesin dynamics could be useful for parameterizing future polymer models . While our results indicate that loops are highly dynamic , the question of how they are formed remains . An attractive but not yet verified recent model suggests that loops are formed by cohesin-mediated loop extrusion ( Fudenberg et al . , 2016; Sanborn et al . , 2015 ) , whereby cohesin extrudes a loop by sliding on DNA ( Davidson et al . , 2016; Lengronne et al . , 2004; Nasmyth , 2001; Stigler et al . , 2016 ) until it encounters two convergent and bound CTCF sites ( Figure 4E ) . Our imaging experiments ( Figures 2–3 ) cannot readily distinguish cohesin stably bound at loop anchors from cohesin in the process of extrusion and thus our measured residence time of ~22 min reflects the average total duration of both . In the context of the loop extrusion model , our results suggest a mechanism for boundary permeability through dynamic and stochastic CTCF occupancy at cognate CTCF sites , which may explain the formation of competing loop domains ( Figure 4E ) . This would also explain why DNA-FISH measurements show that most loops are only present in a subset of cells at any given time ( Sanborn et al . , 2015; Williamson et al . , 2014 ) . Finally , the highly dynamic view of frequently breaking and forming chromatin loops presented here may also facilitate dynamic long-distance enhancer-promoter scanning of DNA in cis , which may be important for temporally efficient regulation of gene expression . JM8 . N4 mouse embryonic stem cells ( Pettitt et al . , 2009 ) ( Research Resource Identifier: RRID:CVCL_J962; obtained from the KOMP Repository at UC Davis ) were grown on plates pre-coated with a 0 . 1% autoclaved gelatin solution ( Sigma-Aldrich , St . Louis , MO , G9391 ) under feeder-free condition in knock-out DMEM with 15% FBS and LIF ( full recipe: 500 mL knockout DMEM ( ThermoFisher , Waltham , MA , #10829018 ) , 6 mL MEM NEAA ( ThermoFisher #11140050 ) , 6 mL GlutaMax ( ThermoFisher #35050061 ) , 5 mL Penicillin-streptomycin ( ThermoFisher #15140122 ) , 4 . 6 μL 2-mercapoethanol ( Sigma-Aldrich M3148 ) , 90 mL fetal bovine serum ( HyClone , Logan , UT , FBS SH30910 . 03 lot #AXJ47554 ) ) and LIF . mES cells were fed by replacing half the medium with fresh medium daily and passaged every 2 days by trypsinization . Human U2OS osteosarcoma cells ( Research Resource Identifier: RRID:CVCL_0042; a gift from David Spector’s lab , Cold Spring Harbor Laboratory ) were grown in low-glucose DMEM with 10% FBS ( full recipe: 500 mL DMEM ( ThermoFisher #10567014 ) , 50 mL fetal bovine serum ( HyClone FBS SH30910 . 03 lot #AXJ47554 ) and 5 mL Penicillin-streptomycin ( ThermoFisher #15140122 ) ) and were passaged every 2–4 days before reaching confluency . For live-cell imaging , the medium was identical except DMEM without phenol red was used ( ThermoFisher #31053028 ) . Both mouse ES and human U2OS cells were grown in a Sanyo copper alloy IncuSafe humidified incubator ( MCO-18AIC ( UV ) ) at 37°C/5 . 5% CO2 . For all single-molecule experiments ( both live and fixed ) , cells we grown overnight on 25 mm circular no 1 . 5H cover glasses ( Marienfeld , Germany , High-Precision 0117650 ) . Prior to all experiments , the cover glasses were plasma-cleaned and then stored in isopropanol until use . For U2OS cell lines , cells were grown directly on the cover glasses and for mouse ES cells , the cover glasses were coated with Corning Matrigel matrix ( Corning #354277; purchased from ThermoFisher #08-774-552 ) according to manufacturer’s instructions just prior to cell plating . After overnight growth , cells were labeled with the relevant Halo- or SNAP-dye at the indicated concentration for 15 min ( Halo ) or 30 min ( SNAP ) and washed twice ( one wash: medium removed; PBS wash; replenished with fresh medium ) . At the end of the final wash , the medium was changed to phenol red-free medium keeping all other aspects of the medium the same . For FRAP experiment , cell preparation was identical except cells where grown on glass-bottom ( thickness #1 . 5 ) 35 mm dishes ( MatTek , Ashland , MA , P35G-1 . 5–14 C ) , either directly ( U2OS ) or Matrigel coated ( mESC ) . Mouse ES cell lines stably expressing H2B-Halo , H2B-SNAPf , Fucci reporters or Halo-3xNLS were generated using PiggyBac transposition and drug selection . Briefly , the relevant gene ( e . g . H2B-Halo ) was cloned into a PiggyBac vector co-expressing a drug resistance gene ( G418 or Puromycin ) and this vector was then co-transfected together with a SuperPiggyBac transposase vector into the relevant mouse ES cell line using Lipofectamine 3000 according to manufacturer’s instructions ( 2 μg expression vector and 1 μg PiggyBac transposase vector per well in a 6-well plate ) . The following day , selection was then started by adding 1 mg/mL G418 or 5 μg/mL puromycin . An untransfected cell line was selected in parallel and selection was judged to be complete once no live cells were left in the untransfected cell line . For human U2OS cells , stable cell lines were generated by random integration by transfecting the relevant expression vector with drug selection without using the PiggyBac system . Selection was performed in the same way as for mouse ES cells . Knock-in cell lines were created roughly according to published procedures ( Ran et al . , 2013 ) , but exploiting the HaloTag and SNAPf-Tag to FACS for edited cells . The SNAPf-Tag is an optimized version of the SNAP-Tag , and we purchased a plasmid encoding this gene from NEB ( NEB , Ipswich , MA , #N9183S ) . We transfected both U2OS and mES cells using Lipofectamine 3000 ( ThermoFisher L3000015 ) according to manufacturer’s protocol , co-transfecting a Cas9 and a repair plasmid ( 2 μg repair vector and 1 μg Cas9 vector per well in a 6-well plate; 1:2 w/w ) . The Cas9 plasmid was slightly modified from that distributed from the Zhang lab ( Ran et al . , 2013 ) : 3xFLAG-SV40NLS-pSpCas9 was expressed from a CBh promoter; the sgRNA was expressed from a U6 promoter; and mVenus was expressed from a PGK promoter . For the repair vector , we modified a pUC57 plasmid to contain the tag of interest ( e . g . Halo or SNAPf ) flanked by ~500 bp of genomic homology sequence on either side . For N-terminal FLAG-Halo-tagging of mouse Ctcf and human CTCF , we introduced synonymous mutations ( mCTCF: first nine codons after ATG; hCTCF: first 12 codons after ATG ) , where possible , to prevent the Cas9-sgRNA complex from cutting the repair vector . For C-terminal tagging of mouse Rad21 with SNAPf-V5 , this was not possible . Instead , we designed sgRNAs that overlapped with the STOP codon and , thus , that would not cut the repair vector . For Halo-hCTCF and Halo-mCTCF , we used a TEV linker sequence ( EDLYFQS ) to link the Halo protein to CTCF; for mRad21 , we used the Sheff and Thorn linker ( GDGAGLIN ) ( Sheff and Thorn , 2004 ) . In each case , we designed three or four sgRNAs using the Zhang lab CRISPR design tool ( http://tools . genome-engineering . org ) , cloned them into the Cas9 plasmid and co-transfected each sgRNA-plasmid with the repair vector individually . 18–24 hr later , we then pooled cells transfected with each of the sgRNAs individually and FACS-sorted for YFP ( mVenus ) positive , successfully transfected cells . YFP-sorted cells were then grown for 4–12 days , labeled with 500 nM Halo-TMR ( Halo-Tag knock-ins ) or 500 nM SNAP-JF646 ( SNAPf-Tag knock-in ) and the cell population with significantly higher fluorescence than similarly labeled wild-type cells , FACS-selected and plated at very low density ( ~0 . 1 cells per mm2; mES cells ) or sorted individually into 96-well plates ( U2OS cells ) . Clones were then expanded and genotyped by PCR using a three-primer PCR ( genomic primers external to the homology sequence and an internal Halo or SNAPf primer ) . Successfully edited clones were further verified by PCR with multiple primer combinations , Sanger sequencing and Western blotting . We isolated ~6–10 homozygous knock-in clones for each line . The clones chosen for further study all showed similar tagged protein levels to the endogenous untagged protein in wild-type controls . Sequences for primers and sgRNAs are given in Supplementary file 2 . All plasmids used in this study , including for genome-editing and transient transfections , are available upon request . To verify that genome-edited mES cell lines remain pluripotent , we performed teratoma assays and compared wild-type and C59 FLAG-Halo-mCTCF; mRad21-SNAPf-V5 knock-in cells . Briefly , 350 , 000 cells were injected into the kidney capsule and testis of two 8-week-old Fox Chase SCID-beige male mice ( Charles River ) . Tumors were harvested 27 or 33 days post-injection , fixed with 10% formalin overnight , embedded in paraffin and cut into 5 μm sections and haematoxylin and eosin staining performed . Teratoma assays were performed by Applied Stem Cell , Inc ( Milpitas , CA ) . Wild-type and double FLAG-Halo-mCTCF / mRad21-SNAPf-V5 knock-in mouse ES cell line clone 59 were pathogen tested using the IMPACT II test , which was performed by IDEXX BioResearch ( Westbrook , ME ) . Both the wild-type and C59 cell line were negative for all pathogens including Ectromelia , EDIM , LCMV , LDEV , MAV1 , MAV2 , mCMV , MHV , MNV , MPV , MVM , Mycoplasma pulmonis , Mycoplasma sp . , Polyoma , PVM , REO3 , Sendai , and TMEV . U2OS cell lines were pathogen tested for mycoplasma using a PCR-based assay as described ( Young et al . , 2010 ) ( wild-type U2OS ) and pathogen tested for mycoplasma using an imaging assay ( DAPI staining; C32 knock-in cell line ) . Both were negative for mycoplasma . Both mouse ES cells and human U2OS cells were authenticated by whole-genome sequencing and morphology ( U2OS morphology was compared to U2OS cells obtained from ATCC ) . The wild-type and C32 FLAG-Halo-hCTCF knock-in cell lines were further authenticated using Short Tandem Repeat ( STR ) profiling ( performed by Dr . Alison N . Killilea at the UC Berkeley Cell Culture Facility ) against the following loci: THO1 , D5S818 , D13S317 , D7S820 , D16S539 , CSF1PO , AMEL , vWA and TPOX . Both the wild-type and C32 U2OS cell lines showed a 100% match with U2OS . All single-molecule imaging experiments ( live-cell residence time measurements , live-cell paSMT at 225 Hz , fixed-cell PALM and fixed-cell dSTORM ) were conducted on a custom-built Nikon ( Nikon Instruments Inc . , Melville , NY ) TI microscope equipped with a 100x/NA 1 . 49 oil-immersion TIRF objective ( Nikon apochromat CFI Apo TIRF 100x Oil ) , EM-CCD camera ( Andor , Concord , MA , iXon Ultra 897 ) , a perfect focusing system to correct for axial drift and motorized laser illumination ( Ti-TIRF , Nikon ) , which allows an incident angle adjustment to achieve highly inclined and laminated optical sheet illumination ( Tokunaga et al . , 2008 ) . The incubation chamber maintained a humidified 37°C atmosphere with 5% CO2 and the objective was similarly heated to 37°C for live-cell experiments . Excitation was achieved using the following laser lines: 561 nm ( 1 W , Genesis Coherent , Santa Clara , CA ) for JF549/PA-JF549 and TMR dyes; 633 nm ( 1 W , Genesis Coherent ) for JF646/PA-JF646 dyes; 405 nm ( 140 mW , OBIS , Coherent ) for all photo-activation experiments . The excitation lasers were modulated by an acousto-optic Tunable Filter ( AA Opto-Electronic , France , AOTFnC-VIS-TN ) and triggered with the camera TTL exposure output signal . The laser light is coupled into the microscope by an optical fiber and then reflected using a multi-band dichroic ( 405 nm/488 nm/561 nm/633 nm quad-band , Semrock , Rochester , NY ) and then focused in the back focal plane of the objective . Fluorescence emission light was filtered using a single band-pass filter placed in front of the camera using the following filters: TMR and JF549/PA-JF549: Semrock 593/40 nm band-pass filter; JF646/PA-JF646: Semrock 676/37 nm bandpass filter . The microscope , cameras , and hardware were controlled through the NIS-Elements software ( Nikon ) . For simultaneous two-color experiments ( dSTORM and PALM experiments ) , a custom-built setup using two cameras ( both Andor iXon Ultra 897 EM-CCD ) was used . Cameras were synchronized using a National Instruments ( Austin , TX ) DAQ board ( NI-DAQ PCI-6723 ) . A single-edge dichroic beamsplitter ( Di02-R635−25 × 36 , Semrock ) was used to separate two ranges of wavelengths of emission fluorescence . A 676/37 nm band-pass filter ( FF01-676/37-25 , Semrock ) was placed in front of the first camera and 593/40 nm bandpass filter ( FF01-593/40-25 , Semrock ) in front of the second camera . In ‘slow-tracking’ experiments , to measure residence times , long exposure times ( 300 ms , 500 ms or 800 ms ) and low constant illumination laser intensities ( to minimize photobleaching ) were used . The camera settings were as follows: normal mode; vertical shift speed: 3 . 3 μs; ROI: variable . Generally , each experiment lasted 20 min per cell corresponding to 4000 frames with a 300 ms exposure time , 2400 frames with a 500 ms exposure time and 1500 frames with an exposure time of 800 ms . We recorded 20 min movies from ~6 cells per cell line or condition per day as well as 6 H2B-Halo cells for the photobleaching correction on the same day and all data presented are from at least three independent experiments conducted on different days . In ‘fast-tracking’ stroboscopic paSMT experiments at ~225 Hz , both the main excitation laser ( 633 nm for PA-JF646 or 561 nm for PA-JF549 ) and the photo-activation laser ( 405 nm ) were pulsed . Each frame consisted of a 4-ms camera exposure time followed by a ~447 μs camera ‘dead’ time . The main excitation laser ( 633 nm ) was pulsed for 1 ms starting at the beginning for the 4 ms camera exposure time . The photo-activation laser ( 405 nm ) was pulsed during the ~447 μs camera ‘dead’ time , to minimize fluorescent background signal . This sequence was verified using an oscilloscope . The camera settings were as follows: frame transfer mode; vertical shift speed: 0 . 9 μs; ROI: height 90 pixels , width variable . Each cell was imaged for 20 , 000 frames corresponding to ~1 . 5 min . The photo-activation laser power was optimized to keep an average molecule density of ~0 . 5 localizations per frame , corresponding to ~10 , 000 localization per cell per movie on average . Maintaining a very low density of molecules is necessary to avoid tracking errors . The main excitation laser was used at maximal power . We recorded movies for eight cells per cell line or condition per day , and all data presented are from at least three independent experiments conducted corresponding to at least 24 cells and at least 100 , 000 localizations . In PALM experiments , continuous illumination was used for both the main excitation laser ( 633 nm for PA-JF646 or 561 nm for PA-JF549 ) and the photo-activation laser ( 405 nm ) . However , the intensity of the 405 nm laser was gradually increased over the course of the illumination sequence to image all molecules and at the same time avoid too many molecules being activated at any given frame . The following camera settings were used: 25 ms exposure time; frame transfer mode; vertical shift speed: 0 . 9 μs; ROI: variable . In total , 40 , 000–60 , 000 frames were recorded for each cell ( ~20–25 min ) , which was sufficient to image and bleach all labeled molecules . After overnight growth on 25 mm plasma-cleaned coverslips and dye labeling and washings , cells were fixed in 4% PFA in PBS for 20 min at 37°C , washed with PBS and then imaged in PBS with 0 . 01% ( w/v ) NaN3 on the same day . All PALM images were acquired at room temperature . All analyses presented contain data from at least 20 cells imaged in at least three independent experiments conducted on different days . For two-color dSTORM experiments , cell preparation was similar to PALM . After overnight growth on 25 mm plasma-cleaned coverslips and dye labeling and washings , cells were fixed in 4% PFA in PBS for 20 min at 37°C and washed with PBS . We then added 100 nm fluorescent Tetraspeck beads ( diluted 1:1000 in PBS; T7279 ThermoFisher Scientific ) , allowed the beads to settle and washed three times with PBS . The coverslips were then stored in PBS with 0 . 01% ( w/v ) NaN3 until imaged later on the same day . C59 Halo-mCTCF / mRad21-SNAPf mouse ES cells were labeled with 500 nM Halo-JF646 and 500 nM cp-JF549 . mES cells stably expressing H2B-SNAPf were transfected with a plasmid encoding Halo ( only; without being fused to anything ) and a GFP-NLS protein used for nuclear demarcation . These cells were similarly labeled . Just before imaging , a STORM imaging buffer ( very similar to [Boettiger et al . , 2016] ) was made by mixing 400 μL 50 mM NaCl , 200 mM Tris pH 7 . 9 with 150 μL 50% glucose solution ( w/v ) , 15 μL GLOX solution , 7 . 5 μL COT solution and 50 μL MEA solution . The GLOX solution was made by mixing 100 μL 50 mM NaCl , 200 mM Tris pH 7 . 9 with 7 mg Glucose Oxidase ( Sigma-Aldrich ) and 25 μL catalase ( 16 mg/mL ) . This solution was made the day before imaging . COT solution was made by dissolving 20 . 8 mg of Cyclooctatetraene ( Sigma-Aldrich 138924–1g ) in 1 mL DMSO . COT solution aliquots were stored at −20°C and a fresh aliquot used each time . MEA solution was made by dissolving 77 mg cysteamine ( Sigma-Aldrich ) in 1 mL water . A few drops of 1 M HCl were added to dissolve the cysteamine . STORM imaging buffer was added to the coverslip with fixed cells , the imaging chamber sealed with parafilm and then immediately loaded on the microscope . Both JF549 and JF646 could be converted into a rapidly blinking state in STORM buffer upon high-intensity laser illumination . For each cell , we exposed cells to high-power 405 nm , 561 nm and 633 nm excitation for ~5–10 s . We then acquired 50 , 000 frames of simultaneous two-color images with constant low-intensity 405 nm excitation and high-intensity 561 nm and 633 nm excitation using 25 ms exposure time on both EM-CCD cameras ( Andor iXon Ultra 897 ) . Before imaging , we aligned the two cameras using fluorescent beads ( 100 nm TetraSpeck beads; T7279 ThermoFisher Scientific ) to a registration offset below 50 nm . Before imaging each cell , we imaged a cell-adjacent bead . Similarly , after imaging each cell we also imaged a different cell-adjacent bead ( 1000 frames at 25 ms each time ) . We then used the mean offset from the bead measurements before and after imaging a cell for two-color registration for that cell . We estimate a chromatic shift registration error of ~10 nm . The pair cross correlation data presented are from around ~12–18 cells measured on 3 different days . All PALM and dSTORM experiments on fixed cells were conducted at room temperature to minimize drift . All single-molecule imaging data were processed using a custom-written MATLAB implementation of the MTT algorithm ( Sergé et al . , 2008 ) . A GUI of this implementation , SLIMfast ( Normanno et al . , 2015 ) , is available at https://elifesciences . org/content/5/e22280/supp-material1 ( Teves et al . , 2016 ) . Briefly , single molecules are localized using bi-dimensional Gaussian fitting ( approximating the microscope PSF ) subject to a generalized log-likelihood ratio test with a ‘localization error’ threshold ( in the range of 10−6-10−7 ) , with the option of allowing deflation to detect molecules partially obscured by others . Tracking , that is connecting localizations between consecutive frames , was limited by setting a maximal expected diffusion constant , and takes the trajectory history into account as well as allowing for gaps due to blinking or missed localizations . For analysis of ‘slow-tracking’ experiments , to measure residence times , the following algorithm parameters were used: Localization error: 10−7; deflation loops: 1; Blinking ( frames ) : 2; maximum number of competitors: 1; maximal expected diffusion constant ( μm2/s ) : 0 . 1 . For analysis of ‘fast-tracking’ stroboscopic paSMT experiments at ~225 Hz , the following algorithm parameters were used: Localization error: 10-6 . 25; deflation loops: 0; Blinking ( frames ) : 1; maximum number of competitors: 3; maximal expected diffusion constant ( μm2/s ) : 20 . For analysis of PALM experiments , the following algorithm parameters were used: Localization error: 10−6; deflation loops: 0; Blinking ( frames ) : 1; maximum number of competitors: 3; maximal expected diffusion constant ( μm2/s ) : 0 . 05 . For analysis of dSTORM experiments , we used the same algorithm parameters as for PALM analysis for both color channels . All subsequent analyses of trajectories were performed using custom-written code in MATLAB as described in detail in the following sections . To extract kinetic information from fast stroboscopic paSMT at approximately 225 Hz , we developed and fit a mathematical model to the jump length or displacement distributions . Our approach is largely inspired by an elegant modeling approach previously introduced by Mazza et al . ( Mazza et al . , 2012 ) , but with a number of significant differences and modifications that we will highlight below . The evolution over time of a concentration of particles located at the origin as a Dirac delta function and which follows free diffusion in two dimensions with a diffusion constant D can be described by a propagator ( also known as Green’s function ) . Properly normalized , the probability of a particle starting at the origin ending up at a location r= ( x , y ) after a time delay , Δτ , is then given by:P ( r , Δτ ) =Nr2DΔτe−r24DΔτ Here , N is a normalization constant with units of length . In practice , we compare this distribution to binned data . Thus , in practice , we integrate this distribution over a small histogram bin window , Δr , to obtain a normalized distribution to compare to the empirically measured distribution . For simplicity , we therefore leave out this normalization constant of subsequent expressions . Furthermore , in practice , we are unable to determine the precise localization of a single molecule . Instead , it is associated with a certain localization error , σ , which under our stroboscopic paSMT conditions is approximately 35 nm . Correcting for localization errors is important because it will otherwise appear as if molecules move further between frames than they actually did . Thus , we obtain the following expression for the jump length distribution taking localization error , σ , into account ( Matsuoka et al . , 2009 ) :P ( r , Δτ ) =r2 ( DΔτ+σ2 ) e−r24 ( DΔτ+σ2 ) DNA-binding molecules such as CTCF can generally exist in either a bound or a freely diffusing state . The bound state exhibits very short jump lengths ( presumably due to slow chromatin diffusion ) and has an associated diffusion constant , DBOUND , whereas the freely diffusing population tends to exhibit much longer jump lengths and has its own associated diffusion constant , DFREE . Next , we assume that binding to chromatin and unbinding from chromatin are both first-order processes with rate constants kON∗ and kOFF . We denote kON∗ with a ‘*' because it is really a pseudo first-order process since it depends on the concentration of free binding sites: kON∗=[BSFREE]kON . Thus , the steady-state jump length distribution of a population of molecules that can exist in either their bound or free state is then given by:P ( r , Δτ ) =FBOUNDr2 ( DBOUNDΔτ+σ2 ) e−r24 ( DBOUNDΔτ+σ2 ) + ( 1−FBOUND ) r2 ( DFREEΔτ+σ2 ) e−r24 ( DFREEΔτ+σ2 ) where FBOUND is the fraction of the population that is bound to chromatin and , FFREE=1−FBOUND , is the fraction of the population that is exhibiting free 3D diffusion . These fractions are related to the first-order rate constants:FBOUND=kON∗kON∗+kOFFFFREE= ( 1−FBOUND ) =kOFFkON∗+kOFF These expressions assume that molecules do not change between their bound and free states during the time delay between frames , Δτ . Previous studies have derived analytical expressions to account for this ( Mazza et al . , 2012; Yeung et al . , 2007 ) . However , implementing these expressions numerically greatly slows down fitting the model to the raw jump length distributions . Accounting for state-changes between the free and bound states was necessary in the previous study by Mazza et al . ( 2012 ) because relatively long exposure times ( 40 ms or 25 Hz ) and lag times , Δτ , ( up to 800 ms ) were considered . In this study , we are imaging at a much higher frame-rate ( 4 . 4477 ms exposure or ~225 Hz ) and only consider much shorter lag times , Δτ , ( up to seven jumps , i . e . 31 . 5 ms ) . Thus , in our case , the probability of observing a state-change is much lower . Moreover , the residence time of CTCF ( ~60–75 s ) is much longer than the residence time of p53 ( ~1 . 8 s ) ( Mazza et al . , 2012 ) . Thus , we can calculate the probability that a bound CTCF molecule unbinds during the longest lag times considered ( Δτ = 31 . 5 ms ) as:PSWITCH=1−e−kOFFΔτ≈7⋅10−5 Thus , accounting for state changes during the lag time , Δτ , makes a negligible difference for CTCF . Even if we consider short-lived non-specific interactions , the probability of a state-change is still negligible with our short lag times . Single-molecule tracking ( SMT ) is heavily biased toward bound molecules and against freely diffusing molecules for two major reasons . First , almost all single-molecule localization algorithms , including the MTT-algorithm ( Sergé et al . , 2008 ) used here , achieve sub-diffraction limit resolution ( super-resolution ) by treating individual fluorophores as point-source emitters , which generate blurred images that are described by the Point-Spread Function ( PSF ) of the microscope . Two-dimensional Gaussian modeling of the PSF allows extraction of the particle centroid with sub-pixel resolution . In SMT experiments , this works well for bound molecules , which exhibit negligible movement during the laser exposure time . However , fast moving molecules will tend to ‘motion-blur’ because they can move several pixels during the long exposure times typically used in SMT experiments . ‘Motion-blurred’ particles will thus spread their photons over multiple pixels in the direction of their movement . Therefore , they tend to be missed by most PSF-fitting localization algorithms , which results in a large bias toward bound molecules and a general bias against fast-moving molecules . This means that the bound fraction will be overestimated . To minimize this bias against fast-moving molecules , we use stroboscopic illumination where although we have a time delay of Δτ = 4 . 4477 ms , we only laser-illuminate the sample for 1 ms per frame . For a molecule like CTCF where the freely diffusing population has an apparent DFREE ~2 . 5 μm2/s , we can calculate the fraction of the population which moves more than a certain length during the 1 ms laser illumination time . Using our imaging setup ( pixel size: 160 nm ) , less than ~0 . 0036% ( ~3 . 6 molecules per 100 , 000 molecules ) of the free CTCF population move more than two pixels during the 1 ms laser exposure time . Thus , while we cannot eliminate all bias against moving molecules , our fast stroboscopic SMT methods greatly reduce bias against fast-moving molecules compared to previous approaches . Second , fast-moving molecules are likely to move out of the focal plane or axial detection window ( Δz ) during 2D image acquisition . Even though we consider short lag times Δτ ~4 . 5–31 . 5 ms , this is still long enough for a large fraction of the free population to be lost . As a consequence , bound molecules tend to have much longer trajectories than do free molecules . Again , this means that we are oversampling the bound population and undersampling the free population . To correct for this , we consider the probability that a freely diffusing molecule with diffusion constant , DFREE , will move out of the axial detection window , Δz , during a lag time , Δτ . This problem has also been previously considered by Kues and Kubitscheck ( Kues and Kubitscheck , 2002 ) . If we consider the extreme case of a population of molecules equally distributed one-dimensionally along an axis , z , with an absorbing boundary at ZMAX=ΔZ/2 and ZMIN=−ΔZ/2 , the fraction of molecules remaining at lag time , Δτ , is given by:PLEFT ( Δτ ) =1Δz∫−Δz/2Δz/2{1−∑n=0∞ ( −1 ) n[erfc ( ( 2n+1 ) Δz2−z4DFREEΔτ ) +erfc ( ( 2n+1 ) Δz2+z4DFREEΔτ ) ]}dz However , this expression significantly overestimates how many freely diffusing molecules are lost since it assumes absorbing boundaries – any molecules that comes into contact with the boundary at ± Δz/2 are permanently lost . In reality , there is a significant probability that a molecule , which has briefly contacted or exceeded the boundary , re-enters the axial detection window , Δz , during a lag time , Δτ . Moreover , since we allow trajectory gaps of one during in our tracking algorithm ( i . e . a molecule present in frame n and n+2 can still be tracked even if it was not localized in frame n+1 ) , we must consider the probability that a lost molecule re-enters the axial detection window during twice the lag time , 2Δτ . This results in the somewhat counter-intuitive effect , which was also noted by Kues and Kubitscheck , that the decay rate depends on the microscope frame rate – in other words , the fraction lost depends on how often one ‘looks’ . One approach ( Mazza et al . , 2012 ) of accounting for this is to use a corrected axial detection window larger than the true axial detection window: ΔzCORR>Δz . To find the corrected axial detection window , we first measured the true empirical axial detection window , Δz . We labeled C59 Halo-mCTCF mouse embryonic stem cells and C32 Halo-hCTCF human U2OS cells grown on plasma-cleaned 25 mm #1 . 5 cover glasses with JF646 at a low enough density to clearly observe single molecules and fixed them in 4% PFA in PBS for 20 min . We then collected an extensive z-stack throughout the nucleus with a range of 6 μm and a step size of 20 nm ( 301 frames ) and imaged single molecules at a signal-to-background ratio comparable to the one used during our fast 225 Hz paSMT experiments . We tracked molecules using the MTT algorithm ( Sergé et al . , 2008 ) and the same parameters used for our paSMT experiments . We then analyzed the survival curve , corrected for photobleaching , of single JF646-labeled Halo-CTCF molecules as a function of the step size and found the axial detection window to be approximately Δz ≈ 700 nm and highly similar in U2OS and mES cells under HiLo-illumination ( Tokunaga et al . , 2008 ) . Next , we performed Monte Carlo simulations following the Euler-Maruyama scheme . For a given diffusion constant , D , we randomly distributed 50 , 000 molecules one-dimensionally along the z-axis from ZMIN=−Δz/2 = −350 nm to ZMAX=Δz/2 = 350 nm , where Δz ≈ 700 nm . Next , using a time-step of Δτ = 4 . 4477 ms , we simulated one-dimensional Brownian diffusion along the z-axis by randomly picking Gaussian-distributed numbers from a normal distribution with parameters: μ=0; σ=2DΔτ using the function normrnd in MATLAB . For time gaps from 1 Δτ to 15 Δτ , we then calculated the fraction of molecules that were lost , allowing for one missing frame as in our tracking algorithm . We repeated these simulations for particles with diffusion constants in the range of D = 1 μm2/s to D = 12 μm2/s to generate a comprehensive dataset over a range of biologically plausible diffusion constants . We then performed least-squares fitting of this dataset to the equation for PLEFT ( Δτ ) using a corrected ΔzCORR:ΔzCORR=Δz+aD+b The simulated data were well fit using this corrected axial detection window , and we found the following best-first parameters: a = 0 . 15716 s-1/2; b = 0 . 20811 μm . Practically , we evaluated the equation for PLEFT ( Δτ ) using numerical integration in MATLAB and aborted the infinite sum once the absolute value of another iteration fell below 10−12 . We performed non-linear least-squares fitting in MATLAB by stochastically generating random parameter guesses for a and b as a starting point for the least-squares fitting routine lsqcurvefit and iterating using multiple random input guesses to avoid local minima . Having derived an analytical expression for the probability of a free molecule being lost due to axial diffusion during the imaging time , we can now thus write down the final equations used for fitting the raw jump length distributions:P ( r , Δτ ) =FBOUNDr2 ( DBOUNDΔτ+σ2 ) e−r24 ( DBOUNDΔτ+σ2 ) +ZCORR ( Δτ ) ( 1−FBOUND ) r2 ( DFREEΔτ+σ2 ) e−r24 ( DFREEΔτ+σ2 ) where:ZCORR ( Δτ ) =1Δz∫−Δz/2Δz/2{1−∑n=0∞ ( −1 ) n[erfc ( ( 2n+1 ) Δz2−z4DFREEΔτ ) +erfc ( ( 2n+1 ) Δz2+z4DFREEΔτ ) ]}dz and:Δz=0 . 700 μm+ 0 . 15716 s−1/2D+0 . 20811 μm In practical terms , we consider the jump length or displacement distributions for timepoints 1 to 8 , corresponding to seven jumps with delays from 1Δτ to 7Δτ ( i . e . this includes 6 jumps of 1Δτ , 5 jumps of 2Δτ , and so on ) . Thus , the probability of seeing a free molecule present in the first frame is higher in the second frame than in the seventh frame according the ZCORR equation above . While we have many trajectories that are much longer than eight localizations , we refrain from using the entire trajectories since almost all very long trajectories ( e . g . >100 localizations ) are highly biased toward bound molecules . While the above ZCORR equation should in principle correct for this , at long time lags the probability of still seeing a moving molecule approaches zero and thus small errors in the ZCORR equation , which is an approximation , is likely to strongly affect the estimation of the bound fraction . We note that a question arises of whether to use the entire trajectory or not . One bias against moving molecules is that frequently , freely diffusing molecules will translocate through the axial detection window , Δz , yielding only a single detectable localization and thus no jumps to be counted . Conversely , one bias against bound molecules , is that moving molecules can re-enter the axial detection window multiple times resulting in the same molecule appearing as multiple distinct trajectories and thus being over-counted . Clearly , the extent of the bias will depend on the photobleaching rate – in the limit of no photobleaching , a single freely diffusing molecule could yield a very high number of different trajectories , leading to large over-counting of the free population . However , in practice , under our stroboscopic paSMT conditions , the average dye lifetime is quite short . We note that dye disappearance is both due to photobleaching and blinking , but note that blinking should not affect estimates of the fraction bound . The actual mean number of frames depends on the fraction bound and diffusion constant – proteins with slow diffusion constants and a high bound fraction stay in the axial detection volume for longer and thus yield longer trajectories . Accordingly , for Halo-mCTCF , the mean number of frames per trajectory is ~3–4 , whereas for Halo-3xNLS it is less than two , even though the photobleaching rate is the same . We took two approaches to test whether the fraction of the trajectory that is included in the modeling would strongly affect the fraction bound estimate: analysis of our raw data and Monte Carlo simulations according to the Euler-Maruyama scheme . First , in the case of our raw data , the difference between using only the first seven jumps and using the entire trajectory only affects the fraction bound estimate by a few percentage points , suggesting that it makes a minor difference under conditions where photobleaching and blinking results in relatively short trajectories . Second , we performed Monte Carlo simulations following the Euler-Maruyama scheme and with the following assumptions: 50% of molecules are bound and the free diffusion constant is 2 . 5 μm2/s; the axial detection volume is 700 nm and the laser excitation beam under highly inclined and laminated optical sheet illumination ( HiLo ) illuminates ~4 μm ( Tokunaga et al . , 2008 ) , corresponding to half the nucleus ( nuclear diameter: 8 μm ) ; molecules within the HiLo sheet photobleach with a constant rate ( thus molecules can photobleach outside of the detection slice as in our experiments ) ; the 2D localization error is 35 nm and the timestep is 4 . 5 ms; since the vast majority of trajectories lasts no more than tens of milliseconds , but both the CTCF unbinding rate ( ~1 min ) and re-binding rate ( ~1 min ) are much slower , we ignore changes in state ( bound vs . free ) during the trajectory lifetime; Brownian motion was simulated for 500 , 000 trajectories in three dimensions enclosed within the nucleus by picking random numbers in each dimension from a normal distribution defined as: N∼ ( 0 , 2DΔτ ) . Our simulations showed that our paSMT modeling approach could accurately infer both the free diffusion constant ( slight overestimate of D , but error less than 5% ) and the fraction bound and that using the entire trajectory leads to a very small overestimate of the bound fraction ( one percentage point ) and that using the first seven jumps only leads a small underestimate of the bound fraction ( ~3 percentage points ) under conditions where the mean trajectory length ( ~3 ) was similar to the mean trajectory length for Halo-mCTCF in mESCs under our experimental conditions . However , under conditions with negligible photobleaching and extremely long trajectories of a mean length of ~100 frames , using only the first seven jumps leads to a serious underestimate of the bound fraction . We note that it is not experimentally realistic to obtain trajectories of this length with currently available dyes and microscope modalities and thus not relevant in this case , but we nevertheless note that generalizing the approach to trajectories of any length is an interesting future direction . Finally , because of the numerous other biases against free molecules noted above , we only use the first seven jumps and ignore all subsequent jumps in longer trajectories for our model fitting in this case . We then fit the above equation for , P ( r , Δτ ) , to the raw jump lengths distributions for time gaps of 1Δτ to 7Δτ corresponding to 4 . 5 ms to 31 . 5 ms . Although we show the fit function to the probability density , that is histograms ( Figure 3A–E ) , since this is more intuitive , this introduces binning artifacts ( bin: 10 nm ) . Thus , for quantitative analysis , we instead fit the model to the cumulative distribution function ( CDF ) calculated from the data . The model has three fit parameters , DBOUND , DFREE and FBOUND , and is fit to the combined jump length CDFs ( from 1Δτ to 7Δτ ) using least squares fitting . We constrain DBOUND to a range of [0 . 0005 , 0 . 08] μm2/s , but note that slight errors in the estimation of the localization error would make it appear as if the bound molecules move faster or slower than they actually do . FBOUND is of course constrained to a range of [0 , 1] and we only constrain DFREE to be greater than 0 . 15 μm2/s . We randomly generated initial parameter guesses for DBOUND , DFREE and FBOUND and then fit the model to the seven CDFs through non-linear least squares minimization implemented in MATLAB through the function lsqcurvefit . We then repeat this for multiple iterations of random initial parameter guesses and record the best-fit parameters . Thus , from the kinetic modeling , we obtain DBOUND , DFREE and FBOUND , from which we can also calculate FFREE=1−FBOUND . We note that although the previous study on p53 by Mazza et al . ( 2012 ) required two freely diffusive states and one bound state to fit the jump length distributions , in our case a single free diffusion state and one bound state were sufficient to accurately fit the raw jump length distributions . Thus , we did not consider the possibility of additional diffusive states . Next , we sought to further extend our knowledge of the nuclear target search mechanism in vivo using the parameters inferred from our kinetic modeling of the fast paSMT data as well as our residence time measurements . First , we illustrate the approach using CTCF as an example . We will continue with the steady-state two-state model ( bound or free ) introduced above , but further distinguish specific and non-specific binding . From the kinetic model fitting above , we determine the total bound fractions for CTCF to be: mESC C59 Halo-mCTCF , 68 . 0 ± 3 . 3%; mESC C87 Halo-mCTCF , 68 . 4 ± 2 . 1%; U2OS C32 Halo-hCTCF , 58 . 9 ± 2 . 0% . However , this total bound fraction contains both CTCF molecules bound specifically to their cognate binding sites and non-specific interactions . For example , sliding on DNA would be indistinguishable from stable binding to a cognate site under our paSMT conditions ( localization error ~35 nm ) . We estimate the fraction that is non-specifically bound using a mutant CTCF , 11ZF-mut-Halo-mCTCF , where we have introduced mutations into the DNA-binding domain . This mutant contains a His-to-Arg mutation in each of the 11 zinc-finger domains . Since the mutant , by design , is unable to interact specifically with chromatin through its zinc-finger domains , we reason that this mutant interacts only non-specifically . From our kinetic model fitting of the 11ZF-mut-Halo-mCTCF jump length histograms , we estimate the bound fraction for this mutant to be 19 . 1 ± 4 . 1% in mouse ES cells and 17 . 7% in human U2OS cells . Thus , the specifically bound fraction can be calculated according to:FBOUND , specific=FBOUND , total−FBOUND , non−specific Using the numbers above , we then obtain the following estimates for the specifically bound fraction: mESC C59 Halo-mCTCF , 48 . 9%; mESC C87 Halo-mCTCF , 49 . 3%; U2OS C32 Halo-hCTCF , 41 . 2% . We note that this estimation is associated with definitional uncertainty as well measurement uncertainty . It is difficult to define exactly what a non-specific interaction is , but it likely involves transient binding and/or sliding on DNA . It is also difficult to define precisely for how long a molecule has to associate with DNA for that to be reasonably counted as a non-specific interaction . Nevertheless , if we operationally define non-specific interaction here as an interaction present after mutation of the DNA-binding domain , we can proceed with investigating the target search mechanism . Next , we would like to determine the average time it takes a single CTCF protein to find another specific binding site . In the following , we will use ‘s’ and ‘ns’ , as abbreviations for specific and non-specific , respectively . The pseudo-first-order rate constant for specific binding sites , kON , s∗ , is related to the fraction bound by:FBOUND , S=kON , s∗kON , s∗+kOFF , s∗⟺kON , s∗=FBOUND , skOFF , s1−FBOUND , s We determined the off-rate for a specific interaction in our residence time measurements ( Figure 2 ) . Thus , from the previously determined values of FBOUND , s and kOFF , s , we can calculate kON , s∗ . kON , s∗ is an interesting constant because it is directly related to the average search time for a specific CTCF-binding site:τsearch , s=1kON , s∗=1−FBOUND , sFBOUND , skOFF , s When we plug in the previously determined values of FBOUND , s and kOFF , s , we thus obtain total search times of: mESC C59 Halo-mCTCF , ~65 . 9 s; mESC C87 Halo-mCTCF , ~62 . 6 s; U2OS C32 Halo-hCTCF , ~102 . 8 s . We note that the search times depend sensitively on kOFF , s , such that if a CTCF residence time of ~4 min is used instead , the search time also increases to around 4 min in mES cells and to ~5 . 7 min in U2OS cells . Regardless of the total search time , CTCF molecules spend roughly 50% of their time searching for binding sites in mES cells and roughly 60% of their time searching for binding sites in human U2OS cells . This search time contains intermittent periods of free 3D diffusion interrupted by brief non-specific binding or sliding interactions on chromatin . For example , for mESC C59 Halo-mCTCF , 51 . 1% of the total time is spent searching - 19 . 1% of the total time is spent in 1D sliding on DNA or transient interactions and 32 . 0% of the total time is spent on free 3D diffusion . Since we know the average search time to be ~65 . 9 s , we can thus calculate that during this average search time , ~41 . 3 s are spent in free 3D diffusion and ~24 . 6 s are spent in non-specific DNA interactions such as sliding . Thus , for mESC C59 Halo-mCTCF roughly 37% of the total search time is spent in non-specific DNA interactions and roughly 63% of the time is spent on free 3D diffusion . Similar analysis of C32 Halo-hCTCF in human cells show that 58 . 8% of the total time is spent searching , with 17 . 7% of the total time in non-specific chromatin association ( e . g . 1D sliding ) and 41 . 1% of the total time in free 3D diffusion . Thus , with an average search time of ~102 . 8 s , human Halo-hCTCF spends on average ~30 . 9 s on non-specific chromatin association and ~71 . 9 s on free 3D diffusion . We can apply the same approach to cohesin as measured by following mRad21 in mES cells . We note that the above approach assumes a single bound state and a single free state . This is certainly too simplistic in S/G2 , since our FRAP experiments suggest that the chromatin residence time of cohesin involved in sister chromatid cohesion is likely much longer than the cohesin involved in chromatin looping . Moreover , it is far from clear that the ON-rate , that is topological loading of cohesin onto chromatin , would be similar for cohesin involved in chromatin looping and in sister chromatid cohesion . Thus , we restrict our analysis to G1 . Even then , we stress that this analysis assumes that all topologically engaged G1 cohesin has the same ON- and OFF-rates . We estimated the G1 cohesin residence time to be 19 . 51 min ( C45 mRad21-Halo ) and 24 . 16 min ( C59 mRad21-SNAPf ) . In the following , we will use the mean: 21 . 8 min . Using stroboscopic paSMT , we estimated the G1 total fraction bound of cohesin to be 53 . 5 ± 4 . 1% and the non-specifically bound fraction to be 13 . 7 ± 3 . 1% using a mutant ( F601R , L605R , Q617K ) that is reported to be unable to form cohesin complexes ( Haering et al . , 2004 ) . Thus , 39 . 8% of cohesin is topologically bound to chromatin , 13 . 7% non-specifically associated with chromatin and 46 . 5% in free 3D diffusion in G1-phase of the cell cycle . Non-specific chromatin association may include non-productive topological loading attempts . This yields a search time of ~33 . 0 min of which around 7 . 51 min is spent on non-specific chromatin association ( e . g . sliding ) and 25 . 49 min is spent on free 3D diffusion . We note that this description of the cohesin search mechanism is somewhat simplified since assisted topological loading is a bit more complicated than finding a cognate-binding site for a typical sequence-specific transcription factor . Rather , it is likely that the cohesin search mechanism is regulated by other protein interaction partners and by post-translational modifications ( Skibbens , 2016 ) . Nevertheless , even if topological loading involves multiple steps , the process can be described as a single first-order reaction if there is a single rate-limiting step . To extract residence times from SMT data recoded at long exposure time , we took a hybrid approach related to that of Chen et al . ( 2014 ) and Mazza et al . ( 2012 ) . Briefly , we took advantage of long exposure times ( 300 ms , 500 ms or 800 ms ) as previously described ( Chen et al . , 2014 ) : this causes freely-diffusing molecules to motion-blur into the background such that they are generally missed by our detection algorithm ( Sergé et al . , 2008 ) . We then recorded the trajectory length of each ‘bound’ molecule and used these to generate a survival curve ( 1-CDF ) . However , as previously reported there are multiple contributions to this survival curve beyond specific binding , which is what we are interested in , such as non-specific binding ( Chen et al . , 2014 ) and slow-diffusing molecules ( Mazza et al . , 2012 ) . Beyond these two , localization errors can cause both false-positive and false-negative detections . False negative detections especially occur for molecules close to being out-of-focus . This can cause a single long trajectory to appear as many short ones . Thus , we performed double-exponential fitting ( corresponding to specific and non-specific binding ) using:P ( t ) =Ae−knst+Be−kst where kns corresponds to the unbinding rate for non-specific binding and ks corresponds to the unbinding rate constant for specific binding . We note that the first rate constant , kns , is likely to be contaminated by localization errors ( e . g . from molecules close to being out-of-focus ) and experimental noise and we therefore caution against over interpreting it . To filter out contributions from tracking errors and slow-diffusing molecules , we applied an objective threshold as previously described to consider only particles tracked for at least Nmin frames ( Mazza et al . , 2012 ) . To determine Nmin , we plotted the inferred residence time as a function of Nmin and observed convergence to a single value after ~2 . 5 s ( i . e . 8 frames at 300 ms exposure time , 5 frames at 500 ms exposure time , 3 frames at 800 ms exposure time; Figure 2—figure supplement 1A ) . We thus used this threshold to determine the value of ks . The measured ks , however , reflects both unbinding from chromatin as well as photobleaching etc . :ks=ks , true+kbias Photobleaching clearly needs to be corrected for . But several other factors also contributed faster apparent unbinding . Among these were axial cell drift , lateral cell drift , fluctuating background and others . Axial cell drift can cause a single molecule to move gradually out-of-focus , which appears as unbinding . We also observe significant lateral cell drift , especially for mES cells due to cell movement , which can appear as unbinding if particle movement exceeds the threshold . Drift is especially an issue for molecules exhibiting relatively stable binding such as CTCF , where we occasionally , but very rarely , observe single molecules for around 10 min under constant laser illumination . To correct for all these factors including photobleaching , we reasoned that , if we assume that all these processes are Poisson processes , then the sum of independent Poissons is also a Poisson . If we further assume that these processes will affect H2B-Halo to the same extent as CTCF ( i . e . photobleaching depends only on the dye used and the laser intensity; axial chromatin or cell drift is the same for Halo-CTCF cells as for H2B-Halo cells ) , then we can measure an apparent unbinding rate for H2B-Halo and use this as kbias . This analysis assumes that any apparent unbinding of H2B will be due to photobleaching or drift etc . , which is consistent with our FRAP data . However , we note that although H2B molecules are no doubt occasionally evicted from chromatin ( e . g . during chromatin remodeling ) , as long as the rate is much smaller than the unbinding rate of CTCF , this makes a negligible contribution . Thus , to estimate kbias , we repeated the experiments on mES or U2OS cells stably expressing H2B-Halo and estimated kbias as the slow component from double-exponential fitting as described above . We always performed the H2B-Halo control experiment on the same day as the other experiments . Having measured kbias , we then calculated the residence time asτs=1ks , true We note that the above analysis assumes that the unbinding rate for all CTCF sites is identical , which is clearly an approximation , although the ability of the model to fit the data suggests it is a reasonable approximation . However , this analysis would miss a very small CTCF fraction ( <3% ) showing different residence times . So the above-calculated residence time should be interpreted as an average residence time , which holds for most CTCF sites , but may not hold for all . We extracted single-molecule x , y coordinates from single-color PALM images using the following pipeline . We took advantage of the high photostability of the PA-JF549 and PA-JF646 dyes to increase localization accuracy and to perform drift correction . Similar fiducial marker independent drift-correction algorithms have been described previously ( Elmokadem and Yu , 2015; Wang et al . , 2014 ) . At the laser intensity used and an exposure time of 25 ms , each JF549/JF646 molecule lasted ~5–10 frames on average before photobleaching . Thus , after localizing molecules in each frame and tracking them between frames , we obtaining several estimates of the true x , y coordinates for each molecule , which improves the localization precision . Moreover , since each frame contained 5–10 molecules on average this allowed us to perform drift correction by tracking the average drift of particles over time after binning to average out noise in individual localizations . For spatial clustering analysis , we segmented the nucleus by convolving the PSF with the single-molecule localizations and then blurring the image using iterative Gaussian smoothing followed by thresholding or by manual polygon segmentation . We then divided the nucleus into partially overlapping 3 μm squares and performed clustering analysis on these squares using a recently reported Bayesian algorithm ( Rubin-Delanchy et al . , 2015 ) . We used the same prior as published ( Rubin-Delanchy et al . , 2015 ) and performed cluster identification and characterized clusters according to their cluster radius and fraction of molecules in clusters as described ( Rubin-Delanchy et al . , 2015 ) . A major concern in clustering analysis of PALM images is photo-blinking , where a dye turns off for some frames and the re-appears . Since we track single molecules across frames and allow for gaps of 1 frame , most molecules that exhibit multiple appearances will be collapsed into a single localization . However , it is not possible to unambiguously distinguish two different co-localizing molecules that appear many frames apart , from a single molecule that exhibits a long photo-blink . Thus , to investigate to what extent the apparent clustering that we observe is due to uncorrected photo-blinking we took the following approaches . First , we compared our Halo-CTCF and Rad21-Halo PALM reconstructions to H2B-Halo and Halo-3xNLS . While there is no known protein whose nuclear organization perfectly exhibits complete spatial randomness , we reasoned that Halo-3xNLS should exhibit a relatively uniform distribution . Thus , by using the same dye and imaging conditions as for CTCF and Rad21 , we treat the level of clustering observed for Halo-3xNLS as being largely due to blinking , and thus generate a ‘blinking floor’ . Since both CTCF and Rad21 exhibits much higher clustering than Halo-3xNLS ( Figure 4—figure supplement 1C ) , we conclude that most of the observed clustering is not due to photo-blinking . We note that both the H2B-Halo and Halo-3xNLS transgenes are expressed at very high levels . Thus , we empirically adjusted the PA-JF549 concentration so as to get similar numbers of localizations as for CTCF , so as to exclude any bias coming from the number of molecules . Second , in mES cells CTCF and H2B exhibit comparable levels of clustering , but Ripley’s L ( r ) -r curves are qualitatively different , with H2B showing clustering at larger length scales . This further suggests that our PALM approach is measuring real clustering and that the relatively small clusters observed for CTCF and Rad21 are not merely photo-blinking artefacts . Third , we performed two-color labeling and imaging to unambiguously distinguish true clusters from photo-blinking . We labeled Halo-hCTCF in C32 U2OS cells with approximately equimolar concentrations of PA-JF549 and PA-JF646 dyes and performed two-color PALM . Since each Halo-Tag can only bind one dye , any cluster composed of N molecules should under ideal circumstances exhibit a binomial distribution of JF549 and JF646 molecules . That is , the probability of a cluster composed of N CTCF molecules having k JF549-conjugated CTCF molecules should follow:P ( XJF549=k ) = ( Nk ) pJF549k ( 1−pJF549 ) N−k wherepJF549=NJF549NJF549+NJF646 is the fraction of all dye-labeled nuclear CTCF molecules that was labeled with JF549 . Conversely , consider the other extreme case where all clusters are exclusively due to photo-blinking artifacts . In this extreme scenario , all apparent clusters should be exclusively composed of JF549-conjugated CTCF molecules or exclusively composed of JF646-conjugated CTCF molecules . If we plot the probability density function for the fraction of JF549-labeled molecules in clusters , the idealized case should show a binomial distribution with a peak at pJF549 . On the other hand , the extreme ‘photo-blinking only’ case should show a probability density function for the fraction of JF549-labeled molecules in clusters with peaks at 0 and 1 and nothing in between , corresponding to exclusive JF549 and exclusive JF646 clusters . Thus , to apply this analysis , we merged all JF549 and JF646 localizations and applied the Bayesian cluster identification algorithm to the merged dataset . We then analyzed all the called clusters that were composed of at least 10 detections . We consider only these clusters since for very small clusters the probability of finding clusters exclusively in one color is significant even in the ideal binomial case . In a given nucleus , hundreds of clusters fulfilled this criterion ( >10 detections ) . To robustly compare this to the ideal binomial case , for each cluster of size N , we generated binomial random clusters using binornd in MATLAB . Finally , we compared the distribution of cluster compositions for the observed clusters and the binomial random clusters in Figure 4—figure supplement 1E . Since each nucleus had a slightly different fraction of molecules labeled with JF549 and JF646 , we only show the distribution for a single nucleus . As can be seen , the deviation from the binomial case is small . Essentially , all clusters at this size contain molecules of both colors demonstrating that clustering is not exclusively a photo-blinking artifact . Thus , although some clustering is clearly due to photo-blinking , the majority of clusters are composed of multiple distinct molecules . To summarize the results for multiple cells , we also calculated the Kullback-Leibler divergence between the expected binomial and observed distributions for each cell . The mean Kullback-Leibler divergence was ~0 . 3 bits further demonstrating that most clusters are not a photo-blinking artifact . Finally , we note that a recent paper demonstrates that PA-JF549 shows limited photo-blinking ( Grimm et al . , 2016 ) . We processed two-color dSTORM data essentially identically to PALM data . After chromatic registration , blinking-correction and drift-correction using the same approach as for PALM analysis , nuclei were manually segmented using polygon segmentation based on a rough image generated by convolving the PSF with the single-molecule localizations and then blurring the image . We note that SNAP-tag dye-labeling is somewhat less specific than HaloTag labeling ( Figure 1—figure supplement 1 ) – in particular , when we label wild-type cells that do not express a SNAP-tag protein with cp-JF549 ( or any other SNAP dye ) we observe enrichment along the nuclear envelope that does not disappear even after extensive washings . Labeling inside the nucleus , however , appears to be specific with cp-JF549 , but less so with SNAP-TMR ( compare Figure 1—figure supplement 1B and C ) . To avoid this affecting our dSTORM analysis , we segmented out the nuclear envelope during segmentation of the nucleus . Images ( such as Figure 4A ) were generated by binning single-molecule localizations into square pixel-bins of 10 nm and then false-color rendering JF549 localizations in green and JF646 localizations in magenta , such that saturating co-localization appears white . We note that co-localization of two single molecules are therefore not visible in these rendered images . Only overlap of clusters with saturating brightness appear white . Thus , most co-localizing CTCF and Rad21 molecules are not visible in Figure 4A . Thus , as a much more quantitative analysis we performed pair cross correlation analysis . Like pair correlation analysis , which quantifies the spatial interaction of proteins with themselves ( i . e . clustering ) , pair cross correlation analysis quantifies spatial interactions between two different proteins . Thus , C ( r ) quantifies enrichment between two different proteins as a function of interparticle distance , r . When the two proteins are independent ( Complete Spatial Randomness ( CSR ) ) , C ( r ) =1 for all r . We calculate C ( r ) using the whole nucleus and edge-correction as previously described ( Stone and Veatch , 2015 ) using bins of 10 nm . The main way in which pair cross correlation can cause false-positive pair cross correlation is through fluorophore bleedthrough during simultaneous two-color imaging . E . g . if 561 nm excited J549 molecules emit enough far-red photons to be detected in the JF646 channel , this would result in high , but false-positive , pair cross correlation at small r . To rule out bleedthrough and any other bias , we also imaged a mES cell line stably expressing H2B-SNAPf transfected with a plasmid encoding a free Halo protein . We expect no significant co-localization between these proteins beyond mild exclusion from certain nuclear regions ( e . g . nucleolar regions ) . In agreement , their experimentally observed pair cross correlation was not significantly different from CSR at any r . Since these cells were imaged under the same conditions as C59 Halo-mCTCF/mRad21-SNAPf , this rules out the possibility that the observed pair cross correlation at small r between CTCF and cohesin is due to fluorophore bleedthrough or any other technical artifact . Antibodies were as follows: ChromPure rabbit and mouse normal IgG from Jackson ImmunoResearch ( West Grove , PA ) ; anti-CTCF for Western Blot ( WB ) from Millipore ( Temecula , CA ) ( EMD 07–729 ) , for ChIP and Co-IP from Abcam ( ab128873 ) ; anti-Rad21 for WB and ChIP from Abcam ( Cambridge , MA ) ( ab154769 ) , for CoIP from Millipore ( EMD 05–908 ) ; anti-SMC1 and anti-SMC3 from Bethyl ( Montgomery , TX ) ( A300-055A , A300-060A ) ; anti-FLAG from Sigma-Aldrich ( F7425 ) ; anti-TBP , anti-H3 , and anti-V5 from Abcam ( ab51841 , ab1791 , ab9116 ) . ChIP assays in wild-type and double CTCF/Rad21 knock-in ( clone C59 ) mouse JM8 . N4 mES cells were performed essentially as described ( Testa et al . , 2005 ) with minor modifications . Cells were cross-linked for 5 min at room temperature with 1% formaldehyde-containing medium; cross-linking was stopped by PBS-glycine ( 0 . 125 M final ) . Cells were washed twice with ice-cold PBS , scraped , centrifuged for 10 min at 4000 rpm , resuspended in cell lysis buffer ( 5 mM PIPES , pH 8 . 0 , 85 mM KCl , and 0 . 5% NP-40 , 1 ml/15 cm plate ) and incubated for 10 min on ice . During the incubation , the lysates were repeatedly pipetted up and down every 5 min . Lysates were then centrifuged for 10 min at 4000 rpm . Nuclear pellets were resuspended in six volumes of sonication buffer ( 50 mM Tris-HCl , pH 8 . 1 , 10 mM EDTA , 0 . 1% SDS ) , incubated on ice for 10 min , and sonicated to obtain DNA fragments below 2000 bp in length ( Covaris ( Woburn , MA ) S220 sonicator , 20% Duty factor , 200 cycles/burst , 100 peak incident power , 50 cycles of 30’ on and 30’ off ) . Sonicated lysates were cleared by centrifugation and 400–1600 μg of chromatin was diluted in RIPA buffer ( 10 mM Tris-HCl , pH 8 . 0 , 1 mM EDTA , 0 . 5 mM EGTA , 1% Triton X-100 , 0 . 1% SDS , 0 . 1% Na-deoxycholate , 140 mM NaCl ) to a final concentration of 0 . 8 μg/μl , precleared with Protein A sepharose ( GE Healthcare , Pittsburgh , PA ) for 2 hr at 4°C and immunoprecipitated overnight with 8–16 μg of normal rabbit IgGs , anti-Rad21 or anti-CTCF antibodies . About 15% of the precleared chromatin was saved as input . Immunoprecipitated DNA was purified with the Qiagen ( Germantown , MD ) QIAquick PCR Purification Kit , eluted in 60 μl of water and analyzed by qPCR together with 2% of the input chromatin prior to ChIP-seq library preparation ( SYBR Select Master Mix for CFX , ThermoFisher , see Supplementary file 2 for primer sequences ) . ChIP-seq libraries were prepared independently from two ChIP biological replicates using the Illumina ( San Diego , CA ) TruSeq DNA sample preparation kit according to manufacturer instructions with few modifications . We used 100 ng of ChIP input DNA ( as measured by Fragment analyzer ) and 50 μl of immunoprecipitated DNA as a starting material; Illumina adapters were diluted 1:50 , and library samples were enriched through 18 cycles of PCR amplification . We assessed library quality and fragment size by qPCR and Fragment analyzer , and when necessary we performed an additional size selection step on agarose gel after PCR amplification to enrich for fragments between 150 and 500 bp . We sequenced four to eight multiplexed libraries per lane on the Illumina HiSeq4000 sequencing platform ( single end-reads , 50 bp long ) at the Vincent J . Coates Genomics Sequencing Laboratory at UC Berkeley , supported by NIH S10 OD018174 Instrumentation Grant . Input , IgG , Rad21 and CTCF ChIP-seq raw reads from wild type and knock-in ESCs from two biological replicates ( 18 libraries total , see Supplementary file 1 ) were quality-checked with FastQC and aligned onto the mouse genome ( mm10 assembly ) using Bowtie ( Langmead et al . , 2009 ) , allowing for two mismatches ( -n 2 ) and no multiple alignments ( -m 1 ) . Enriched regions were visualized on the mm10 genome with the Integrative Genomics Viewer ( IGV ) ( Robinson et al . , 2011; Thorvaldsdóttir et al . , 2013 ) , after creating tiled data files from alignment files ( igvtools count -w 50 -e 200 ) . Peaks were called with MACS2 ( --nomodel --extsize 250 ) ( Zhang et al . , 2008 ) combining inputs from the two replicates as a control , first for each biological replicate separately , and then , after having verified that results were highly reproducible , for the merged replicates ( Supplementary file 1 ) . Coverage and overlap between ChIP-seq peaks across samples and with previously published CTCF and Rad21 datasets were computed through Galaxy ( Blankenberg et al . , 2010; Giardine et al . , 2005; Goecks et al . , 2010 ) , requiring a minimum 1 bp overlap between peak intervals ( Supplementary file 1 ) . To create heatmaps , we used deepTools ( version 2 . 4 . 1 ) ( Ramírez et al . , 2016 ) . We first ran bamCoverage ( --binSize 50 --normalizeTo1 × 2150570000 extendReads 250 --ignoreDuplicates -of bigwig ) and normalized read numbers of WT and C59 IgG , CTCF and Rad21 merged replicates to 1x sequencing depth , obtaining read coverage per 50 bp bins across the whole genome ( bigWig files ) . We then used the bigWig files to compute read numbers across 6 kb centered on either WT CTCF or WT Rad21 peak summits as called by MACS2 ( computeMatrix reference-point --referencePoint=TSS --upstream 3000 --downstream 3000 --missingDataAsZero --sortRegions=no ) . We sorted the output matrices by decreasing WT enrichment , calculated as the total number of reads within a MACS2 called ChIP-seq peak . Finally , heatmaps were created with the plotHeatmap tool ( --averageTypeSummaryPlot=mean --colorMap='Blues' --sortRegions=no ) . Total RNA was purified from cell pellets using RNeasy Plus Mini kit ( Qiagen ) and quantified by Nanodrop . For RT-qPCR , 1 μg of total RNA was retrotranscribed to cDNA with oligo ( dT ) primers ( Ambion , Life Technologies , ThermoFisher ) and Superscript III ( Invitrogen , ThermoFisher ) . 2 μl of 1:40 cDNA dilutions were used for quantitative PCR ( qPCR ) with SYBR Select Master Mix for CFX ( Applied Biosystems , ThermoFisher ) on a BIO-RAD CFX Real-time PCR system ( see Supplementary file 2 for primer sequences ) . Cells were collected by scraping from plates in ice-cold phosphate-buffered saline ( PBS ) , pelleted , and flash-frozen in liquid nitrogen . For Western blot analysis , cell pellets where thawed on ice , resuspended to 1 mL/10 cm plate of low-salt lysis buffer ( 0 . 1 M NaCl , 25 mM HEPES , 1 mM MgCl2 , 0 . 2 mM EDTA , 0 . 5% NP-40 and protease inhibitors ) , with 125 U/mL of benzonase ( Novagen , EMD Millipore ) , passed through a 25G needle , rocked at 4°C for 1 hr and a NaCl solution was added to reach a final concentration of 0 . 2 M . Lysates were then rocked at 4°C for 30 min and centrifuged at maximum speed at 4°C . Supernatants were quantified by Bradford . Between 15 and 60 μg of proteins were loaded onto 9% Bis-Tris SDS-PAGE gel , transferred onto nitrocellulose membrane ( Amershan Protran 0 . 45 um NC , GE Healthcare ) for 2 hr at 100V , blocked in TBS-Tween with 10% milk for at least 1 hr at room temperature and blotted overnight at 4°C with primary antibodies in TBS-T with 5% milk . HRP-conjugated secondary antibodies were diluted 1:5000 in TBS-T with 5% milk and incubated at room temperature for an hour . For Co-IP experiments , cell pellets where thawed on ice , resuspended to 1 ml/10 cm plate of cell lysis buffer ( 5 mM PIPES pH 8 . 0 , 85 mM KCl , 0 . 5% NP-40 and protease inhibitors ) , and incubated on ice for 10 min . Nuclei were pelleted in a tabletop centrifuge at 4°C , at 4000 rpm for 10 min , and resuspended to 0 . 5 mL/10 cm plate of low salt lysis buffer with benzonase as above . For each sample , 1 mg of proteins was diluted in 1 mL of Co-IP buffer ( 0 . 2 M NaCl , 25 mM Hepes , 1 mM MgCl2 , 0 . 2 mM EDTA , 0 . 5% NP-40 and protease inhibitors ) , pre-cleared for 2 hr at 4°C with protein-G-sepharose beads ( GE Healthcare Life Sciences ) before overnight immunoprecipitation with 4 μg of either normal serum IgGs or specific antibodies as listed above . Some pre-cleared lysate was kept at 4°C overnight as input . Protein-G-sepharose beads precleared overnight in CoIP buffer with 0 . 5% BSA were then added to the samples and incubated at 4°C for 2 hr . After extensive washes in Co-IP buffer , proteins were eluted from the beads by boiling for 5 min in 2X SDS-loading buffer and analyzed by SDS-PAGE and Western blot . The ChIP-seq data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus ( Edgar et al . , 2002 ) and are accessible through GEO Series accession number GSE90994 . We compared our ChIP-seq to previous ChIP-Seq studies of Rad21 and CTCF: ( Handoko et al . , 2011; Nitzsche et al . , 2011; Shen et al . , 2012 ) and GSE29218 . FRAP was performed on an inverted Zeiss ( Germany ) LSM 710 AxioObserver confocal microscope equipped with a motorized stage , a full incubation chamber maintaining 37°C/5% CO2 , a heated stage , an X-Cite 120 illumination source as well as several laser lines ( only the 561 nm laser was used here ) . Images were acquired on a 40x Plan NeoFluar NA1 . 3 oil-immersion objective at a zoom corresponding to a 100 nm x 100 nm pixel size and the microscope controlled using the Zeiss Zen software . In most FRAP experiments , except where otherwise noted , 300 frames were acquired at either one frame per second allowing 20 frames to be acquired before the bleach pulse to accurately estimate baseline fluorescence or 330 frames at one frame per 2 s again allowing 20 frames to be acquired before the bleach pulse . A circular bleach spot ( r = 10 pixels ) was chosen in a region of homogenous fluorescence at a position at least 1 μm from nuclear or nucleolar boundaries . The spot was bleached using maximal laser intensity and pixel dwell time corresponding to a total bleach time of ~1 s . We note that because the bleach duration was relatively long compared to the timescale of molecular diffusion , it is not possible to accurately estimate the bound and free fractions from our FRAP curves . We generally collected data from 6 to 10 cells per cell line per condition per day , and all presented data are from at least three independent replicates on different days . To quantify and drift-correct the FRAP movies ( cell movement is an issue , especially for mES cells ) , we custom-wrote a pipeline in MATLAB . Briefly , we manually identify the bleach spot . The nucleus is automatically identified by thresholding images after Gaussian smoothing and hole-filling ( to avoid the bleach spot as being identified as not belonging to the nucleus ) . We use an exponentially decaying ( from 100% to ~85% of initial over one movie ) threshold to account for whole-nucleus photobleaching during the time-lapse acquisition . Next , we quantify the bleach spot signal as the mean intensity of a slightly smaller circle ( r = 0 . 6 μm ) , which is more robust to lateral drift . The FRAP signal is corrected for photobleaching using the measured reduction in total nuclear fluorescence ( ~15% over 300–330 frames at the low laser intensity used after bleaching ) and internally normalized to its mean value during the 20 frames before bleaching . We correct for drift by manually updating a drift vector quantifying cell movement during the experiment . Finally , drift- and photobleaching corrected FRAP curves from each single cell were averaged to generate a mean FRAP recovery . We used the mean FRAP recovery in all figures and for model-fitting . Model selection is a crucial step in FRAP experiments and has been studied extensively ( Mueller et al . , 2008 , 2010; Sprague et al . , 2004 ) . A full FRAP model considers both diffusion , the shape of the bleach spot and reactions ( e . g . binding and unbinding ) . However , Sprague et al . identified circumstances under which simpler models are applicable ( Sprague et al . , 2004 ) . Importantly , minimizing the number of fitted parameters is desirable because FRAP modeling tends to otherwise be prone to overfitting . Sprague et al . showed that when:kON∗w2DFREE≪1andkOFFkON∗≲1 Then a ‘reaction dominant’ FRAP model is most appropriate ( w is the radius of the bleach spot ) . In the case of the second condition , for CTCF in both mES and U2OS cells , kOFF≈kON∗ . Likewise , for mRad21-Halo in mESCs kOFF≈kON∗ . Thus , the second condition suggests a reaction dominant model . For the first condition , we find: Halo-mCTCF in mESCs: kON∗w2DFREE=0 . 015s−1⋅ ( 0 . 6 μm ) 22 . 5 μm2s−1=0 . 0022≪1 mRad21 in mESCs ( G1 phase ) : kON∗w2DFREE=0 . 0005s−1⋅ ( 0 . 6 μm ) 21 . 5 μm2s−1=0 . 00012≪1 Thus , both CTCF and Rad21 lie within the reaction dominant parameter space and a reaction-dominant FRAP model is therefore the most appropriate choice . As has been demonstrated previously ( Sprague et al . , 2004 ) , in the reaction-dominant parameter range , the FRAP recovery depends only on kOFF and we fit the FRAP recoveries to the reaction-dominant model below:FRAP ( t ) =1−Ae−kat−Be−kbt After model-fitting ( Figure 2—figure supplements 2D and 3C ) , we used the slower off rate to estimate the residence time according to τs=1koff . In FRAP modeling , an important question is whether or not it is justifiable to ignore diffusion ( as the above model does ) and the radial shape of the bleach spot . Mueller et al . previously showed that ignoring diffusion can lead to serious errors for typical transcription factors which show rapid FRAP recovery ( in the seconds to tens of seconds range ) ( Mueller et al . , 2008 ) . To test whether diffusion must be taken into account we plotted the radial shape of the bleach spot as a function of time . In general , if recovery is due to binding , the recovery should be mostly uniform across the bleach area , since all binding sites are equally likely to be sampled . If on the other hand diffusion dominates the recovery , the outer edges of the circle will recover first and the center of the circle last , since unbleached molecules are diffusing in from the outside . As can be seen ( Figure 2—figure supplement 3E ) , the radial profile of the bleach spot is flat and thus diffusion can be ignored in the FRAP modeling . We note that in previous studies on typical transcription factors , complete or near-complete FRAP recovery was generally observed in the 10–20 s range and here diffusion is critical ( Mazza et al . , 2012; Mueller et al . , 2008; Sprague et al . , 2004 ) . But in the case of CTCF and cohesin , FRAP recovery is about two orders of magnitude slower , and thus , it is not surprising that diffusion can be ignored . Finally , Mueller et al . modeled the shape of the bleach spot as a Gaussian ( Mueller et al . , 2008 ) , but showed that if the flat part of the bleach spot is used instead , equivalent results are obtained . Thus , in our case , we bleach a circle with a 1 μm radius but use a circle with a 0 . 6 μm radius to calculate the FRAP recovery , which is in the uniform area of the radial bleach profile . In addition to being equivalent to the full Gaussian description of the radial bleach profile , it has the advantage of being much more robust to cell drift , which is extensive for mES cells over the 11 min that most of our FRAP experiments last . Finally , it came to our attention that during extended FRAP experiments ( in the multi hour range ) , incomplete washout of Halo- or SNAP-dye can lead to artifactual FRAP recovery ( Rhodes et al . , 2017 ) . This is most likely through dye binding to new protein produced after the bleach pulse . This can be corrected for by adding an excess of ‘dark’ Halo- or SNAP-ligand , such that any newly synthesized protein binds the dark ligand . However , this is unlikely to contribute significantly to FRAP recoveries on the minute timescale since we estimate that only around 1% of the total protein is replenished during our longest FRAP experiments . Consistently , we could not detect a difference in FRAP recovery after adding excess dark ligand ( Figure 2—figure supplement 3F ) . We conclude that our FRAP experiments were not affected by this .	A human cell contains about 2 meters of DNA tightly packed in a compartment called the nucleus . Within the space inside the nucleus , different parts of the DNA fold into distinct bundles known as domains . These domains are important for organising the genome and are crucial for regulating gene expression , by stimulating specific DNA segments to activate certain genes . Previous research has shown that DNA segments within the same domain frequently interact , whereas DNA segments in different domains rarely do . The domains are often folded into loops that are held together by a ring-shaped protein complex called cohesin , while another protein called CTCF positions cohesin and thereby sets the boundaries between the domains . Some mutations are known to disrupt these boundaries , which allows certain DNA segments to activate the wrong genes . This can lead to cancer or cause defects when embryos are developing . However , we do not currently understand how these domains are formed or maintained . In particular , it was unclear whether these loop domains are stable or dynamic structures . Hansen et al . addressed these questions in embryonic stem cells from mice and human cancer cells . It was found that cohesin and CTCF form a complex that binds to the DNA and likely holds the loops together . In further experiments , single molecules of cohesin and CTCF were tracked inside cells using super-resolution microscopy . The results showed that CTCF and cohesin bind to DNA with different dynamics: CTCF binds the DNA for about a minute , whereas cohesin binds the DNA for about 20–25 minutes . Once CTCF detaches from DNA , it quickly rebinds DNA at another site , but cohesin takes much longer . These observations suggest that rather than remaining static , chromatin domains are held together by a dynamic protein complex , with a molecular composition that exchanges over time . This results suggests that DNA loop domains , which were generally assumed to be very stable anchor points , are in fact highly dynamic structures that frequently fall apart and reform . The next challenge will be to understand how the dynamic nature of these loop domains contribute to gene regulation . This may , one day , enable us to manipulate the domains to correct faulty folding of DNA in cancer and other diseases .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "structural", "biology", "and", "molecular", "biophysics" ]	2017	CTCF and cohesin regulate chromatin loop stability with distinct dynamics
Exhaustion of stem cells is a hallmark of aging . In the Drosophila testis , dedifferentiated germline stem cells ( GSCs ) derived from spermatogonia increase during lifespan , leading to the model that dedifferentiation counteracts the decline of GSCs in aged males . To test this , we blocked dedifferentiation by mis-expressing the differentiation factor bag of marbles ( bam ) in spermatogonia while lineage-labeling these cells . Strikingly , blocking bam-lineage dedifferentiation under normal conditions in virgin males has no impact on the GSC pool . However , in mated males or challenging conditions , inhibiting bam-lineage dedifferentiation markedly reduces the number of GSCs and their ability to proliferate and differentiate . We find that bam-lineage derived GSCs have significantly higher proliferation rates than sibling GSCs in the same testis . We determined that Jun N-terminal kinase ( JNK ) activity is autonomously required for bam-lineage dedifferentiation . Overall , we show that dedifferentiation provides a mechanism to maintain the germline and ensure fertility under chronically stressful conditions . A robust stem cell pool is critical to the maintenance of highly proliferative tissues during an organism’s lifetime . Adult tissue stem cells typically reside in niches , anatomically defined as microenvironments that support their ‘stemness’ and promote their proliferation , resulting in some daughter cells that retain stem cell characteristics and some that begin differentiation ( Morrison and Spradling , 2008; Blanpain et al . , 2014; Spradling et al . , 2011 ) . Depletion or reduction of stem cells results in tissue atrophy , and the exhaustion of stem cell function is a hallmark of aging ( López-Otín et al . , 2013; Wang and Jones , 2011 ) . Thus , the mechanisms that maintain stem cells during an individual’s lifespan are of critical importance to understanding the relationship between stem cells and aging and to develop therapies against aging-related clinical conditions like infertility and Parkinson’s . The stem cell pool is dynamic and responds to insults , including injury and starvation in both invertebrate and mammalian model organisms ( Angelo and Van Gilst , 2009; Tetteh et al . , 2016; van Es et al . , 2012; Yang and Yamashita , 2015; McLeod et al . , 2010; Li and Jasper , 2016 ) . Recently , dedifferentiation has emerged as a conserved mechanism underlying the replenishment of the stem cell pool after stem cell depletion ( Merrell and Stanger , 2016 ) . In the mouse intestine , extensive radiation ablates Lgr5-positive crypt stem cells and lineage-tracing experiments revealed that secretory cells dedifferentiate into Lgr5-positive stem cells following this insult ( van Es et al . , 2012 ) . In the Drosophila intestine , complete starvation induces the loss of all intestinal stem cells , and polyploid enterocyte cells undergo a reduction in ploidity ( called amitosis ) and transform into intestinal stem cells ( Lucchetta and Ohlstein , 2017 ) . In Drosophila gonads , after forced differentiation of all germline stem cells ( GSCs ) , differentiating spermatogonia revert to the stem cell state and become functional GSCs ( Brawley and Matunis , 2004; Kai and Spradling , 2004; Sheng et al . , 2009 ) . While these previous studies showed that dedifferentiation indeed occurs after acute insults or injuries , they did not address its functional significance in these events . Here , we test the functional importance of dedifferentiation through a new genetic approach . We have developed a genetic technique to indelibly mark the cells undergoing dedifferentiation , while at the same time functionally inhibiting the process . We used the Drosophila testis for these studies because of the powerful genetic techniques available in this organism and the broad knowledge about the biology of this organ and its various cell types . In this tissue , approximately 8–14 GSCs reside in a quiescent niche ( Greenspan et al . , 2015 ) . GSCs adhere to niche cells and undergo oriented mitosis , resulting in one daughter cell that retains the stem cell state and remains in contact with the niche ( Figure 1A ) . The other GSC daughter cell ( the gonialblast ) is physically displaced from the niche . After encapsulation by somatic support cells , this latter daughter cell begins differentiation through four rounds of mitotic divisions with incomplete cytokinesis , resulting in 2- , 4- , 8- and 16-cell spermatogonial cysts , the lattermost of which undergoes meiosis to produce 64 spermatids . At the 4- and 8-cell cyst stage , germ cells express bag of marbles ( bam ) , which is necessary and sufficient for their differentiation ( Sheng et al . , 2009; Gönczy et al . , 1997 ) . The testis niche also supports a somatic stem cell population called cyst stem cells ( CySCs ) that produces somatic support cells , which exit the cell cycle and ensheath differentiating GSC daughter cells . During aging , the population of GSCs declines such that at 50 days of adulthood ~35% of GSCs are lost from the niche and the remaining GSCs have reduced proliferation ( Boyle et al . , 2007; Wallenfang et al . , 2006 ) . The 35% reduction in the GSC pool in aged males is much smaller than predicted . The average half-life of a GSC is 14 days , and for a testis with 10 GSCs at day 0 of adulthood , there should be <1 GSC at 50 days ( Boyle et al . , 2007; Wallenfang et al . , 2006 ) . In other words , the reduction in the total GSC pool should be more than 90% at 50 days . This discrepancy in predicted vs observed size of the GSC pool raised the possibility that a mechanism such as spermatogonial dedifferentiation could be responsible for the apparent resistance of the GSC pool to the deleterious effects of aging ( Wang and Jones , 2011; Wallenfang et al . , 2006; Cheng et al . , 2008 ) . However , to date no study has tested this hypothesis by specifically inhibiting dedifferentiation in spermatogonia . Certain genetic manipulations ( transient removal of responses to niche signals or transient mis-expression of the key differentiation factor bam ) cause all GSCs to differentiate . However , upon silencing of these triggers , spermatogonia break apart , migrate back to the niche , outcompete the resident CySCs and become functional GSCs by transducing JAK/STAT signals and repressing bam expression ( Brawley and Matunis , 2004; Sheng et al . , 2009; Sheng and Matunis , 2011 ) . Interestingly , these studies revealed that the 8-cell spermatogonial cyst is the oldest stage still competent to dedifferentiate . bam-lineage labeling analysis of 4- and 8-cell spermatogonial cysts revealed that the proportion of dedifferentiated cells in the GSC pool increases with aging; in 50 day old males , ~40% of the GSCs are derived from bam-lineage spermatogonia that dedifferentiated ( Cheng et al . , 2008 ) . Here , we have developed a methodology that enabled us to inhibit specifically dedifferentiation of bam-expressing , 4- and 8-cell spermatogonial cysts without apparent side effects , while at the same time lineage-tracing these cells . This has allowed us to address the long-term effects of dedifferentiation . Surprisingly and contrary to predictions , we find that bam-lineage dedifferentiation is not required to maintain the GSC pool during aging under normal laboratory conditions . However , it is critical to maintain a robust GSC pool under chronically stressful conditions . Our methodology also allows the identification of dedifferentiated GSCs from their non-dedifferentiated siblings , facilitating the comparison of their characteristics . We find that bam-lineage dedifferentiated GSCs have a higher proliferative rate compared to their sibling GSCs in the same testis . Finally , we show that Jun N-terminal kinase ( JNK ) signaling is activated in germ cells during recovery from stress . By inhibiting JNK activity in bam-expressing spermatogonia , we demonstrate that this pathway is essential for dedifferentiation of these cells . In order to study the contribution of dedifferentiation to the maintenance of the GSC pool , we lineage-traced spermatogonial cells . Similar to a previous study ( Cheng et al . , 2008 ) , we used a bam-Gal4 line , expressed specifically in 4- and 8-cell spermatogonial cysts , to drive UAS-Flippase ( Flp ) expression , and this Flp in turn excises an FRT-stop-FRT from the ubiP63E-FRT-stop-FRT-GFP cassette ( Evans et al . , 2009 ) . After recombination , GFP becomes an indelible marker of differentiating germ cells that had expressed bam-Gal4 , and GFP persists even if the cell turns off the bam promoter . Since transient germline mis-expression of bam is sufficient to induce germline differentiation ( Sheng et al . , 2009; Ohlstein and McKearin , 1997 ) , we speculated that mis-expression of additional Bam protein in these bam-Gal4-positive spermatogonia ( referred to as bam > bam ) should prevent them from undergoing dedifferentiation ( Figure 1—figure supplement 1 ) . As a control and to maintain similar titration of the Gal4 protein , we mis-expressed a neutral construct UAS-LacZ by bam-Gal4 in a second set of flies ( referred to as bam > LacZ ) . For reasons unknown to us but also observed by another group ( Cheng et al . , 2008 ) , some somatic support cells are labeled for real-time and lineage bam expression ( Figure 1B and Figure 1—figure supplement 2C ) . We note that this methodology will likely not label all dedifferentiating germ cells , as it has been speculated that gonialblasts and 2-cell spermatogonia can revert to become GSCs . We also note that the efficiency of Flp is not 100% , and so we are not labeling all bam-lineage cells . In this study , we refer to GFP-positive germ cells as ‘bam-lineage positive’ and ‘dedifferentiated’ , and GFP-negative germ cells as ‘bam-lineage negative’ and ‘wild type siblings’ . We first analyzed the role of dedifferentiation in control bam > LacZ and experimental bam > bam males during standard aging conditions , that is , maintaining flies at a low density , in the absence of females and on standard food . In control flies , we observed a significant increase in the percentage of GFP-positive GSCs derived from dedifferentiated bam-lineage spermatogonia , from 7 . 0% in young flies to 21 . 6% in 45-day-old males ( Figure 1B , C , F ) , consistent with a prior study ( Cheng et al . , 2008 ) . As expected , GFP-positive germ cells lose bam expression in the process of reverting to a GSC identity ( Figure 1—figure supplement 2A–B’’ ) . However , bam mis-expression in the bam lineage effectively blocked lineage dedifferentiation , as there was no significant increase in the percentage of lineage-dedifferentiated GSCs in aged males , from 3 . 3% in young flies to 3 . 6% in 45-day-old flies ( Figure 1D , E , F ) . These results demonstrate that bam mis-expression is an effective way to prevent dedifferentiation . Prior work has shown that the number of GSCs decreases slightly during aging under normal laboratory conditions , and this has led to the model that dedifferentiation provides a means to offset normal GSC loss during lifespan ( Wang and Jones , 2011; Wallenfang et al . , 2006; Cheng et al . , 2008 ) . If this hypothesis is correct , we would expect a further reduction in GSC number after bam mis-expression by bam-Gal4 . Because the starting number of GSCs varies from strain to strain , we compared the relative number of GSCs between bam > LacZ and bam > bam flies . To our surprise , we found that the relative number of GSCs decreases significantly by 10% after 45 days in both genotypes ( from 10 . 4 to 9 . 4 cells in bam > LacZ , and from 7 . 9 to 7 . 1 cells in bam > bam ) ( Figure 1G and Supplementary file 1 ) . Furthermore , we did not observe any notable differences between aged bam > LacZ and aged bam > bam testes , as all stages of spermatogenesis appeared similar between the two genotypes ( see Figure 4A–B and G–H ) . We note , however , that due to the fact that some of the bam-expressing cells might not undergo recombination ( due to incomplete Flp efficiency ) , we cannot rule out the possibility that unlabeled cells may dedifferentiate and help to maintain the GSC pool under normal laboratory conditions . Additionally , prior work has shown that symmetric renewal , whereby a gonialblast swivels to gain direct access to the niche , occurs at low levels in testes from wild type males ( Sheng and Matunis , 2011 ) . It is possible that by blocking bam-lineage dedifferentiation , we are shifting the equilibrium towards symmetric renewal , but live imaging will be needed to test this hypothesis . Nevertheless , although unexpected and contrary to our predictions , our results strongly suggest that dedifferentiation of the bam-lineage does not play an important role in maintaining the GSC pool during aging under normal laboratory conditions . We speculated that dedifferentiation could be vital during challenging life conditions , such as starvation . Previous work has shown that 6–15 days of protein deprivation ( also generally referred to in the literature as ‘starvation’ ) causes a 25% reduction in total GSC number ( Yang and Yamashita , 2015; McLeod et al . , 2010 ) . Furthermore , the number of GSCs recovered to original levels after 5 days of refeeding on standard food ( McLeod et al . , 2010 ) . We hypothesized that the recovery of GSCs during refeeding could result from dedifferentiating spermatogonia . To test this model , we subjected bam > LacZ flies to 15 days of protein starvation , followed by a 9-day time course of refeeding ( Figure 2A ) . We assessed the proportion of dedifferentiated GSCs at the end of the starvation period and at different time points during the refeeding phase . Similar to these prior reports , we observed a 33% decrease in the relative number of GSCs after starvation ( 10 . 4 at 0 days and 7 . 0 after 15 days of starvation ) and a full recovery of the GSC pool after 5 days of refeeding ( Figure 2B , B’ and Supplementary file 1 ) . In accordance with previous studies , we found that after the 15-day protein starvation period , the percentage of dedifferentiated GSCs did not increase compared to day 0 ( compare Figure 2A–G and [Yang and Yamashita , 2015; McLeod et al . , 2010] ) . However , after the 5 day refeeding period , this proportion significantly increased from 4 . 2% at day 0 after refeeding to 19 . 7% at day five after refeeding and did not increase further after longer refeeding times ( Figure 2A and Supplementary file 1 ) . We note that this increase in dedifferentiation during the refeeding period strongly correlates with the recovery in the size of the GSC pool ( Figure 2B’ , gray dashed line ) . Among the bam > LacZ controls , we observed variable numbers of bam-lineage dedifferentiated GSCs per testis . Most testes contained both GFP-negative wild type and GFP-positive dedifferentiated GSCs at the niche , however , some testes did not contain any dedifferentiated GSCs and some testes contained all dedifferentiated GSCs . We hypothesized that testes with more dedifferentiated GSCs had a faster recovery of the GSC pool after starvation and refeeding . Comparing testes with at least one dedifferentiated GSC ( Figure 2B’ , light blue line ) to those with 0 dedifferentiated GSCs ( Figure 2B’ , dark blue line ) , we found that the former fully recovered the pool of GSCs after 3 days , while the latter required significantly longer , up to 7 days . Although we cannot exclude other variables such as germ cell death , GSC loss , and GSC gain through symmetric renewal , this correlation strongly suggests that bam-lineage dedifferentiation accelerates the recovery of the GSC pool under challenging conditions . Thus , we hypothesized that preventing dedifferentiation using bam > bam flies would mimic this delay in recovery of the GSC pool observed in control flies lacking dedifferentiation . In bam > bam testes , the proportion of dedifferentiated GSCs did not increase ( 0% at day 0 after refeeding and 1 . 8% at day 5 after refeeding ) , in contrast to the bam > LacZ controls ( Figure 2C–G and Supplementary file 1 ) . Indeed , we observed that blocking dedifferentiation ( i . e . , bam > bam ) retards the recovery of the GSC pool in a similar manner to control bam > LacZ testes devoid of bam-lineage dedifferentiation ( Figure 2H , compare red to dark blue line ) . Compared to bam > LacZ testes with dedifferentiated GSCs , both bam > bam testes and bam > LacZ testes lacking dedifferentiated GSCs were significantly delayed in the recovery of the GSC pool ( Figure 2H , compare red and dark blue lines to the light blue line ) . Centrosome mis-orientation in GSCs increases during aging and after irradiation ( Cheng et al . , 2008 ) . We found that both dedifferentiated GSCs and their ‘wild type’ siblings displayed equally high rates ( ~35–40% ) of centrosome mis-orientation after starvation and refeeding ( Figure 2—figure supplement 1 , third set of bars ) , similar to maximal rate of 40% reported in a previous study ( Cheng et al . , 2008 ) . These results suggest that centrosome mis-orientation was not specific to bam-lineage dedifferentiated GSCs . Taken together , these results suggest that dedifferentiation of the bam-lineage promotes recovery of the GSC pool after challenging conditions . Our results suggest a functional role of dedifferentiation in maintaining a robust GSC pool during transient insults or challenging conditions . We further speculated that dedifferentiation could have additional biological roles when such conditions become chronic . We wondered whether males always housed with ( and presumably mating with ) females could force GSCs in the testis to cope with a higher demand for sperm production and thus accelerate the age-related reduction of the GSC pool . We note that while mating is a normal physiological event , it is stressful as mating significantly decreases both male and female lifespans ( Branco et al . , 2017; Fowler and Partridge , 1989 ) . However , little is known about the effects of mating on the GSC pool in males . To determine this , we monitored the mitotic index in unmated vs continuously mated bam > LacZ males after 15 days of adulthood . We found significantly more GSCs in M-phase in testes from mated males compared to age-matched unmated males ( Figure 3—figure supplement 1A and Supplementary file 2 ) . We next assessed if both GFP-negative wild type and GFP-positive dedifferentiated GSCs in these two conditions had similar M-phase distribution . Indeed , in mated males there were more wild type and dedifferentiated GSCs in M-phase compared to unmated males ( Figure 3—figure supplement 1B and Supplementary file 2 ) . These data suggest that there is a systemic or non-autonomous effect of mating on GSC proliferation . In a second set of experiments , we addressed whether mating increased dedifferentiation of the bam lineage . We aged bam > LacZ and bam > bam flies for 45 days in the constant presence of wild type OregonR females . The rate of dedifferentiation in control bam > LacZ testes significantly increased from 7 . 0% at 0 days to 30 . 8% at 45 days when males are housed with females ( Figure 3—figure supplement 1C and Supplementary file 1 ) . As expected , the rate of dedifferentiation upon mating does not increase in bam > bam flies ( Figure 3—figure supplement 1C and Supplementary file 1 ) . Moreover , while the relative number of GSCs is not diminish in testes from mated bam > lacZ controls , suggesting the pool is preserved , it is significantly ( 18% ) smaller in mated bam > bam flies ( Figure 3—figure supplement 1D and Supplementary file 1 ) . Therefore , bam-lineage dedifferentiation is important in maintaining the size of the GSC pool under mating conditions . This result is in contrast to that obtained in unmated conditions ( see Figure 1G ) . Since both GSC proliferation and bam-lineage dedifferentiation are significantly increased upon mating , we speculated that this would affect GSC dynamics in the niche by increasing GSC turnover . Specifically we hypothesized that neutral GSC clones should be lost more frequently in testes from mated males compared to unmated ones . To test this , we generated neutral MARCM clones that expressed GFP but were otherwise wild type and measured clone residency in the niche at 2 days post clone induction ( dpci ) to establish clone induction frequency and at 15 dpci to monitor clone retention . As expected , clones were generated at equal frequency in both unmated and mated conditions ( Figure 3—figure supplement 1E , first set of bars ) . However , at 15 dpci , there was a significant difference in clone retention with fewer clones retained in mated testes ( Figure 3—figure supplement 1E , second set of bars ) . These results are consistent with the model in which increased proliferation of GSCs and outcompetition of resident GSCs by dedifferentiating germline cells alter stem cell dynamics in the testis niche . After establishing the stressful effects of starvation and mating , we designed an aging protocol that might resemble sub-optimal conditions faced by flies in the wild , including intermittent access to food and to mating partners . This protocol consists of cycles of 6 days of protein starvation , which is sufficient to induce a significant reduction in the GSC pool ( Yang and Yamashita , 2015 ) , followed of 4 days of refeeding . In each cycle , during the last 2 days of the 4 days of refeeding , we added virgin females to the vials , which were subsequently removed for the next cycle . Our 10-day cycle of starvation and refeeding allowed us to accommodate 4 cycles of challenging conditions ( see Materials and methods ) . In control testes , after 4 cycles of starvation , refeeding and mating , we observed a significant increase in the rate of bam-lineage dedifferentiation , from 6 . 9% at day 0% to 43 . 9% at day 41 ( Figure 3A , B , E and Supplementary file 1 ) . As expected in bam > bam testes , the rate of dedifferentiation was completely blocked , 3 . 3% at day 0% to 2 . 1% at day 41 ( Figure 3C , D , E and Supplementary file 1 ) . Importantly , when we blocked dedifferentiation , the relative number of GSCs was significantly reduced by 33% compared to the controls after four cycles ( Figure 3F and Supplementary file 1 ) . These results indicate that , under these chronic challenging conditions , bam-lineage dedifferentiation is important to preserve the GSC pool . We speculated that the reduced number of GSCs in testes lacking bam-Gal4 lineage dedifferentiation might impact spermatogenesis . Indeed , after 2 and 4 cycles of starvation and refeeding , bam > bam testes were thinner and appeared to lack cells in intermediate stages of spermatogenesis ( Figure 4B , D , F ) when compared to matched bam > lacZ controls ( Figure 4A , C , E ) . This phenotype is specific to chronic challenging conditions , because testes from 45 day old , fed and unmated flies displayed no apparent differences between the two genotypes ( Figure 4G , H ) . We quantified the number of early-stage spermatogonia in the two genotypes: goniablasts ( or 1-cell ) , 2-cell , 4-cell and 8-cell cysts . At 0 days , the number of gonia at each stage is indistinguishable between bam > LacZ and bam > bam flies ( Figure 4I and Supplementary file 3 ) , while as noted above , the absolute number of GSCs was distinct between the two genotypes . However , after 4 cycles of starvation , refeeding and mating , the number of gonia was significantly reduced up to 33% at each gonial stage when dedifferentiation was blocked ( Figure 4J and Supplementary file 3 ) . The reduced number of gonia in bam > bam testes is unlikely to be a direct result of the bam mis-expression , as the bam-Gal4 line used is only active in the 4- and 8-cell stage , and bam-Gal4 activity disappears once the cells have dedifferentiated ( Figure 1—figure supplement 2 ) . Instead , we observed a decrease in all the stages of spermatogenesis , including the pre-meiotic cysts , with a low-magnification inspection ( Figure 4A–F ) . The reduced number of gonia in bam > bam testes after four cycles ( Figure 4J ) could result from the reduced GSC pool in these testes ( Figure 3F ) . However , an alternative explanation is that bam > bam testes lack the dedifferentiated GSCs found in control bam > lacZ testes and that these dedifferentiated GSCs have a higher proliferation rate than ‘wild type’ siblings and produced more gonial offspring . To assess whether bam-lineage dedifferentiated GSCs are more proliferative than their lineage-negative , wild type siblings , we directly compared the number of offspring of GFP-positive GSCs to those from GFP-negative GSCs in the same bam > lacZ control testis . We divided the number of spermatogonia at each stage by the number of labeled GSCs that were producing them . This experiment is analogous to a clonal analysis , but here we scored the contribution of an entire type of GSC rather than of a single GSC . Strikingly , we observed that after 4 cycles of starvation , dedifferentiated GSCs contribute up to 45% more offspring than their wild type siblings in the same testis ( Figure 5A and Supplementary file 3 ) . Additionally , we directly measured proliferation by scoring the proportion of GSCs positive for the S-phase marker EdU and the M-phase marker phospho-Histone3 ( pH3 ) . We found that there were significantly more bam-lineage positive GSCs in S-phase and in M-phase compared to bam-lineage negative GSCs ( Figure 5B , C ) . Centrosome mis-orientation in GSCs slows the rate of proliferation during aging ( Cheng et al . , 2008 ) . However , both bam-lineage-positive and lineage-negative GSCs had equally high rates of centrosome mis-orientation after 4 cycles of starvation and refeeding ( Figure 2—figure supplement 1 , fourth set of bars ) . These data suggest that in some challenging conditions , factors independent of centrosome orientation may regulate GSC proliferation rates . We reasoned that JNK signaling could be promoting spermatogonial dedifferentiation because of its established roles in cellular reprogramming and stress responses . An evolutionarily conserved kinase cascade , the JNK pathway plays essential roles in regeneration of numerous organs , including imaginal discs and intestine , and can trigger changes in cell identity and transdifferentiation ( Jiang et al . , 2009; Bergantiños et al . , 2010; Herrera et al . , 2013; Smith-Bolton et al . , 2009; Sun and Irvine , 2014; Herrera and Morata , 2014; Lee et al . , 2005; Gettings et al . , 2010 ) . JNK signaling is activated in a variety of stress responses; it is detected in the somatic support cells in the testis during protein starvation and contributes non-autonomously to the maintenance of the GSC pool ( Yang and Yamashita , 2015 ) . We were unable to detect real-time JNK activity reporters puckered ( puc ) -LacZ ( Martín-Blanco et al . , 1998 ) or TRE-GFP ( Chatterjee and Bohmann , 2012 ) in the germline during or after starvation ( data not shown ) . Since the frequency of dedifferentiation is on average 2 . 2 GSCs per testis during 5 days of refeeding ( Supplementary file 1 ) , the chances for detecting this JNK activation are low , especially if it is transient and/or if only low levels are required . For this reason , we decided to use a more sensitive assay: lineage labeling of puc-Gal4-positive cells . puc is a transcriptional target and repressor of the pathway ( Martín-Blanco et al . , 1998 ) . Similar to the bam-Gal4 lineage labeling , Flp under the control of puc-Gal4 recombines an ubiP63E-FRT-stop-FRT-GFP cassette so that cells that have expressed puc in the past , even at low levels , will become permanently labeled by GFP ( genotype: puc > GFP ) . Two puc-Gal4 lines with independent origins were used for these experiments ( see Materials and methods ) , showing similar results . At 0 days old , testes from puc > GFP males had GFP expression in a fraction of hub cells . However , GFP expression was largely absent from somatic cyst cells and the germline , with 20% and 12% of testes , respectively , labeled at 0 days in fed conditions ( Figure 6A , F , F’ ) . Testes from puc > GFP males that were maintained under fed conditions and aged for 15 or 20 days showed a similar trend of low rates of labeling ( Figure 6B , C , F , F’ ) . After 15 days of protein starvation , 51% of puc > GFP testes displayed GFP labeling of somatic cyst cells , while GFP expression in the germline was observed in less than 11% of the cases ( Figure 6D , F , F’ ) . This increase of JNK signaling in the somatic cells after starvation is consistent with a previous study ( Yang and Yamashita , 2015 ) . After 15 days of starvation and 5 of refeeding , 82% of puc > GFP testes had GFP labeling of somatic cells , and 55% now displayed germline labeling as well , including GSCs ( Figure 6E–F’ ) . These results indicate that germline cells acquire JNK activity specifically during the refeeding phase , when the burst of dedifferentiation takes place ( Figure 2A ) . We analyzed germline labeling at 2 and 3 days post refeeding to gain insight into how spermatogonial cysts dedifferentiate ( Figure 6F ) . Because prior work has documented fragmenting of 4- and 8-cell gonia prior to dedifferentiation into GSCs ( Brawley and Matunis , 2004; Sheng et al . , 2009; Cheng et al . , 2008; Sheng and Matunis , 2011 ) , we predicted that we would observe a majority of testes with only gonia but not GSCs labeled with puc > GFP at these time points . Of the 97 testes scored at 2 and 3 days post refeeding , 17 ( 17 . 5% ) had germline labeling ( defined as any GSC , 1- , 2- , 4- , 8- and/or 16-cell gonia expressing puc > GFP ) . Of these 17 testes , most ( n = 15 ) had at least 1 GSC labeled as well as various gonia; only two testes had gonia but no GSCs labeled . We speculate that high number of cases where an entire germline lineage is labeled results from the considerable time ( >24 hr ) required to activate puc , recombine the lineage-tracing cassette and induce GFP expression . Our observations are consistent with JNK-activated gonia breaking apart and liberating gonial cells to dedifferentiate into GSCs . However , we cannot rule out the possibility that during refeeding JNK signaling is directly activated in resident GSCs that have not dedifferentiated . In support of our model that spermatogonia fragment and revert to the GSC state after autonomous JNK activation , we observed one testis at 3 days of refeeding where a 4-cell gonia labeled with puc > GFP appeared to fragment into a 3-cell gonia and a GSC at the niche ( Figure 6—figure supplement 1 ) . Each germ cell in the 3-cell gonia had a dot spectrosome , characteristic of fragmenting germ cells during dedifferentiation and very early germ cells ( GSCs and gonialblasts ) in wild type ( Cheng et al . , 2008; de Cuevas et al . , 1997 ) . To functionally test the relevance of JNK activity in spermatogonial dedifferentiation , we blocked its activity in the bam-Gal4 lineage by mis-expressing the JNK inhibitor puc or a dominant negative form of the JNK basket ( bsk ) . Concomitantly , we lineage traced bam-Gal4 cells . If JNK activity is necessary for dedifferentiation , we predict that blocking its activity would reduce dedifferentiation upon refeeding . Indeed , after one cycle of starvation and refeeding , the increase in bam-lineage GSCs was no longer detectable in bam > puc or bam > bskDN testes compared with bam > LacZ controls ( Figure 6G–K ) . We note that neither mis-expression of puc nor bskDN appears to adversely affect germ cell differentiation as robust spermatogonia and spermatocytes expressing puc or bskDN survive and continue to progress towards meiotic stages ( Figure 6—figure supplement 2 ) . Taken together , these results indicate that JNK pathway activity is autonomously necessary for germline cells to undergo dedifferentiation . While spermatogonial dedifferentiation increases during aging ( this study and ( Cheng et al . , 2008 ) ) , the biological role of this process has remained unknown . Surprisingly , through a combination of genetic methodologies , we demonstrate that dedifferentiation of the bam lineage plays no role in maintaining the GSC pool during aging under standard conditions ( abundant food and no females ) . Instead , we find that under normal but stressful conditions such as mating or under challenging conditions such as starvation and refeeding , dedifferentiation is important for both the quick recovery of the GSC pool in the short term and the preservation of the stem cell number in the long term . These results lead us to propose that bam-lineage dedifferentiation is akin to a regenerative response aimed to preserve the number of gonadal stem cells under adverse situations but is dispensable under optimal life conditions . This model is consistent with previous results demonstrating an increase in dedifferentiation after damage induced by irradiation ( Tetteh et al . , 2016; van Es et al . , 2012; Cheng et al . , 2008 ) . Furthermore , our results suggest that spermatogonial cysts can fragment and the liberated germ cells migrate back to the niche to become functional GSCs , similar what was observed upon regeneration of the germline after GSC depletion ( Figure 7 and [Brawley and Matunis , 2004; Sheng et al . , 2009] ) . Since germ cells from fragmented cysts have to compete against resident GSCs as well as resident somatic stem cells in order to re-occupy the niche , germ cells that successfully dedifferentiate must have increased competitive properties , which are currently not understood . Future work using live imaging will be needed to uncover the dynamics of this process . Our results reveal a critical role of the JNK pathway in spermatogonial dedifferentiation . We have detected its activation during the refeeding phase after starvation and have proven its requirement in promoting this phenomenon ( Figure 7 ) . As mentioned above , we show that dedifferentiation is critical to maintaining a robust GSC pool during challenging conditions and as such is similar to a regenerative response . Recently , several studies have demonstrated the importance of the JNK pathway for proliferation , for triggering other signaling pathways , and for cellular reprogramming during regenerative events in imaginal discs ( Bergantiños et al . , 2010; Herrera et al . , 2013; Smith-Bolton et al . , 2009; Sun and Irvine , 2014; Herrera and Morata , 2014; Lee et al . , 2005 ) . It is unclear what JNK signaling regulates during spermatogonial dedifferentiation , but this phenomenon involves some of the features of regeneration , particularly reprogramming which reverts the cell identity to ‘stemness’ . Taken together with our results , these studies suggest that JNK activity may be a universal feature of regenerative responses in Drosophila whether it is to reconstruct an appendage or recover a pool of stem cells . Similar to our results , two previous studies did not detect bam-lineage-traced , dedifferentiated GSC in the testis during protein starvation ( Yang and Yamashita , 2015; McLeod et al . , 2010 ) . Therefore , there is consensus in the field that germline dedifferentiation does not occur during starvation . Our work shows that dedifferentiation occurs during the refeeding period with the maximal percentage of dedifferentiated GSCs being observed only after 5 days . We note that another group did not observe dedifferentiation during the refeeding period . However , they only scored this event at 2 days after refeeding and not at later time points ( McLeod et al . , 2010 ) . We think the discrepancy between our studies is due to the different lengths of the refeeding period . One caveat of our results is the possibility that spurious activity of bam-Gal4 in GSCs could be forcing the recombination of the GFP flip-out cassette and thus marking them indistinguishably from truly dedifferentiated cells . There are , however , three arguments against this being responsible for the increase in bam-lineage labeled GSCs: ( 1 ) we failed to detect real time expression of bam-Gal4 in GSCs ( Figure 1—figure supplement 2 ) ; ( 2 ) preventing dedifferentiation ( bam > bam ) blocked the increase in labeled GSCs in every scenario we tested; and ( 3 ) under challenging conditions , the GFP-positive GSCs were demonstratively more proliferative than their unlabeled siblings , a result that is inconsistent with spurious labeling . Another caveat of our experiments is that the extent of dedifferentiation may be underscored because our labeling methodology precludes us from lineage-tracing very early germ cells; the driver we used ( bam-Gal4 ) is restricted to 4- and 8-cell spermatogonia and does not label gonialblasts ( i . e . , 1-cell ) and 2-cell spermatogonia . To the best of our knowledge , there are no Gal4 lines specific to the lattermost cell types and at the same time excluded from GSCs . This limitation makes it impossible for us to track dedifferentiation of gonialblasts and 2-cell spermatogonia and likely causes undersampling of all dedifferentiation events . We note that examples of dedifferentiation from 2-cell gonia have been documented by live imaging ( Sheng and Matunis , 2011 ) ; this study also showed that ‘symmetric renewal’ occurs when the gonialblast of a GSC-gonialblast pair swivels into the niche resulting in two stem cell offspring instead of 1 stem cell and 1 differentiating offspring . ‘Reversion’ occurs when spermatogonial cysts break apart and germ cells return to the niche to become functional GSCs ( Sheng and Matunis , 2011 ) . Both symmetric renewal and reversion could be at play in challenging conditions and may underscore why , following one cycle of starvation , the GSC pool does indeed recover albeit in a delayed fashion even when dedifferentiation is blocked ( Figure 2B’ ) . The inability to track dedifferentiation of gonialblasts and 2-cell gonia , combined with the less than 100% efficiency of Flp/FRT , may underlie the increased mis-oriented centrosomes in bam-lineage negative GSCs under a variety of conditions tested here . Although we observed a significantly higher proportion of bam-lineage GSCs with mis-oriented centrosomes compared to lineage-negative GSCs at day 0 , there was no statistically significant difference between these two populations after 45 days of ‘normal’ aging ( Figure 2—figure supplement 1 ) , consistent with a prior report ( Cheng et al . , 2008 ) . Additionally , we assessed mis-oriented centrosomes after challenging conditions like 1 cycle of starvation and refeeding and 4 cycles of challenging conditions and found no difference in bam-lineage positive vs negative GSCs ( Figure 2—figure supplement 1 ) . It is possible that the lineage-negative GSCs are in fact dedifferentiated from gonialblasts and 2 cell gonia . While we cannot experimentally test this hypothesis , if both types of GSCs were in fact derived from dedifferentiated gonia , this could account for the roughly equal rates of mis-oriented centrosomes in both pools ( Figure 2—figure supplement 1 ) . Despite similarly high rates of centrosome mis-orientation in bam-lineage dedifferentiated GSCs and lineage-negative siblings , the former proliferate faster . This result was unexpected , as it has been reported that dedifferentiated GSCs have decreased proliferation rates due to increased centrosome mis-orientation ( Cheng et al . , 2008 ) . We speculate that downstream effectors of JNK signaling could be responsible for this increased cycling . The transient activity of JNK in germline cells that dedifferentiate could elicit a change in the epigenetic landscape by modulating Polycomb-group ( Pc-g ) and trithorax-group ( trx-g ) genes , as has been shown in several regenerative contexts ( Herrera and Morata , 2014; Lee et al . , 2005; Roumengous et al . , 2017 ) . Such epigenetic changes , possibly downstream of JNK/Pc-g or JNK/trx-g , may endow dedifferentiated GSCs with increased proliferation compared to their wild type siblings . For example , potential cell reprogramming of dedifferentiated GSCs may ‘refresh’ a stem cell’s genetic landscape , compared with its wild type siblings that may have acquired genetic damage or imprinting . Future experiments will be necessary to directly test these hypotheses . The following fly stocks were obtained from the Bloomington Drosophila Stock Center ( BDSC ) : Oregon-R ( Ore ) R , Ubi-p63E ( FRT . STOP ) Stinger , UAS-RedStinger , UAS-LacZ , UAS-bskK53R , pucE69-Gal4 , puc-IT . Gal4 . Additionally , we used the following fly stocks: UAS-puc ( Martín-Blanco et al . , 1998 ) ; bam-Gal4:VP16 ( Chen and McKearin , 2003 ) ; UAS-bam:GFP ( Chen and McKearin , 2003 ) ; FRT40A and y , w , hs-Flp112; tub-Gal80 , FRT40A ( Amoyel et al . , 2016 ) . The primary antibodies used were: goat anti‐Vasa ( Santa Cruz , 1:200 ) , mouse anti-FasIII ( Developmental Studies Hybridoma Bank ( DSHB ) , 1:50 ) , mouse anti-α-Spectrin ( DSHB , 1:50 ) , mouse anti-γTubulin ( Sigma , 1:100 ) , rabbit anti-phospho-Histone3-Ser10 ( Millipore , 1:200 ) . 5‐ethynyl‐2′‐deoxyuridine ( EdU , Invitrogen ) labeling was carried out as previously described ( Amoyel et al . , 2014 ) . In all the experiments , flies were raised at 25°C . For standard aging conditions , virgin males were collected and kept isolated from females . Flies were kept in vials with food at a density of 20 males per vial ( 1-inch width ) and transferred into fresh new vials every 2 days . Flies were maintained on standard fly food . For starvation ( protein deprivation ) periods , males were transferred into vials with 10% sucrose/1% agar , replaced by fresh ones every 2 days . For aging in the presence of females ( mated males ) , a maximum of 20 males were placed in the same vial with 40 young ( no older than 1 week ) wild type ( Oregon-R ) virgin females . Flies were transferred to new vials every 2 days , and the females were replaced by new young virgins every 2 weeks . For aging in challenging conditions , we subjected flies to a regime of cycles of the following composition: 6 days of protein deprivation , followed by 2 days of refeeding in standard food and two additional days of refeeding in standard food in the presence of virgin wild type Oregon-R females in a 2:1 ratio of females to males . At the end of each cycle , females were discarded . For the last cycle , we extended the refeeding time one extra day , so the refeeding phase encompassed 5 days instead of 4 , in order to collect testes with a degree of recovery comparable to the single-cycle experiments . We scored a cell as a GSC if it met these conditions: ( 1 ) it is Vasa-positive ( Vasa is expressed only in germline cells in the testis ) ; ( 2 ) it is a single cell that is not part of a spermatogonial cyst; ( 3 ) it makes direct contact with the niche ( FasIII-positive cells ) . In some experiments , the second criterion was evaluated by assessing the presence of dot fusomes when stained with α-Spectrin , as this is a hallmark of GSCs . In the course of our experiments , we realized that distinct genotypes have a different number of GSCs at 0 days ( see Figure 4I for example ) . To compare total GSCs in each genotype over time , we normalized the total number of GSCs at each time point to that of the start ( i . e . , day 0 ) . For this reason , most of the data are shown as ‘relative GSC number’ , thus enabling direct comparisons between genotypes after the same treatment . The percentage of dedifferentiation is calculated as ( 1 ) the proportion of GFP-positive ( i . e . , bam-lineage-positive ) GSCs divided by the total number of GSCs in each individual testis and ( 2 ) each testis in a particular genotype was averaged . The number of cysts at each stage in Figures 4I , J and 5A was scored using the α-Spectrin antibody , which labels the fusome that connects all the cells in a spermatogonium . The percentage of GSCs with mis-oriented centrosomes was calculated by means of an established methodology ( Cheng et al . , 2008 ) . Images were acquired on a Zeiss LSM 510 confocal microscope . Image analysis and quantifications were performed with Fiji-ImageJ ( Schindelin et al . , 2012 ) and Adobe Photoshop software . Fisher’s exact tests were used for Figures 5B , C and 6F–F’ and Figure 3—figure supplement 1A , B , E . The rest of the statistical tests were performed with Student’s t tests . Data were analyzed with Microsoft Excel and GraphPad Prism . A summary of results , including averages and sample sizes is included in Supplementary file 1–3 . Paint3D was used in Figure 6—figure supplement 1 to illustrate the position of the confocal slice in the z-stack as well as to indicate the position of the fragmenting germ cyst .	From the heart to the brain , our bodies are made of a collection of cells that are specialized to perform precise roles . Yet , certain organs host ‘stem cells’ , which can become any kind of tissue . For example , the testicles of the fruit fly contain germline stem cells; when one of these cells divides , a daughter remains unspecialized , while the other specializes – or differentiates – to become sperm . Despite previous beliefs , a cell that is undergoing specialization can dedifferentiate to become a stem cell again . As the organism gets older , stem cells become ‘exhausted’: they divide less , and lose their ability to remain unspecialized . Scientists therefore proposed that dedifferentiation could be a way to replenish a dwindling pool of stem cells , and ward off the effects of age . However , this line of thought has not been tested in the laboratory . Here , Herrera and Bach tried to test this assumption by creating two populations of male fruit flies . One was genetically intact and the other was modified so that the cells that would become sperm cells could not dedifferentiate to become germline stem cells again . The insects were then raised in either a standard environment ( plenty of food and no females ) or in stressful conditions ( periods of starvation with or without mating ) . The experiments showed that dedifferentiation was important to maintain a robust germline stem cell pool , both in the short and long term . This , however , was only the case in the difficult environment; the ability to dedifferentiate made no difference in the easier living conditions . In addition , Herrera and Bach observed that , in the flies’ testicles , stem cells obtained through dedifferentiation divided much more often than the ‘original’ stem cells . Finally , further analyses highlighted a series of genes that are required for dedifferentiation . That stem cells coming from dedifferentiated cells divide at a higher rate could be relevant to scientists across various fields . For example , this knowledge may help those who study how tissues regenerate after injury , a process that involves dedifferentiation . It also may be used as a cautionary tale for researchers who work on induced pluripotent stem cells – which are created in the laboratory by dedifferentiating specialized cells . This may be especially important because these cells could one day be put in patients to treat diseases such as Parkinson’s or Alzheimer’s .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine", "developmental", "biology" ]	2018	JNK signaling triggers spermatogonial dedifferentiation during chronic stress to maintain the germline stem cell pool in the Drosophila testis
Peripheral neural sensory mechanisms play a crucial role in metabolic regulation but less is known about the mechanisms underlying vagal sensing itself . Recently , we identified an enrichment of liver X receptor alpha and beta ( LXRα/β ) in the nodose ganglia of the vagus nerve . In this study , we show mice lacking LXRα/β in peripheral sensory neurons have increased energy expenditure and weight loss when fed a Western diet ( WD ) . Our findings suggest that the ability to metabolize and sense cholesterol and/or fatty acids in peripheral neurons is an important requirement for physiological adaptations to WDs . Vagal afferent neurons innervate the gastrointestinal ( GI ) tract , pancreas , liver , and portal vein and link peripheral levels of GI nutrients as well as circulating and stored fuels ( Berthoud , 2008 ) . Peripheral sensory mechanisms play a crucial role in the regulation of satiation ( Schwartz , 2004; Bello and Moran , 2007; Grill , 2010 ) , but mechanisms underlying vagal sensing itself are still unknown . Recently , we identified an enrichment of ‘lipid sensing’ nuclear receptors ( NRs ) , including liver X receptor alpha and beta ( LXRα/β ) in the nodose ganglia ( NG ) of the vagus nerve ( Liu et al . , 2014 ) . Notably , these neurons and their processes reside outside the blood–brain barrier , enabling the potential for direct sensing of molecules released by adipose tissue or liver . LXRs are oxysterol-sensitive NRs that direct cholesterol uptake , transport , and excretion in various tissues . LXRα and LXRβ are encoded by the Nr1h3 and Nr1h2 genes , respectively ( NR subfamily 1 , group H , member 3 and 2 ) . LXRs regulate target genes encoding for ATP-binding cassette proteins , and apolipoproteins ( Repa et al . , 2000a; Venkateswaran et al . , 2000; Chawla et al . , 2001; Bradley et al . , 2007; Hong et al . , 2012 ) . Ligand activation of LXRs also stimulates de novo lipogenesis of triglycerides in liver via sterol regulatory element-binding protein 1c ( Repa et al . , 2000b; Zhang et al . , 2012 ) . Hepatic LXRα/β regulate whole lipid and glucose homeostasis , and LXRα/β in macrophages regulate inflammation ( Chawla et al . , 2001; Rong et al . , 2013; A-Gonzalez et al . , 2013; Zelcer and Tontonoz , 2006 ) . In the central nervous system , they regulate local inflammation , differentiation , and neuron survival by orchestrating cholesterol uptake and efflux ( Wang et al . , 2002 ) . Further studies in LXR null mice revealed the rather surprising finding that these mice were resistant to obesity when challenged with a diet containing both high fat and cholesterol ( Kalaany et al . , 2005 ) . This study showed that the LXR−/− response was due to abnormal energy dissipation resulting in part from ectopic expression of uncoupling proteins in white adipose . Here , we show that Western diet ( WD ) -fed mice that lack LXRα/β in sensory neurons of the NG have altered neuronal cholesterol content and increased white adipose tissue browning , leading to changes in energy expenditure and body weight . This unexpected function of LXRs in vagal sensory neurons provides a plausible mechanism that may in part explain the role LXRs on metabolism in response to a diet containing fat and cholesterol . We first assessed the effects on LXR-target gene expression following pharmacological administration of LXR agonists . Canonical LXR target genes , including ATP binding cassette protein A1 ( Abca1 ) and SREBP-1c ( Srebf1 ) , were up-regulated in the NG of mice treated with LXR agonists ( Figure 1A ) . These results agree with reports of expression of Abca1 in the central nervous system ( CNS ) and its up-regulation in cultured neurons and sciatic nerves following LXR agonist treatment ( Fukumoto et al . , 2002; Cermenati et al . , 2010 ) . Previous data have also shown SREBP-1c to be expressed in Schwann cells and CNS neurons , and LXR agonist stimulated Srebp1f expression has also been reported in various tissues ( Cermenati et al . , 2010 ) . Carbohydrate-responsive element-binding protein ( Chrebp ) expression remained unchanged in response to LXR agonist ( Figure 1A ) . 10 . 7554/eLife . 06667 . 003Figure 1 . LXRs signaling regulates cholesterol metabolism in nodose ganglia neurons . ( A ) Regulation of selected target genes in nodose ganglia ( NG ) following liver X receptor ( LXR ) agonist treatment in vivo ( left panel ) n = 5–8 per group . *p < 0 . 005 . ( B ) NG organotypic slices were prepared from LXRsNav or LXRsfl/fl treated with GW3965 ( 5 μM ) or vehicle for 4 hr . Quantitative PCR ( qPCR ) data are expressed as average fold-change relative to vehicle ± S . E . M . , n = 3 independent experiments . # and indicates p < 0 . 5 , p < 0 . 005 . ( C ) Quantification of total cholesterol . Values were expressed as ng of cholesterol per NG , n = 6 . ## and p < 0 . 001 . ( right panel ) . Neurons isolated from LXRsNav or control mice NG were subjected to Filipin staining ( representative images of staining from 3 individual mice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06667 . 00310 . 7554/eLife . 06667 . 004Figure 1—figure supplement 1 . Ablation of LXR in the NAV1 . 8 positive neurons . qPCR analysis detecting the expression of truncated isoforms of Nr1h3 ( A ) and Nr1h2 ( B ) in NG , liver , white adipose , brown adipose tissues ( BATs ) , and muscle . Error bars show S . E . M . *indicates p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06667 . 004 These results suggest that Abca1 and Srepb1f are targets of LXR in the NG . To confirm these findings , we established NG organotypic cultures from mice lacking LXRα and LXRβ in vagal sensory neurons expressing the sodium ion channel NAV1 . 8 encoded by the Scn10a gene . The resulting LXRsNav mice were generated by breeding double floxed Nr1h3 and Nr1h2 mice ( LXRsfl/fl mice ) with the Nav1 . 8::Cre mice , which selectively expresses Cre-recombinase in peripheral sensory neurons ( Stirling et al . , 2005; Gautron et al . , 2011 ) . As expected , LXRα/β were deleted in the majority of NG neurons in LXRsNav mice compared to LXRsfl/fl littermate controls ( Figure 1—figure supplement 1 ) . No difference in LXRs expression was observed in liver , white adipose tissue , brown adipose tissues ( BATs ) , and muscle ( Figure 1—figure supplement 1 ) . In LXRsfl/fl controls , Abca1 and Srebp1f mRNA levels increased significantly after agonist treatment . This stimulation was fully ( Abca1 ) or partially ( Srebp1f ) blunted in NG from LXRsNav littermates ( Figure 1B ) . Collectively , these results suggest that in the NG , the regulation by LXRs of Abca1 is restricted to sensory neurons , but that its action on Srebp1f likely occurs in multiple cell types . Interestingly , cholesterol assays on whole NG taken from WD ( 42% fat , 0 . 2% cholesterol ) -fed LXRsNav knockout mice showed a 60% increase in total cholesterol compared to littermate controls ( Figure 1C ) . An increase of intracellular cholesterol level was also observed in LXRsNav dispersed neurons ( Figure 1C ) . Collectively , our results suggest that in a WD setting , LXRs regulate NG sensory neuron cholesterol levels though ABCA1-dependent signaling . A role for ABCA1 in cholesterol , lipid distribution has previously been extensively observed in the brain ( Tam et al . , 2006; Yang et al . , 2006; Kruit et al . , 2010 ) . In addition , LXR agonist or ABCA1 over-expression has been shown to alter cholesterol content in degenerating neurons or cancer cells ( Smith and Land , 2012 ) . Our results demonstrate that in sensory neurons , NG cholesterol metabolism is also regulated via LXR . To measure the physiological impact of LXRs loss in peripheral sensory neurons expressing Nav1 . 8 , LXRsfl/fl , LXRsNav , Nav1 . 8::Cre were maintained for 16 weeks on a standard rodent chow diet ( normal chow , NC , 4% fat ) or WD . All mice had similar body weights at weaning . However , LXRsNav mice weighed significantly less than controls after 11 weeks of NC ( Figure 2A ) . Similarly , LXRsNav mice fed WD were resistant to diet-induced obesity . NMR analysis revealed that LXRsfl/fl mice had twofold more body fat after 10 weeks of WD than LXRsNav mice ( Figure 2B ) . However , the differences in body fat do not completely reflect the body weight difference , despite no difference in lean mass ( Figure 2B ) . WD-fed LXRsNav mice had higher energy expenditure than littermate controls ( Figure 2C , D ) . However , no differences in food intake were noted in metabolic chambers . Furthermore , no important changes were found in plasma glucose , serum cholesterol , free fatty acids and liver triglycerides or cholesterol ( Figure 2—figure supplement 1A-F ) . A decrease in adiposity in the setting of nutrient excess is sometimes due to ectopic lipid deposition in liver . However , histological examination using Hematoxylin and Eosin staining ( H&E staining ) showed that when fed WD , both LXRsfl and LXRsNav mice developed hepatosteatosis and increase in hepatic lipid droplets ( Figure 2—figure supplement 1G ) . Our results suggest that LXRsNav mice have an exacerbated response to the WD with increased diet-induced thermogenesis . This suggests that LXRs may regulate the whole-body energy expenditure according to the amount of fat or cholesterol present in the diet . 10 . 7554/eLife . 06667 . 005Figure 2 . Ablation of LXRs from NAV1 . 8 expressing neurons exacerbates high-fat diet induced thermogenesis . ( A ) Body weight of LXRsNav mice and littermate controls on chow and Western Diet ( WD ) as followed over time ( n = 12 per genotype ) for WD , # for normal chow ( NC ) . ( B ) Adipose tissue as a percentage of total body weight in mice fed WD for 9 weeks , measured by NMR ( n = 6 ) . ( C , D ) Energy expenditure in weight-matched mice . ( C ) Calorimetry trace before and after a switch from NC to WD . ( D ) Energy expenditure during light and dark cycles before and after a switch from NC to WD . Error bars show S . E . M . indicates p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06667 . 00510 . 7554/eLife . 06667 . 006Figure 2—figure supplement 1 . Blood chemistry and lipid levels of LXRsNav mice . Metabolite levels as measured by the UTSW metabolic core from LXRsNav and LXRsfl/fl mice fed 16 weeks with NC or WD . ( A ) Glucose . ( B ) Serum cholesterol . ( C ) Fatty acid ( FFA ) . ( D ) Serum triglycerides . ( E ) Liver triglycerides . ( F ) Liver cholesterol . DOI: http://dx . doi . org/10 . 7554/eLife . 06667 . 006 Expectedly , we found that WD-fed control mice accumulated large lipid droplets in their BAT . Notably , this accumulation was considerably attenuated in LXRsNav mice ( Figure 3A ) . To evaluate the BAT activity , markers for mitochondrial metabolism and thermogenesis regulation in adipose tissue were assessed by Western blot or real-time PCR . Uncoupling protein 1 ( Ucp1 ) and peroxisome proliferator-activated receptor gamma coactivator 1-alpha ( Pgc1α ) were higher in LXRsNav mice BAT ( Figure 3B , C ) . A more striking threefold UCP1 increase was observed in LXRsNav mice subcutaneous fat as assessed by immunohistochemistry ( IHC ) and whole fat pad Western blot ( Figure 3D , E ) . These results suggest LXRs in sensory neurons may be important for BAT activity regulation and subcutaneous white fat conversion to a ‘beige’ phenotype in response to WD . 10 . 7554/eLife . 06667 . 007Figure 3 . Adipose tissue and muscle reprogramming in LXRsNav mice vs control mice . ( A ) Immunohistochemistry for UCP1 in BAT . ( B ) UCP1 Western blot on whole BAT pads ( upper panel , n = 4 ) UCP1 molecular weight ( MW ) is 33 kDa , Actin served as the loading control , MW 42 kDa , the graph represents blot quantification . ( C ) mRNA levels in the BAT of LXR ablated mice vs control mice ( lower panel , n = 4 ) . indicates p < 0 . 05 . ( D ) UCP1 staining and ( E ) Western-blot analysis of UCP1 protein levels in whole , individual dorsal subcutaneous fat pads . Actin served as the loading control . The signal is quantified in the graph below ( n = 4 ) Scale bar = 100 μm . ( F , G ) Oxygen-consumption rates were determined using the XF24 Extracellular Flux Analyzer following the manufacturers' protocols ( n = 3 ) . indicates p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 06667 . 00710 . 7554/eLife . 06667 . 008Figure 3—figure supplement 1 . Electron-flow experiments . Isolated skeletal muscle mitochondria were seeded at 10 μg of protein per well in XF24 V7 cell-culture microplates and analyzed with the XF24 Extracellular Flux Analyzer ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06667 . 008 To establish whether sensory neuron-specific deletion of LXRs alters skeletal muscle mitochondrial respiration , we examined mitochondrial electron transport chain activity by performing mitochondrial electron-flow ( EF ) and electron-coupling ( EC ) experiments to assess oxygen-consumption rates ( OCRs ) . During EF analyses , we observed that skeletal muscle mitochondria derived from LXRsNav mice exhibit markedly higher OCRs in response to the substrates pyruvate , malate , succinate , and ascorbate ( Figure 3F , G ) , when compared from control mice skeletal muscle mitochondria . Furthermore , EC experiments to gage mitochondrial coupling and integrity revealed no defects in skeletal muscle mitochondrial function in either genotype ( Figure 3—figure supplement 1 ) . Taken together , these data indicate that deletion of LXR specifically in peripheral sensory neurons enhances skeletal muscle mitochondrial oxidative respiration . To gain more insights into the signaling downstream of LXR in NG sensory neurons , we surveyed genes important for vagal neuronal function , including: cholecystokinin A and B receptor ( Cckar and Cckbr ) , which regulate vagal nerve activity and feeding; neuregulin 1 ( Nrg1 ) , involved in axon/Schwann cell communication and sensory nerve structure ( Gambarotta et al . , 2013; Stassart et al . , 2013 ) ; and α , β , and γ synuclein ( Syna , Synb , Syng ) , which are involved in intraneuronal trafficking/cell–cell communication and known to be LXR targets in the brain ( Golovko et al . , 2009 ) . Nrg1 and Syng were decreased twofold in both chow and WD-fed LXRsNav mice , interestingly Syna was only significantly up-regulated in control mice in response to WD ( Figure 4A ) . Interestingly , in recent reports , γ synuclein has been linked to metabolism ( Oort et al . , 2008; Golovko et al . , 2009; Millership et al . , 2012 ) . γ synuclein is a protein that modulates synaptic trafficking in neurons but also lipid droplets generating intracellular fatty acids . Notably , the γ synuclein whole body knockout is protected against diet-induced obesity ( Oort et al . , 2008; Millership et al . , 2012 ) . 10 . 7554/eLife . 06667 . 009Figure 4 . NG gene expression in LXRsNav mice vs control mice . ( A ) Gene expression in NG from LXRsNav and control mice fed with NC or WD . ( B ) Gene expression in NG from LXRsNav and control mice fasted 20 hr or fed . All genes show a significant difference between both genotypes . indicates p < 0 . 05 ( n = 6–10 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06667 . 009 Starvation reduces triglyceride levels in serum , but increases circulating fatty acids , which are rapidly taken up by the liver . To study the ability of NG to acutely respond to nutrient changes ( including triglycerides and fatty acids ) , we studied the expression of LXR targets in the NG of fed or fasted mice . We asked whether fasting-induced increases in fatty acid availability modified NG gene expression . Abca1 , Srepb1f , and Syng mRNA levels were significantly changed in LXRsfl/fl mice fasted 20 hr ( Figure 4B ) . Notably , the fasting-induced increase in Abca1 and Syng was blunted in LXRsNav mice ( Figure 4B ) . We also exposed NG cultures to serum obtained from mice fed or fasted for 20 hr . These data suggest that NG neurons may sense circulating secreted starvation cues and respond by regulating unique LXR-dependent genes . Despite its well-documented role in regulating the transcription of genes crucial for lipid synthesis and storage upon cholesterol sensing , little is known about how LXRs function in peripheral neurons and further investigation is necessary to completely understand the role of these NRs in the NG neurons . Our study suggests that LXRs in vagal sensory neurons potentially regulate vagal synaptic transmission to ultimately affect the gating of information to adipose tissues and muscle . Since WAT , BAT , and muscle receive innervation from sympathetic neurons , we suspect that the increased sympathetic tone ( secondary to altered input from the vagal sensory neuron activity ) may underlie the increased energy expenditure observed in LXRsNav mice . Interestingly , Kalaany et al . , also described LXR null mice as resistant to obesity when challenged with a Western-style diet containing high fat and cholesterol . This phenotype was surprisingly independent of SREBP-1c and due to a net increase in the energy utilization in white adipose and muscle ( Kalaany et al . , 2005 ) . Our study of LXR function in NG neurons is consistent with this previous report and provides a partial explanation of how LXRs can affect the whole body thermogenesis via sensory neurons . Diano et al . , ( 2011 ) , previously showed induction of LXRs in the hypothalamus in response to high-fat feeding . This finding suggests that LXRs may also regulate energy homeostasis beyond the nodose nucleus neurons . Based on the aforementioned findings , we postulate that the LXR pathway may mediate certain aspects of lipid sensing in peripheral sensory neurons . Our findings also suggest that the ability to metabolize and sense cholesterol and/or fatty acids in peripheral neurons may be an important requirement for physiological adaptations to high-fat/high-cholesterol ( Western ) diets , such including thermogenesis leading to changes in key metabolic processes . All ‘Materials and methods’ were approved by the Institutional Animal Care and Use Committee at UT Southwestern Medical Center . All mice were housed in a temperature-controlled room with a 12-hr light/dark cycle in the animal facility of University of Texas Southwestern Medical Center . Mice were fed with either NC ( #2916 , Harlan-Teklad , Madison , WI; 4 . 25% kcal from fat ) or WD ( #88137 , Harlan Teklad , cholesterol ( 0 . 2% total cholesterol ) , fat ( 42% kcal from fat ) , high in saturated fatty acids ( >60% of total fatty acids ) ) . LXRsfl/fl mice ( floxed Nr1h3 and Nr1h2 genes ) on a mixed C57BL6 and 129SV background were bred and kept as a mixed background in a closed colony in UTSW germ-free facility . LXRsfl/fl mice were then backcrossed to C57BL/6J mice for six generations prior to experiments . Transgenic mice ( C57/BL6 background ) carrying Cre recombinase driven by a Scn10a promoter ( called Nav1 . 8::Cre mice ) were bred with backcrossed LXRsfl/fl mice . LXRsfl/fl mice were bred with LXRsfl/fl Nav+/− to generate cohorts of littermate LXRsNav and LXRsfl that were used and compared in at least three cohorts for each experiment . We have generated cohort of Nav1 . 8-Cre/ tdTomato reporter mice to examine the hypothalamus . Our current findings show that tdTomato Nav1 . 8 expression is absent from the entire hypothalamus ( Data not shown ) . Body weight was monitored weekly from weaning ( at 4 weeks of age ) to 28 weeks of age . In the WD studies , mice were maintained on NC until 6 weeks old before being fed WD . Food intake , meal patterns , energy expenditure , and physical activity were continuously monitored using a combined indirect calorimetry system ( TSE Systems GmbH , Bad Homburg , Germany ) . Experimental animals ( 11-week-old ) were acclimated in the metabolic chambers for 5 days before data collection . Mice were initially maintained on NC during the acclimation period and the first two days of data collection and then fed WD for the next three days . O2 consumption and CO2 production were measured to determine the energy expenditure . In addition , physical activity was measured using a multi-dimensional infrared light bean system . C57Bl/6 males were treated with vehicle ( 1% methylcellulose ) and LXR agonist ( 10 milligram/kg body weight ( mpk ) GW3965 ) , by oral gavage at 14 hr and 2 hr prior to sacrifice . NG were rapidly dissected and frozen in liquid nitrogen . Metabolic parameters were continuously monitored using a combined indirect calorimetry system ( TSE Systems GmbH ) as described in the supplemental methods . The data were represented as mean ± S . E . M . Statistical analyses were performed using GraphPad PRISM version 6 . 0 . Single comparisons were made using 1- or 2-tailed t tests , as appropriate , and multiple comparisons were performed using 1-way analysis of variance ( ANOVA ) followed by Dunnett's post hoc test . For repeated measures , 2-way repeated-measures ANOVA was performed , with Bonferroni post hoc tests . A p value less than 0 . 05 was considered significant . Real-time quantitative PCR ( qPCR ) gene expression analysis was performed using inventoried TaqMan Gene Expression Assays ( Applied Biosystems ) . 18 s was used as normalizer . TaqMan probes used for qPCR include 18 s ( ABI , Hs99999901_s1 ) , Adrb3 ( ABI , Mm02601819_g1 ) , Cckar ( ABI , Mm00438060_m1 ) , Cckbr ( ABI , Mm00432329_m1 ) , Ucp1 ( ABI , Mm01244861_m1 ) , Pgc1a ( 01016719_m1 ) , Nrg1 ( Mm01212130_m1 ) , Syna ( Mm01188700_m1 ) , Synb ( Mm00504325_m1 ) , Synb Mm00488345_m1 ) , Abca1 ( Mm00442646_m1 ) , Two Sybr green-based primer sets located in the first exons of the floxed Nr1h3 and Nr1h2 genes were used to specifically detect the truncated form of LXRα and β . Mouse pups between 8 and 11 day old were decapitated , and the NG were quickly removed and cultured in chilled Gey's Balanced Salt Solution ( Invitrogen ) enriched with glucose ( 0 . 5% ) and KCl ( 30 mM ) . The NG were then placed on Millicell-CM filters ( Millipore; pore size 0 . 4 μm ) and then maintained at the air-media interface in minimum essential medium ( Invitrogen ) supplemented with heat-inactivated horse serum ( 25% , Invitrogen ) , glucose ( 32 mM ) , and GlutaMAX ( 2 mM , Invitrogen ) . Cultures were typically maintained for 10 days in standard medium , which was replaced three times a week . After an overnight incubation in low serum , ( 1 . 5% ) MEM supplemented with GlutaMAX ( 2 mM ) , slices were stimulated with vehicle , 5 μM GW3965 for 4 hr . RNA was harvested using Acturus PicoPure RNA Extraction kit ( Applied Biosystems ) . For IHC , sections were deparaffinized and the wax at the surface was removed with xylenes . After antigen retrieval and blockage of endogenous peroxidase activity , sections were stained with primary antibodies against UCP-1 ( Cat# ab10983 , Abcam ) followed by biotinylated secondary antibodies ( anti-rabbit; Dako , Glostrup , Denmark ) . Secondary antibodies were detected using a DAB chromogen A kit ( Dako ) following the manufacturer's protocol . The slides were also counterstained with Hematoxylin . Filipin staining for unesterified cholesterol was performed according to manufacturer's instructions ( FilipinIII cholesterol detection , Abcam ) . After measuring the fasting glucose levels , mice were given an i . p . dose of glucose ( 1 . 5–2 g/kg body weight ) . Blood glucose levels were then monitored using an AlphaTrak glucometer ( Abbott Laboratories , North Chicago , IL ) designed for use in rodents . Adult tissue was resuspended in 1× lysis buffer placed on ice 30 min and homogenized in 2 ml tubes , glass bead-containing tubes . Samples were then directly used to quantify total cholesterol . The reaction was performed in 96-well plates by adding Amplex Red reagent , horseradish peroxidase , cholesterol oxidase , and cholesterol esterase ( Amplex Red Cholesterol Assay Kit; Life technologies ) . The reactions were incubated for 30 min at 37°C . Results presented here were obtained from individual mice ( n = 6 ) . To isolate mitochondria , skeletal muscle tissues were homogenized using a motorized Dounce homogenizer in ice-cold MSHE buffer ( 70 mM sucrose , 210 mM mannitol , 5 mM HEPES , 1 mM EDTA ) containing 0 . 5% FA-free Bovine Serum Albumin ( BSA ) . Homogenates then underwent low centrifugation ( 800×g for 10 min ) to remove nuclei and cell debris , followed by high centrifugation ( 8000×g for 10 min ) to obtain the mitochondrial pellet , which was washed once in ice-cold MSHE buffer and was resuspended in a minimal amount of MSHE buffer prior to determination of protein concentrations using a BCA assay ( Pierce ) . OCRs were determined using the XF24 Extracellular Flux Analyzer ( Seahorse Bioscience , MA ) following the manufacturers' protocols . For the EF experiments , isolated skeletal muscle mitochondria were seeded at 10 μg of protein per well in XF24 V7 cell-culture microplates ( Seahorse Bioscience ) , then pelleted by centrifugation ( 2000×g for 20 min at 4°C ) in 1× MAS buffer ( 70 mM sucrose , 220 mM mannitol , 10 mM KH2PO4 , 5 mM MgCl2 , 2 mM 4- ( 2-hydroxyethyl ) -1-piperazineethanesulfonic acid ( HEPES ) , 1 mM ethylene glycol tetraacetic acid ( EGTA ) in 0 . 2% FA-free BSA; pH 7 . 2 ) supplemented with 10 mM pyruvate , 10 mM malate , and 4 μM carbonyl cyanide 4- ( trifluoromethoxy ) phenylhydrazone ( FCCP ) ( for EF experiments ) , with a final volume of 500 μl per well . For EC experiments , 1× MAS buffer was supplemented with 10 mM succinate and 2 μM rotenone . The XF24 plate was then transferred to a temperature-controlled ( 37°C ) Seahorse analyzer and subjected to a 10-min equilibration period and 2 assay cycles to measure the basal rate , comprising a 30-s mix , and a 3-min measure period each; and compounds were added by automatic pneumatic injection followed by a single assay cycle after each; comprising a 30-s mix and 3-min measure period . For EF experiments , OCR measurements were obtained following sequential additions of rotenone ( 2 μM final concentration ) , succinate ( 10 mM ) , antimycin A ( 4 μM ) , and ascorbate ( 10 mM ) ( the latter containing 1 mM N , N , N′ , N′-tetramethyl-p-phenylenediamine [TMPD] ) . For EC experiments , OCR measurements were obtained post sequential additions of ADP ( 4 mM ) , oligomycin ( 2 μM ) , FCCP ( 4 μM ) , and antimycin-A ( 2 μM ) . OCR measurements were recorded at set interval time-points . All compounds and materials above were obtained from Sigma–Aldrich .	The vagus nerves run from the brainstem to the heart and the digestive system and help to control several processes including digestion and heart rate . Because of their role in regulating food intake , these nerves are attractive targets for scientists hoping to develop treatments for obesity . There are two types of fat tissue found in mammals: white fat , which is used as an energy store and makes up most of the extra fat seen in obese individuals; and brown fat , which can generate body heat . The vagus nerves monitor fat and cholesterol levels in the body via receptor proteins that respond to messages sent from the fat tissues and the liver . Previous research unexpectedly found that mice genetically engineered to lack these receptor proteins—called LXRα and LXRβ—do not become obese even when fed a high-fat , high-cholesterol diet that would make normal mice gain excessive weight . Mansuy-Aubert et al . have now investigated in more detail why mice without these receptor proteins are resistant to obesity . When fed a high-fat , high-cholesterol diet , mice that lacked the LXRα and LXRβ receptors in sensory neurons had higher cholesterol levels in their nerve cells than normal mice on the same diet . Mice lacking these receptors also burned more energy and gained less weight than normal mice . Next , Mansuy-Aubert et al . examined fat tissue from both types of mice . This revealed that the heat-generating brown fat was more active in mice lacking the LXRα and LXRβ receptors . Some of the white fat in these mice had also become more like brown fat , allowing the mice to burn more energy and so gain less weight . In many Western countries , many people also eat a diet that is high in fat and cholesterol . This raises the possibility that drugs that block the LXRα and LXRβ receptors in sensory neurons in humans could help to treat or prevent obesity , although further work will be needed to investigate this .	[ "Abstract", "Introduction", "Results", "Materials", "and", "methods" ]	[ "short", "report", "neuroscience" ]	2015	Loss of the liver X receptor LXRα/β in peripheral sensory neurons modifies energy expenditure
Protection against malaria in humans can be achieved by repeated exposure to infected mosquito bites during prophylactic chloroquine treatment ( chemoprophylaxis and sporozoites ( CPS ) ) . We established a new mouse model of CPS immunization to investigate the stage and strain-specificity of malaria immunity . Immunization with Plasmodium chabaudi by mosquito bite under chloroquine cover does not generate pre-erythrocytic immunity , which is acquired only after immunization with high sporozoite doses . Instead , CPS immunization by bite elicits long-lived protection against blood-stage parasites . Blood-stage immunity is effective against a virulent , genetically distinct strain of P . chabaudi . Importantly , if exposure to blood-stage parasitemia is extended , blood-stage parasites induce cross-stage immunity targeting pre-erythrocytic stages . We therefore show that CPS immunization can induce robust , long-lived heterologous blood-stage immunity , in addition to protection against pre-erythrocytic parasites following high dose sporozoite immunization . Cross-stage immunity elicited by blood-stage parasites may further enhance efficacy of this immunization regimen . Protective immunity against microorganisms is developed after repeated infection and recovery ( Mutapi et al . , 2013 ) . Vaccines are usually successful if they mimic these naturally acquired immune responses ( Fox , 1984; Mielcarek et al . , 2006 ) . While protection against viruses and bacteria can be induced by vaccination with killed ( inactivated ) or live-attenuated pathogens ( Delany et al . , 2014 ) , there is no licensed vaccine for human parasitic diseases like malaria that pose a major global health burden . The apicomplexan malaria parasite Plasmodium is transmitted by bites of female anopheline mosquitoes . In the vertebrate host , sporozoites injected into the dermis migrate to the liver , where they establish a clinically silent infection of hepatocytes . Merozoites are then released from the liver and invade erythrocytes , leading to an exponential asexual replication cycle that is entirely responsible for the clinical signs and symptoms associated with malaria . Immunity against severe disease can be acquired following repeated infection , but sterile parasite clearance is rarely achieved ( Goncalves et al . , 2014 ) . Clinically immune adults in endemic areas still harbor parasites in their blood-stream ( Okell et al . , 2009 ) . These asymptomatic carriers also develop gametocytes , the form transmissible to mosquitoes , thereby allowing the parasite to complete its life cycle . To achieve malaria control and eventually eradication , transmission must be blocked ( Kappe et al . , 2010 ) . A vaccine that protects against pre-erythrocytic parasites , and thus outperforms naturally acquired immunity , would greatly facilitate this aim . Parasites that arrest during the liver stage , either because of irradiation ( Nussenzweig et al . , 1967 ) or targeted gene deletion ( Mueller et al . , 2005 ) , can provide immunity against challenge infection . Indeed , immunization of human volunteers with irradiated sporozoites can induce sterile protection in experimental settings ( Clyde et al . , 1973; Seder et al . , 2013 ) . However , in the absence of acquired immunity to the blood-stage parasite a pre-erythrocytic vaccine that is only partially effective , and therefore permits breakthrough erythrocytic infections , will provide no protection against severe malaria ( Bejon et al . , 2011 ) . The inclusion of a blood-stage component together with an effective pre-erythrocytic vaccine is therefore preferred to provide a multi-stage malaria vaccine that minimizes both transmission and disease ( Ellis et al . , 2010; Goodman and Draper , 2010 ) . A recently described experimental malaria immunization protocol using chemoprophylaxis and sporozoites ( CPS ) ( Roestenberg et al . , 2009 ) ensures exposure to pre-erythrocytic and blood-stage parasites , and hence has the unique potential to induce protection against all Plasmodium life cycle stages in the vertebrate host . Three immunizations with bites of 10–15 Plasmodium falciparum-infected mosquitoes under chloroquine chemoprophylaxis are sufficient to elicit sterile protection against homologous challenge in human volunteers ( Roestenberg et al . , 2009; Bijker et al . , 2013 , 2014 ) . Although it is not possible to measure liver parasite burden in human volunteers directly , it appears that immunity exclusively targets pre-erythrocytic parasite life cycle stages , as there is no protection against direct blood challenge ( Bijker et al . , 2013 ) . CPS immunization is thus substantially more effective than immunization with irradiated sporozoites , which requires 1000 mosquito bites ( Clyde et al . , 1973 ) or five intravenous ( iv ) injections of more than 100 , 000 sporozoites for sterile protection ( Seder et al . , 2013 ) . We therefore hypothesize that transient blood-stage parasitemia , before abrogation by chloroquine , may contribute to immunity following CPS immunization . In this study , we have investigated the stage- and strain-specificity of protection in a novel mouse model of CPS immunization using Plasmodium chabaudi . P . chabaudi establishes a chronic , non-lethal blood-stage infection , which has been used extensively to characterize the immune response to blood-stage parasites in vivo ( Stephens et al . , 2012 ) . A recently optimized protocol for P . chabaudi mosquito transmission ( Spence et al . , 2012 ) allows us now to also study pre-erythrocytic stages of this rodent parasite . Heterologous protection can readily be assessed since many genetically distinct P . chabaudi isolates displaying a variety of virulence phenotypes are available ( Mackinnon and Read , 1999; Otto et al . , 2014 ) . We immunized C57BL/6 mice three times with bites of P . chabaudi-infected mosquitoes under oral chloroquine chemoprophylaxis similar to human clinical trials ( Roestenberg et al . , 2009; Bijker et al . , 2013 , 2014 ) . This approach is unique amongst all published animal models of CPS immunization ( Beaudoin et al . , 1977; Golenser et al . , 1977; Orjih et al . , 1982; Belnoue et al . , 2004; Friesen and Matuschewski , 2011; Inoue et al . , 2012; Nganou-Makamdop et al . , 2012b; Doll et al . , 2014; Lewis et al . , 2014; Peng et al . , 2014 ) , which have ( without exception ) used iv injection of high numbers of Plasmodium berghei or Plasmodium yoelii sporozoites for immunization . Furthermore , rather than evaluating effector mechanisms by challenging shortly after immunization ( Beaudoin et al . , 1977; Belnoue et al . , 2004; Friesen and Matuschewski , 2011 ) , we performed the challenge 100 days after the final immunization to test the generation and maintenance of long-term immunological memory . CPS immunization with P . chabaudi by mosquito bite does not generate pre-erythrocytic immunity , which is acquired only after immunization with high doses of sporozoites . Instead , immunization by bite elicits blood-stage immunity that is effective against the immunizing strain and also a more virulent , genetically distinct P . chabaudi . Moreover , extended exposure to blood-stage parasitemia elicits robust pre-erythrocytic immunity , comparable to protection afforded by high dose sporozoite immunization . Exposure to blood-stage parasites thus elicits heterologous blood-stage immunity and can contribute to the pre-erythrocytic efficacy of this immunization regimen . Therefore , these findings add significantly to advances from previous CPS immunization mouse models by evaluating the generation of immune memory after immunization with P . chabaudi by mosquito bite . This is relevant for our understanding of acquired immunity in a malaria endemic setting , and can inform multi-stage malaria vaccine development . Transient blood-stage parasitemia is a key feature of CPS immunization . We used quantitative RealTime ( qRT ) PCR to measure erythrocytic parasite burden after each immunization with P . chabaudi AS-infected mosquito bites under chloroquine cover . After the first immunization , approximately 50 , 000 parasites per ml whole blood were detected within the first erythrocytic cycle ( Figure 2 ) . The amount of blood-stage parasites within the first erythrocytic cycle varied extensively ( median 47 , 596 , range 67–222 , 699 ) , reflecting the stochastic inoculation of sporozoites during mosquito bite ( Beier et al . , 1991; Ponnudurai et al . , 1991; Medica and Sinnis , 2005 ) . Thereafter , chloroquine reduced parasitemia by 86–96% every 24 hr . After the fourth cycle , the majority of erythrocytic parasites were cleared . Similarly , after the second and third immunization mice experienced a substantial number of circulating blood-stage parasites for 48–72 hr . However , although blood-stage parasites were detected in all but one mouse , parasitemia in the first erythrocytic cycle was reduced by 5- and 13-fold after the second and third immunizations , respectively , when compared to infection controls ( Figure 2 ) . Consequently , one CPS immunization is sufficient to reduce blood-stage parasite burden within the first erythrocytic cycle , which indicates either pre-erythrocytic or blood-stage immunity . 10 . 7554/eLife . 05165 . 004Figure 2 . Chloroquine permits transient blood-stage parasitemia during each immunization . The number of parasitized erythrocytes ( pE ) per ml of whole blood was enumerated by quantitative RealTime PCR after each CPS immunization ( i1 , i2 , i3 ) with P . chabaudi AS-infected mosquito bites under chloroquine ( CQ ) cover . The number of pE ( at the late trophozoite stage ) was quantified immediately before merozoite egress from the liver , at 48 hr post mosquito transmission ( erythrocytic cycle 0 ) , and then every 24 hr until erythrocytic replication cycle 4 . Daily parasitemia of 10 CPS immunized mice ( each color represents an individual mouse ) are shown . Blood-stage parasites were detected within the first erythrocytic cycle after every immunization in all but one mouse after the final immunization . Gray bars represent the mean parasitemia in the first erythrocytic cycle of naive mice infected as controls for mosquito transmission efficiency separate with each immunization ( n = 3–5 ) . Significant differences in the number of blood-stage parasites in the first erythrocytic cycle between naive and CPS immunized mice are indicated ( Mann Whitney test , *p ≤ 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05165 . 004 In order to assess directly whether pre-erythrocytic immunity was generated by this CPS immunization protocol , liver parasite burden was analyzed after mosquito bite challenge . Surprisingly , there was no difference in the liver parasite burden between mice given three CPS immunizations with P . chabaudi AS-infected mosquito bites and infection controls ( Figure 3A ) . Therefore , immunization by mosquito bite under chloroquine cover according to this protocol failed to elicit pre-erythrocytic immunity . This is in contrast to results from other animal models of CPS immunization where pre-erythrocytic immunity is induced after high numbers of sporozoites are injected iv ( Belnoue et al . , 2004; Friesen and Matuschewski , 2011; Inoue et al . , 2012; Nganou-Makamdop et al . , 2012b; Lewis et al . , 2014; Peng et al . , 2014 ) . To test if pre-erythrocytic immunity can be induced after CPS immunization with large numbers of sporozoites , we used P . chabaudi CB , since mosquitoes infected with this parasite strain harbor an increased number of sporozoites in their salivary glands compared to P . chabaudi AS ( Spence et al . , 2012 ) , making these experiments technically feasible . In agreement with previous studies ( Belnoue et al . , 2004; Friesen and Matuschewski , 2011; Inoue et al . , 2012; Nganou-Makamdop et al . , 2012b; Lewis et al . , 2014; Peng et al . , 2014 ) , mice immunized iv three times with 10 , 000 P . chabaudi CB sporozoites under chloroquine cover did show reduced liver parasite burden ( up to 90% ) after mosquito bite challenge , compared to infection controls ( Figure 3B ) . Conversely , mice immunized iv three times with a low dose of 100 P . chabaudi CB sporozoites ( representative of the estimated number of P . chabaudi sporozoites that initiate infection via mosquito bite , Spence et al . , 2012 ) do not acquire pre-erythrocytic immunity ( Figure 3B ) . It also appears that CPS immunization with 10 , 000 live sporozoites was more effective at inducing pre-erythrocytic immunity than immunization with 10 , 000 irradiated P . chabaudi CB sporozoites , which arrest during hepatic development ( Suhrbier et al . , 1990 ) and do not establish a blood-stage infection ( Figure 3B ) . This suggests that complete liver-stage maturation and the increased blood-stage parasitemia that accompanies immunization with 10 , 000 sporozoites ( as compared to 100 sporozoites or mosquito bite ) could contribute to pre-erythrocytic protection . 10 . 7554/eLife . 05165 . 005Figure 3 . Pre-erythrocytic immunity following CPS immunization requires high doses of sporozoites . Liver parasite burden was determined 42 hr after mosquito bite challenge as copy number of P . chabaudi-specific 18S rRNA . ( A ) Mice were CPS immunized three times ( i1 , i2 , i3 ) with P . chabaudi AS-infected mosquito bites under chloroquine ( CQ ) cover ( CPS ( Bite ) ) and challenged ( C ) 96–104 days after immunization by bites of P . chabaudi AS-infected mosquitoes ( pooled data from three independent experiments; naive infection controls ( − ) n = 25 , CPS ( Bite ) n = 35 ) . ( B ) 100 or 10 , 000 untreated or irradiated ( Irr . ) P . chabaudi CB sporozoites ( spz ) were injected iv three times under CQ cover . Mice were challenged 96 days after immunization by bites of P . chabaudi CB-infected mosquitoes ( naive infection controls ( − ) n = 12 , all other groups n = 20 ) . All data are displayed relative to the mean of corresponding liver parasite burden of naïve infection controls and presented as mean ± SEM , ( A ) Mann–Whitney test: no significant difference between the groups; ( B ) Kruskal Wallis with Dunn's multiple comparisons test p ≤ 0 . 05 , **p ≤ 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 05165 . 005 To test whether the transient blood-stage infection resulting from CPS immunization by mosquito bite ( Figure 2 ) is sufficient to induce protection against erythrocytic parasites , we challenged mice approximately 100 days after the last immunization and measured blood-stage parasitemia ( Figure 4 ) . Mice that were CPS immunized with P . chabaudi AS had similar numbers of parasitized erythrocytes as compared to mock immunized controls within the first five erythrocytic cycles following mosquito bite challenge ( Figure 4A ) . However , from erythrocytic cycle 6 parasitemia was significantly reduced and blood-stage parasites were cleared more rapidly in immunized mice . This was reflected in a 6-fold reduction of total area under the curve ( AUC ) ( right inset , Figure 4A ) . Protection against erythrocytic parasites was also evaluated by direct blood challenge , using blood-stage parasites obtained from a donor mouse infected by mosquito bite . Similar to the results of mosquito bite challenge , blood-stage parasitemia was significantly reduced ( Figure 4B ) , but the infection was still chronic in some mice ( Figure 4—figure supplement 1 ) . However , blood-stage protection was abrogated when CPS immunized mice were challenged with serially blood passaged ( SBP ) P . chabaudi; blood-stage parasites with increased virulence following multiple passages through naive mice ( Spence et al . , 2013 , Figure 4C ) . In this case , CPS immunization reduced blood-stage parasite burden only between erythrocytic cycle 6 and 8 , as compared to mock immunized controls . This was reflected in only a 1 . 2-fold reduction in total AUC ( right inset , Figure 4C ) . Therefore , CPS immunization by mosquito bite elicits homologous blood-stage immunity , which is most effective in the context of mosquito transmission . 10 . 7554/eLife . 05165 . 006Figure 4 . CPS immunization elicits blood-stage immunity . Mice were CPS immunized three times ( i1 , i2 , i3 ) using chloroquine ( CQ ) and P . chabaudi AS-infected or uninfected mosquito bites ( mock immunized ) . Approximately 100 days after the final CPS immunization , mice were challenged ( C ) with P . chabaudi AS . Erythrocytic parasitemia was evaluated daily by quantitative RealTime PCR ( cycle 1–3 , displayed as parasitized erythrocytes ( pE ) per ml whole blood; left ) and from cycle 3–14 by thin blood-film ( expressed as % parasitized erythrocytes [parasitemia] 0 . 01% parasitemia corresponds to 1 , 000 , 000 pE per ml; middle ) . The total area under the curve ( AUC ) was calculated for each mouse between erythrocytic cycle 3 and 14 ( right ) . ( A ) Mosquito bite challenge: parasitemia from 1st to 3rd ( n = 10 ) and between 3rd and 14th erythrocytic cycle ( representative of three independent experiments , n = 12–19 ) , total AUC between cycle 3 and 14 ( n = 19 ) . ( B ) Direct blood challenge using 10 , 000 erythrocytic parasites obtained from a donor mouse infected by mosquito bite; injected intraperitoneal ( ip ) : parasitemia between 1st and 3rd ( n = 10 ) and 3rd and 14th erythrocytic cycle ( representative of three independent experiments , n = 10 ) , total AUC between cycle 3 and 14 ( n = 10 ) . ( C ) Blood challenge using 10 , 000 serially blood passaged parasites; injected ip: parasitemia from 1st to 3rd ( n = 10 ) and between 3rd and 14th erythrocytic cycle ( representative of two independent experiments , n = 8–10 ) , total AUC between cycle 3 and 14 ( n = 10 ) . All data are presented as mean ± SEM , Mann–Whitney test per time point p ≤ 0 . 05 , p ≤ 0 . 01 , p ≤ 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 05165 . 00610 . 7554/eLife . 05165 . 007Figure 4—figure supplement 1 . Chronic blood-stage infection in CPS immunized mice . Naive infection controls or mice that were CPS immunized three times ( i1 , i2 , i3 ) using chloroquine ( CQ ) and P . chabaudi AS-infected mosquito bites were challenged ( C ) by intraperitoneal ( ip ) injection of 100 , 000 parasitized erythrocytes derived from a mosquito bite initiated infection 100 days later . Parasitemia was evaluated by thin blood film for 96 erythrocytic cycles . Different colors represent individual CPS immunized mice ( n = 8 ) , naive infection controls are presented in grey with dashed lines ( n = 6 ) . The limit of detection by thin blood film is 0 . 01% parasitemia ( 1 parasite in 10 , 000 erythrocytes or 1 , 000 , 000 parasites per ml blood ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05165 . 007 CPS immunization has so far not been shown to induce protection against challenge with genetically distinct strains of Plasmodium; a situation that would be encountered in human malaria-endemic areas . The genetic diversity amongst strains of P . chabaudi ( Mackinnon and Read , 1999; Otto et al . , 2014 ) allows us to investigate heterologous immunity in this model of CPS immunization . Mice that were CPS immunized with P . chabaudi AS-infected mosquito bites had reduced peak parasitemia , and blood-stage parasites were cleared faster , when compared to mock immunized mice after homologous ( P . chabaudi AS ) and heterologous ( P . chabaudi CB ) mosquito bite challenge ( Figure 5A–D ) . A direct comparison between P . chabaudi AS and CB challenge revealed nonetheless that homologous blood-stage immunity is more effective than heterologous immunity . In this experiment , CPS immunization reduced total AUC by 25-fold following homologous challenge , as compared to 6-fold following heterologous challenge ( Figure 5E ) . Nevertheless , CPS immunization elicits blood-stage protection against a robust heterologous challenge with the genetically distinct , and more virulent , CB strain of P . chabaudi . 10 . 7554/eLife . 05165 . 008Figure 5 . CPS immunization elicits heterologous blood-stage immunity . Mice were CPS immunized three times ( i1 , i2 , i3 ) under chloroquine ( CQ ) cover by P . chabaudi AS-infected mosquito bites or mock immunized with uninfected mosquito bites , and challenged ( C ) 96–107 days later by mosquito bite . Erythrocytic parasitemia was evaluated daily by quantitative RealTime PCR for blood-stage parasites ( cycle 1–3 , displayed as parasitized erythrocytes ( pE ) per ml whole blood ) and from cycle 3–20 by thin blood-film ( expressed as % parasitized erythrocytes [parasitemia] , 0 . 01% parasitemia corresponds to 1 , 000 , 000 pE per ml ) . ( A/B ) Heterologous challenge with P . chabaudi CB infected mosquitoes ( A ) Parasitemia between 1st and 3rd ( n = 10 ) and ( B ) from 3rd to 20th erythrocytic cycle ( n = 20 ) . ( C/D ) Homologous challenge with P . chabaudi AS-infected mosquitoes ( C ) Parasitemia between 1st and 3rd ( n = 10 ) and ( D ) from 3rd to 20th erythrocytic cycle ( CPS immunized n = 20 , mock immunized n = 19 ) . ( E ) Total AUC comparing mock and CPS immunized mice receiving heterologous or homologous mosquito bite challenge . Data are presented as mean ± SEM , ( A–D ) Mann–Whitney test per time point ( E ) Kruskal Wallis with Dunn's multiple comparisons test p ≤ 0 . 05 , p ≤ 0 . 01 , p ≤ 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 05165 . 008 Blood-stage parasites appear to be both the source and target of protection following CPS immunization with P . chabaudi AS-infected mosquito bites . It was shown that immunization with serially blood passaged P . yoelii parasites with prophylactic chloroquine treatment can elicit pre-erythrocytic immunity ( Belnoue et al . , 2008 ) . We wanted to assess whether blood-stage parasites have the potential to induce pre-erythrocytic protection also in the context of mosquito transmission . We therefore asked whether a fulminant blood-stage infection could elicit cross-stage immunity against pre-erythrocytic parasites . We infected mice with P . chabaudi AS by mosquito bite or intraperitoneal ( ip ) injection of recently mosquito transmitted blood-stage parasites . The two groups of mice were not drug-treated , and therefore experienced a low-grade chronic , recrudescing blood-stage infection for up to 90 days ( Spence et al . , 2013 ) . After mosquito bite challenge , both groups of mice demonstrated reduced liver parasite burden ( up to 85% ) , compared to infection controls ( Figure 6 ) . Cross-stage immunity is therefore a powerful mechanism for protection against pre-erythrocytic parasites , which may be absent during CPS immunization with small sporozoite numbers as the blood-stage infection is curtailed by the use of chloroquine . 10 . 7554/eLife . 05165 . 009Figure 6 . Extended blood-stage parasitemia elicits pre-erythrocytic immunity . Mice received a first infection with P . chabaudi AS either by mosquito bite ( Bite ) or intraperitoneal ( ip ) injection of 10 , 000 parasitized erythrocytes obtained from a donor mouse infected by mosquito bite ( Blood ( MT ) ) . Blood-stage parasitemia was self-cured before challenge ( C; 98 days post-infection ) using P . chabaudi AS-infected mosquito bites . Liver parasite burden was determined 42 hr after mosquito bite challenge as copy number of P . chabaudi-specific 18S rRNA . Data are displayed relative to the mean of corresponding liver parasite burden of naive infection controls ( − ) . Pooled data from two independent experiments ( Naive ( − ) and Bite n = 30 , Blood ( MT ) n = 20 ) are presented as mean ± SEM , Kruskal Wallis with Dunn's multiple comparisons test p ≤ 0 . 05 , **p ≤ 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 05165 . 009 Timing , route of infection , and antigen dose play major roles in determining the initial priming of the antimalarial immune response ( Legorreta-Herrera et al . , 2004; Elliott et al . , 2005; Belnoue et al . , 2008; Guilbride et al . , 2012; Nganou-Makamdop et al . , 2012a ) . We incorporated many of these aspects from human clinical trials ( Roestenberg et al . , 2009; Bijker et al . , 2013 , 2014 ) in a new P . chabaudi mouse model of CPS immunization to investigate the stage- and strain-specificity of CPS-induced protection against malaria . Our results highlight the complexity of immunity against the different life cycle stages of the malaria parasite ( Table 1 ) . Pre-erythrocytic immunity appears to depend on the number of immunizing sporozoites . In this study , we find no evidence of pre-erythrocytic immunity after CPS immunization with P . chabaudi infected mosquito bites , which inoculate an estimated maximum of 100 sporozoites per immunization ( Spence et al . , 2012 , 2013 ) . Sterile pre-erythrocytic protection was however reported in human CPS immunization trials ( Roestenberg et al . , 2009; Bijker et al . , 2013 , 2014 ) . Anopheles stephensi mosquitoes experimentally infected with P . falciparum 3D7 or NF54 habor 50–200 times more sporozoites in their salivary glands ( Bijker et al . , 2013 ) compared to P . chabaudi AS ( Spence et al . , 2012 ) . However , only few sporozoites are injected into the skin during mosquito bite and this number is independent of salivary gland sporozoites load ( Beier et al . , 1991; Ponnudurai et al . , 1991; Medica and Sinnis , 2005 ) . Since the number of sporozoites establishing a liver-stage infection can further be influenced by inherent sporozoite infectivity ( Khan and Vanderberg , 1991 ) , our best estimate of immunizing sporozoite dose is the number of infected erythrocytes observed directly after egress from the liver . Assuming that 10 , 000 merozoites are released from one infected liver cell ( White et al . , 2014 ) , we estimate approximately 400 infected hepatocytes ( 95% CI 137–1250 ) in human volunteers ( Bijker et al . , 2013 ) compared to 5 in P . chabaudi AS immunized mice ( 95% CI 1–31 ) . Therefore , the number of infected hepatocytes after CPS immunization by mosquito bite in the P . chabaudi mouse model is approximately 100-fold lower than in CPS immunized humans . We show that this differences in infected hepatocyte numbers by a factor of 100 can be significant for the development of pre-erythrocytic immunity: three immunizations with 10 , 000 P . chabaudi sporozoites iv induce long-lasting protection against mosquito bite challenge , while three immunizations with 100 P . chabaudi sporozoites fail to do so . This is in general agreement with other rodent malaria studies using P . berghei ( Beaudoin et al . , 1977; Golenser et al . , 1977; Orjih et al . , 1982; Friesen and Matuschewski , 2011; Nganou-Makamdop et al . , 2012b; Lewis et al . , 2014 ) or P . yoelii ( Belnoue et al . , 2004; Inoue et al . , 2012; Doll et al . , 2014; Peng et al . , 2014 ) , in which sterile pre-erythrocytic immunity is observed after immunization with high sporozoite doses ( typically 10 , 000–50 , 000 sporozoites per immunization ) , while a reduction in sporozoite dose or the number of immunizations leads to breakthrough blood-stage infections upon challenge ( Belnoue et al . , 2004; Inoue et al . , 2012 ) . Reduction of the number of P . falciparum-infected mosquitoes also reduces the frequency of sterilely protected volunteers ( Bijker et al . , 2014 ) , indicating that the number of sporozoites establishing a liver-stage infection fails to surpass the protective threshold to elicit sterile pre-erythrocytic immunity . This may also explain why in malaria-endemic areas pre-erythrocytic immunity is thought to be absent ( Tran et al . , 2013 ) . In addition to maximizing specific responses against immunodominant antigens , CPS immunization with high numbers of sporozoites may broaden the immune repertoire by including protective responses against subdominant antigens , which could enhance heterologous pre-erythrocytic protection ( Trieu et al . , 2011 ) . This may further be enhanced by the longer liver-stage development of P . falciparum ( egress 6 . 8 days after mosquito bite , Roestenberg et al . , 2012 ) compared to P . chabaudi ( egress after 52 hr , Stephens et al . , 2012 ) . Longer liver stage development may positively influence the generation of pre-erythrocytic immunity by allowing time for protective immune responses to develop . 10 . 7554/eLife . 05165 . 010Table 1 . Relationship between the dose of immunizing pre-erythrocytic and blood-stage parasites and the acquisition of immunity . DOI: http://dx . doi . org/10 . 7554/eLife . 05165 . 010Low doseHigh doseSporozoites ( liver-stage parasites ) Mosquito bite or 100 sporozoites iv: no pre-erythrocytic immunity ( Figure 3A/B ) 10 , 000 sporozoites iv: pre-erythrocytic immunity ( Figure 3B ) Blood-stage parasitesBlood-stage infection curtailed by chloroquine: partial blood-stage immunity ( Figure 4 ) Fulminant , self-cured blood-stage infection: blood-stage immunity ( Spence et al . , 2013 ) ; pre-erythrocytic immunity ( Figure 6 ) A key feature of CPS immunization is that it permits exposure to all Plasmodium life cycle stages in the vertebrate host , including parasitized erythrocytes . It is well known that repeated exposure to blood-stage parasites , for example , in rodent models ( Jarra and Brown , 1985; Legorreta-Herrera et al . , 2004; Elliott et al . , 2005 ) , in humans exposed to ultra-low doses of parasitized erythrocytes while receiving drug treatment ( Pombo et al . , 2002 ) , during malaria-therapy of neurosyphilis patients ( Collins and Jeffery , 1999 ) , and in people living in malaria-endemic areas ( Bull et al . , 1998 ) , induces blood-stage immunity . Our results also clearly show that during CPS immunization repeated transient blood-stage infection ( less than 0 . 01% parasitemia for 48–72 hr ) elicits long-lasting blood-stage immunity . Protection against challenge infection was only apparent after multiple erythrocytic replication cycles and patent blood-stage parasite densities , which could indicate that blood-stage protection was not observed in CPS immunized human volunteers after direct blood-challenge because drug treatment is required at low parasite densities as soon as patency is reached ( typically between the third and fourth erythrocytic cycle , Bijker et al . , 2013 ) . Blood-stage parasites are however recognized in human volunteers , which was demonstrated by an earlier increase of IFNγ and monokines induced by IFNγ ( MIG ) concentrations in CPS immunized volunteers compared to infection controls after direct blood challenge ( Bijker et al . , 2013 ) . It will be of value to investigate whether the observed blood-stage protection in this mouse model is also detected in CPS immunized primates , where a longer blood-stage infection than that allowed in human clinical trials is possible . Despite the reduced peak parasitemia and faster clearance of blood-stage parasites during the acute phase of infection in CPS immunized mice , recrudescent parasitemia could still be observed in the chronic phase of infection after challenge , suggesting a parasite variant escapes the protective immune response ( McLean et al . , 1982 ) . There are very few reports on protective efficacy of CPS immunization ( or indeed any sporozoite-based vaccine ) against direct blood-challenge ( Belnoue et al . , 2004; Inoue et al . , 2012; Peng et al . , 2014 ) . Doll et al . ( 2014 ) reported that sustained , subpatent blood-stage infection after treatment with a commonly used dose of chloroquine can induce partial blood-stage protection . Low-grade transient blood-stage parasitemia , achieved by attenuation of blood-stage parasites using an antimalarial drug ( Pombo et al . , 2002; Elliott et al . , 2005 ) , a DNA alkylating agent ( Good et al . , 2013 ) or genetic tools ( Ting et al . , 2008; Aly et al . , 2010 ) similarly provides protection against homologous and heterologous blood-stage challenge . We could show that CPS-induced blood-stage immunity is effective against heterologous mosquito bite challenge with a more virulent and genetically distinct strain of P . chabaudi ( Mackinnon and Read , 1999; Lamb and Langhorne , 2008 ) . In agreement with the observed cross-species protection after CPS immunization with mefloquine ( Inoue et al . , 2012 ) , and one study using chemically attenuated sporozoites for immunization ( Purcell et al . , 2008 ) , heterologous protection is less effective than homologous immunity . Nevertheless , CPS immunization can elicit long-lived protection against both homologous and heterologous blood-stage parasites , which will be important to minimize disease severity in the case of breakthrough blood-stage infections . This is essential for the development of an effective multi-stage malaria vaccine ( Ellis et al . , 2010; Bejon et al . , 2011 ) . In stark contrast to the observed heterologous blood-stage protection after mosquito bite challenge infection , protection was almost completely abrogated following direct blood-challenge with virulent parasites obtained after continuous serial blood passage . This suggests that blood-stage parasites immediately after mosquito transmission express antigens not present on serially blood passaged parasites and that these antigens may be the target of protective immunity following CPS immunization . Serially blood passaged parasites can hence escape from CPS-induced blood-stage immunity . One group of Plasmodium genes , whose expression is altered by mosquito transmission during blood-stage infection is the Plasmodium interspersed repeat gene family ( pir ) ; termed cir in P . chabaudi ( Lawton et al . , 2012 ) . Transcription of more than half of all cir genes is increased in blood-stage parasites after mosquito transmission compared with their transcription after serial blood passage . This diversification of cir transcription is associated with a more effective host immune response , which in turn attenuates parasite virulence ( Spence et al . , 2013 ) . The cir genes could also be candidate targets for cross-stage immunity , as P . berghei pir genes are also transcribed during the liver stages ( personal communication BM Franke-Fayard and CJ Janse , Leiden University Medical Center , The Netherlands ) . An investigation into shared PIR proteins between liver and blood-stage parasites may hence provide valuable information for multi-stage malaria vaccine development . Furthermore , the absence of blood-stage protection in previous CPS models may have been due to challenge with serially blood passaged rather then recently mosquito transmitted blood-stage parasites . It is therefore always essential to evaluate blood-stage immunity in the context of mosquito transmission . As shared antigenic targets between liver- and blood-stage parasites have been described ( Tarun et al . , 2008 ) , the exciting possibility of cross-stage protection has been considered but only rarely assessed . Genetically modified fabb/f- sporozoites that arrest late in liver-stage development protect against iv challenge with 100 blood-stage parasites ( Butler et al . , 2011 ) . On the other hand , a blood-stage infection with serially blood passaged P . yoelii , drug-treated with chloroquine after 4–5 days , reduces liver parasite load upon iv challenge with 35 , 000 sporozoites ( Belnoue et al . , 2008 ) . In our model of CPS immunization by mosquito bite , it is likely that repeated transient blood-stage infection during immunization elicits the observed blood-stage protection , although we cannot yet exclude that responses acquired against pre-erythrocytic antigens , which are shared with blood-stage parasites ( Tarun et al . , 2008 ) , contribute as well . Because of the low number of infected hepatocytes after CPS immunization with P . chabaudi-infected mosquito bites this seems however unlikely . Nevertheless , we demonstrate unequivocally that extended exposure to blood-stage parasites is an effective stimulator of pre-erythrocytic immunity . Exposure to blood-stage parasites during CPS immunization may thus significantly contribute to the observed pre-erythrocytic protection in human volunteers ( Roestenberg et al . , 2009; Bijker et al . , 2013 , 2014 ) . Indeed , cross-stage immunity could be responsible for the unprecedented efficacy of CPS immunization compared to immunization with irradiated sporozoites ( Clyde et al . , 1973; Seder et al . , 2013 ) , which arrest early during liver-stage development and never establish a blood-stage infection . While extending exposure to replicating blood-stage parasites by delayed drug administration is not possible in humans , the incorporation of chemically ( Good et al . , 2013 ) or genetically ( Ting et al . , 2008; Aly et al . , 2010 ) attenuated blood-stage parasites should be considered to further enhance the generation and maintenance of both pre-erythrocytic and blood-stage immunity . This makes CPS immunization a powerful tool for the development of an effective multi-stage malaria vaccine . Inbred C57BL/6 mice , originally obtained from Jackson Laboratories ( Bar Harbor , USA ) , were bred under specific pathogen-free conditions at the MRC National Institute for Medical Research ( NIMR ) for over 30 years . All experiments were performed in accordance with UK Home Office regulations ( PPL 80/2358 ) and approved by the ethical review panel at the MRC NIMR . Mice were housed under reverse light conditions ( light 19 . 00–07 . 00 , dark 07 . 00–19 . 00 GMT ) at 20–22°C and 50% relative humidity , with continuous access to mouse breeder diet and water . Plasmodium chabaudi chabaudi ( P . chabaudi ) AS and CB were cloned at the University of Edinburgh and sent to the NIMR in 1978 and 1982 , respectively . Both parasite lines were routinely serially blood-passaged ( SBP ) through mice between 26 and 32 times by ip injection of parasitized erythrocytes or mosquito transmitted according to a recently published protocol ( Spence et al . , 2012 ) . In brief C57BL/6 mice were injected ip with 100 , 000 parasitized erythrocytes and 14 days post infection gametocytemia was assessed on Giemsa-stained ( VWR , Lutterworth , UK ) thin blood film . A . stephensi mosquitoes , pre-treated with 50 μg/ml gentamicin ( Sigma , Gillingham , UK ) and starved for 24 hr before transmission , were fed on anaesthetised mice with >0 . 1% gametocytes of total erythrocytes at a ratio of >1 mouse per 100 mosquitoes . Mosquitoes were kept at 26 . 0°C ( ± 0 . 5°C ) in an ultrasonic humidity cabinet and provided with 8% Fructose and 0 . 05% 4-Aminobenzoic acid ( both Sigma , Gillingham , UK ) feeding solution . After 8 days , a sample of 20 mosquitoes were dissected to assess development of P . chabaudi oocyts in the midgut . For infection of experimental mice 20–23 mosquitoes were transferred into 25 cl paper cups after 14 days , starved for 24 hr and fed on anaesthetized mice for 20–25 min at room temperature . Typically , mice were exposed to 9 . 15 ( median , range 6 . 9–13 . 6 ) P . chabaudi-infected mosquito bites ( Spence et al . , 2012 ) . Sporozoites were isolated from P . chabaudi-infected mosquito salivary glands 15 or 16 days post gametocyte feed . Salivary glands were dissected under a stereomicroscope , transferred to a glass homogenizer and kept in RPMI supplemented with 0 . 2% Glucose , 0 . 2% Sodium bicarbonate ( both Sigma , Gillingham , UK ) , 2 mM L-Glutamine ( Gibco , Paisley , UK ) and 10% fetal bovine serum ( GE Healthcare Life Sciences , Pittsburgh , Pennsylvania ) , on ice for maximum 2 hr . Sporozoites were released from the glands by gentle homogenization and washed three times before enumeration . The number of sporozoites per infected mosquito was enumerated for each mosquito transmission experiment . For iv injection P . chabaudi CB sporozoites were used , since mosquitoes infected with this parasite strain harbor an increased number of sporozoites per infected mosquito in their salivary glands ( median 1638 , range 175–2576 ) compared to P . chabaudi AS ( median 438 , range 43–956 ) ( Spence et al . , 2012 ) . To arrest parasite development in the early stages of liver development P . chabaudi CB infected A . stephensi were exposed to 16 Gray ( =16 , 000 rad ) ( Nganou-Makamdop et al . , 2012b ) of Caesium-137 γ-irradiation 15 or 16 days post gametocyte feed just prior to sporozoite dissection . Female age-matched 8–10 week old C57BL/6 mice were infected three times in 2-week intervals with P . chabaudi: either by P . chabaudi AS-infected mosquito bites or iv injection of P . chabaudi CB sporozoites . Mice were treated after each immunization with 100 mg/kg chloroquine diphosphate salt ( chloroquine , Sigma , Gillingham , UK ) by gavage daily for 10 days , starting from the day of mosquito transmission . Mock immunized mice received uninfected mosquito bites and chloroquine treatment . 100 days after the last CPS immunization mice were challenged with P . chabaudi AS or P . chabaudi CB-infected mosquito bites , or via ip injection of 100 , 000 parasitized erythrocytes ( direct blood-challenge ) that were obtained from either a donor mouse infected by mosquito bite or after serial blood passage . Since each erythrocytic cycle of P . chabaudi is approximately 24 hr long ( Stephens et al . , 2012 ) development of blood-stage parasitemia was monitored daily by microscopy of Giemsa-stained thin blood films , from erythrocytic cycle 3 to 14 and every other day thereafter . The limit of detection was 0 . 01% parasitemia , which equals 1 parasitized red blood cell in 10 , 000 erythrocytes or 1 , 000 , 000 parasitized erythrocytes per ml of blood . Liver and blood-parasitemia was assessed by quantifying 18S rRNA using qRT PCR . 42 hr after mosquito bite challenge mice were terminally anaesthetized and immediately upon cessation of respiration their livers were perfused with 5 ml RNAse-free Phosphate buffered saline ( PBS , Gibco , Paisley , UK ) . Using the Gentle MACS homogenizer ( Miltenyi , Bisley , UK ) the whole liver was homogenized in 4 ml Guanidinium thiocyanate ( Sigma , Gillingham , UK ) solution ( Chomczynski and Sacchi , 2006 ) , and 600 μl aliquots were stored at −80°C . To assess blood-parasite burden during CPS immunization and in the first erythrocytic cycles following bite challenge 10 μl of blood were isolated from the tip of the mouse-tail and after two washes in RNAse-free PBS stored at −80°C in 100 μl Guanidinium thiocyanate solution ( Chomczynski and Sacchi , 2006 ) . The first sample was taken either just before ( erythrocytic cycle 0 ) or 20 hr after liver merozoite egress ( erythrocytic cycle 1 ) and then every 24 hr for 4 days . Since P . chabaudi displays a synchronous infection ( Stephens et al . , 2012 ) all blood-stage parasites analyzed were therefore at the late trophozoite stage of development . RNA was extracted from liver- as well as blood-samples using the Guanidinium-thiocyanate-phenol-chlorophorm method ( all Sigma , Gillingham , UK; Chomczynski and Sacchi , 2006 ) . RNA was thereafter reverse transcribed by PCR ( temperature profile: 25°C for 10 min , 42°C for 20 min , 98°C for 5 min ) using 75U MuLV Reverse Transcriptase , 30U RNAse Inhibitor , and 2 . 5 μM Random Hexamer primers ( all Applied Biosystems , Paisley , UK ) per sample . The amount of 18S rRNA copies was quantified by Real-Time PCR using TaqMan Universal PCR Master Mix ( Applied Biosystems , Paisley , UK ) , 300 ηM forward primer ( 5′-AAGCATTAAATAAAGCGAATACATCCTTAT-3′ ) , 300 ηM reverse primer ( 5′-GGGAGTTTGGTTTTGACGTTTATGCG-3′ ) , and 50 ηM probe ( [6FAM]CAATTGGTTTACCTTTTGCTCTTT[TAM] ) . All reactions were performed in the ABI 7900 HT Real Time PCR machine ( temperature profile: 50°C for 2 min , 95°C for 10 min , 40 cycles of 95°C for 15 s and 60°C for 1 min ) . The amount of parasite 18S rRNA in the liver was calculated based on a Standard curve of known copy numbers of 18S rRNA . For every experiment liver-parasite burden was normalized to the mean burden of controls infected in the same experiment . Blood-stage parasitemia was quantified based on a Standard curve of 10-fold dilutions of mosquito transmitted P . chabaudi AS late trophozoites prepared identically to the samples . Data were analyzed using GraphPad Prism v7 . Unpaired data between two groups at a specific time point were analyzed by Mann–Whitney test ( two-tailed , non-parametric ) . Differences between more than two groups were analyzed by non-parametric Kruskal–Wallis test with Dunn's multiple comparisons test . Significant differences are indicated by asterisks with p ≤ 0 . 05 , p ≤ 0 . 01 , *p ≤ 0 . 001 .	Malaria is a life-threatening infectious disease in humans that is caused by a single-celled parasite called Plasmodium . The parasite is carried between people by mosquitos; when an infected mosquito bites a human , the parasite is injected into the bloodstream with the mosquito's saliva . Plasmodium first infects liver cells but then re-enters the bloodstream , where it infects red blood cells leading to symptoms of disease . If another mosquito bites the infected individual at this so-called ‘blood-stage’ , the parasite can be passed to this mosquito and the cycle of transmission continues . Currently there are no vaccines available that can effectively protect against malaria . Although an experimental vaccine containing a weakened form of the parasite can protect against the liver-stage parasites , it fails to prevent the parasite from multiplying in the red blood cells . Therefore , the individuals remain susceptible to severe malaria . Recently , researchers have developed a new strategy for immunization that provides exposure to both liver-stage and blood-stage parasites . Human volunteers taking an anti-malarial drug were deliberately exposed to mosquitos carrying the parasite on three separate occasions . Although the volunteers were infected with the parasite , the anti-malarial drug killed the parasites inside the red blood cells . After the end of the drug treatment , the volunteers were exposed to mosquitos carrying the parasite and they were still protected from infection . These results are promising , but it is not clear if the volunteers have acquired immunity to liver-stage or blood-stage parasites , or even both . To answer this important question , Nahrendorf et al . developed a similar immunization strategy in mice . Just like the human volunteers , the mice were treated with an anti-malarial drug and exposed to mosquitos carrying Plasmodium on three separate occasions . Although the immunizations did not protect the mice against early infection in the liver , they did provide long-term protection against parasites multiplying in the red-blood cells . The immunity generated by this immunization strategy also protected the mice against another strain of Plasmodium , different to the one used in the immunizations . The experiments also show that prolonged exposure to the blood-stage parasites can even lead to immunity against the liver-stage parasites . Nahrendorf et al . 's findings show that this immunization strategy can protect individuals against both the liver-stage and blood-stage parasites . The next challenges are to find out how the immunity generated by one stage of infection can protect against the other stages , and to discover which molecules on the parasite the immune system targets .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "immunology", "and", "inflammation" ]	2015	Blood-stage immunity to Plasmodium chabaudi malaria following chemoprophylaxis and sporozoite immunization
Methods for measuring gut microbiota biochemical activities in vivo are needed to characterize its functional states in health and disease . To illustrate one approach , an arabinan-containing polysaccharide was isolated from pea fiber , its structure defined , and forward genetic and proteomic analyses used to compare its effects , versus unfractionated pea fiber and sugar beet arabinan , on a human gut bacterial strain consortium in gnotobiotic mice . We produced ‘Microbiota Functional Activity Biosensors’ ( MFABs ) consisting of glycans covalently linked to the surface of fluorescent paramagnetic microscopic glass beads . Three MFABs , each containing a unique glycan/fluorophore combination , were simultaneously orally gavaged into gnotobiotic mice , recovered from their intestines , and analyzed to directly quantify bacterial metabolism of structurally distinct arabinans in different human diet contexts . Colocalizing pea-fiber arabinan and another polysaccharide ( glucomannan ) on the bead surface enhanced in vivo degradation of glucomannan . MFABs represent a potentially versatile platform for developing new prebiotics and more nutritious foods . Increasing effort is being directed to determining how components of diets consumed by various human populations impact the composition and expressed functional features of their gut microbial communities ( e . g . , Johnson et al . , 2019; Ghosh et al . , 2020 ) . A hoped-for benefit is to gain new insights about how food ingredients and their biotransformation by the microbiota are linked to various aspects of human physiology , and new ways to both define and improve nutritional status . However , there are many formidable challenges . The gut microbiota is complex and dynamic and exhibits considerable intra- and interpersonal variation in its configurations ( Lloyd-Price et al . , 2017 ) . The chemical compositions of food staples are being cataloged at ever-deepening levels of detail using higher throughput analytical methods , such as mass spectrometry . Even as this knowledge is being acquired , how food components are recognized by members of the microbiota , and the pathways through which these chemical entities are metabolized by community members to influence their functions and those of the host remain challenging to define . Furthermore , much needs to be learned about the effects of current methods of food processing on the representation of bioactive food components ( Carmody et al . , 2015; Wolf et al . , 2019 ) , and about the mechanisms that determine whether and how microbes compete and/or cooperate for them . Plant fibers epitomize these challenges . Plant fibers are complex mixtures of biomolecules whose composition varies depending upon their source , method of initial recovery , and the food processing techniques used to incorporate them into food products that have satisfactory organoleptic properties ( texture , taste , smell ) ( Shahidi and Ho , 1998; Caffall and Mohnen , 2009 ) . Plant fibers are composed of but not limited to polysaccharides , proteins , fatty acids , polyphenols , and other plant-derived small molecules ( Nicholson et al . , 2012; Scalbert et al . , 2014 ) . Isolating and/or purifying component polysaccharides from crude plant fiber mixtures for studies of the mechanisms by which they influence members of the microbiota can be very challenging; even if separation is achieved , painstaking analysis is required to define their structures ( Pettolino et al . , 2012 ) . Nonetheless , knowledge of the genetic underpinnings of how various members of the gut microbiota recognize and metabolize glycans has increased substantially in recent years through a combination of in vitro and in vivo approaches ( Brown and Koropatkin , 2020; Kaoutari et al . , 2013; Ndeh and Gilbert , 2018 ) . Knowing that a given microbiota member has a suitable complement of genes for acquiring and processing a given glycan structure in vitro does not necessarily predict whether that organism will be a consumer in vivo ( Flint et al . , 2015 ) . For example , an individual’s microbiota may harbor a number of organisms with the capacity to compete or cooperate with one another for utilization of a given type of glycan . Additionally , the physical and chemical structure of a plant fiber ( e . g . , its size , surface properties , nutrient composition ) in a given region of the gut could influence which set of microbes attach to its surface , how its associated microbes prioritize consumption of its component glycans , and how/whether fiber particle-associated microbes can share products of glycan metabolism with one another . The current study illustrates an approach for identifying bioactive molecules from plant fibers and defining how they affect and how they are processed by members of the human gut microbiota . Pea fiber was selected based on results obtained from a screen we conducted of 34 types of food-grade crude plant fibers obtained from various sources , including the byproducts of food manufacturing ( Patnode et al . , 2019 ) . The screen was performed in gnotobiotic mice colonized with a defined consortium of cultured sequenced human gut bacterial strains , including several saccharolytic Bacteroides species . Mice were fed a low-fiber diet formulated to represent the upper tertile of saturated fat consumption and lower tertile of fruit and vegetable consumption by individuals living in the USA , as reported in the NHANES database ( Ridaura et al . , 2013 ) . Supplementation of this ‘HiSF-LoFV’ diet with fiber generated from the seed coat of the pea , Pisum sativum , produced a significant increase in the abundance of Bacteroides thetaiotaomicron ( Patnode et al . , 2019 ) . We have now isolated a bioactive arabinan-enriched fraction from pea fiber , defined its structure , and characterized how a model human gut community , containing human gut Bacteroides established in gnotobiotic mice , responds to the isolated arabinan-enriched fraction versus unfractionated pea fiber . We go on to describe a generalizable method for covalently attaching different glycans to microscopic paramagnetic glass beads with different covalently bound fluorophores , eliminating the proteinaceous component of an approach published by our group that relies on streptavidin-coated beads and bifunctional biotin-conjugated polysaccharides ( Patnode et al . , 2019 ) . Introduction of these ‘Microbiota Functional Activity Biosensors’ ( MFABs ) into gnotobiotic mice fed the HiSF-LoFV diet with or without glycan supplementation followed by their recovery from the gut allowed us to directly compare the capacity of these glycans to be degraded by this community . Chemically colocalizing pea fiber arabinan with glucomannan , another type of polysaccharide not found in the diet , in a monolayer on an MFAB surface enhanced the efficiency of microbial community degradation of bead-associated glucomannan when animals were given a pea-fiber-supplemented HiSF-LoFV diet . These findings illustrate how knowledge of the bioactive components of fibers , and the capacity to directly measure microbiota function with MFABs , could provide new approaches for designing ‘next-generation’ prebiotics and foods that are more accessible to , and have a greater impact on , the gut microbiota ( and by extension , the host ) . Raw pea fiber was subjected to serial extraction with aqueous buffers ( Pattathil et al . , 2012 ) . Eight soluble fractions were assayed for total nucleic acid , protein , and carbohydrate content plus monosaccharide composition ( Supplementary file 1 ) . The fraction isolated under the harshest conditions ( fraction 8; 4M KOH ) possessed high carbohydrate and low protein content , high monosaccharide diversity , and a high relative proportion of arabinose . We subsequently developed a scalable abbreviated procedure for preparing this fraction from pea fiber ( see Materials and methods ) . The mole percent representation of arabinose in the end product was 71% ( corresponding values for xylose , galactose , and glucose were 12% , 11% and 6% , respectively ) ( Figure 1A ) . Glycosyl-linkage analysis of the isolated glycan revealed a 2-O-branched arabinan ( 5-substituted arabinose , 2 , 5-substituted arabinose , 2 , 3 , 5-substituted arabinose ) ( Figure 1B ) attached to small rhamnogalacturonan-I ( RGI ) pectic fragments ( 2-substituted and 2 , 4-substituted rhamnose ) and small galactan oligomers ( 4-substituted galactose ) ( Supplementary file 1 ) . Xylose was present as a linear xylan polysaccharide ( 4-substituted xylose ) and glucose in the form of residual starch . We named the isolated polysaccharide fraction ‘pea fiber arabinan’ ( PFABN ) based on ( 1 ) these linkage results , ( 2 ) the fact that arabinose comprises the majority of its monosaccharide content ( 71 mole percent ) , and ( 3 ) our observation that ~90% of all non-starch carbohydrate is represented by what is likely a single species of polysaccharide , with the remaining being xylan . The biological activity of PFABN was compared to that of unfractionated pea fiber and arabinan isolated from sugar beet . Unlike PFABN , sugar beet arabinan ( SBABN ) is primarily 3-O-branched as revealed by glycosyl-linkage analysis ( Figure 1B , Supplementary file 1 ) . PFABN also contains twofold more triply branched 2 , 3 , 5-substituted arabinose monomers , suggesting it is more sterically encumbered than SBABN . The galactan portion of each arabinan is also unique: while both polymers contain 4-substituted galactose , 4 , 6-substituted galactose is enriched more than twofold in PFABN , while 6- and 3 , 6-substituted galactose are enriched threefold and fourfold in SBABN , respectively ( Supplementary file 1 ) . In vitro assays performed in a minimal defined medium ( McNulty et al . , 2013 ) with Bacteroides type strains established that B . ovatus ATCC 8483 , B . cellulosilyticus WH2 , and B . thetaiotaomicron VPI-5482 grew on isolated PFABN and SBABN , although less rapidly during the exponential phase of growth and to a lower cell density than in medium containing an equivalent concentration of d-glucose ( results based on measurements of OD600; see Figure 1C , Supplementary file 1 ) . In contrast , B . vulgatus ATCC 8482 grew as rapidly or faster and to the same or higher density on PFABN and SBABN , respectively , compared to d-glucose ( p<0 . 05 , one-way analysis of variance [ANOVA] with Tukey’s honest significant difference , FDR corrected ) . B . vulgatus ATCC 8482 also reached exponential growth more quickly than when d-glucose was present in this medium ( Figure 1C ) . Adult germ-free C57Bl/6J mice were colonized with a 14-member consortium of sequenced bacterial strains containing 58 , 537 known or predicted protein-coding genes . The consortium included the four Bacteroides type strains , another strain of B . thetaiotaomicron , three additional Bacteroides species ( B . caccae , B . massiliensis , B . finegoldii ) plus six other types of bacteria ( Figure 2A ) . Five of the Bacteroides ( B . thetaiotaomicron strains VPI-5482 and 7330 , B . vulgatus ATCC 8482 , B . cellulosilyticus WH2 , B . ovatus ATCC 8483 ) were each represented by previously described libraries of tens of thousands of transposon ( Tn ) insertion mutants ( Hibberd et al . , 2017; Wu et al . , 2015 ) . These studies had shown that collectively the overall change in abundance of each mutant library in response to various diet manipulations was similar to the corresponding parental wild-type strain . Two days post-gavage ( dpg 2 ) of the defined bacterial consortium , mice were switched from the unsupplemented HiSF-LoFV diet to a HiSF-LoFV diet supplemented with either unfractionated pea fiber , isolated PFABN , or isolated SBABN . Animals were fed these diets ad libitum for 10 days and then euthanized . A control group was maintained on unsupplemented HiSF-LoFV chow ( Figure 2A; n = 5–6 mice/treatment group ) . The amount of supplementation was calibrated so that mice in all treatment arms would receive the same daily dose of arabinan ( 100 mg; see Materials and methods ) . The absolute abundances of each strain were defined by shotgun sequencing of DNA isolated from feces collected on dpg 2 , 6 , 8 , and 11 and from the cecum at the time of euthanasia ( two independent experiments , labeled Experiment one and Experiment two in Supplementary file 2; McNulty et al . , 2011; Stämmler et al . , 2016 ) . Of the 14-member consortium , only B . thetaiotaomicron 7330 failed to colonize mice at levels above a ( relative ) abundance of 0 . 1% under any of the diet conditions tested . Five community members exhibited statistically significant , albeit distinct , differences in their absolute abundances in response to specific glycan preparations: ( 1 ) B . thetaiotaomicron VPI-5482 responded to all three supplements; ( 2 ) B . ovatus ATCC 8483 increased after exposure to both intact pea fiber and PFABN but not to SBABN; ( 3 ) B . vulgatus ATCC 8482 exhibited significant increases with SBABN; and ( 4 ) B . cellulosilyticus WH2 and Ruminococcaceae sp . TSDC 17 . 2 increased significantly when exposed to intact pea fiber but not to either of the isolated arabinan preparations ( Figure 2B , Supplementary file 2 ) ( p<0 . 01 , generalized linear mixed-effects model [Gaussian]; two-way ANOVA with Tukey’s HSD , FDR corrected ) . We calculated the sum total of the changes in absolute abundances of the four responsive Bacteroides relative to dpg two under each diet condition and expressed the results as bacterial cells per gram of feces per gram of supplement provided to each mouse daily . This metric of specific activity , which takes into account the absolute mass of supplement consumed daily by members of the different groups of mice , revealed that both of the isolated arabinan preparations had a significantly greater effect compared to intact pea fiber at the doses tested ( Figure 2C ) ( p<0 . 05 mixed-effects linear model [Gaussian]; one-way ANOVA with Tukey’s honest significant difference , FDR corrected ) . We took advantage of the fact that the gene content of the community was known and performed mass spectrometry-based metaproteomic analysis on feces , collected on dpg 6 , to define the responses of community members to the different glycan preparations . Bacteroides spp . possess multiple polysaccharide utilization loci ( PULs ) ; a shared feature of PULs is an adjacent pair of susC and susD homologs responsible for binding extracellular polysaccharide fragments and importing them into the periplasm . PUL genes also encode various carbohydrate active enzymes ( CAZymes ) involved in polysaccharide depolymerization , as well as transcriptional regulators that allow the locus to be induced in the presence of glycans it can recognize/utilize ( Anderson and Salyers , 1989; Terrapon et al . , 2018 ) . Regulated expression of PULs allows bacteria to acquire nutrients within the highly competitive environment of the gut ( Martens et al . , 2011; Tuncil et al . , 2017 ) . Supplemental results , Figure 2—figure supplement 1 , Figure 2—source data 1 , and Supplementary file 3 summarize the results of our analysis of PUL protein expression from the two independent experiments ( total of 10–11 mice/treatment arm ) . Based on gene set enrichment analysis ( Luo et al . , 2009 ) , we identified 14 , 12 , 11 , and 8 PULs that we deemed ‘responsive’ to at least one of the diet supplements in B . thetaiotaomicron VPI-5482 ( BT ) , B . ovatus ATCC 8483 ( Bovatus ) , B . cellulosilyticus WH2 ( BcellWH2 ) , and B . vulgatus ATCC 8482 ( BVU ) , respectively ( adjusted p-value<0 . 05 , unpaired one-sample Z-test , FDR-corrected ) . Additionally , we used multi-taxon insertion site sequencing ( INSeq , Wu et al . , 2015 ) of the five strains represented as Tn mutant libraries to identify genes with significant contributions to bacterial fitness in each diet context . Fitness was calculated as ( 1 ) the log2 ratio of the number of sequencing reads originating from the site of insertion of the Tn in the organism in fecal communities sampled on dpg 6 versus dpg 2 , relative to ( 2 ) the same ratio calculated in mice monotonously fed the unsupplemented HiSF-LoFV diet . A negative score indicates that a gene is important for fitness . The score of each gene was parameterized using linear models generated with limma ( Ritchie et al . , 2015 ) to identify those whose effects on fitness were significantly different compared to when the unsupplemented HiSF-LoFV diet was being consumed . The results disclosed that the fitness scores of a total of 39 genes in B . thetaiotaomicron VPI-5482 , 135 genes in B . ovatus ATCC 8483 , 346 genes in B . cellulosilyticus WH2 , and 82 genes in B . vulgatus ATCC 8482 were significantly decreased during diet supplementation with either pea fiber , PFABN , or SBABN ( Supplementary file 4 ) ( adjusted p-value<0 . 05 , FDR corrected ) . Plots of fitness score versus change in protein abundance were generated for all genes in each of these Bacteroides . Supplemental results and parts A–D of Figure 2—source data 1 summarize how expression and the fitness contribution of specific PULs vary for individual Bacteroides across the dietary contexts tested . Together , these community configurational and functional responses to diet supplementation provided evidence that PFABN is a key bioactive component of pea fiber utilized by B . thetaiotaomicron VPI-5482 , B . vulgatus ATCC 8482 , B . cellulosilyticus WH2 , and B . ovatus ATCC 8483 . However , these results do not directly establish that it is consumed , nor do they offer a direct comparison of the efficiency of metabolism of PFABN and SBABN . To produce such evidence , we developed a bead-based method for quantifying polysaccharide degradation within the intestinal tracts of colonized gnotobiotic mice . PFABN and SBABN were immobilized onto amine plus phosphonate-derivatized beads . Beads acetylated with acetic anhydride after fluorophore labeling were used as controls ( Figure 4A ) . Each of these three bead types contained a unique fluorophore . The three bead types were pooled , and the mixture was introduced by oral gavage into four groups of mice 10 days after they received the 14-member consortium: one group of recipient animals had been fed the unsupplemented HiSF-LoFV diet , while the other groups had received HiSF-LoFV containing unfractionated pea fiber , PFABN , or SBABN ( n = 5 animals/group ) . Germ-free mice fed HiSF-LoFV supplemented with PFABN served as controls ( n = 5; Experiment 1 ) . The bead mixtures were harvested using a magnet from the cecums of animals four hours after their introduction by oral gavage; the individual bead types were then purified by fluorescence-activated cell sorting ( FACS ) . Polysaccharide degradation was quantified by GC–MS of neutral monosaccharides released after acid hydrolysis of the purified beads . Results were referenced to the masses of monosaccharides released from aliquots of each input bead type ( i . e . , the same bead preparation but never introduced into mice ) . The quantities of neutral monosaccharides liberated by acid hydrolysis from the surfaces of beads recovered from the cecums of germ-free mice were not significantly different from the amounts liberated from the input bead preparations with one exception – a slight , albeit statistically significant , increase in galactose from beads coated with PFABN and SBABN ( Figure 4—figure supplement 3; p<0 . 05 , Mann–Whitney U test ) . This result established the stability and utility of cyanate-ester–coupled MFABs for studying polysaccharide degradation within the mouse gut , and the recalcitrance of both PFABN and SBABN to host digestive enzymes . In contrast to germ-free controls ( see Figure 4—figure supplement 3 ) , the mass of arabinose was significantly decreased when PFABN- or SBABN-coated beads were recovered from colonized mice fed the unsupplemented HiSF-LoFV diet ( Figure 4B , C; p<0 . 05 , Mann–Whitney U test ) . Compared to the base HiSF-LoFV diet , supplementation with unfractionated pea fiber induced a community configuration associated with significantly increased capacity to degrade both PFABN and SBABN as judged by the amount of arabinose remaining on recovered beads ( Figure 4B , C; p<0 . 05 , Mann–Whitney U test ) . Loss of arabinose from either PFABN- or SBABN-beads was not significantly different between the two bead types when the HiSF-LoFV diet was supplemented with either of these isolated arabinan preparations , demonstrating functional equivalence in the capacity of each community to utilize either arabinan ( see Figure 4B , C and Supplementary file 5 which provides evidence that results from cecal samples [Experiment 1] and fecal samples [Experiment 2] were comparable ) . Beads coated in PFABN revealed that xylan ( xylose monosaccharide remaining on PFABN beads ) was more efficiently processed by the microbiota in all three supplemented diet contexts ( Figure 4B; p<0 . 05 , Mann–Whitney U test ) . Our group previously explored the importance of arabinoxylan utilization from the base HiSF-LoFV diet ( Patnode et al . , 2019 ) . In contrast to xylan , the galactan content from both arabinan preparations was not utilized under any of the diet conditions tested ( Figure 4B , C and Supplementary file 5 ) , suggesting that β ( 1-4 ) galactan degradation in vivo has lower priority compared to the available arabinan ( Tuncil et al . , 2017 ) . As noted in Introduction , plant fibers have complex physical–chemical properties manifest in part by their mixtures of different glycan structures and by their varying shapes and surface properties . Plant fiber particles are impacted by methods , such as extrusion , that are commonly used to incorporate plant fibers into food products so that these products have acceptable organoleptic properties ( Caffall and Mohnen , 2009; Gualberto et al . , 1997 ) , and by the mechanical forces and digestive enzymes ( both host and microbial ) that are encountered as food passes through the gastrointestinal tract . In foods , plant fibers exist mostly as micro-particles . Although it would be desirable to be able compare the dynamics of degradation of glycans on MFABs with glycans in food particles with similar dimensions/characteristics , robust methods for reproducibly recovering food particles from luminal contents are not currently available . Therefore , we reasoned that the MFAB platform could provide a way of testing whether deliberately colocalizing distinct polysaccharides , akin to natural plant fiber particles , would result in ‘synergistic’ polysaccharide degradation by microbial community members . To explore this notion , we turned to glucomannan , a hemicellulosic linear β ( 1-4 ) polysaccharide composed of d-mannose and d-glucose . We found that among the pea fiber-responsive Bacteroides identified above , only B . ovatus ATCC 8483 and B . cellulosilyticus WH2 were able to grow in minimal medium containing glucomannan as the sole carbon source ( Figure 5A ) . Both organisms have PULs known to be induced by glucomannan in vitro ( PUL28 in B . cellulosilyticus WH2; PULs 52 and 80 in B . ovatus ATCC 8483 ) ; each of these PULs encodes at least one GH26 enzyme with β-mannanase activity ( Bågenholm et al . , 2017; Martens et al . , 2011 ) . Multiple genes in the glucomannan-responsive PUL28 of B . cellulosilyticus WH2 were consistently expressed , but not at significantly different levels , when mice were fed the unsupplemented HiSF-LoFV and pea fiber supplemented HiSF-LoFV diets ( Supplementary file 3 ) . Only two B . ovatus ATCC 8483 genes from its glucomannan-responsive Bovatus_PUL52 were expressed , albeit at the very limit of detection , under both diet conditions , and none from Bovatus_PUL80 ( Supplementary file 3 ) . Neither B . thetaiotaomicron VPI-5482 nor B . vulgatus ATCC 8482 , which fail to grow on glucomannan as the sole carbon source , contain GH26 , GH2 , or GH130 genes with known or predicted β-mannanase or β-mannosidase activities that were induced during pea fiber supplementation ( Supplementary file 3 ) ( among the two organisms , only the protein products of B . thetaiotaomicron BT_0458 [GH2] and BT_1033 [GH130] were detected under either diet conditions , and only at the very threshold of detection ) . Based on these considerations , we hypothesized that supplementing the diet with pea fiber would induce expression of PULs in community members , so that they could readily utilize bead-associated PFABN; moreover , those community members that could utilize PFABN and express β-mannanases would be able to more efficiently access/metabolize glucomannan positioned on the same bead . To test this hypothesis , we synthesized beads coated with PFABN alone , glucomannan alone , or both glycans together , as well as control acetylated beads that lack a bound polysaccharide ( Figure 5B ) . These four bead types , each labeled with a distinct fluorophore , were simultaneously introduced into two groups of mice colonized with the 14-member community – one group was fed the unsupplemented HiSF-LoFV diet , while the other group received a pea fiber-supplemented diet ( n = 7–8 mice/group ) ( Supplementary file 2 ) . Beads were recovered from their cecums 4 hr after gavage; the different bead-types were then isolated using FACS ( Figure 5C ) and subjected to acid hydrolysis and neutral monosaccharide analysis by GC–MS . We used the amount of mannose remaining on the bead as a proxy of glucomannan degradation because it represents the bulk of monosaccharide present in glucomannan and is absent in PFABN . The results revealed that glucomannan on beads coated with glucomannan alone was degraded to a similar extent in mice receiving the unsupplemented or pea fiber-supplemented HiSF-LoFV diets ( Figure 5D and Supplementary file 5; p=0 . 87 , Mann–Whitney U test ) . However , when presented with PFABN on the same bead , significantly more glucomannan was degraded by the microbiota of mice receiving the pea fiber-supplemented diet as compared to the unsupplemented diet ( Figure 5D; p<0 . 05 , Mann–Whitney U test ) . The amount of arabinose remaining on beads coated with PFABN and glucomannan , and PFABN alone , was also significantly reduced ( degradation increased ) with pea fiber supplementation ( Figure 5D; p<0 . 05 , Mann–Whitney U test ) . These results show that deliberate physical colocalization can result in quantitatively modest , albeit statistically significant synergistic degradation of polysaccharides during fiber supplementation ( Supplementary file 5; p<0 . 05 , linear model; diet supplement by bead-type interaction term ) . This finding , and the approach used to obtain these results , have implications for food science and prebiotic/synbiotic discovery efforts . We have used gnotobiotic mice colonized with a defined model human gut microbial community containing 58 , 537 known or predicted protein-coding genes together with metaproteomic and forward genetic analyses to demonstrate the sensitivity of a microbiota to structural differences in glycans ( arabinans ) isolated from two distinct plant sources . Robust isolation procedures and analytical characterization are required to isolate grams of polysaccharide and test their biological effects in gnotobiotic mice . The approach we employed allowed us to isolate 50 g of material to ~85% purity with the remaining material comprised of starch and xylan . Starch can be degraded by host enzymes . The xylan present in our PFABN preparation could , in principle , influence the microbiota response . However , there is arabinoxylan already present in the base HiSF-LoFV diet , and we found that expression of known xylan-responsive PULs in B . cellulosilyticus WH2 and B . ovatus ATCC 8483 was not induced by PFABN supplementation ( Figure 2—figure supplement 1; Patnode et al . , 2019; McNulty et al . , 2013; Martens et al . , 2011 ) . The specific activity of each isolated arabinan ( in this case the increase in absolute abundance of the responsive Bacteroides per unit mass of diet ingredient consumed per day ) was superior to unfractionated pea fiber ( Figure 2C ) , emphasizing , as others have , the importance of characterizing the effects of plant fibers on both the microbiota and host in order to inform future studies at the intersection of food science , nutrition , and human microbial ecology research ( Gidley and Yakubov , 2019 ) . Although each of the isolated arabinans had distinct effects on gene expression and organismal fitness , our development of orally administered recoverable and chemically modifiable probes to quantitate community metabolic function revealed their equivalent degradation in specified diet contexts . These findings illustrate how in vivo functional assays can complement and expand current approaches for characterizing microbiomes . The bead-based MFABs described in this report represent a technology for measuring biochemical activities expressed by a microbial community of interest in vivo or ex vivo . Future studies where MFABs are used in conjunction with genetic tools that enable rapid deletion of genes or entire multi-kilobase loci could provide new insights about the mechanisms by which community members acquire/degrade nutrients ( polysaccharides ) . We emphasize that in this study , we use MFABs to quantify community degradative activity and not to analyze community consumption of MFAB-bound polysaccharides and its relationship to bacterial growth . Determining the latter is challenging in our animal model; a typical gavage of 107 beads contained 100 µg of bead-immobilized polysaccharide , 1000 times more polysaccharide was consumed daily in the diet by mice in the polysaccharide-supplemented arms of our experiments . Installing specific functional groups on the surfaces of microscopic paramagnetic glass beads using commercially available organosilane reagents allows ‘modular’ incorporation of different biomolecules . This approach represents an alternative to a procedure we described recently , where bifunctional biotinylated ligands are generated prior to immobilization on glass beads coated with streptavidin ( Patnode et al . , 2019 ) . Each of these approaches has distinct advantages and disadvantages . By immobilizing ligand directly on the bead surface , MFABs possess a greater number of sites for ligand attachment than streptavidin-coated beads and include no protein constituents . Higher ligand density enables higher levels of ligand loading , which reduces the absolute number of beads that need to be administered , recovered , and processed , and increases the dynamic range of a functional activity readout . A unifying question for food science , microbiome science , and nutrition research is how to decipher the effects of a nutrient and the matrix in which it resides within a food on modulation of microbiota functions ( Gidley and Yakubov , 2019; Leitch et al . , 2007 ) . Crude plant fibers contain various polysaccharides densely intercalated within a cellulose-lignin matrix ( Caffall and Mohnen , 2009 ) . We found that the absolute abundances of B . cellulosilyticus WH2 and Ruminococcaceae sp . TSDC 17 . 2 increase significantly when animals were consuming the crude pea fiber-supplemented diet , but not during supplementation with either of the isolated arabinans . The chemistry for covalent polysaccharide attachment to MFABs not only allows for dense ligand presentation , but also enables multiple ligands to be simultaneously immobilized to create ‘hybrid’ beads to model the effects of physical colocalization of different fiber components on microbial degradation . In principle , a wide range of different glycan combinations with varying stoichiometries can be explored owing to the fact that different hybrid bead types , each with its own fluorophore , can be created and tested simultaneously in vitro and in vivo to investigate the mechanism , generality , and biological significance of colocalization of glycans on their degradation by gut microbes . The selection of gut bacterial taxa for in vivo tests of fiber effects in gnotobiotic animal models is critical for yielding ecologically and physiologically relevant results . Characterizing defined communities comprised of organisms beyond those represented in the 14-member consortium will undoubtedly be informative . For example , members of Lactobacillus , Bifidobacterium , and Faecalibacterium are more tolerant of acidic conditions after carbohydrate fermentation and can outcompete acid intolerant bacteria such as Bacteroides; their inclusion would likely change the response of community members to the presence of different plant fiber preparations and the dynamics of plant fiber degradation documented by various collections of MFABs . The approach we describe for ligand immobilization does not require the synthesis of bifunctional ligands ( or fluorophores ) ; instead , custom functional groups can be incorporated into the probe through modification of the organosilane donor molecule . As such , the MFAB platform provides an opportunity to develop chemistries for nondestructively releasing ligands for analysis ( Bielski and Witczak , 2013 ) . For example , characterizing microbial degradation of polysaccharides needs to move beyond relatively ‘simple’ GC–MS measurements of monosaccharides released from the surface of recovered beads to readouts of glycan structures recovered from the bead surface ( prior to and after exposure to microbes ) . This information would provide a more informed view of functional properties ( saccharolytic activities ) expressed by model communities comprised of cultured sequenced members of a microbiota representing a population of interest , or their intact uncultured microbiota , as well as greater insights about structure/activity relationships of existing or new candidate prebiotic and synbiotic formulations . Bacterial stocks , previously stored at −80°C , were struck onto Brain-heart infusion ( BHI; Becton Dickinson ) agar plates supplemented with 10% ( vol:vol ) horse blood . Plates were incubated in an anaerobic growth chamber ( Coy Laboratory Products; atmosphere 3% hydrogen , 20% CO2 , and 77% N2 ) . Single colonies were picked and grown overnight on a defined Bacteroides minimal medium ( BMM ) ( McNulty et al . , 2013 ) containing 5 mg/mL d-glucose . Bacteria were then diluted 1:500 ( vol:vol ) into BMM supplemented with a carbon source at a final concentration of 0 . 5% ( wt:wt ) and distributed into the wells of a 96-well half-area plate ( Costar; Cat . No . : 3696 ) . Plates were sealed with an optically clear membrane ( Axygen; Cat . No . : UC500 ) and growth at 37°C was monitored by measuring optical density at 600 nm every 15 min ( Biotek Eon instrument with a BioStack 4 ) . Carbon sources tested include d-glucose , PFABN , SBABN , and glucomannan ( Megazyme ) . All conditions were tested in quadruplicate . Readings obtained from control wells inoculated with bacteria but lacking a carbon source were averaged and subtracted from data obtained from carbon-supplemented cultures to generate background subtracted OD600 growth curves . All experiments involving mice were carried out in accordance with protocols approved by the Animal Studies Committee of Washington University in Saint Louis . COPRO-Seq and INSeq datasets are deposited at the European Nucleotide Archive ( ENA ) under study accession: PRJEB38095 . Proteomic data are available in the MassIVE database under project number: MSV000085341 . COPRO-Seq analysis software can be accessed at https://gitlab . com/hibberdm/COPRO-Seq and INSeq analysis software at https://github . com/mengwu1002/Multi-taxon_analysis_pipeline; a copy has been archived at swh:1:rev:fac437a7d35ecfd53600ff4dc667563dfb251d25 .	Tens of trillions of microbes living in the gut help humans and other animals digest their food . In the process , the microbes provide necessary nutrients for themselves and the animal . Learning more about the interaction of food components and gut bacteria could help scientists to better understand how different diets affect human health . Currently , studying these complex interactions is challenging , but new technologies that measure microbial nutrient processing in the gut could help . Now , Wesener et al . show that swallowable microscopic biosensors can measure how gut bacteria break down nutrients from food . To make the biosensors , Wesener et al . attached complex carbohydrates extracted from peas and fluorescent tags to microscopic beads . In the experiments , mice colonized with human gut microbes were fed the beads along with a traditional low fiber , Western diet . Some of the animals also received fiber supplements . The microscopic beads were then recovered from the intestines after digestion and the remaining carbohydrates on the beads were measured . The genetic makeup of the gut microbiome and the expression of microbial genes was also examined . The experiments revealed which pea carbohydrates the gut microbes consumed and showed that pairing certain carbohydrates together on the microbead surface increased their digestion in mice that received fiber supplements . If future studies prove that the microbead biosensors created by Wesener et al . are safe for humans to ingest , they could be used to help diagnose how well a person’s gut microbiota can process different foods . Studies using the microbead sensors may also help scientists develop more nutritious foods or supplements that promote the growth of microbes important for health .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2021	Microbiota functional activity biosensors for characterizing nutrient metabolism in vivo
Social behavior in mammals is often studied in pairs under artificial conditions , yet groups may rely on more complicated social structures . Here , we use a novel system for tracking multiple animals in a rich environment to characterize the nature of group behavior and interactions , and show strongly correlated group behavior in mice . We have found that the minimal models that rely only on individual traits and pairwise correlations between animals are not enough to capture group behavior , but that models that include third-order interactions give a very accurate description of the group . These models allow us to infer social interaction maps for individual groups . Using this approach , we show that environmental complexity during adolescence affects the collective group behavior of adult mice , in particular altering the role of high-order structure . Our results provide new experimental and mathematical frameworks for studying group behavior and social interactions . Understanding the nature and impact of the interactions that underlie the behavior of groups of organisms is a central question , shared across biology , physics , psychology , and mathematics . The coherence of ‘collective behavior’ patterns of large groups of animals , such as insect swarms ( Buhl et al . , 2006; Simpson et al . , 2006 ) fish schools ( Sumpter et al . , 2008; Katz et al . , 2011 ) , bird flocks ( Cavagna et al . , 2010; Nagy et al . , 2010 ) , and human crowds ( Song et al . , 2010; Gallup et al . , 2012 ) , presents us with fundamental questions relating to distributed information processing , computation , and learning . The adequacy of different mathematical models of the interactions between animals in describing the behavior of large groups has therefore been of great interest ( Winfree , 1967; Vicsek et al . , 1995; Couzin et al . , 2002; Lathe , 2004; Couzin et al . , 2005; Ben-Jacob , 2009; Cavagna et al . , 2010; Lukeman et al . , 2010; Nagy et al . , 2010; Bialek et al . , 2012 ) . Smaller groups of animals present an interesting and sometimes more difficult scenario , where individual traits may play an important role in shaping the group behavior ( Lathe , 2004 ) . This may be especially true in mammals , where both individual behavior and interactions are often assumed to be more complex . It has therefore been common to study ‘social behavior’ in small groups and explore the interplay of individual and group relations in decision-making ( Couzin et al . , 2011 ) , information transfer ( Leadbeater and Chittka , 2007 ) , learning ( Couzin et al . , 2005 ) , and more . Yet , much of our understanding of social behavior has come from studies of just pairs of animals under artificial settings ( Insel and Fernald , 2004; Langford et al . , 2006; Moy et al . , 2008; Branson et al . , 2009; Dankert et al . , 2009; Blumstein et al . , 2010; Silverman et al . , 2010; Ben-Ami Bartal et al . , 2011; de Chaumont et al . , 2012 ) . It is not clear , however , what the detailed analysis of social interaction at the level of a single pair implies for larger groups . In particular , new features may emerge that characterize the group as a whole that cannot be inferred from the study of individuals or pairs ( Cavagna et al . , 2010 ) . To study the nature of interactions underlying social behavior in a group , we used a novel automatic system for tracking individuals in small groups of mice over long periods of time , in an environment that is ethologically relevant . Systems for tracking individual animals in simple and enriched environments have been used in recent years to characterize individual behavior ( Branson et al . , 2009; Green et al . , 2012; Freund et al . , 2013 ) , and even to relate individual behavior to neurogenesis ( Freund et al . , 2013 ) . We focused here on the nature of group behavior , and in particular how group behavior is the result of individuality , pairwise , and potentially higher-order interactions between the animals . We then used a maximum entropy-based modeling framework ( Jaynes , 1957; Schneidman et al . , 2003 , 2006 ) to quantify the nature of correlated group behavior and map the social interactions between individuals . Finally , we compared the joint activity patterns and social contacts of mice subjected to environmental manipulations . To characterize the behavior of the mice as a group , we studied their joint spatial configurations over long periods of time . We defined the ‘state’ of the group at time t , as a vector , ( x1 , x2 , x3 , x4 ) , where xi denotes the location of mouse i at that time , and xi = 1 , … , 10 , denote the regions defined in Figure 1 . An example of these state vectors as a function of time is shown in Figure 2A , with time bin of Δt = 240 ms ( this choice did not affect the results over a wide range of values , see ‘Materials and methods’ ) . We then compared the empirical probability to find the group in a given spatial configuration pempirical ( x1 , x2 , x3 , x4 ) with the prediction of a model that assumes that the mice choose their locations independently , based on their individual preferences , pind ( x1 , x2 , x3 , x4 ) = p ( x1 ) p ( x2 ) p ( x3 ) p ( x4 ) . This difference is exactly the extent to which the group is different from the case of a collection of independent individuals . We found that the two distributions were very distinct , that is the group behavior is very different from what one would expect from studying single mouse properties . In particular , Figure 2B shows the distribution of observed states for a typical group , where out of the 104 possible states ( of 4 mice in 10 zones ) , only approximately 2000 occurred in the experiment , whereas the independent model predicts that approximately 4000 states would typically occur in our experiment . In other words , the correlations between mice contract the space of ‘allowed’ configurations , such that many of them are socially avoided . 10 . 7554/eLife . 00759 . 006Figure 2 . Characterization of group behavior patterns , and signatures of strong group correlations between mice . ( A ) The joint configuration of the mice at each time frame was represented by a 4-dimensional vector , where each dimension denotes the location of a particular mouse in 1 of the 10 regions of interest . ( B ) Comparison between the empirical probability distribution of the observed configurations and a predicted distribution from a model that assumes independence between mice . Configurations were ranked from the most to the least prevalent . ( C ) Fraction of uncertainty about the location of a mouse that can be read from the location of other group members ( mutual information about location , divided by location entropy ) . Every dot shows the fraction of information about the location of mouse i that can be read from the joint location of the other three vs the sum of pairwise information terms between i and each of the other mice . Each of the 32 dots corresponds to 1 mouse ( 4 mice in 8 groups ) , and the information that can be read from his group members . The results are for day 2 of the test . DOI: http://dx . doi . org/10 . 7554/eLife . 00759 . 006 To quantify the strength of dependencies between the mice , we first asked how much does knowing the location of one mouse tells us about that of the others . ( If the mice were completely individual and ignored one another , then knowing the location of one mouse would tell us nothing about that of the others . ) Since the entropy of the distribution of locations of a mouse , H ( xi ) =−∑xip ( xi ) log2p ( xi ) measures how much we do not know about its location , then the dependency between mice can be measured in terms of how much of this uncertainty is reduced by knowing the location of another mouse . This is exactly the mutual information I ( xi;xj ) =H ( xi ) −H ( xi\|xj ) between the location of one mouse xi and that of the other mouse xj . To get a normalized measure of the fraction of uncertainty about the location of mouse i that can be ‘read’ from mouse j , we divided I ( xi;xj ) by the entropy of location of mouse i , H ( xi ) . We found that knowing the location of one mouse gives relatively little information about the location of another—typically less than 5% ( averaged on all pairs , Figure 5—figure supplement 1 ) . However , knowing the joint locations of three mice gives much more information about that of the fourth one—over all groups and mice I ( xi;{xj , xk , xl} ) was , on average , 25% of H ( xi ) . Figure 2C shows that this information was highly synergistic—namely that the information about the location of one mouse that can be read from the locations of the other mice can be more than double the sum of pairwise information values of that mouse with all the others: I ( xi;xj ) + I ( xi;xk ) + I ( xi;xl ) . Thus , the group is not only more complex than a collection of individuals , but much more complex than even the full collection of pairs . The remaining 75% of uncertainty about the location of each mouse is exactly the level of individuality of each mouse , which cannot be explained in terms of the location of the other mice . If there were more information about the location of one mouse from the locations of the others , then that mouse would be less ‘free’ to decide on its location . We therefore turned to characterize group behavior in terms of the combination of individual mouse traits and the dependencies between mice . Dissecting the role of individual behavior and the dependencies between animals in shaping the group’s behavior is difficult , since we need to infer from the joint behavior what the underlying contributions of ‘pure’ mouse individuality and the nature of the interactions are . The difficulty arises since in general , for any given set of observable features of the behavior , there can be multiple models that will be consistent with these observables . We therefore used the idea of maximum entropy ( ME ) models from physics ( Jaynes , 1957 ) to construct minimal models of the group , based on different order of dependencies between the animals . Since the entropy of a distribution measures its randomness of lack of structure , then among all models that are consistent with some desired feature of the data , the maximum entropy model is the most parsimonious explanation that does not make any additional assumptions beyond the required features . This minimal model is mathematically unique and can be found numerically ( Schneidman et al . , 2003 ) . Such models have been successfully used to infer functional dependencies between neurons , genes , proteins and more ( e . g . , Schneidman et al . , 2006; Ganmor et al . , 2011; Lezon et al . , 2007; Marre et al . , 2009; Mora et al . , 2010; Stephens et al . , 2010; Stephens et al . , 2013; Bialek et al . , 2012 ) . We built a hierarchy of maximum entropy models to describe the group configurations , based on successive orders of correlations between the mice ( one that relies only on individual behavior of the mice , one that adds pairwise dependencies between mice , third order , etc . ) . The relationship between these models then allowed us to dissect exactly the contribution of each order to the total group behavior . The first-order model is one that relies only on the individual behavior of each of the mice , but assumes no dependencies among them at all . The maximum entropy model in this case is built on the observed probability of finding each mouse in one of the regions in the arena , and is exactly the independent model ( that we used above ) , namely p ( 1 ) ( x1 , x2 , x3 , x4 ) = p ( x1 ) p ( x2 ) p ( x3 ) p ( x4 ) . As it was clear already from Figure 2B , the independent model is insufficient to describe the behavior of the group . Next , we tried to describe the group configurations using a model that takes into account both the individual behavior and the pairwise relations between mice . The minimal pairwise-based model is then given by the maximum entropy distribution that is consistent with the distribution of states that we observe for each mouse individually ( i . e . , first-order statistics ) , and the pairwise correlations between them ( i . e . , second-order statistics ) . Unlike the independent case , this cannot be performed by a simple factorized probability distribution and must be found numerically by solving an optimization problem in which we maximize the entropy with a given set of constraints . The solution of this optimization problem ( see ‘Materials and methods’ ) is given byp ( 2 ) ( x1 , x2 , x3 , x4 ) =1Zexp ( ∑iαi ( xi ) fi ( xi ) +∑i<jβij ( xi , xj ) fij ( xi , xj ) ) , where the parameters , αi ( xi ) for each mouse i for location xi and βij ( xi , xj ) for each pair i and j ( one for every combination of locations , xi and xj ) , are set such that the marginal probabilities of the model agree with the empirically observed p ( xi ) and p ( xi , xj ) ; fi ( x ) is an indicator function , which equals 1 when mouse i is in location xi and 0 otherwise , and fij ( xi , xj ) is an indicator function , which equals 1 when mouse i is in xi and mouse j is in xj; Z is the normalization factor , or partition function . We can build more complex models of group behavior by adding orders of interactions between mice . Thus , the third-order model is given by a maximum entropy distribution of a similar form , which has the same single mouse , and pairwise statistics , but also the empirically third-order statistics . This third-order model , p ( 3 ) , has , in addition to αi ( xi ) ’s and βij ( xi , xj ) ’s , interaction parameters for each triplet and locations γijk ( xi , xj , xk ) . The fourth-order model , p ( 4 ) , uses all possible correlations among mice . We emphasize that the maximum entropy models give the most parsimonious explanation of the data for each order , and therefore are not just an arbitrary ‘modeling approach’ but rather the least structured models one could build for the observed data . This hierarchy of maximum entropy models allows us to dissect the role of individual behavior , pairwise relations , triplets and so on , since every model adds a unique set of independent constraints . Figure 3A shows the accuracy of the maximum entropy models of different orders in describing the empirical distribution of the spatial configurations of the mice ( see ‘Materials and methods’ ) . The top left panel shows how poorly the independent model p ( 1 ) describes the empirical distribution of configurations of the group pempirical . This discrepancy ( which was already apparent in Figure 2B ) reflects the effect of the correlations among mice on the group behavior . The top right panel shows that the pairwise model p ( 2 ) was a much better model of the group behavior , and captured much of these correlations in the group , but still shows considerable differences from the empirical data . Thus , the group correlations have a significant higher-order contribution . We see that p ( 3 ) gave a very good approximation to the empirical data ( left bottom panel ) , and was very close to the accuracy of p ( 4 ) that relies on all orders of correlations among mice ( bottom right panel ) . We emphasize that the comparison was performed through cross-validation , namely , ME models were fit to a randomly chosen half of the data ( train data ) , and then compared to the empirical distribution based on the other half ( test data ) . 10 . 7554/eLife . 00759 . 007Figure 3 . High-order maximum entropy models show the role of pairwise and triplewise interactions in shaping the group configurations of one representative group . ( A ) In each panel we present the accuracy of the corresponding ME model , from first to fourth order , in describing the empirical data for the group . Each dot corresponds to one configuration state of the group , and its probability is shown for the data ( x-axis ) and the prediction of the model ( y-axis ) . The grey funnel shows the 95% confidence interval of estimation of the empirical distribution of configurations . Examples of two specific configurations are highlighted in all graphs ( green and blue dot ) , to show improvement of model accuracy over orders . ( B ) Breakdown of the total group correlations , or multi-information IN , to the contribution of the pairwise interactions between mice , I ( 2 ) , triplet interactions between them , I ( 3 ) , and fourth-order contribution I ( 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00759 . 007 To quantify the accuracy of each of these models in capturing the full group configurations , and the relative contribution of each interaction order in explaining the group behavior , we first estimated the total correlations of all orders in the group that go beyond the individual behavior of the mice . We used the ‘multi-information’ of the group ( Schneidman et al . , 2003 ) , which is a generalized form of the mutual information between two variables that measures dependencies among a group of variables IN ( {x1 , x2 , . . . , xN} ) . It measures by how much the dependencies between mice change the configurations of the mice , compared to what would be expected if the mice were independent , through the difference between the entropy of the independent ( first-order ) model and the entropy of the empirical distribution , ( see ‘Materials and methods’ ) . IN can be uniquely broken down to the sum of exact contributions of each order of correlations , when the contribution of order k to IN , is given by I ( k ) =H[p ( k−1 ) ]−H[pk] , where H[pk] is the entropy of the ME models of order k ( Schneidman et al . , 2003 ) . Here , IN = I ( 2 ) + I ( 3 ) + I ( 4 ) , which then gives the second-order , third-order and fourth-order contributions to the correlation in the group beyond what the individuality predicts . We found that over all groups , the contribution of the pairwise ME model , given by I ( 2 ) was 57 . 2% ± 10 . 2% of IN . We found that I ( 3 ) carried nearly a third of the total correlations , and so p ( 3 ) that relies on individual traits , pairwise and triple interactions between the mice explains 92 . 8% ± 2 . 9% of the correlations ( Figure 3B shows as an example the results for the group shown in Figure 3A ) . Thus , for a group of four mice even using all pairwise interactions is not enough to capture the group behavior; the third-order interactions are therefore necessary and capture about a third of the correlation structure . This strong high-order dependency is consistent with the synergistic effect seen in Figure 2 in terms of the information that can be read about the location of a mouse from that of the other mice . The maximum entropy models we have used to describe the group behavior are a generalized form of the Potts model from statistical physics , which describes the behavior of spins in a lattice in terms of the interactions between them ( Landau and Lifshitz , 1980 ) . We therefore interpreted the parameters of our ME models as the interactions between the mice at the different locations; these reflect functional ( rather than physical ) dependencies between the animals . To give as compact an explanation of the social interaction between animals as possible ( and to avoid overinterpretation of the parameters we found in the ME model ) , we tried to identify the dominant and irreducible dependencies between animals . We therefore constructed a third-order ME model , where we tried to minimize the number of parameters of the model . Specifically , we fitted the maximum entropy model but added a constraint in the form of a cost for every interaction term that is not zero ( see ‘Materials and methods’ and Figure 4—figure supplement 1 ) . This standard regularization approach gives a model , p* , that is nearly as accurate as the full third model ( Figure 4A ) , but uses far less parameters ( Figure 4B and Figure 4—figure supplement 1 ) . Figure 4C shows all the pairwise interactions between the mice for one of the groups ( Figure 4—figure supplement 2 ) . We found that most pairwise parameters ( or functional interactions ) were negative , making the corresponding configurations less common than predicted from single mouse preferences , and positive interactions were less common and weaker , making the corresponding configurations occur more than expected from single mouse traits . Figure 4D , E show the most dominant pairwise and triplewise interactions for one group , respectively , overlaid on a drawing of the arena . Importantly , the interaction maps show that mere physical limitations do not play a key factor in shaping the group configurations . In particular , we did not find strong high-order interactions for configurations in which more than two mice are in the same location . ( i . e . , there is no exclusion of these configurations that requires a special interaction that would ‘prevent’ this from happening ) . 10 . 7554/eLife . 00759 . 008Figure 4 . Functional social interaction maps between mice . ( A ) Accuracy of a ‘regularized’ third-order maximum entropy model of the spatial configurations of the same groups of mice from Figure 3A . Model predictions are plotted against the empirical distribution . For details of parameter selections for the regularized model see Figure 4—figure supplement 1 . ( B ) The distribution of ME parameters according to the order of interactions in the regularized p* model ( shown above the horizontal line ) , compared to the model without regularization ( shown below the line ) . The distribution is over parameters of all eight groups of SE mice taken together . ( C ) Full pairwise interaction maps for four representative groups . ( Group S2 is magnified as it is used in following panels . ) In each of these maps , the colored dots represent the location of a mouse according to the color coding in the bottom of the figure . The colors of the mice are depicted near their corresponding locations . The color of a vertex shows whether the interaction is positive ( red ) or negative ( blue ) and its width reflects the interaction strength . An alternative presentation of all the pairwise interaction parameters is shown in Figure 4—figure supplement 2 . ( D ) The dominant positive and negative pairwise interactions are shown overlaid on a diagram of the arena . ‘Filled mice’ show positive interactions , and ‘empty mice’ show negative interactions . A star denotes that the mouse is on the nest . The value of the corresponding interaction is shown on the bottom of each panel . ( E ) The dominant positive and negative triplewise interactions for the same group as in D , overlaid on a diagram of the arena . DOI: http://dx . doi . org/10 . 7554/eLife . 00759 . 00810 . 7554/eLife . 00759 . 009Figure 4—figure supplement 1 . Tradeoff between generalization and accuracy of the maximum entropy model . We found the 3rd order maximum entropy model for the mice configurations , with an additional penalty term that minimized the number of non-zero parameters ( see Materials and methods ) . The balance between maximizing the model's entropy and minimizing the penalty is controlled by parameter ∈ . ( A ) The effect of the tradeoff parameter on the accuracy of the model is shown as the Jensen–Shannon divergence ( DJS ) between the third order maximum entropy model with the penalty term and the model without the penalty term ( as in Figure 3 ) . The Jensen–Shannon divergence equals 0 when the two models are identical , and would be 1 at its maximal value—when the two distributions are distinct . The results are from the second day of the same group as in Figure 3a . ( B ) The fraction of parameters that equal zero is shown for each order ( 1st , 2nd and 3rd order parameters of the maximum entropy model ) is shown as a function of ∈ . The chosen value ∈0=2−16 , which we used in Figure 4 is marked by a dashed line on the graphs . DOI: http://dx . doi . org/10 . 7554/eLife . 00759 . 00910 . 7554/eLife . 00759 . 010Figure 4—figure supplement 2 . All pairwise interactions of a typical group . These interactions are the weights of the second-order interactions in the regularized third-order maximum entropy model . Each panel corresponds to one pair of mice , and its rows and columns correspond to the locations within the arena according to the legend at the bottom of the figure . DOI: http://dx . doi . org/10 . 7554/eLife . 00759 . 010 Since the nature of the environment and availability of resources determine population density , aggression , dominance , and territoriality in mice ( Bronson , 1979; Haemisch et al . , 1994; Van Loo et al . , 2001 ) , we asked how raising mice in a complex and more populated environment ( Sztainberg and Chen , 2010; Sztainberg et al . , 2010 ) might affect their group behavior . We found that groups raised in standard laboratory conditions environment ( SE , n = 8 ) and those raised in a complex environment ( CE , n = 9; Figure 5A ) , already showed distinct behavior at the individual level , as CE mice spent significantly more time inside the large nest and less time outside ( Figure 5B ) . But more importantly , we found clear differences between SE groups and CE ones in terms of the overall group behavior and , in particular , the nature of the interactions in the group that go beyond single mouse individuality . Given their accuracy in describing the group behavior , we used the third-order models that we fitted for each group separately to compare the distribution of the spatial configurations in the arena . We found that CE groups were more similar to other CE groups than to SE groups ( and vice versa ) in terms of the overall distribution of observed configurations of the mice ( Figure 5C ) . 10 . 7554/eLife . 00759 . 011Figure 5 . Environmental background changes group behavior and interactions . ( A ) Experimental design . At the age of 4 weeks ( day 0 of the experiment ) , mice were separated into two different housing conditions: standard environment and complex environment . After 6 weeks , groups of four mice from both treatments were color marked , returned to their cages for an additional week , and then put in the arena for recording their group social behavior . ( B ) Behavioral ethograms of two representative groups from each treatment ( left ) . Data shown in these panels is for the 12 hr of the second day . Average percentage of time spent at the different regions over all groups for each treatment ( right ) . ( C ) Similarity of group behavior between all SE and CE groups . For each pair of groups , the Jensen–Shannon divergence between the third-order maximum entropy models of the groups was calculated . Matrix entries are ordered by their corresponding SE or CE label . ( D ) Total group correlation ( multi-information , IN ) of the SE and the CE groups over 4 consecutive days . ( E ) The contribution of each order of interaction to the total correlation in the groups . Figure 5—figure supplement 1 presents the differences in the distribution of the fraction of information about the location of one mouse that can be ‘read’ from the location of the other mice for SE and CE groups . DOI: http://dx . doi . org/10 . 7554/eLife . 00759 . 01110 . 7554/eLife . 00759 . 012Figure 5—figure supplement 1 . Histogram of the fraction of information about the location of one mouse that can be ‘read’ from the location of the other mice . Top: mice in standard conditions . We estimated pairwise mutual information between all pairs , and normalized by mouse entropy ( main text ) . Histogram of values over all mice is shown as a horizontal bar in blue , with median as central line , and the left and blue boxes show the range of 25% and 75% values , correspondingly . The histogram of the fraction of information that can be read from the joint location of the other three mice is shown in green . The sum of pairwise information for each mouse with the others is shown in red . Bottom: same as top , but for complex environment groups . Green star denotes the significant difference between the histograms of the group information values of complex and standard groups ( Klomogorov–Smirnoff , p<0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00759 . 012 We found that the total group correlation IN values of the SE and CE groups were similar on the first day , but then the SE groups became more correlated . In contrast , the correlations among the CE mice remained unchanged ( Figure 5D ) . In other words , there was a progressive increase in social correlation , or ‘socialization’ in the SE groups , which was absent in the CE groups . We emphasize that these are differences at the level of group behavior that go beyond the differences between the individual ( single animal ) behavior patterns of SE and CE mice . Finally , we found that the contribution of the different orders of interactions to the group behavior differed significantly between the SE and CE groups . In the CE groups the contribution of pairwise interaction to the full group correlations was higher than in SE groups ( 74 . 6% ± 2 . 5% in CE groups compared to 61 . 9% ± 2 . 4% for the SE ones , averaged over all four days ) . The dominance of the low-order interactions was also reflected by the virtual lack of contribution of fourth-order interaction in the CE groups , whereas the SE groups showed more complex high-order structure ( fourth-order interactions contribution to the total correlation in the SE groups was 6 . 0% ± 0 . 7% [Figure 5E] ) . The larger role of pairwise interactions in shaping the group’s behavior is also reflected in the information about the location of one mouse that can be read from the locations of the other three mice . In the CE groups the information from three mice about the fourth was far less synergistic than in the case of SE groups ( Figure 5—figure supplement 1 ) . Thus , mice exposed to a complex environment during adolescence were more individual , and their weaker group correlations relied more on pairwise rather than higher-order dependencies . Quantifying social interactions presents an ethological challenge both experimentally and theoretically ( Adolphs , 2010 ) . While even solitary species display social behavior such as mating , aggression , and maternal care , species that live in groups display profoundly more complex social repertoires ( Silk , 2013 ) . This group behavior , ideally dissected into individual and group parts , is also likely to depend on the environmental context ( Insel and Fernald , 2004 ) . Thus , understanding group social behavior requires a framework combining an experimental system for recording group behavior with high resolution both spatially and temporally in a reliable manner over long time windows , and a mathematical formalism to quantify the nature of interactions and their contribution to the group’s behavior . Mice are an ideal model organism for investigating social behavior in mammals and the neural mechanisms that underlie it . Together with the ability to manipulate their genomic make-up ( Lewandoski , 2001 ) , and record neural activity ( electrophysiologically or optically ) , mice live in groups and form diverse societies with different characteristics such as group size , hierarchy , aggression , and social tolerance ( Bronson , 1979 ) . However , despite the complex nature of their social organization , the current methodologies used for analysis of social behavior in mice have focused mainly on dyadic interactions such as in the classical three-chamber social approach test and the partition test ( Silverman et al . , 2010 ) . One common approach has been to record , via video , the interaction between two animals and then have defined behaviors scored by trained human observers ( Moretti et al . , 2005 ) . This allows for identifying intricate social behaviors and can provide new insights about underlying features of social interaction , but demands immense human resources and is prone to human error . Another approach has focused on the behavior of one individual towards other restricted conspecifics ( Nadler et al . , 2004; Moy et al . , 2008; Ben-Ami Bartal et al . , 2011 ) , which allows for an automated behavioral scoring system . However , since only one animal is free to roam , its behavior might be altered due to the synthetic dynamics of interactions . de Chaumont et al . ( de Chaumont et al . , 2012 ) reported an automated video tracking system for social interaction between two rats , which were analyzed based on their relative locations in a 10 min test . This method holds the advantages of both rats roaming free and the use of an automated system; however , the ability to distinguish between different behaviors is limited . The system we introduce here enables automatic tracking of group behavior of mice in the dark , over long periods of time , and in a semi-natural environment , with high spatiotemporal resolution while maintaining individual identities . Similar systems for tracking multiple individuals ( Freund et al . , 2013 ) allow for tracking animals over long periods of time , using radio-frequency identification ( RFID ) tagging of individual animals . We note that RFID-based systems allow for the tracking of a very large number of animals , whereas our tracking capacity depends on the number of distinguishable dyes and spatial marking patterns on the mice ( our preliminary results suggest we can expand even the current system to more than 10 mice ) . However , the strength of our system is in its much higher spatial and temporal resolution , and the ability to track and analyze details of individual behavior of the animals and between them . Clearly , these kinds of systems would change the way individual and social behavior can be studied and quantified . The recent work by Freund et al . used the tracking of many mice over several months to study the individual mouse behavior within a large group and showed correlation between the roaming behavior of a mouse and the level of neurogenesis in its hippocampus ( Freund et al . , 2013 ) . The work we have presented here addresses the complementary question of the nature of group behavior that goes beyond individual traits , focusing on the interactions between animals . Our results show the limitations of individual-based and even pairwise-based approaches , and identify irreducible high-order structure among mice . Moreover , although every individual group is likely to have its own unique nature , hierarchy , traits and rules , we were still able to identify universal features of the groups that govern their behavior and distinguish different behavioral contexts . Combining detailed behavioral and genetic analysis at the level of individuals as seen in Freund et al . ( 2013 ) , in association with the kind of group analysis used , may enable the identification of genetic and neuronal correlates of complex social interactions . Our analysis of the groups relied on a representation of the mice in their preferred locations in the arena . This is a discretized version of the full physical space , but even at this level the number of potential group configurations , which is exponential in the number of animals , is very large . We found strongly correlated group structure among the animals , which dictate which configurations are permitted and which are not . Moreover , we found that more information was obtained from the joint position of the other mice than from summing all the information provided by the interactions between the pairs of mice . To assess the contribution of individuality and of pairwise and higher-order interactions among the mice , we used tools from information theory to quantify any kind of dependency , of any order , be it linear or non-linear . Intriguingly , the pairwise-based model of the group that assumed no higher-order contribution could only explain approximately 60% of the correlation structure in the group , whereas models that included also third-order dependencies ( but not fourth-order ones ) captured approximately 90% of the group correlation structure . How should one interpret these results together ? The ME model shows how well we can describe the distribution , whereas the information about location reflects how deterministic the behavior of one mouse is given the others . What we can read about the location of a mouse from the location of the others is much more than what one would naively expect from the pairwise relations between mice , which amounts to approximately 5% . This strong synergistic effect is the result of high-order dependency between the animals ( which the maximum entropy models reflect ) , but it is still the case that three-quarters of the uncertainty we have about the location of a mouse comes from its own individuality . That is , the mice still have a significant individual component , even given the other mice . The need for models that include high-order interactions is surprising , since intuitively one might have expected that it would be possible to construct a mathematically accurate description of the group once all the interactions between pairs of mice were known . After all , most social behaviors , such as fights , chases , courtship , and grooming , are usually observed in pairs . Our analysis of the social interaction network underlying the group behavior relied on a family of maximum entropy models , which enabled us to uniquely dissect the contribution of different orders of correlations in the group . This approach has been useful for different biological networks , from small to large networks of neurons ( Schneidman et al . , 2006; Shlens et al . , 2006; Marre et al . , 2009; Ohiorhenuan et al . , 2010; Ganmor et al . , 2011 ) , genes ( Lezon et al . , 2007 ) , T cells ( Mora et al . , 2010 ) , letters in words ( Stephens and Bialek , 2010 ) , the structure of images ( Stephens et al . , 2013 ) , and birds in large flocks ( Bialek et al . , 2012 ) . Interestingly , in almost all these cases the contribution of pairwise interactions was very large and dominated the network structure , especially in small networks—in clear contrast to what we found for the mice . The parameters of the maximum entropy models can be interpreted as functional social interactions between the animals ( similar to the parameters of the corresponding Potts models from physics ) . We emphasize that these functional interactions reflect statistical dependencies , and will probably differ from explicit physical interactions between the animals that one could measure . Yet , these statistical dependencies highlight the most prominent relations that underlie the patterns of group behavior . The strongest functional interactions corresponded not to the most frequent events , but rather to those events that are most surprising or not predicted from lower orders of interactions , thus presenting interactions of a truly social nature . Our results show that the dominant interactions in the group are negative ones , namely compound configurations that tend not to happen compared to expectations based on individual behavior . This may suggest competition as a dominant force in the social structure . In addition , the relative sparseness of the interaction maps indicates that even a small number of social events can have a strong , macroscopic impact on the group . As an example of how the combination of group tracking and the analysis based on information theory tools can enhance our understanding of the effect of external factors on group behavior , we compared the effects of different environmental exposures on social behavior . We found that growing up in a complex environment with more mice , which better resembles a natural habitat , resulted in groups that were far less correlated as a group , and their social structure could be explained to a much larger degree based on pairwise interactions . We suggest that this approach could now enable the quantitative characterization of many different aspects of group behavior that have so far only been studied in much more restricted set-ups , such as the effects of stress , rewards , and learning on the group . Several technical and mathematical issues should be further explored to allow the extension of our approach to other groups of animals and contexts . First , we reiterate that our analysis has focused on a reduced description of the mice configurations ( regions of interest ) , and not absolute or relative coordinate space . While it is not immediately clear how to construct such models , they have the potential to reveal new features and dependencies in the group , and with respect to cues from the environment as well . Moreover , it would be interesting to consider how our analysis might be related to more standard hierarchy models in groups . Second , we have focused on the joint configurations of the mice at given time points , and have not included temporal correlations between them . Third , it will be interesting to consider how the number of animals in the group affects the nature of group correlations and the contributions of the different interaction orders . Preliminary results suggest that our system can be expanded in terms of tracking more mice using additional dyes and using spatial color patterns on the mice . We expect that mapping of the social interactions among other and larger groups of animals , and their dynamics , will change our understanding of group behavior in terms of the interplay between genetics , individuality , environment , and social hierarchy . Of particular interest would be the extension of our approach to study animal models of maladaptive social behavior . For example , our analysis would allow identifying mutants that rely on different kinds of low or high-order interactions compared to wild type littermates; such analysis would be useful for studying mechanisms underlying social intolerance and group stability , as well as models of autism and schizophrenia . Because our approach is based on high throughput as well as high spatiotemporal resolution , it may also be useful in detecting subtle changes in social behavior in mice that may not be detectable in standard social behavior paradigms even for standard parameters such as exploration , feeding or locomotor activity . Mice were studied in a specialized arena designed for automated tracking of individual and group behavior . The arena consisted of an open 70 cm × 50 cm × 50 cm box and included the following objects: Z-shaped wall , a water dispenser , two feeders , a small nest and a large nest , an elevated block , and two elevated ramps ( Figure 1—figure supplement 1 ) . Food and water were given ad libitum . Two UVA fluorescent lamps ( 18 W ) were placed 3 m above the arena’s floor to illuminate the surrounding area with 370–380 nm ‘black light’ . To avoid reflections from white objects in the room , a black curtain was drawn from the fluorescent lamps down to the arena . A color sensitive camera ( Panasonic Color CCTV , WV-CL924AE ) was placed 1 m above the arena . The camera analog input is converted to digital information with a digitizer ( Picolo Diligent frame grabber board ) , and recoded on a standard computer . Mice trajectories were automatically detected offline using specially written software in Matlab ( Mathworks , Natick , MA ) . Adult male ICR mice ( Harlan Laboratories , Jerusalem , Israel ) were used for the standard and complex environment experiment . Throughout the experiments , the animals were maintained in a temperature-controlled mouse facility ( 22°C ± 1 ) on a reverse 12 hr light–dark cycle . Food and water were given ad libitum . All experimental protocols were approved by the Institutional Animal Care and Use Committee of The Weizmann Institute of Science . Mice were mildly anesthetized with a mix of ketamine ( 70 mg/kg ) and xylazin ( 7 mg/kg ) . Their eyes were protected against drying using eye gel ( viscotears liquid gel; Alcon ) . The fur of the mice was stained using a regular brush with fluorescent semi-permanent hair dyes that glow under black light . The fur was dried with a fan ( low power and heat ) for 3 min . After awakening , mice were kept in separate carton boxes for 4 hr before reunion . Mice were introduced to the arena for tracking 5 days after the fur staining . The dyes used were Electric banana , ( HCR 11012 ) , composed of natural ingredients , Virgin snow white ( HCR 11033 ) , and Raven ( HCR 11007 ) , from Tish & Snooky’s ( manicpanic . com ) , and High octane orange , from specialeffectsusa . com . Color under black lightColor under regular lightDyes ratioGreenYellow100% Electric bananaPurpleWhite100% Virgin snow whiteWhiteYellowish20% Electric banana , 80% Virgin snow whiteRedRed80% High octane orange , 20% Electric bananaOrangeOrange20% High octane orange , 80% Electric bananaBlackBlack100% Raven At the age of weaning ( 4 weeks ) , mice were randomly distributed into 2 types of groups: standard environment ( SE ) mice that were housed in groups of 4 in standard laboratory cages , and complex environment ( CE ) mice that were housed in groups of 16 male mice in a relatively spacious and complex cage , with a variety of objects such as shelters , tunnels , running wheels , and mouse nest boxes ( Sztainberg and Chen , 2010 ) . After a period of 6 weeks , CE mice were randomly divided into groups of four , color marked , and introduced to the novel arena , as the SE mice , for analysis of group social behavior . Mice were identified and tracked automatically , according to their fur colors , which were learned from labeled data . Because of low signal-to-noise ratio , due to the dim lighting and the camera’s sensitivity , some frames had reflection artifacts or missing parts . To overcome this noise , we used a Bayesian model to infer the most likely location of a mouse given the observed location of connected colored blobs . Validation of the tracking algorithm was performed by comparing the algorithm’s performance to human labeling of 500 randomly chosen frames , which gave 99 . 6% accuracy . The raw camera acquisition rate was 25 frames/s ( 40 ms per frame ) . In analyzing the configurations of the mice we used a lower resolution of 240 ms per frame , as this did not have a major effect on the state distribution , but was more robust to single frame noise . The uncertainty about the location of mouse i is given by the entropy of its location distribution , H ( xi ) =−∑xip ( xi ) log2p ( xi ) . The mutual information between the location of mouse i and that of mouse j is given by I ( xi;xj ) =H ( xi ) −H ( xi\|xj ) , where H ( xi\|xj ) =−∑xi , xjp ( xi , xj ) log2p ( xi\|xj ) is the average conditional entropy or uncertainty about mouse i given the location of mouse j . The fraction of the uncertainty about the location of mouse i that can be extracted from the location of mouse j is then given by I ( xi;xj ) H ( xi ) . The fraction of uncertainty about the location of mouse i that can be ‘read’ from the joint location of the three other mice is given by ( 1 ) I ( xi;{xj , xk , xl} ) H ( xi ) The naive additive pairwise information fraction was defined as ( 2 ) I ( xi;xj ) H ( xi ) +I ( xi;xk ) H ( xi ) +I ( xi;xl ) H ( xi ) The total correlation of all orders between mice was quantified by the multi-information of the group ( Schneidman et al . , 2003 ) ( 3 ) IN ( {xi} ) =∑jH ( xj ) −H ( {xi} ) =∑{xi}p ( {xi} ) log2p ( {xi} ) ∏jp ( xj ) where the joint entropy of the mice configurations is defined by ( 4 ) H ( {xi} ) =−∑{xi}p ( {xi} ) log2p ( {xi} ) ( where {xi}={x1 , x2 , x3 , x4} ) and the entropy of the independent mice model or the sum of the entropies of the mice is ∑iH ( xi ) . For a given set of observed average functions of the group , <fi ( {xi} ) > , the maximum entropy model , which is the minimally structured model that is consistent with these measured functions , is given by ( 5 ) p ( {xi} ) =1Zexp ( ∑iλifi ( {xi} ) ) where λi are set such that the average of each fi under the model , <fi>p is identical to the empirical expectation value , <fi>empirical , and Z is the normalization or partition function ( Jaynes , 1957 ) . For each group of mice , we then find a hierarchy of maximum entropy models that gives the minimal description of the mice configurations , relying only on pairwise correlations between mice ( p ( 2 ) ) , pairwise and triplewise correlations ( p ( 3 ) ) , and all correlations ( p ( 4 ) ; pairs , triplets , and quadruplet correlations ) . The constraints of each of these maximum entropy models are the empirical marginal of different orders , that is , single mice pempirical ( xi ) , pairs pempirical ( xi , xj ) , and so on . For example , the pairwise model is the maximum entropy distribution over all mice , such that the marginal probabilities p ( xi ) and p ( xi , xj ) , the pairwise marginal probability to find mouse i and mouse j in location xi and xj , are the same as empirically found in the data . Formally , we seek p ( {xi} ) that maximizes ( 6 ) L ( p ( {xi} ) , {αi ( xi ) } , {βij ( xi , xj ) } ) =−∑{xi}p ( {xi} ) log2p ( {xi} ) −∑i∑xiαi ( xi ) ( p ( xi ) −pempirical ( xi ) ) −∑i<j∑xi , xjβij ( xi , xj ) ( p ( xi , xj ) −pempirical ( xi , xj ) ) −λ0 ( ∑{xi}p ( {xi} ) −1 ) The resulting maximum entropy distribution is given by ( 7 ) p ( 2 ) ( {xi} ) =1Z2exp ( ∑i∑xiαi ( xi ) fi ( xi ) +∑i<j∑xi , xjβij ( xi , xj ) fij ( xi , xj ) ) where the Lagrange multipliers αi ( xi ) and βij ( xi , xj ) have to be chosen to satisfy the constraints , and fi ( xi ) is an indicator function which equals 1 if mouse i is in location xi , and 0 otherwise; the partition function Z2 is a normalization factor . The maximum entropy models were fit using a combination of the generalized iterative scaling algorithm ( Darroch and Ratcliff , 1972 ) , and a maximum-likelihood optimization using a variant of the gradient descent algorithm with line search ( Nesterov , 2005 ) . The maximum entropy models of different orders form a hierarchy of correlation-based descriptions of the mice , from p ( 1 ) where all mice are independent , to p ( 4 ) which is an a description that allows arbitrary complex interactions; their entropies decrease monotonically toward the true entropy . ( 8 ) H[p ( 1 ) ]≥H[p ( 2 ) ]≥…≥H[p ( N ) ]=H The multi-information IN = H[p ( 1 ) ] − H[p ( N ) ] can be broken down to the sum of contributions of each order of correlation , where the k’th order contribution is given by I ( k ) = H[p ( k-1 ) ] − H[p ( k ) ] , and IN = I ( 2 ) + I ( 3 ) + … + I ( N ) . To build a more compact model for the mice configurations and isolate the significant functional correlations between the mice , we constructed a model , p* , for the mice configurations that has the maximal entropy given a set of constraints , but also minimizing the total sum of the non-zero parameters of the model . Thus we added a penalty term ( ‘regularization’ ) , to the standard maximum entropy optimization problem from equation 6 , and maximize ( 9 ) L ( p ( {xi} ) , {αi ( xi ) } , {βij ( xi , xj ) } , {γijk ( xi , xj ) } ) −ε0 ( ∑i\|αi\|+∑i<j\|βij\|+∑i<j<k\|γijk\| ) where ε0 is an adjustable parameter that controls the trade off between maximizing the entropy and minimizing the total sum of absolute values of the parameters or the L1 norm , also known as lasso regularization ( Dudık et al . , 2007 ) . Similarity between groups was quantified by the Jensen–Shannon divergence ( DJS ) between the regularized third-order models of the groups ( Lin , 1991 ) . Since the mice were arbitrarily labeled , we used the permutation of mouse identities that gave the smallest value of DJS between two groups . Thus the distance between groups i and j is ( 10 ) d ( i , j ) =minπDjs ( pi ( {xk} ) , pj ( π{xk} ) ) , where π is a permutation of the mice labels such that ( 11 ) π ( x1 , x2 , x3 , x4 ) = ( xk1 , xk2 , xk3 , xk4 ) where k1∈{1 , . . . , 4} and are unique .	All animals need to interact with others of the same species , even if it is only to mate . To date , social behavior has been studied mainly at two extremes: detailed observation of pairs; and studies of the collective behavior of large groups , such as flocks of birds . However , to gain an understanding of social behavior in mammals will require an approach that falls between these two extremes . It will be necessary to study animals in larger groups , rather than in pairs , but also to track individuals rather than looking at the activity of the group as a whole . Now , Shemesh et al . have developed a system that can track the behavior of each of four mice with high spatial and temporal resolution as they move around freely in an arena containing ramps , nest boxes , and barriers . Because mice are largely nocturnal , Shemesh et al . dyed the animals’ fur with compounds that produced different coloured fluorescence under ultraviolet light , and then employed an automated system to accurately track each mouse during 12 hr of darkness , over a number of days . Using these data it was possible to estimate the extent to which the behavior of the group is determined by the characteristics of individual mice and how much is determined by interactions between animals . Models based solely on the behavior of individuals could not accurately describe the behavior of the group . Surprisingly , neither could models that focused on interactions between pairs of mice . Only models that included interactions between three mice gave a good approximation of the observed behavior . This shows that , even in a small group , social behavior is determined by relatively complex interactions . Shemesh et al . also found that the behavior of the mice depended on the environment in which they had been raised . Animals that had lived in larger groups and in more interesting enclosures were influenced more by pairwise interactions , and less by three-way interactions , than mice that had been raised in a standard laboratory environment . This suggests that being raised in a complex environment strengthens mouse ‘individuality’ . The approach developed by Shemesh et al . could be extended to study larger groups of animals and could also be used to examine the interplay between genes , environment and other factors in shaping social interactions .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2013	High-order social interactions in groups of mice
Populations are often divided categorically into distinct racial/ethnic groups based on social rather than biological constructs . Genetic ancestry has been suggested as an alternative to this categorization . Herein , we typed over 450 , 000 CpG sites in whole blood of 573 individuals of diverse Hispanic origin who also had high-density genotype data . We found that both self-identified ethnicity and genetically determined ancestry were each significantly associated with methylation levels at 916 and 194 CpGs , respectively , and that shared genomic ancestry accounted for a median of 75 . 7% ( IQR 45 . 8% to 92% ) of the variance in methylation associated with ethnicity . There was a significant enrichment ( p=4 . 2×10-64 ) of ethnicity-associated sites amongst loci previously associated environmental exposures , particularly maternal smoking during pregnancy . We conclude that differential methylation between ethnic groups is partially explained by the shared genetic ancestry but that environmental factors not captured by ancestry significantly contribute to variation in methylation . Race , ethnicity , and genetic ancestry have had a complex and often controversial history within biomedical research and clinical practice ( Risch et al . , 2002; Cooper et al . , 2003; Yudell et al . , 2016; Burchard et al . , 2003; Phimister , 2003 ) . For example , race- and ethnicity-specific clinical reference standards are based on population-based sampling on a given physical trait such as pulmonary function ( Hankinson et al . , 1999; Quanjer et al . , 2012 ) . However , because race and ethnicity are social constructs and poor markers for genetic diversity , they fail to capture the heterogeneity present within racial/ethnic groups and in admixed populations ( Borrell , 2005 ) . To account for these heterogeneities and to avoid social and political controversies , the genetics community has grouped individuals by genetic ancestry instead of race and ethnicity ( Yudell et al . , 2016 ) . Indeed , recent work from our group and others have demonstrated that genetic ancestry improves diagnostic precision compared to racial/ethnic categorizations for specific medical conditions and clinical decisions ( Kumar et al . , 2010; Udler et al . , 2015; Nalls et al . , 2008 ) . However , racial and ethnic categories also reflect the shared experiences and exposures to known risk factors for disease , such as air pollution and tobacco smoke , poverty , and inadequate access to medical services , which have all contributed to worse disease outcomes in certain populations ( Nguyen et al . , 2014; Evans and Kantrowitz , 2002 ) . Thus , it is unclear whether defining groups through genetic ancestry can capture these shared exposures . In this work we seek to explore the contributions of genetically defined ancestry and social , cultural and environmental factors to understanding differential methylation between ethnic groups . Epigenetic modification of the genome through methylation plays a key role in the regulation of diverse cellular processes ( Smith and Meissner , 2013 ) . Changes in DNA methylation patterns have been associated with complex diseases , including various cancers ( Kulis and Esteller , 2010 ) , cardiovascular disease ( Udali et al . , 2013; Kato et al . , 2015 ) , obesity ( Bell et al . , 2010 ) , diabetes ( Chambers et al . , 2015 ) , autoimmune and inflammatory diseases ( Liu et al . , 2013 ) , and neurodegenerative diseases ( Lardenoije et al . , 2015 ) . Epigenetic changes are thought to reflect influences of both genetic ( Bell et al . , 2011 ) and environmental factors ( Feil and Fraga , 2011 ) , and have been shown to vary between racial groups ( Barfield et al . , 2014 ) . The discovery of methylation quantitative trait loci ( meQTL’s ) across populations by Bell et al . established the influence of genetic factors on methylation levels in a variety of tissue types ( Bell et al . , 2011 ) , with meQTL’s explaining between 22% and 63% of the variance in methylation levels . Multiple environmental factors have also been shown to affect methylation levels , including endocrine disruptors , tobacco smoke ( Joubert et al . , 2012 , 2016 ) , polycyclic aromatic hydrocarbons , infectious pathogens , particulate matter , diesel exhaust particles ( Jiang et al . , 2014 ) , allergens , heavy metals , and other indoor and outdoor pollutants ( Ho et al . , 2012 ) . Psychosocial factors , including measures of traumatic experiences ( Chen et al . , 2013; Ressler et al . , 2011; van der Knaap et al . , 2014 ) , socioeconomic status ( Lam et al . , 2012; Borghol et al . , 2012 ) , and general perceived stress ( Vidal et al . , 2014 ) , also affect methylation levels . Since both genetic and environmental exposures affect methylation , this represents an ideal phenotype to explore the contributions of these two factors on differential methylation between ethnic groups . In this work , we leveraged genome-wide methylation data in 573 Latino children of diverse Latino sub-ethnicities enrolled in the Genes-Environment and Admixture in Latino Americans ( GALA II ) study ( Oh et al . , 2012 ) whose genetic ancestry had been determined from dense genotyping arrays . This allowed us to explore the extent to which the differences in methylation between Latino sub-groups could be explained by their shared genetic ancestry . We found that many of the methylation differences associated with ethnicity could be explained by shared genetic ancestry . However , even after adjusting for ancestry , significant differences in methylation remained between the groups at multiple loci , reflecting social and environmental influences upon methylation . Our findings have important implications for both the use of ancestry to capture biological changes and of race/ethnicity to account for social and environmental exposures . Epigenome-wide association studies in diverse populations may be susceptible to confounding due to environmental exposures in addition to confounding due to population stratification ( Michels et al . , 2013 ) . The findings also have implications for the common practice of considering individuals of Latino descent , regardless of origin as a single ethnic group . Differences in ethnicity and ancestry resulted in discernible patterns in the global methylation profile as demonstrated in a multidimensional scaling analysis ( Figure 2A ) . As expected ( Houseman et al . , 2012; Lam etal . , 2012 ) , the first few principal coordinates are strongly correlated to imputed cell composition ( Figure 2B–C ) . There are also significant associations of self-identified sub-ethnicity with PC2 ( p-ANOVA = 0 . 003 ) , PC3 ( p-ANOVA = 0 . 004 ) , PC6 ( p-ANOVA = 0 . 0001 ) , PC7 ( p-ANOVA = 0 . 0003 ) ( Figure 3A ) , and PC8 ( p-ANOVA = 0 . 0003 ) , after adjusting for age , sex , disease status , cell components , and technical laboratory factors ( plate and position ) . Genetic ancestry was associated with PC3 ( p=0 . 002 ) , PC7 ( p=0 . 0004 ) ( Figure 3B ) and PC8 ( p=0 . 001 ) in a two degree of freedom ANOVA test , adjusting for age , sex , disease status , cell components , technical factors , and ethnicity . Supplementary file 1A summarizes the results of the simple correlation analysis of methylation with ethnicity and ancestry , as well as the adjusted nested ANOVA models described above and the mediation results described below . 10 . 7554/eLife . 20532 . 005Figure 2 . Patterns of global methylation . ( A ) Distribution of the first 10 principal coordinates of the methylation data . Plots in the diagonal show the univariate distribution; those in the lower left triangle show bivariate relationship between each pair of PCs , while those in the upper right show the bivariate density . ( B ) Bivariate or ANOVA associations between principal coordinates and technical factors ( chip , position ) , cell counts , genetic ancestry ( European , Native American , African ) , recruitment site ( New York , NY , San Francisco , CA , Chicago , IL , Houston , TX , and Puerto Rico ) , demographic factors ( ethnicity , age , sex ) , and case status . ( C ) Correlation coefficients between the various factors and principal coordinates . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 00510 . 7554/eLife . 20532 . 006Figure 3 . Associations between ethnicity , ancestry and global methylation . ( A ) Association between ethnicity and principal coordinate 7 . ( B ) Association between Native American ancestry proportion and PC7 , colored by ethnicity . Native American ancestry explains approximately 81% of the association between PC7 and ethnicity . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 006 A mediation analysis ( Tingley et al . , 2014 ) revealed that the associations between ethnicity and PCs 3 , 7 , and eight were significantly mediated by Native American ancestry , which explained ~100% ( 95% CI: 37–100% , p=0 . 01 ) of PC3 , 83% ( 95% CI 37–100% , p<0 . 001 ) of PC7 and 66% ( 95% CI: 25% to 100% , p<0 . 001 ) of PC8 . Inclusion of Native American ancestry in the regression model of PCs 3 , 7 , and eight caused the ethnicity associations to be non-significant . However , the associations of ethnicity with PCs 2 and 6 were not explained by Native American , African or European ancestry ( mediation p>0 . 05 ) , suggesting that the ethnic differences in these principal components are associated with global methylation patterns not captured by the shared genetic ancestry of each ethnic group . When genetic ancestry was regressed on the methylation data with the principal coordinates recalculated using the residuals of the regression between methylation and ancestry , there was an association between ethnicity and PC6 ( p-ANOVA = 0 . 003 ) . However , there was no association with any of the other principal coordinates . These observations suggest that while shared genetic ancestry can explain over 50% of the association between ethnicity and global methylation patterns in three PC’s , other non-genetic factors , such as environmental and social exposure differences associated with ethnicity influence methylation and are not captured by measures of genetic ancestry in two others . An epigenome-wide association study of self-identified ethnicity ( see Materials and methods for details of ascertainment of ethnicity ) and methylation identified a significant difference in methylation M-values between ethnic groups at 916 CpG sites at a Bonferroni-corrected significance level of less than 1 . 6 × 10−7 ( Figure 4A and Supplementary file 1B ) . The most significant association with ethnicity occurred at cg12321355 in the ABO blood group gene ( ABO ) on chromosome 3 ( p-ANOVA 6 . 7 × 10−22 ) ( Figure 4B ) . A two degree of freedom ANOVA test for genomic ancestry was also significantly associated with methylation level at this site ( p=2 . 3×10−5 ) ( Figure 4C ) , and when the analysis was stratified by ethnic sub-group , showed an association in both Puerto Ricans and Mexicans ( p=0 . 001 for Puerto Ricans , p=0 . 003 for Mexicans ) . Although adjusting for genomic ancestry attenuated the effect of ethnicity , a significant association between ethnicity and methylation remained ( p=0 . 04 ) . Recruitment site , an environmental exposure proxy , was not significantly associated with methylation at this locus ( p=0 . 5 ) , suggesting that environmental differences associated with ethnicity beyond geography and ancestry are driving the association . 10 . 7554/eLife . 20532 . 007Figure 4 . Associations between ethnicity and methylation ( A ) Manhattan plot showing the associations between ethnicity and methylation at individual CpG loci . ( B ) Violin plot showing one such locus , cg19145607 . Mexicans are relatively hypermethylated compared to Puerto Ricans ( p=1 . 4×10–19 ) . ( C ) Plot showing the association between Native American ancestry at the locus and methylation levels at the locus colored by ethnicity; Native American ancestry accounts for 58% of the association between ethnicity and methylation at the locus . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 007 To determine the contribution of shared genetic ancestry and other factors associated with ethnicity , we repeated the analysis adjusting for ancestry . A significant association remained in 314 of the 834 ( 37 . 8% , p=1 . 7×10−183 for enrichment ) CpG sites associated with ethnicity ( Figure 5A and Supplementary file 1B ) ( 82 sites were excluded because they demonstrated unstable coefficient estimates and inflated standard errors due to strong correlations between ethnicity and ancestry , especially Native American ancestry [see Figure 1] ) . Table 2 and Figure 5b show the proportion of variance explained by ethnicity , genomic ancestry , and their joint effect in the 916 CpG’s associated with ethnicity , as well as the 314 CpG’s that remained associated with ethnicity after adjustment for ancestry and the 520 CpG’s whose association with ethnicity was no longer significant when ancestry terms were introduced into the model . Even after adjusting for genomic ancestry , ethnicity explained 1 . 7% ( IQR 0 . 785% to 3 . 0% ) but as much as 13 . 4% of the variance in methylation across these loci . Genomic ancestry explained a median of 4 . 2% ( IQR 1 . 8% to 8 . 3% ) of the variance in methylation at all loci associated with ethnicity and accounts for a median of 75 . 7% ( IQR 45 . 8% to 92% ) of the total variance in methylation explained jointly by ethnicity and ancestry ( median of 6 . 8% , IQR 4 . 5% to 10 . 0% ) ( Figure 5B ) . 10 . 7554/eLife . 20532 . 008Figure 5 . Relationship between genomic ancestry and the association between ethnicity and methylation . ( A ) Venn diagram showing the effect of adjustment for ancestry on the association between ethnicity and methylation . The components of the diagram represent the number of CpG’s that remained associated with ethnicity after adjustment for ancestry and the number of CpG’s that were associated with ancestry . ( B ) Relative proportion of variance in methylation explained by ethnicity and genomic ancestry across loci significantly associated with ethnicity . Mediation analysis of associations between ethnicity and methylation M-values for ( C ) Native American ancestry and ( D ) African ancestry . For simplicity , only significant mediation effects are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 00810 . 7554/eLife . 20532 . 009Table 2 . Proportion of variance in methylation explained by ethnicity and ancestry . Numbers represent the median and interquartile range . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 009ComponentAll CpG’s associated with ethnicity ( n = 916 ) CpG’s associated with ethnicity after adjusting for ancestry ( n = 314 ) CpG’s whose association with ethnicity is explained by ancestry ( n = 520 ) Joint6 . 8% ( 4 . 5% to 10% ) 6 . 2% ( 4 . 4% to 8 . 8% ) 7 . 8% ( 5 . 3% to 11 . 1% ) Ethnicity1 . 7% ( 0 . 78% to 3 . 0% ) 3 . 5% ( 2 . 2% to 5 . 1% ) <1% Ancestry4 . 2% ( 1 . 8% to 8 . 3% ) 1 . 8% ( 0 . 8% to 4 . 0% ) 6 . 6% ( 4 . 0% to 10 . 2% ) Ethnicity and ancestry jointly explained as much as 38 . 5% of the variance in methylation in one CpG ( cg0966827 ) and there were 17 CpG’s where ethnicity and ancestry jointly explain more 25% of the variance . Among the 314 CpG’s that remained associated with ethnicity after adjustment for ancestry , ethnicity accounted for a larger share of the joint variance than genomic ancestry ( 3 . 5% , IQR 2 . 2% to 5 . 1% versus 1 . 8% , IQR 0 . 8% to 4 . 0% ) . We saw a moderate amount of correlation between the 314 methylation sites associated with ethnicity after adjusting for ancestry ( median R2 of 0 . 044 , IQR 0 . 01 to 0 . 13 ) . Sensitivity tests for departures from linearity , fine scale population substructure and the exclusion of the 16 participants who self-identified as ‘Mixed Latino’ sub-ethnicity , did not meaningfully affect our results ( See Supplementary file 1B–F ) . To rule out any residual confounding due to recruitment sites , we conducted an additional analysis on the effect of recruitment site on methylation both for the overall study and for the Mexican participants ( the largest study population in this analysis ) . We observed no significant independent effect of recruitment site suggesting that confounding due to recruitment region was limited , at least within the United States . To explore the effect of departures from a linear association between ancestry and methylation , we incorporated both higher order polynomials and cubic splines of ancestry into our models . We observed a significant departure from linearity ( p<0 . 05 ) in only 26 ( for splines ) and 25 ( for polynomials ) of the 314 CpG’s where an association between ethnicity and methylation remained after adjusting for ancestry; however , the association between ethnicity and methylation remained even after adjusting for non-linearity at all sites ( Supplementary file 1C , D ) . Environmental differences between geographic locations or recruitment sites are a potential non-genetic explanation for ethnic differences in methylation . We investigated the independent effect of recruitment site on methylation by analyzing the associations between recruitment site and individual methylation loci after adjusting for ethnicity . We did not find any loci significantly associated with recruitment site at a significance threshold of 1 . 6×10−7 . We then performed an analysis to assess the effect of recruitment sites on methylation stratified by ethnicity . We did not find any loci significantly associated with recruitment site and methylation among Mexican participants . We were underpowered to perform a similar analysis for Puerto Ricans because there were only 27 Puerto Rican participants recruited outside of Puerto Rico . To ensure that the absence of association in Mexicans was not due to the loss of power from the smaller sample size , we repeated our analysis of the association between ethnicity and ancestry randomly down-sampling to 276 participants to match the sample size in the analysis of geography in Mexicans . While down-sampling the study to this degree resulted in a loss of power , 128 methylation sites were still associated with ancestry . We conclude that recruitment site was unlikely to be a significant confounder of our associations between ethnicity and methylation and was not a significant independent predictor of methylation . While most population substructure in Latinos would be expected to arise from differences in continental ancestry ( Galanter , 2012; Bryc et al . , 2010 ) , there is evidence of finer scale ( sub-continental ) ancestry in Latino populations ( Moreno-Estrada et al . , 2014 ) . We tested for the effect of fine scale substructure by calculating principal components for all participants with genotyping data using Eigensoft ( Patterson et al . , 2006 ) . We found significant associations between principal components 3–10 ( PC’s 1 and 2 were almost perfectly collinear with ancestry , with an adjusted R2 > 0 . 998 for all three ancestry proportions , and were therefore excluded ) and ethnicity . We therefore added these 8 PC’s to models of ethnicity and methylation , and found an association between these genetic PC’s and methylation in 63/314 CpG’s that had remained associated with ethnicity after adjusting for ancestry . Adjusting for higher order substructure in these CpG’s explained the association between ethnicity and methylation in 51 additional loci . This left 263 loci associated with ethnicity after adjustment for ancestry where there was either no association between PC’s 3–10 and methylation or the inclusion of these PC’s did not affect the association between ethnicity and methylation . ( Supplementary file 1E ) At these 314 loci , the median total variance accounted for by ethnicity , ancestry , and fine-scale substructure was 10 . 4% ( IQR 6 . 6% to 16 . 1% ) , of which ethnicity explained a median of 1 . 7% ( IQR 0 . 8% to 3 . 8% ) , ancestry explained a median of 2 . 9% ( IQR 1 . 0 to 4 . 6% ) and fine scale substructure explained a median of 3 . 4% ( IQR 2 . 0% to 4 . 2% ) . Among the 263 CpG’s whose association with ethnicity could not be explained by fine-scale substructure , ethnicity explained a median of 1 . 9% ( IQR 1 . 0% to 4 . 0%; max 26 . 7% ) , ancestry explained 2 . 8% ( IQR 1 . 0% to 6 . 2% ) , and fine scale ancestry explained 3 . 2% ( IQR 1 . 9% to 4 . 7% ) . As only 16 participants self-identified as ‘Mixed Latino’ , we performed a sensitivity analysis to test the effect of excluding these participants from the analysis and only examining Puerto Ricans , Mexicans , and ‘Other Latinos’ . We found that excluding self-identified ‘Mixed Latino’ participants from the analysis did not significantly alter the results in most cases ( Supplementary file 1F ) . Of the 916 CpG’s associated with ethnicity at a genome-wide scale ( p<1 . 6×10–7 ) in models including individuals self-identified as ‘Mixed Ethnicity’ , 894 ( 97 . 5% ) were still significant at a genome-wide scale when ‘Mixed Latinos’ were excluded . All but two of the CpG’s that did not meet genome-wide significance were significant when correcting for 916 tests ( p<5×10–5 ) . In addition , an additional 290 CpG loci that did not meet genome-wide significance in the original analysis were significant at a genome-wide scale when self-identified ‘Mixed Latinos’ were excluded . While these loci did not meet genome-wide significance in the original analysis that included Mixed Latinos , they all had p-values lower than 2 × 10−6 . Thus we conclude that a sensitivity test excluding individuals of mixed Latino ethnicity did not significantly alter the conclusions . We conclude that shared genetic ancestry explains much but not all of the association between ethnicity and methylation . Other , non-genetic factors associated with ethnicity likely explain the ethnicity-associated methylation changes that cannot be accounted for by genomic ancestry alone . Methylation at CpG loci that had previously been reported to be associated with environmental exposures whose exposure prevalence differs between ethnic groups were tested for association with ethnicity in this study . A recent meta-analysis of maternal smoking during pregnancy , an exposure that varies significantly by ethnicity ( Oh et al . , 2012 ) , identified associations with methylation at over 6000 CpG loci ( Joubert et al . , 2016 ) . We found 1341 of 4404 that passed QC in our own study ( 30 . 4% ) were nominally associated with ethnicity ( p<0 . 05 ) , which represented a highly significant ( p<2×10−16 ) enrichment . Using a Bonferroni correction for the 4404 loci tested , 126 maternal-smoking related loci were associated with ethnicity ( p<1 . 1×10−5 ) , and 27 loci were among the 916 CpG’s reported above as associated with ethnicity ( Supplementary file 1G ) . Of these , 14 were among the 314 CpG’s whose association with ethnicity could not be explained by ancestry and 12 were among the 263 CpG’s whose association with ethnicity could not be explained by ancestry or fine-scale substructure . We also examined methylation loci from an earlier study of maternal smoking in Norwegian newborns ( Joubert et al . , 2012 ) as well as studies of diesel exhaust particles ( Jiang et al . , 2014 ) and exposure to violence ( Chen et al . , 2013 ) . These results are supportive of our hypothesis that environmental exposures may be responsible for the observed differences in methylation between ethnic groups and are presented in Supplementary file 1H . In an earlier study of maternal smoking in Norwegian newborns ( Joubert et al . , 2012 ) that identified 26 loci associated with maternal smoking during pregnancy , 19 passed quality control ( QC ) in our own analysis , and the association between methylation and ethnicity was found to be nominally significant ( p<0 . 05 ) at 6 ( 31 . 6% ) CpG loci . Adjusting for 19 tests ( p<0 . 0026 ) , cg23067299 in the aryl hydrocarbon receptor repressor ( AHRR ) gene on chromosome five remained statistically significant ( Supplementary file 1H ) . These results suggest that ethnic differences in methylation at loci known to be responsive to tobacco smoke exposure in utero may be explained in part by ethnic-specific differences in the prevalence of maternal smoking during pregnancy . We also found that CpG loci previously reported to be associated with diesel-exhaust particle ( DEP ) exposure ( Jiang et al . , 2014 ) were significantly enriched among the set of loci whose methylation levels varied between ethnic groups . Specifically , of the 101 CpG sites that were significantly associated with exposure to DEP and passed QC in our dataset , 31 were nominally associated with ethnicity ( p<0 . 05 ) , and five were associated with ethnicity after adjusting for 101 comparisons ( p<0 . 005 ) . Finally , we found that methylation levels at cg11218385 in the pituitary adenylate cyclase-activating polypeptide type I receptor gene ( ADCYAP1R1 ) , which had been associated with exposure to violence in Puerto Ricans ( Chen et al . , 2013 ) and with heavy trauma exposure in adults ( Ressler et al . , 2011 ) , was significantly associated with ethnicity ( p=0 . 02 ) . We also found 194 loci with a significant association between global genetic ancestry and methylation levels ( after adjusting for ethnicity ) at a Bonferroni corrected association p-value of less than 1 . 6 × 10−7 ( Figure 6 and Supplementary file 1I ) , including 48 that were associated with ethnicity in our earlier analysis . Of these significant associations , 55 were driven primarily by differences in African ancestry , 94 by differences in Native American ancestry , and 45 by differences in European ancestry . The most significant association between methylation and ancestry occurred at cg04922029 in the Duffy antigen receptor chemokine gene ( DARC ) on chromosome 1 ( ANOVA p-value 3 . 1 × 10−24 ) ( Figure 6B ) . This finding was driven by a strong association between methylation level and global African ancestry; each 25 percentage point increase in African ancestry was associated with an increase in M-value of 0 . 98 , which corresponds to an almost doubling in the ratio of methylated to unmethylated DNA at the site ( 95% CI 0 . 72 to 1 . 06 per 25% increase in African ancestry , p=1 . 1×10−21 ) . There was no significant heterogeneity in the association between genetic ancestry and methylation between Puerto Ricans and Mexicans ( p-het = 0 . 5 ) . Mexicans have a mean unadjusted methylation M-value 0 . 48 units lower than Puerto Ricans ( 95% CI 0 . 35 to 0 . 62 units , p=1 . 1×10−11 ) . However , adjusting for African ancestry accounts for the differences in methylation level between the two sub-groups ( p-adjusted = 0 . 4 ) , demonstrating that ethnic differences in methylation at this site are due to differences in African ancestry . 10 . 7554/eLife . 20532 . 010Figure 6 . Associations between genomic ancestry and individual methylation loci . ( A ) Manhattan plot showing the associations between genomic ancestry and methylation at individual CpG loci . ( B ) Plot showing one such locus , cg04922029 , and genomic African ancestry , showing a strong correlation between African ancestry and hypermethylation at that site . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 010 The distribution of methylation M-values at cg04922029 is tri-modal , raising the possibility that a SNP whose allele frequency differs between African and non-African populations may be driving the association . We therefore looked at the association between methylation at cg0422029 and ancestry at that locus . We found almost perfect correlation between methylation and African ancestry at the locus ( p=6×10−162 ) ( Figure 7A ) . Each African haplotype at the CpG site was associated with an increase in methylation M-value of 2 . 7 , corresponding to a 6 . 5-fold increase in the ratio of methylated to unmethylated DNA per African haplotype at that locus . We then looked for SNPs within 10 , 000 base pairs of the CpG site that explained the admixture mapping association . We found that methylation at cg04922029 was significantly correlated with SNP rs2814778 ( Figure 7B ) , the Duffy null mutation , 212 base pairs away; each copy of the C allele was associated with an increase in M-value of 1 . 5 , or a 2 . 9-fold increase in the ratio of methylated to unmethylated DNA ( p=3 . 8×10−90 ) ( Figure 7C ) . 10 . 7554/eLife . 20532 . 011Figure 7 . Association between local ancestry and methylation . ( A ) Association between cg04922029 on the DARC locus and African ancestry , color coded by ethnic group . There is near perfect correlation between the two . ( B ) Association between SNPs located within 1 Mb of cg04922029 and methylation levels at that CpG . ( C ) Association between rs2814778 ( Duffy null ) genotype and methylation at cg04922029 , color coded by the number of African alleles present . There is near perfect correlation between genotype , ancestry and methylation at the locus . ( D ) Allele frequency of rs2814778 by 1000 Genomes population . The C allele is nearly ubiquitous in African populations and nearly absent outside of African populations and their descendants . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 011 When we examined the effect of local ancestry at the other 194 CpG’s we find that a substantial proportion of the effect of global ancestry on local methylation levels is due to local ancestry acting in –cis . Among the 194 CpG sites associated with global ancestry , local ancestry at the CpG site explained a median of 10 . 4% ( IQR 3 . 0% to 19 . 4% ) of the variance in methylation , accounting for a median of 52 . 8% ( IQR 20 . 3% to 84 . 9% ) of the total variance explained jointly by local and global ancestry ( Figure 8 ) . 10 . 7554/eLife . 20532 . 012Figure 8 . Relative proportion of variance in methylation explained by global and local ancestry across loci significantly associated with global ancestry . DOI: http://dx . doi . org/10 . 7554/eLife . 20532 . 012 In a diverse population of Latinos , we have shown that a substantial number of loci are differentially methylated between ethnic sub-groups . While genomic ancestry can explain the association between ethnicity and methylation at 66% of the 916 loci associated with ethnicity , factors other than shared ancestry that correlate with ethnicity , such as social , economic , cultural and environmental exposures account for the association between ethnicity and methylation at 34% ( 314/916 ) of loci . We conclude that systematic environmental differences between ethnic subgroups likely play an important role in shaping the methylome for both individuals and populations . Loci previously associated with diverse environmental exposures such as in utero exposure to tobacco smoke ( Joubert et al . , 2012 , 2016 ) , as well as diesel exhaust particles ( Jiang et al . , 2014 ) and psychosocial stress ( Chen et al . , 2013 ) were enriched in our set of loci where methylation was associated with ethnicity . Twenty-seven of the loci associated with maternal smoking during pregnancy in a large consortium meta-analysis ( Joubert et al . , 2016 ) were differentially methylated between Latino sub-groups at a genome-wide significance threshold of 1 . 6 × 10−7 . Interestingly , this included both loci whose association persisted after adjustment for ancestry and fine-scale population substructure and are thus presumed to be due to environmental differences between ethnic groups and loci in which the association between ethnicity and methylation could be fully explained by genetically defined ancestry . There are a number of plausible reasons for overlap between CpG’s associated with ancestry and those associated with environmental exposure . It is possible that this represents a gene-environment interaction , and that individuals with certain genetic backgrounds are more susceptible to the effects of environmental exposures such as in utero tobacco smoke than those of other genetic backgrounds . It has been previously reported that Hispanic smokers with high Native American ancestry had reduced risk of methylation across 12 genes , suggesting an ancestry by smoking interaction ( Leng et al . , 2013 ) . Because the majority of studies that comprised the consortium that identified differentially methylated regions enrolled participants of European descent , such interactions might not have been evident in their study . It is also possible that environmental exposures correlate with ancestry and that participants with certain ancestral backgrounds may have been more exposed to in utero tobacco smoke than those of other backgrounds . Several studies have shown correlations between genetic ancestry and environmental exposures , including socioeconomic status ( Florez et al . , 2011 ) , overweight and obesity ( Ziv et al . , 2006 ) , and birth site and country of residence ( González Burchard et al . , 2005 ) . Though our analysis of global ancestry showed that a majority of the variance explained jointly by local and global ancestry can be traced to specific loci in the genome acting in –cis , a substantial proportion cannot . Some of the residual association between global ancestry and methylation may be due to genetic effects acting in –trans; however , the possibility that some of it may be due to environmental exposures correlating with global ancestry cannot be excluded . Thus , it is plausible that genomic ancestry is acting as a proxy for both genetic and environmental effects in our study . If this is the case , our study likely underestimates the degree to which environmental factors explain differential methylation between ethnic groups . Finally , it is possible that our analysis identified DMRs that are independently modifiable by both genetic and environmental exposures . Thus , regions of the genome that are differentially methylated due to genetic polymorphisms may also be more susceptible to differential methylation due to environmental exposures . Thus , inclusion of relevant social and environmental exposures in studies of methylation may help elucidate racial/ethnic disparities in disease prevalence , health outcomes and therapeutic response . However , in many cases , a detailed environmental exposure history is unknown , unmeasurable or poorly quantifiable , and race/ethnicity may be a useful , albeit imperfect proxy . However , if a comprehensive catalog of the effects of exposures can be compiled , it may be possible to use genome-wide methylation analysis as a biomarker of exposure long after the exposure has passed and can no longer be measured . Our comprehensive analysis of high-density methyl- and genotyping from genomic DNA allowed us to investigate the genetic control of methylation in great detail and without the potential destabilizing effects of EBV transformation and culture in cell lines ( Grafodatskaya et al . , 2010 ) . The strongest patterns of methylation are associated with cell composition in whole blood ( Lam et al . , 2012 ) . However , the specific type of Latino ethnic-subgroups ( Puerto Rican , Mexican , other , or mixed ) is also associated with principal coordinates of genome-wide methylation . Our approach has some potential limitations . It is possible that fine-scale population structure ( sub-continental ancestry ) within European , African , and Native American populations may contribute to ethnic differences in methylation , as we had previously reported in the case of lung function ( Moreno-Estrada et al . , 2014 ) . However , despite the presence of additional substructure among the GALA II participants , PC’s 3–10 explained the association between ethnicity and ancestry at only 51 loci . PCs from chip-based genotypes will not capture all forms of genetic variation . Clusters of ethnicity specific rare variants of large effect or strong ethnicity-specific selective sweeps in the last 8–12 generations ( Galanter et al . , 2012 ) could also give rise to methylation differences , but these are inconsistent with existing rare variant and selection analyses ( Hernandez et al . , 2011; Tang et al . , 2007 ) . Our models of genetic ancestry assumed a linear effect of ancestry on methylation , whereas a nonlinear association or other model misspecification could have led to incomplete adjustment for genetic ancestry , and thus , led to a residual association between ethnicity and methylation . However , when we added second and third order polynomials or cubic splines to our models , we found evidence for a nonlinear association between ancestry and methylation at only 25 and 26 loci , respectively , and it did not affect the association between ethnicity and methylation . Although it is impossible to account for all types of non-linearity and non-additivity ( such as gene by gene or gene by environment interaction ) , our analysis suggests that non-linear effects are unlikely to be significant . Since our study was geographically diverse , recruiting participants at five recruitment sites in the United States and Puerto Rico , it is possible that systematic differences associated with site of recruitment might have influenced observed methylation differences between ethnic groups . However , when we included recruitment site as a covariate , we found no significant effect on methylation independent of ethnicity . The presence of a strong association between genetic ancestry and methylation raises the possibility that epigenetic studies can be confounded by population stratification , similar to genetic association studies , and that adjustment for either genetic ancestry or selected principal components is warranted . This possibility was first demonstrated in a previous analysis of the association between self-described race and methylation ( Barfield et al . , 2014 ) . However , the study only evaluated two distinct racial groups ( African Americans and Whites ) , while the present study demonstrates the possibility of population stratification in an admixed and heterogeneous population with participants from diverse Latino national origins . The tendency to consider Latinos as a homogenous or monolithic ethnic group makes any analysis of this population particularly challenging . Our finding of loci whose methylation patterns differed between Latino ethnic subgroups , even after adjusting for genetic ancestry , suggests that any analysis of these populations in disease-association studies without adjusting for ethnic heterogeneity is likely to result in spurious associations even after controlling for genomic ancestry . However , the methylation loci identified in this study , as well as studies of environmental exposures , could be particularly interesting loci for the study of biomedical outcomes , particularly those with disparate prevalence between racial/ethnic groups , such as asthma ( Barr et al . , 2016 ) . If methylation loci associated with ethnicity or ancestry were shown to be associated with a biomedical outcome , it could help explain racial/ethnic disparities in disease . In summary , this study provides a framework for understanding how genetic , social and environmental factors can contribute to systematic differences in methylation patterns between ethnic subgroups , even between presumably closely related populations such as Puerto Ricans and Mexicans . Methylation QTL’s whose allele frequency varies by ancestry lead to an association between local ancestry and methylation level . This , in turn , leads to systematic variation in methylation patterns by ancestry , which then contributes to ethnic differences in genome-wide patterns of methylation . However , although genetic ancestry has been used to adjust for confounding in genetic studies , and can account for much of the ethnic differences in methylation in this study , ethnic identity is associated with methylation beyond the effects of shared genetic ancestry . This is likely due to social and environmental effects captured by ethnicity . Indeed , we find that CpG sites known to be influenced by social and environmental exposures are also differentially methylated between ethnic subgroups . These findings called attention to a more complete understanding of the effect of social and environmental variables on methylation in the context of race and ethnicity to fully understanding this complex process . Our findings have important implications for the independent and joint effects of race , ethnicity , and genetic ancestry in biomedical research and clinical practice , especially in studies conducted in diverse or admixed populations . Our conclusions may be generalizable to any population that is racially mixed such as those from South Africa , India , and Brazil , though we would encourage further study in diverse populations , and likely has implications for all studies of diverse populations . As the National Institutes of Health ( NIH ) embarks on a precision medicine initiative , this research underscores the importance of including diverse populations and studying factors capturing the influence of social , cultural , and environmental factors , in addition to genetic ones , upon disparities in disease and drug response . All research on human subjects was approved by the Institutional Review Board at the University of California and each of the recruitment sites ( Kaiser Permanente Northern California , Children’s Hospital Oakland , Northwestern University , Children’s Memorial Hospital Chicago , Baylor College of Medicine on behalf of the Texas Children’s Hospital , VA Medical Center in Puerto Rico , the Albert Einstein College of Medicine on behalf of the Jacobi Medical Center in New York and the Western Review Board on behalf of the Centro de Neumologia Pediatrica ) , and all participants/parents provided age-appropriate written assent/consent . Latino children were enrolled as a part of the ongoing GALA II case-control study ( Oh et al . , 2012 ) . A total of 4702 children ( 2374 participants with asthma and 2328 healthy controls ) were recruited from five centers ( Chicago , Bronx , Houston , San Francisco Bay Area , and Puerto Rico ) using a combination of community- and clinic-based recruitment . Participants were eligible if they were 8–21 years of age and self-identified as a specific Latino ethnicity and had four Latino grandparents . Asthma cases were defined as participants with a history of physician diagnosed asthma and the presence of two or more symptoms of coughing , wheezing , or shortness of breath in the two years preceding enrollment . Participants were excluded if they reported any of the following: ( 1 ) 10 or more pack-years of smoking; ( 2 ) any smoking within 1 year of recruitment date; ( 3 ) history of lung diseases other than asthma ( cases ) or chronic illness ( cases and controls ) ; or ( 4 ) pregnancy in the third trimester . Further details of recruitment are described elsewhere ( Oh et al . , 2012 ) . Latino sub-ethnicity was determined by self-identification and the ethnicity of the their four grandparents . Due to small numbers , ethnicities other than Puerto Rican and Mexican were collapsed into a single category , ‘other Latino’ . Participants whose four grandparents were of discordant ethnicity were considered to be of ‘mixed Latino’ ethnicity . Trained interviewers , proficient in both English and Spanish , administered questionnaires to gather baseline demographic data , as well as information on general health , asthma status , acculturation , social , and environmental exposures . Genomic DNA ( gDNA ) was extracted from whole blood using Wizard Genomic DNA Purification Kits ( Promega , Fitchburg , WI ) . A subset of 573 participants ( 311 cases with asthma and 262 healthy controls ) was selected for methylation . Methylation was measured using the Infinium HumanMethylation450 BeadChip ( Illumina , Inc . , San Diego , CA ) following the manufacturer’s instructions . 1 µg of gDNA was bisulfite-converted using the Zymo EZ DNA Methylation Kit ( Zymo research , Irvine , CA ) according to the manufacturer’s instructions . Bisulfite converted DNA was isothermally amplified overnight , enzymatically fragmented , precipitated , and re-suspended in hybridization buffer . The fragmented , re-suspended DNA samples were dispensed onto Infinitum HumanMethylation450 BeadChips and incubated overnight in an Illumina hybridization oven . Following hybridization , free DNA was washed away , and the BeadChips were extended through single nucleotide extensions with fluorescent labels . The BeadChips were imaged using an Illumina iScan system , and processed using the Illumina GenomeStudio Software . Failed probes were identified using detection p-values using Illumina’s recommendations . Probes on sex chromosomes and those known to contain genetic polymorphisms in the probe sequence were also excluded , leaving 321 , 503 probes for analysis . Raw data were normalized using Illumina’s control probe scaling procedure . Beta values of methylation ( ranging from 0 to 1 ) were converted to M-values via a logit transformation ( Du et al . , 2010 ) . Details of genotyping and quality control procedures for single nucleotide polymorphisms ( SNPs ) and individuals have been described elsewhere ( Galanter et al . , 2014 ) . Briefly , participants were genotyped at 818 , 154 SNPs on the Axiom Genome-Wide LAT 1 , World Array 4 ( Affymetrix , Santa Clara , CA ) ( Hoffmann et al . , 2011 ) . We removed SNPs with >5% missing data and failing platform-specific SNP quality criteria ( n = 63 , 328 ) , along with those out of Hardy-Weinberg equilibrium ( n = 1845; p<10–6 ) within their respective populations ( Puerto Rican , Mexican , and other Latino ) , as well as non-autosomal SNPs . Subjects were filtered based on 95% call rates and sex discrepancies , identity by descent and standard Affymetrix Axiom metrics . The total number of participants passing QC was 3804 ( 1902 asthmatic cases , 1902 healthy controls ) , and the total number of SNPs passing QC was 747 , 129 . The number of participants with both methylation and genotyping data was 524 . GALA II participants were combined with ancestral data from 1000 Genomes European ( CEU ) and African ( YRI ) populations and 71 Native American ( NAM ) samples genotyped on the Axiom Genome-Wide LAT one array . A final sample of 568 , 037 autosomal SNPs with relevant ancestral data was used to estimate local and global ancestry . Global ancestry was estimated using the program ADMIXTURE ( Alexander et al . , 2009 ) , with a three population model . Local ancestry at all positions across the genome was estimated using the program LAMP-LD ( Baran et al . , 2012 ) , assuming three ancestral populations . Principal components for the genetic data were determined using the program EIGENSTRAT ( Patterson et al . , 2006 ) . Using a variance in methylation m-value of 0 . 2 units , which corresponded to approximately the 90th percentile of the variance in m-value in our pilot data , we determined that in order to have an 80% power to detect a difference in mean methylation between the two major ethnic groups of 0 . 25 units , using a Bonferroni significance threshold of 1 . 6 × 10−7 a sample , a sample size of 251 participants in each group was required . That total sample size of 502 participants gave us 80% power to detect correlations between ancestry and methylation of medium ( Pearson r > 0 . 25 ) effect , meaning that we had 80% power to detect loci where ancestry accounted for at least 6 . 25% of the variance in methylation . Unless otherwise noted , all regression models were adjusted for case status , age , sex , estimated cell counts , and plate and position . To account for possible heterogeneity in the cell type makeup of whole blood we inferred white cell counts using the method by Houseman et al ( Houseman et al . , 2012 ) . Indicator variables were used to code categorical variables with more than two categories , such as ethnicity . In these cases , a nested analysis of variance ( ANOVA ) was used to compare models with and without the variables to obtain an omnibus p-value for the association between the categorical variable and the outcome . For analyses of dependent beta-distributed variables ( such as African , European , and Native American ancestries ) , or cell proportion , k-1 variables were included in the analysis , and a nested analysis of variance ( ANOVA ) was used to compare models with and without the variables to obtain an k-1 degree of freedom omnibus p-value for the association between predictor ( such as ancestry ) and the outcome variable . The Bonferroni method was used to adjust for multiple comparisons . For methylome-wide associations , the significance threshold was adjusted for 321 , 503 probes , resulting in a Bonferroni threshold of 1 . 6 × 10−7 . Analyses were performed using R version 3 . 2 . 1 ( The R Foundation for Statistical Computing ) ( R Core Team ) and the Bioconductor package version 2 . 13 . Multidimensional scaling of the logit transformed methylation data ( M-values ) was performed by first calculating the Euclidian distance matrix between each pair of individuals and then calculating the first 10 principal coordinates of the data ( Figure 2A ) . We performed both a simple correlation analysis of these principal coordinates to demographic factors ( age , sex , ethnicity ) , estimated cell counts and technical factors ( batch , plate , and position ) to identify factors that correlated with global methylation patterns [see Figure 2B ) . In addition , we performed a multiple regression analysis of methylation principal coordinates by ethnicity and ancestry , adjusting for case status , age , sex , estimated cell counts , and plate and position ( Supplementary file 1A ) . We also sought to establish the extent to which global differences in methylation between Puerto Ricans and Mexicans could be explained by differences in ancestry between the two groups . We estimated the proportion of the ethnicity association that was mediated by genomic ancestry using the R package ‘mediation’ ( Tingley et al . , 2014 ) for methylation principal coordinates , which demonstrated a significant association with ethnicity . We also sought to correlate ethnicity and methylation at a locus-specific level . We thus performed a linear regression between methylation at each CpG site and self-reported ethnicity ( Mexican , Puerto Rican , Mixed Latino , and Other Latino ) , followed by a three degree of freedom analysis of variance to determine the overall effect of ethnicity on methylation We repeated the analysis excluding the 16 participants that were self-described as ‘Mixed Latino’ , and tested for non-linearity in two ways: by adding second and third order polynomials to the model , and by adding a 3-degree of freedom cubic spline and comparing models with the non-linear terms to those without using a nested ANOVA . At loci where there was evidence for non-linearity , we tested whether ethnicity remained associated with methylation after adjusting for ancestry as well as the deviations from linearity . Finally , we tested for the presence of population sub-structure beyond that conveyed through ancestry by adding the genetic principal components 3–10 ( PCs 1 and 2 were co-linear with ancestry with a coefficient of determination R2 > 0 . 998 ) and comparing models with those PCs to those without . At loci where there was evidence for association between PC’s 3–10 and methylation , we tested whether ethnicity remained associated with methylation after adjusting for ancestry as well as the PC’s 3–10 . We calculated the proportion of variance in methylation explained by ethnicity and genomic ancestry at each site where ethnicity was significantly associated with methylation . To do this , we fit a model that included both ethnicity and global ancestry as well as the confounders described above and calculated the proportion of variance explained by multiplying the ratio of the variance between predictors ( ethnicity and genomic ancestry ) and outcome ( methylation ) by the square of the effect magnitude ( ß ) . We also examined whether differences in methylation patterns by ethnicity could be associated with known loci that had previously been reported to vary based on common environmental exposures , including maternal smoking during pregnancy ( Joubert et al . , 2012 ) , diesel exhaust particles ( DEP ) ( Jiang et al . , 2014 ) , and exposure to violence ( Chen et al . , 2013 ) . We have previously shown that exposure to these common environmental exposures or similar exposures varied by ethnicity within our own GALA II study populations ( Oh et al . , 2012; Nishimura et al . , 2013; Thakur et al . , 2013 ) . In addition , we examined the association between global ancestry and methylation across all CpG loci using a two-degree of freedom likelihood ratio test as well as by examining the association between individual ancestral components ( African , European , and Native American ) and methylation at each CpG site . At each site where methylation was significantly associated with genomic ancestry proportions , we determined the relative effect of global ancestry ( θ , theta ) and local ancestry ( γ , gamma ) in a joint model by calculating the proportion of variance explained as above . To determine whether ancestry associations with methylation were due to variation in local ancestry , we correlated local ancestry at each CpG site with methylation at the site . Because ancestry LD is much stronger than genotypic LD , it is possible to accurately interpolate ancestry at each CpG site based on the ancestry estimated at the nearest SNPs ( Galanter et al . , 2014; Rosenberg et al . , 2010 ) . Measures of locus-specific ancestry were correlated with local methylation using linear regression . We performed a two-degree of freedom analysis of variance test evaluating the overall effect of all three ancestries as well as single-ancestry associations comparing methylation at a given locus with the number of African , European and Native American chromosomes at that CpG site .	Whether a person develops a particular disease can depend on both genetic and environmental factors . Many studies have found that people of different races and ethnicities have different likelihoods of acquiring certain diseases . Race and ethnicity are social constructs; that is , they are not necessarily defined biologically . However , shared ancestry will produce genetic links between members of a group . In addition , members of an ethnic group often share a culture or environment that may influence their risk of disease . For example , the ‘Mediterranean diet’ inspired by the dietary habits of Southern Italians has been shown to reduce the risk of heart disease , diabetes and cancer . The addition of chemical groups – such as methyl groups – to DNA strands can affect the activity of nearby genes . Methylation is controlled by both genetic and environmental factors , and altered patterns of DNA methylation are seen in some diseases . It is therefore an ideal biological process to study to determine how race/ethnicity and ancestry contribute to a person’s susceptibility to disease . Galanter et al . have now studied the patterns of methylation found in the blood of 573 people from diverse Latino ethnic sub-groups . The different groups displayed significantly different patterns of methylation at hundreds of locations across the genome . Genetic ancestry explained approximately 75% of the variation in methylation between the sub-groups . In addition , the methylation patterns at DNA locations known to be affected by environmental exposures – for example , by exposure to tobacco while in the womb – were disproportionately likely to be methylated differently in different sub-groups . Now that more is known about the relative effects of race/ethnicity and genetic ancestry on methylation , the next step is to apply this knowledge to disease processes . This will help us to better understand the source of health disparities across different groups of people .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression" ]	2017	Differential methylation between ethnic sub-groups reflects the effect of genetic ancestry and environmental exposures
Spatial learning requires the hippocampus , and the replay of spatial sequences during hippocampal sharp wave-ripple ( SPW-R ) events of quiet wakefulness and sleep is believed to play a crucial role . To test whether the coordination of VTA reward prediction error signals with these replayed spatial sequences could contribute to this process , we recorded from neuronal ensembles of the hippocampus and VTA as rats performed appetitive spatial tasks and subsequently slept . We found that many reward responsive ( RR ) VTA neurons coordinated with quiet wakefulness-associated hippocampal SPW-R events that replayed recent experience . In contrast , coordination between RR neurons and SPW-R events in subsequent slow wave sleep was diminished . Together , these results indicate distinct contributions of VTA reinforcement activity associated with hippocampal spatial replay to the processing of wake and SWS-associated spatial memory . Hippocampal dependent learning and memory formation are influenced by reward and are believed to occur during distinct behavioral states . As animals explore the environment , hippocampal place cells fire sequentially under the modulation of the theta rhythm . Subsequently , these sequences of neuronal activity are replayed in association with hippocampal sharp wave-ripple ( SPW-R ) events of quiet wakefulness and sleep ( O'Keefe and Dostrovsky , 1971; Lee and Wilson , 2002; Foster and Wilson , 2006; Diba and Buzsáki , 2007 ) . SPW-R events contribute to spatial learning ( Jadhav et al . , 2012; Girardeau et al . , 2009; Ego-Stengel and Wilson , 2010 ) , and the capacity for reward to influence reactivation of CA3 place cell pairs in SPW-R events ( Singer and Frank , 2009 ) suggests that reward-related neural activity is likely to play an important role in this process . It has been unclear whether replay events of quiet wakefulness and sleep differ in their contribution to learning and memory , but the observation that replay events during slow wave sleep ( SWS ) are lower fidelity than replay events during quiet wakefulness ( Karlsson and Frank , 2009 ) supports this possibility . Dopamine neurons of the VTA represent reward prediction error Schultz , 1998 and appear to be an important brain substrate for reinforcement learning ( Montague et al . , 1996 ) . Optogenetic activation of dopamine cells during spatial learning has recently been demonstrated to increase the reactivation of CA1 place cell pairs in sleep and stabilize subsequent spatial learning ( Mcnamara et al . , 2014 ) . In addition , electrical stimulation of the medial forebrain bundle triggered on a hippocampal place cell’s spikes has recently been shown to drive goal-directed behavior toward its place field ( de Lavilléon et al . , 2015 ) . However , it is unclear how under normal physiological conditions dopamine neuronal activity engages with the hippocampus . Dopamine cells could coordinate with and reinforce replayed hippocampal sequences . In addition , the fast-onset , slowly decaying profile of dopamine synaptic release has led to the suggestion that dopamine could implement the propagation of expected value across reactivated hippocampal sequences ( Foster and Wilson , 2006 ) . We hypothesized that replayed hippocampal spatial sequences would coordinate with reward-related representations of VTA neurons during tasks that place demands on spatial memory . Here , we acquired simultaneous multi-electrode ( tetrode ) recordings of neurons of the hippocampus and the VTA as rats performed appetitive spatial tasks and subsequently slept to determine the relationship between VTA neuronal activity , hippocampal SPW-R-associated activity , and sequence replay . We show that many reward responsive ( RR ) VTA neurons modulate their firing rate with SPW-R events of quiet wakefulness . Modulation of VTA unit activity was greater in SPW-R events associated with hippocampal replay of task-associated sequences . In contrast to nonRR VTA unit activity , RR unit activity preferentially coordinated with replayed representations of reward sites . In addition , RR VTA units more strongly phase-locked to the hippocampal theta rhythm than nonRR units , and RR VTA units that more strongly coupled to hippocampal theta had greater coordination with replayed reward site representations . In contrast to these findings in the awake state , in post-task epochs of SWS , SPW-R modulation of RR VTA unit activity was significantly reduced . Furthermore , within SWS , RR unit activity decreased during periods of hippocampal SPW-R reactivation . Together , these results indicate distinct contributions of VTA reinforcement activity associated with hippocampal spatial replay to the processing of wake and SWS-associated spatial memory . We recorded the activity of multiple simultaneously isolated units of the hippocampus ( 499 total; for each recording , median of 25 , range 12–37 ) and VTA ( 84 total; median of 5 , range 2–9 ) in five animals , as animals performed a spatial working memory ( SWM ) task ( Jones and Wilson , 2005 ) ( three rats ) ( Figure 1A ) or ran on a linear track ( two rats ) for food reward . The latter task was selected both because the observation of awake replay has been best characterized in a linear environment and because it provides a choice-free spatial task for contrast . Many VTA units modified their firing rate during goal approach and with acquisition of food rewards ( n = 47/84 ) , consistent with prior observations ( Morris et al . , 2006; Roesch et al . , 2007; Totah et al . , 2013 ) ( Figure 1B; Materials and methods ) . These results have been interpreted as the representation of reward prediction error in instrumental tasks ( Morris et al . , 2006; Roesch et al . , 2007; Totah et al . , 2013 ) ( specifically , the Q-associated temporal difference prediction error , where Q-value is the value of selecting a particular action at a given state ) . The mean firing rates for reward responsive ( RR ) and non-reward responsive ( nonRR ) VTA units were 6 . 61 ± 1 . 33 Hz ( mean±s . e . m . , RR units , n = 47 ) and 20 . 59 ± 5 . 49 Hz ( nonRR units , n = 24 ) , respectively . Two populations of cells were observed in a plot of waveform duration versus trough to peak ratio , consistent with prior reports ( Fujisawa and Buzsáki , 2011 ) ( Figure 1C ) . Most RR cells ( 38/47 ) fell in the longer duration cluster ( > 1 ms ) , which appears to be enriched for putative dopamine cells ( Fujisawa and Buzsáki , 2011; Ungless and Grace , 2012 ) . 10 . 7554/eLife . 05360 . 003Figure 1 . Spatial working memory task and VTA unit properties . ( A ) Spatial working memory task . In the force direction ( sample phase ) , rats traverse the central arm for reward ( R ) at either of two pseudorandomly selected left or right force-reward locations . The reward contingency in the choice direction ( test phase ) required that if the rat had been forced to turn left ( or right ) in the sample phase , then the correct response in the test phase was to turn right ( or left , respectively ) . ( B ) Example VTA unit’s average reward site responses for correct trials ( solid line ) and error trials ( dashed line ) . The nosepoke occurs at 0 s . The profile of reward-site associated activity , including differential activity on correct versus error trials during reward approach and during reward acquisition , is consistent with prior observations in instrumental tasks ( Morris et al . , 2006; Roesch et al . , 2007; Totah et al . , 2013 ) . ( C ) Waveform features of 145 VTA units recorded in the sleep box , using the waveform criteria described in ( Fujisawa and Buzsáki , 2011 ) . The waveform duration is defined as the time from waveform major peak to final peak . The trough to peak ratio is defined as the ratio of the waveform trough amplitude to the full amplitude . 84 units that were acquired with adequate task behavior and co-recorded hippocampal activity underwent further analysis . Reward responsive ( RR ) units are shown in blue , and non-reward responsive units ( nonRR ) are shown in red . Waveforms of two units are displayed . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 003 SPW-R events , identified using hippocampal multiunit activity and local field potential ( see Materials and methods ) , were prominent at reward sites during pauses in run behavior between trials ( Figure 2A , B ) and were measured in the period between nosepoke and run initiation to the next reward site . Reward acquisition occurred within the first 1 s of nosepoke . Reward site dwell times were variable and self-paced , with a median of 9 . 3 s ( range: 1 . 5 to 615 . 0 s ) . The frequency of SPW-R events on the SWM task was higher during pauses at reward locations on correct ( rewarded ) trials than on error trials ( correct: 0 . 088 ± 0 . 019 Hz; error: 0 . 029 ± 0 . 009 Hz; p<0 . 01 , signed-rank test ) , consistent with prior results , Singer and Frank , 2009 . 10 . 7554/eLife . 05360 . 004Figure 2 . VTA unit coordination with hippocampal sharp-wave ripples . ( A ) Continuous recordings of hippocampal ( HC ) ( 1 ) single unit activity , ( 2 ) multiunit activity ( MUA , average spike rate per tetrode ) , ( 3 ) local field potential and ripple band , ( 4 ) a simultaneously recorded reward-responsive ( RR ) VTA unit , and ( 5 ) the animal’s position on the track . The hippocampal units are ordered by the position of their place fields on the spatial working memory task . Sharp-wave ripple events ( SPW-R ) are shown in gray . ( B ) A magnified view of 10 s of continuous data . ( C1 ) Rastered RR VTA unit action potentials , ( 2 ) RR VTA unit peri-event time histogram ( PETH; smoothing with a 50ms Gaussian window ) , and ( 3 ) HC multiunit PETH ( 10 ms Gaussian smoothing ) , aligned to the start of SPW-R-associated HC multiunit events . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 00410 . 7554/eLife . 05360 . 005Figure 2—figure supplement 1 . Firing rate distributions of SPW-R modulated VTA units at reward acquisition and at SPW-R events of quiet wakefulness . For units recorded on the SWM task , the average nosepoke triggered PETH for correct trials ( solid blue lines ) and for error trials ( red dashed lines ) are shown . Units acquired on the linear maze have a single nosepoke triggered PETH . Data are aligned to the time of nosepoke ( vertical line ) . For the SPW-R event triggered PETH plots , data are aligned to the start of SPW-R events . Note that VTA unit activity often increases during reward approach and reward acquisition , and that VTA unit activity can be both positively and negatively modulated at SPW-R events . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 005 Many ( 20/84 ) VTA units significantly modulated their firing around SPW-R events ( p<0 . 05 , bootstrapped confidence intervals; median baseline-normalized modulation amplitude of 0 . 15; range 0 . 0003–1 . 41; n = 84; Figure 2 ) . Both positive and negative SPW-R modulations were observed ( positive n = 13; negative n = 7; Figure 2—figure supplement 1 ) . Most SPW-R modulations coincided with SPW-R events; however , some negative SPW-R modulations occurred on a longer timescale , flanking SPW-R events . The majority of VTA units that were significantly modulated at SPW-R events were RR ( 17/20 compared to 47/84 recorded , p=0 . 03 , chi = 4 . 6 , Chi-square test ) , and SPW-R modulation depth was greater for RR units than nonRR units ( RR units 0 . 21 ± 0 . 04 , n = 45; nonRR units 0 . 11± 0 . 02; p=0 . 017 , n = 23 , rank-sum test ) . For RR units , the sign of SPW-R modulation correlated with the sign of firing rate changes associated with reward acquisition ( r = 0 . 55 , p=1 . 2 x× 10-4 ) . Modulation of RR units at SPW-R events did not require active reward consumption , as a similar modulation depth ( 0 . 26 ± 0 . 05 ) was noted when only SPW-R events delayed relative to nosepoke by at least 6 s were considered ( signed-rank test , p=0 . 7 ) . SPW-R modulation depth for RR VTA units was not significantly different on the SWM task compared to the linear track ( SWM task , 0 . 26 ± 0 . 06 , n = 26; linear track , 0 . 13 ± 0 . 02 , n = 19 , p=0 . 3 , rank-sum test ) . We next sought to determine whether RR unit modulation around SPW-R events was related to hippocampal replay . To evaluate spatial information associated with SPW-R replay events , we used a clusterless , probabilistic reconstruction method to maximize decoding fidelity ( Kloosterman et al . , 2014 ) . First , we confirmed that the recorded hippocampal neuronal population conveyed sufficient spatial information to accurately decode the rat’s position on the track . Indeed , a cross-validation procedure showed that decoded hippocampal activity accurately reflected the rat’s location during run behavior ( speed > 10 cm/s ) , with median error of 8 . 3 ± 0 . 5 cm across recording sessions ( Figure 3—figure supplement 1A–C; see Materials and methods ) . We also confirmed that this clusterless reconstruction method resulted in lower median error than a cluster-based approach ( median error 15 . 2 ± 1 . 9 cm , p=1 . 22 × 10-4 , signed-rank test , n = 14 recordings ) . Reconstruction of hippocampal activity during pauses in run behavior ( speed < 10 cm/s; in 25 ms time bins ) identified putative replay events: the representation of a sequence of locations during SPW-R events ( Figure 3—figure supplement 1D ) . For each event , we computed the statistical likelihood that the decoded positions represented a constant-speed traversal of a trajectory on the track ( Davidson et al . , 2009; Kloosterman , 2012 ) and compared it to distributions obtained after two separate randomization procedures ( see Materials and methods ) . Replay events identified with this approach constituted 24 . 8 ± 2% ( 1107/4645 ) of SPW-R events . Modulation of RR VTA unit activity was greater in SPW-R events associated with replay of sequential experience of the task than in SPW-R events that were not ( modulation depth 0 . 28 ± 0 . 04 vs . 0 . 15 ± 0 . 03 , p=4 . 5 × 10-4 , signed-rank test; n = 40; Figure 3 ) . In contrast , for nonRR VTA units , modulation depth was similar across replay and nonreplay events ( replay 0 . 11 ± 0 . 02; nonreplay 0 . 11 ± 0 . 03; p=0 . 6; signed-rank test; n = 22 ) . 10 . 7554/eLife . 05360 . 006Figure 3 . Modulation depth of VTA reward responsive units at hippocampal SPW-R events depends on SPW-R spatial content . ( A ) Rastered reward responsive ( RR ) unit spikes ( 1 ) and RR unit and hippocampal ( HC ) multiunit PETHs ( 2 , 3 ) , aligned to the start of SPW-R events encoding replay sequences . ( B ) As in A , for SPW-R events not encoding replay . ( C ) PETH modulation depth of RR units ( blue ) is greater for replay than nonreplay events; p=4 . 5 × 10-4 , signed-rank test ) . NonRR unit data are shown in red ( p=0 . 6 ) . Solid circles with error bars designate the mean and s . e . m . for RR and nonRR units . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 00610 . 7554/eLife . 05360 . 007Figure 3—figure supplement 1 . Position reconstruction using clusterless hippocampal decoding . ( A ) Bayesian reconstruction of run behavior on the spatial working memory task ( 500 ms time bins ) . The track has been linearized . ( B ) Decomposition of the track into segments for linearization . Maze segments were apposed in the direction of run in the choice direction: from force reward sites ( R3 , R4 ) to force point ( fp ) to choice point ( cp ) to choice reward sites ( R1 , R2 ) . ( C ) Confusion plot for this recording session , using alternating 1 s epochs for training and testing the reconstruction algorithm . ( D1 ) Bayesian reconstruction of a SPW-R event reveals spatial sequence reactivation ( 25 ms time bins ) . ( D2 ) The associated hippocampal multiunit activity . ( D3 ) The action potentials of two simultaneously recorded reward responsive VTA units . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 00710 . 7554/eLife . 05360 . 008Figure 3—figure supplement 2 . Ripple power , SPW-R associated hippocampal activity , and SPW-R event latency in the immediate post-reward period were similar for replay and non-replay events . ( A1 ) For the recording shown in Figure 3A , B , cumulative ripple-band power of replay ( green solid line ) and non-replay ( brown dashed line ) events are displayed . ( A2 ) Across recordings , replay and nonreplay events have similar ripple power ( box and whisker plots , medians with interquartile range; p=1 , sign-rank test ) . ( B1 ) Cumulative SPW-R event peak multiunit activity ( MUA; Hz/tetrode ) for the same example , for replay ( green solid ) and non-replay ( brown dashed ) events . ( B2 ) Across recordings , SPW-R event peak MUA is similar for replay and non-replay events ( medians with interquartile range; p=0 . 6 , sign-rank test ) . ( C ) Cumulative distributions of SPW-R event latencies relative to nosepoke for reward delivery were similar for replay ( green solid line ) and nonreplay ( brown dashed line ) SPW-R events ( p=0 . 2 , Kolmogorov-Smirnov test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 008 We sought to determine whether the greater modulation of RR units around replay-associated SPW-R events compared to nonreplay-associated SPW-R events derived from some difference other than sequence replay . Ripple power and peak hippocampal firing rate at replay and non-replay SPW-R events were not significantly different ( p=0 . 6 , p=1 , respectively , signed-rank tests; Figure 3—figure supplement 2A , B ) . Replay events had longer durations than non-replay events ( 0 . 209 ± 0 . 003 s vs . 0 . 161 ± 0 . 002 s , p<5 . 0 × 10-16 , rank-sum test ) . To address whether SPW-R event duration alone could drive modulation of RR VTA unit activity , we constructed a dataset of replay and nonreplay events matched on their range of durations . Across these matched groups , the greater modulation of RR units at replay events compared to nonreplay events was preserved ( replay 0 . 33 ± 0 . 05; nonreplay 0 . 24 ± 0 . 05 , p=0 . 01 , signed-rank test ) . Similarly , RR units at non-replay events separated by median split into short ( 100 . 6 ± 0 . 0 ms ) and long ( 222 . 6 ± 0 . 1 ms ) events had similar degrees of modulation ( modulation depth of short events 0 . 20±0 . 04; long events 0 . 17 ± 0 . 03; p=0 . 9 , signed-rank test; n = 40 RR units ) ; and RR units at replay events split on median duration also did not differ in their modulation depth ( short events 0 . 29 ± 0 . 04; long events 0 . 39 ± 0 . 06; p=0 . 2 , signed-rank test ) . Despite these similarities , replay events and nonreplay events differed with respect to the fraction of isolated pyramidal units active during each SPW-R event ( replay 35 . 6 ± 0 . 0%; nonreplay 28 . 7 ± 0 . 0% , p=0 . 001 , signed-rank test; n = 14 recordings ) and with respect to the number of single unit action potentials per unit present in each burst ( replay 1 . 06 ± 0 . 07; nonreplay 0 . 72 ± 0 . 09; p=6 . 1 × 10-4 , signed-rank test ) . To address whether the greater activation of RR units around the time of replay events could be due to the greater intensity of hippocampal pyramidal cell spiking seen during these events , we constructed a dataset of spike count matched replay and non-replay events . Across the spike count matched groups , the greater modulation of RR units at replay events compared to nonreplay events was maintained ( replay 0 . 28 ± 0 . 04; nonreplay 0 . 19 ± 0 . 03 , p=0 . 03 , sign rank test ) . Thus , the greater modulation depth of RR units at replay events compared to nonreplay events did not derive from differences in the intensity of hippocampal spiking . We next evaluated whether the difference in RR unit coordination with replay and nonreplay events arose from differences in the timing of these events in the immediate post-reward period , when the activity of RR units often changes . Replay and nonreplay events occurring within a 5 s window from the nosepoke had similar onset latencies ( replay events 2 . 54 ± 0 . 10 s , n = 130; nonreplay events 2 . 67 ± 0 . 07 s , n = 217 , p=0 . 3 , rank-sum test ) , and the temporal distributions of replay and nonreplay events following reward delivery were similar ( p=0 . 2 , Kolmogorov-Smirnov test; Figure 3—figure supplement 2C ) . These data suggest that RR units coordinate preferentially with hippocampal SPW-R events that encode within-session spatial sequences . The coordination of a reward prediction error signal with a hippocampal replay sequence ( for example , one that represents a trajectory towards a reward ) could function to reinforce specific elements of the reactivated sequence , such as a goal location . We therefore explored how VTA unit activity relates to the specific spatial locations contained in replay content . To account for latency between hippocampal SPW-R events and VTA activity , we first examined the hippocampal SPW-R event-triggered VTA LFP . This revealed a prominent negative potential that peaked 84 ± 13 ms after SPW-R onset ( Figure 4A ) . We focused on replay events occurring at forced reward locations , which represent the beginning of choice trials and which comprised the majority of replay events ( 703/876 , 80 . 3%; compared to 173 at choice reward locations ) . For each recording , we constructed a distribution of all decoded locations within these replay events and compared this to the distribution of decoded spatial locations specifically associated with 84 ms delayed RR VTA unit spikes and nonRR VTA unit spikes . 10 . 7554/eLife . 05360 . 009Figure 4 . Reward responsive VTA units coordinate with replayed reward locations . ( A ) The SPW-R triggered VTA local field potential ( LFP ) shows a delayed potential . Time 0 reflects the start of SPW-R events . ( B ) Incorporating this delay between the hippocampus and VTA , across replay events occurring at the forced reward sites , RR unit spikes preferentially coordinated with replayed reward locations compared to SPW-R replay content in general ( p=0 . 048 , chi = 3 . 9 , Chi-square test ) and compared to nonRR units ( p=0 . 016 , nonparametric permutation test ) . Error bars represent s . d . ( C ) Probability distribution of replayed spatial locations for replay events occurring at the forced reward sites on the spatial working memory ( SWM ) ( 1 ) and linear tasks ( 2 ) ( 10 cm bins ) , accumulated across recordings . Dashed boxes designate reward sites . ( D , E ) Distribution of replayed locations coinciding with RR unit spikes ( D1 , 2 ) and nonRR unit spikes ( E1 , 2 ) adjusting for the latency between SPW-R onset and the VTA delayed potential . The probability colorbar for the SWM task ranges from 0 to 0 . 04 and for the linear track ranges from 0 to 0 . 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 009 Across replay events , the probability of decoded spatial locations was biased toward reward sites ( probability/spatial bin of replay content at reward locations 0 . 038 ± 0 . 005 bin-1; non-reward locations 0 . 021 ± 0 . 002 bin-1; p=2 . 4 × 10-4 , signed-rank test; n = 14 recording sessions; see Materials and methods; Figure 4C ) . Incorporating the latency of 84 ms , the timing of RR VTA unit activity within replay events specifically coincided with the replay of reward locations ( probability/spatial bin of VTA unit activity at reward locations 0 . 044 ± 0 . 003 bin-1; non-reward locations 0 . 021 ± 0 . 001 bin-1; p=9 . 0 × 10-8 signed-rank test; n = 41 RR units ) . Across all recordings , RR VTA unit spikes were biased to coincide with replayed reward site locations in excess of the proportion of reward site locations within replay events ( reward site bias for replay events: 0 . 451±0 . 007; excess reward site bias for RR VTA units: 0 . 022 ± 0 . 011 ( mean ± s . d . ) ; p=0 . 048 , chi = 3 . 9 , Chi-square test; see Materials and methods; Figure 4B , D ) . In contrast , nonRR units did not preferentially coordinate with replayed reward site locations ( excess reward site bias for nonRR VTA units: -0 . 013 ± 0 . 010; p=0 . 3 , chi = 1 . 2 , Chi-square test; Figure 4B , E ) . The contrast of excess reward site bias of RR units and nonRR units was significant ( p<0 . 016 , nonparametric permutation test; see Materials and methods ) . The bias in coordination of RR units but not nonRR units with replayed reward locations persisted when the current location of the animal was excluded from the analysis ( reward site bias for replay events: 0 . 331 ± 0 . 009; excess reward site bias for RR VTA units: 0 . 031 ± 0 . 014; p=0 . 041 chi = 4 . 2 Chi-square test; nonRR VTA units: 0 . 008 ± 0 . 013; p=0 . 6 , chi = 0 . 2 , Chi-square test ) . These data suggest that RR VTA units preferentially associate with the hippocampal replayed representation of reward locations . To evaluate the task dependence of this observation , we compared the preferential coordination of RR units with replayed reward site locations on the SWM task and on the linear track . Interestingly , the excess reward site bias of RR units was greater on the SWM task than on the linear track ( excess reward site bias for RR units on the SWM task 0 . 027 ± 0 . 014; excess reward site bias for RR units on the linear track 0 . 017 ± 0 . 015; p=0 . 045 , nonparametric permutation test ) . Because this task-dependence could reflect a role for the coordination of RR units with replayed reward locations in choice behavior , we examined whether the preferential coordination of RR units with replayed reward site locations reflected recent choice behavior or predicted future choice behavior on the SWM task . However , we were unable to detect a difference in the excess reward site bias of RR units at replay events immediately following correct trials versus error trials ( excess reward site bias after correct trials 0 . 022 ± 0 . 020; after error trials -0 . 018 ± 0 . 031; p=0 . 19 , nonparametric permutation test ) . Similarly , the excess reward site bias of RR units at replay events was no greater immediately prior to correct trials than prior to error trials ( excess reward site bias prior to correct trials 0 . 001 ± 0 . 021; prior to error trials 0 . 013 ± 0 . 041 , p=0 . 6 , nonparametric permutation test ) . These results suggest that the preferential coordination of RR unit activity with replayed reward locations is task dependent but may not simply recapitulate or predict immediate reward-associated experience . Previous work has posited a specific coordination between dopamine neuronal activity and replay events comprised of spatial sequences starting locally and replaying away from the animal ( centrifugal events ) in reverse order compared to their order during behavior ( Foster and Wilson , 2006 ) . To address whether the preferential coordination of RR VTA units with replayed reward site locations may derive from a selective engagement of VTA units with centrifugal replay events or with reverse replay events , in distinction from replay sequences that start remotely and replay towards the animal ( centripetal events ) , or forward replay sequences that replay in the same order as they did during behavior , we first differentiated between instantaneous centrifugal and centripetal spatial content , and between forward and reverse spatial content , by reconstructing SPW-R replay events at forced reward sites using both position and run direction data ( Figure 5A–C; see Materials and methods ) . We accumulated centrifugal , centripetal , forward , and reverse replayed spatial distributions separately , and we compared them to each other and to the distribution of decoded spatial locations and run direction specifically associated with RR VTA unit spikes . 10 . 7554/eLife . 05360 . 010Figure 5 . The bias of reward responsive VTA unit activity towards the replay of reward locations is greater for centrifugal than centripetal replay . ( A1 , 2 ) Bayesian reconstruction of run position and run direction on the linear track ( 500 ms time bins ) . Outbound refers to run direction from 0 to 200 cm . ( A3 ) Position confusion plot for this recording session , using alternating 1 s epochs for training and testing the reconstruction algorithm . ( A4 ) Run direction confusion plot . ( B ) Centrifugal , forward replay event occurring while the rat paused at the far reward site ( 190 cm; black circle indicates the rat’s position ) . ( B1 ) Position reconstruction ( 25 ms time bins ) . ( B2 ) direction reconstruction ) . ( B3 ) The associated hippocampal multiunit activity . ( C ) Centripetal , forward replay event occurring while the rat paused at the far reward site . ( C1 ) Position reconstruction . ( C2 ) Direction reconstruction . ( C3 ) Associated hippocampal multiunit activity . ( D ) Across centrifugal replay time bins , RR unit spikes preferentially coordinated with replay of reward locations compared to centrifugal replay content in general ( p=0 . 014 , chi=6 . 0 , Chi-square test ) and compared to nonRR units at centrifugal replay ( p=0 . 05 , nonparametric permutation test ) . Error bars represent s . d . ( E ) Across centripetal replay time bins , RR unit spikes showed no increase in coordination with replay of reward locations compared to centripetal replay content in general ( p=0 . 5 , chi=0 . 5 , Chi-square test ) . Error bars represent s . d . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 01010 . 7554/eLife . 05360 . 011Figure 5—figure supplement 1 . Centrifugal and centripetal replayed locations associated with RR unit activity on the SWM task and the linear track . ( A ) On the SWM task , the distribution of centrifugal replayed locations ( green ) is less concentrated at reward sites ( marked by vertical lines ) than the distribution of RR unit-associated centrifugal replayed locations ( blue ) . See Figure 5 for statistics . Maze segments were aggregated by apposing them in the direction of run in the choice direction: from force reward sites ( R3 , R4 ) to force point ( fp ) to choice point ( cp ) to choice reward sites ( R1 , R2 ) . Spatial bins 1–10 show the average of the force arms of the task ( arms 3 and 4 ) , spatial bins 11–20 show the central arm , and spatial bins 21–30 show the average of the choice arms ( arms 1 and 2 ) . ( B ) The distribution of centripetal replayed locations ( green ) on the SWM task is similar to the distribution of RR unit-associated centripetal replayed locations ( blue ) . ( C ) On the linear track , the distributions of centrifugal replayed locations ( green ) and RR unit-associated centrifugal replayed locations ( blue ) are similar . ( D ) On the linear track , the distribution of centripetal replayed locations ( green ) and the distribution of RR unit-associated centripetal replayed locations ( blue ) are also similar . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 011 RR units preferentially coordinated with the reward site representation of centrifugal replay events but not centripetal replay events ( reward site bias for centrifugal replay spatial content 0 . 493 ± 0 . 005; excess reward site bias for RR units at centrifugal replay 0 . 051 ± 0 . 020; p=0 . 014 , chi = 6 . 0 , Chi-square test; reward site bias for centripetal replay spatial content 0 . 524 ± 0 . 006; excess reward site bias for RR units at centripetal replay 0 . 016 ± 0 . 021; p=0 . 5 , chi = 0 . 5 , Chi-square test; Figure 5D , E ) . In contrast , nonRR units showed no excess reward site bias for replayed spatial content ( excess reward site bias for nonRR units at centrifugal replay 0 . 002 ± 0 . 019 , p=1 , chi = 0 . 001 , Chi-square test; at centripetal replay: 0 . 011 ± 0 . 022; p=0 . 6 , chi = 0 . 2 ) . These results demonstrate that RR units preferentially coordinate with centrifugal replay content . We next examined RR and nonRR unit activity associated with forward and reverse replay events . The probability of decoding spatial locations at reward sites was similar for forward and reverse replay events ( reward site bias for forward replay spatial content 0 . 422 ± 0 . 013; reward site bias for reverse replay spatial content 0 . 419 ± 0 . 014; p=0 . 9 , chi = 0 . 2 , Chi-square test ) . We did not detect a selective engagement of RR units with reward locations of reverse replay over forward replay ( excess reward site bias for RR units at reverse replay 0 . 035 ± 0 . 020; at forward replay 0 . 034±0 . 020; p=1 , nonparametric permutation test ) . To evaluate the task dependence of the preferential coordination of VTA RR units with centrifugal replay spatial content , we analyzed the SWM task and the linear track separately . Similar to the results described above , RR units acquired on the SWM task preferentially coordinated with reward site representations of centrifugal replay events ( excess reward site bias for RR units at centrifugal replay: 0 . 071 ± 0 . 032; p=0 . 028 , chi = 4 . 8 ) in contrast to centripetal replay events ( excess reward site bias for RR units at centripetal replay: 0 . 031 ± 0 . 028 p=0 . 3 , chi = 1 . 0; Figure 5—figure supplement 1 ) . However , this preferential coordination was not observed on the linear track ( excess reward site bias for RR units at centrifugal replay on the linear track: 0 . 040 ± 0 . 024; p=0 . 11 , chi = 2 . 5; excess reward site bias for RR units at centripetal replay: -0 . 004 ± 0 . 031; p=0 . 9 , chi = 0 . 01 ) . Thus , the coordination of VTA RR units with centrifugal replay of reward site locations was stronger on the SWM task . In addition to modulating their activity at hippocampal SPW-R events , many VTA units ( 39/84; 43% ) phase-locked to hippocampal theta during run behavior ( Rayleigh test for uniformity against unimodal alternative p<0 . 05 , phase preference -10 ± 16 degrees , relative to the peak of theta ) , consistent with previous observations ( Fujisawa and Buzsáki , 2011 ) ( Figure 6 ) . The coordination of neural activity with the hippocampal theta rhythm has been proposed to be a mechanism used in spatial working memory ( Jones and Wilson , 2005; Fujisawa and Buzsáki , 2011 ) . We therefore sought to determine the extent to which theta phase-locking of VTA units predicted their coordination with SPW-R events . Circular concentration of VTA unit spikes around the mean preferred hippocampal theta phase was greater for RR units than nonRR units , as measured with the circular concentration coefficient kappa as described previously ( Jones and Wilson , 2005; Siapas et al . , 2005 ) , ( mean ± s . e . m . , RR: 0 . 139 ± 0 . 012 , n = 47; nonRR: 0 . 099 ± 0 . 019 , n = 24; p=0 . 03 , rank-sum test; Figure 6B ) . For theta-modulated RR units but not theta-modulated nonRR units , circular concentration at hippocampal theta positively correlated with the probability that spike-associated replayed spatial content represented reward locations ( RR units: r = 0 . 51 , p=0 . 04 , n = 17; nonRR units: r = 0 . 45 , p=0 . 13 , n = 13; Figure 6C ) . In contrast , the circular concentration coefficient of VTA units at hippocampal theta did not correlate with the firing rate of those units ( RR units: r = -0 . 22 , p=0 . 4 , n = 18; nonRR units: r = -0 . 49 , p=0 . 09 , n = 13 ) . In addition , the circular concentration of RR units at hippocampal theta did not correlate with their modulation depth at SPW-R replay events ( r = 0 . 12 , p=0 . 5 , n = 45 ) . Thus , phase-locking of RR units to hippocampal theta during run behavior was associated with the timing but not the number of spikes of RR units at SPW-R events . 10 . 7554/eLife . 05360 . 012Figure 6 . VTA units coordinate with hippocampal theta . ( A ) Spike times of a reward responsive ( RR ) VTA unit relative to hippocampal theta and raw LFP during running behavior , and spike phase distribution ( circular concentration coefficient , kappa = 0 . 14; Rayleigh statistic p value = 0 . 002 ) . ( B ) Circular concentration at hippocampal theta is greater for RR units than nonRR units ( p=0 . 031 , rank-sum test ) . Error bars represent s . e . m . ( C ) The probability of replayed reward locations coinciding with the spikes of theta-modulated RR units correlates with the circular concentration of those units at hippocampal theta ( r = 0 . 51 , p=0 . 038 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 012 Previous work has demonstrated hippocampal SPW-R replay during SWS ( Lee and Wilson , 2002; Pavlides and Winson , 1989; Wilson and Mcnaughton , 1994; Nádasdy et al . , 1999; Ji and Wilson , 2007 ) and has shown that medial forebrain bundle stimulation triggered on place cell activity in sleep exerts a powerful influence on post-sleep behavior ( de Lavilléon et al . , 2015 ) . Having identified replay-related modulation of VTA unit activity , we therefore sought to evaluate SPW-R-associated VTA unit activity in SWS acquired immediately subsequent to run behavior . Both RR and nonRR VTA units reduced their firing rates in SWS ( RR units: run 8 . 8 ± 1 . 7 Hz , quiet wakefulness 6 . 6 ± 1 . 3 Hz , SWS 4 . 5 ± 0 . 8 Hz; run vs SWS , p=2 . 2 × 10-7; quiet vs SWS , p=2 . 3 × 10-5; signed-rank tests , n = 47; nonRR units: run 26 . 8 ± 7 . 2 Hz , quiet wakefulness 20 . 6 ± 5 . 4 Hz , SWS 10 . 4 ± 2 . 7 Hz; run vs SWS , p=7 . 1 × 10-5; quiet vs SWS , p=1 . 0 × 10-4; signed-rank tests , n = 24 ) . In addition , the modulation depth of RR unit activity at SPW-R events was significantly reduced in SWS ( modulation depth in quiet wakefulness 0 . 21 ± 0 . 04; in SWS 0 . 10 ± 0 . 02; p=0 . 003 , rank-sum test , n = 45 in quiet wakefulness , n = 39 in SWS; Figure 7A–C ) . In contrast to the awake state , RR units in SWS were often negatively modulated at SPW-R events ( 30/39 units , p=0 . 001 , sign test ) . Modulation depths in quiet wakefulness and sleep were significantly correlated ( modulation depth , r = 0 . 47 , p=0 . 002 , n = 39 ) , but the sign of modulation across these states was not ( modulation sign , r = 0 . 13 , p=0 . 4 , n = 39 ) . In contrast to RR units , the SPW-R modulation of nonRR units was not significantly modified by behavioral state ( modulation depth in quiet wakefulness 0 . 11 ± 0 . 02; SWS 0 . 06 ± 0 . 01; p=0 . 13 , rank-sum test , n = 23 in quiet wakefulness , n = 20 in SWS; Figure 7C ) . Thus , RR units coordinated more robustly with hippocampal SPW-R events of quiet wakefulness than with those of SWS . 10 . 7554/eLife . 05360 . 013Figure 7 . SPW-R-associated modulation of VTA units during periods of quiet wakefulness ( QW ) on the task and during subsequent slow wave sleep ( SWS ) . ( A ) Rastered QW-associated reward responsive ( RR ) VTA unit spikes ( 1 ) and RR unit and hippocampal ( HC ) multiunit PETHs ( 2 , 3 ) , aligned to SPW-R events . ( B ) SWS-associated data for the same RR unit . ( C ) SPW-R event modulation depth of RR and nonRR unit activity in QW and SWS ( RR units: QW vs SWS , p=0 . 003 , rank-sum test; nonRR units: QW vs SWS , p=0 . 13 , rank-sum test ) . Error bars represent s . e . m . ( D1 ) Hippocampal multiunit activity , ( 2 ) ripple band , and ( 3 ) two RR VTA units in SWS . ( E ) Distributions of ( 1 ) SWS frame duration and ( 2 ) interframe duration across recordings . ( F ) Cumulative distribution of within-frame SPW-R frequency . ( G ) Within-frame VTA unit activity . RR units are shown separately in dashed line . ( H ) The difference in each VTA unit’s activity at frames of high and low SPW-R rate , defined relative to the mean ( RR units: p=0 . 003 , signed-rank test; nonRR units: p=0 . 5 ) . ( I ) The difference in mean spatial content of frames with high and low VTA unit activity , relative to the mean ( RR units: p=0 . 045 , signed-rank test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 013 The smaller modulation of RR units with SPW-R events in SWS compared to quiet wakefulness could result from a difference in state or from a difference in the spatial content expressed in these states . For example , replayed spatial content in sleep may be less biased by recent experience than replayed spatial content in quiet wakefulness . To explore this question , we set out to examine hippocampal replay in detail in SWS . We first evaluated the prevalence of replay-associated SPW-R events in SWS . Hippocampal replay of recent experience during SWS identified using Bayesian position reconstruction was less prevalent than during quiet wakefulness ( SWS 16 . 0 ± 1 . 7% ( 800/5820 ) , compared with 24 . 8 ± 2% of wake-associated SPW-R events , p=0 . 003 , signed-rank test ) . Hippocampal activity in SWS is characterized by epochs of up-state-like neuronal population activity known as frames , within which SPW-R events occur and during which coordinated replay between the hippocampus and neocortex has been observed ( Ji and Wilson , 2007 ) . Because hippocampal replay in SWS has been associated with SPW-R events within frames , we compared VTA unit activity across frames associated with high versus low SPW-R rates ( SPW-R events per second ) , relative to the mean . RR unit firing rate was lower in high rate SPW-R frames than in low rate SPW-R frames ( high rate SPW-R frames 4 . 23 ± 0 . 74 Hz; low rate SPW-R frames 4 . 64 ± 0 . 77 Hz; signed-rank test , p=0 . 003 , n = 47; Figure 7D–H ) . In contrast , nonRR unit firing rate was similar across these groups of frames ( high rate SPW-R frames 10 . 26 ± 2 . 66 Hz; low rate SPW-R frames 10 . 41 ± 2 . 65 Hz; signed-rank test , p=0 . 45 , n = 24; Figure 7D–H ) . Thus , RR unit activity was biased away from SPW-R rich frames . We next asked whether VTA unit firing rates varied with frame-associated replay of recent experience . For this purpose , we evaluated average spatial content in each frame , measured as the average across time bins of the maximum decoded probability at each time bin . RR unit firing rate was lower in frames associated with higher spatial content ( above the mean spatial content of frames ) than in frames associated with lower spatial content ( below the mean; RR unit firing rate at high spatial content frames 4 . 11 ± 0 . 78 Hz vs . firing rate at low spatial content frames 4 . 83 ± 0 . 91 Hz , p=0 . 001; signed-rank test , n = 42 ) . Frames with higher spatial content were longer than frames with lower spatial content ( median 3 . 98 s vs 1 . 85 s; p=0 . 009; signed-rank test , n = 14 recordings ) . Because differences in frame duration could affect estimates of within-frame firing rates , we also performed an inverse analysis , in which we sorted hippocampal frames by the associated firing rate of VTA units and then compared their spatial content associated with recent experience . Frames associated with high RR unit firing rate ( above the mean firing rate for each unit ) had lower mean spatial content than frames associated with low ( below the mean ) RR unit firing rate ( spatial content at high firing rate frames 0 . 185 ± 0 . 012 ( a . u . ) vs . spatial content at low firing rate frames 0 . 191 ± 0 . 013 ( a . u . ) ; signed-rank test , p=0 . 045 , n = 42; Figure 7I ) . Together , these results demonstrate that SWS is associated with reduced SPW-R modulation of RR unit activity and with reduced RR unit activity in hippocampal frames containing a high rate of SPW-R events and in frames associated with high spatial information about recently explored environments . Hippocampal dependent learning and memory are influenced by reward , and SPW-R events contribute to these functions ( Jadhav et al . , 2012; Girardeau et al . , 2009; Ego-Stengel and Wilson , 2010 ) . As VTA dopamine cells are driven by reward prediction errors ( Schultz , 1998 ) and have been suggested to provide an error signal to guide learning in downstream brain regions ( Montague et al . , 1996 ) , we posited that the coincidence of dopamine neuronal activity with sequence replay in SPW-R events of quiet wakefulness could reinforce spatial experience , mediating the influence of reward on hippocampal dependent processing ( Singer and Frank , 2009 ) and memory formation . Taken collectively , the results of this study support this possibility , demonstrating that during quiet wakefulness but not SWS , RR VTA neurons coordinate selectively with hippocampal replay sequences and are biased in their timing towards the reactivated representation of rewarded locations . In contrast , nonRR VTA neurons did not coordinate with the specific spatial content of replay sequences . RR neurons were also more strongly phase-locked to hippocampal theta than nonRR neurons , and the extent of phase-locking correlated with the coordination of RR unit activity with replayed reward locations . Previous work has demonstrated theta phase-dependent interactions between the hippocampus , prefrontal cortex , and VTA during working memory-dependent , single trial decisions ( Jones and Wilson , 2005; Fujisawa and Buzsáki , 2011 ) . Our data support a model in which these experience-dependent associations , once established , are re-expressed in SPW-Rs of quiet wakefulness , to guide spatial memory across trials . In addition , our results identify two possible endogenous substrates by which optogenetically released dopamine can increase off-line reactivation of hippocampal cells and improve spatial memory performance ( Mcnamara et al . , 2014 ) : direct coordination of dopamine neuronal activity with hippocampal replay of quiet wakefulness , or coordination with hippocampal theta triggering a subsequent reactivation of dopamine neurons that engages with hippocampal replay . The sign of SPW-R modulation varied across RR neurons , often recapitulating their reward-associated modulation of firing rate . This result suggests that as a population , RR neurons replay their reward-related activity in concert with hippocampal sequence replay , to selectively reinforce reward-associated behavior . Coordination with replay has previously been observed in neurons of the primary visual cortex ( Ji and Wilson , 2007 ) and the striatum ( Pennartz et al . , 2004; Lansink et al . , 2009 ) , a major target of the VTA that represents rewards ( Schultz et al . , 1992; Cardinal et al . , 2002 ) . The current results extend these findings , supporting the hypothesis that replay events engage both cortical and subcortical structures to create an accurate memory trace of recent experience . In this study , RR neurons preferentially coordinated with SPW-R events of quiet wakefulness compared to SWS and were least active in SWS frames associated with high spatial content . Although we observed a higher proportion of replay events in quiet wakefulness than in SWS , consistent with prior observations ( Karlsson and Frank , 2009 ) , differences in the prevalence of replay events in quiet wakefulness and SWS are unlikely to underlie the impact of SWS on VTA activity , given that SWS frames with higher spatial content were associated with greater reduction in RR unit activity . These results suggest a functional distinction between brain processes that subserve spatial memory within sessions versus spatial learning across sessions , consistent with prior observations that tie awake hippocampal replay events to within-session performance ( Jadhav et al . , 2012 ) yet associate replay events in post-session epochs rich in SWS to cross-session spatial learning ( Girardeau et al . , 2009; Ego-Stengel and Wilson , 2010 ) . Memory consolidation in SWS is likely to require broad evaluation of behavioral experiences , and the present results suggest that this evaluation can occur in the absence of their reward prediction contingencies , as represented in the activity of VTA neurons . In this regard , introducing anomalous reward prediction-related activity during sleep via medial forebrain bundle stimulation triggered on place cell activity ( de Lavilléon et al . , 2015 ) has been recently demonstrated to drive goal-directed spatial behavior in wakefulness . In neuropsychiatric diseases such as addiction or obsessive compulsive disorders , such anomalous associations could contribute to maladaptive behaviors . In contrast to the state-dependence of VTA-hippocampal interactions , neurons of the ventral striatum have been found to coordinate with hippocampal replay in SWS ( Pennartz , et al . , 2004; Lansink et al . , 2009 ) . One possible explanation for this distinction , consistent with the suggested role of dopamine as a teaching signal , is that VTA dopamine activity stabilizes and links replayed sequences in quiet wakefulness across brain regions for subsequent consolidative processes in SWS . It is notable that replay events in these recordings were biased in their spatial content towards reward sites . The basis for this bias remains to be determined and may be driven by a number of factors not examined here , including the presence or expectation of reward , as well as differences in the dwell times , behavioral states , and behavioral repertoires manifested at reward and nonreward locations . In this study , we observed a preferential engagement of RR units with the reward representation of centrifugal compared to centripetal replayed spatial content , while we did not detect a preference for RR units for the reward representation of reverse compared to forward replayed spatial content . These results are broadly consistent with the previous proposal that dopamine may function to propagate expected value across reactivated hippocampal sequences ( Foster and Wilson , 2006 ) . We also observed greater coordination of RR cells with replayed reward locations in the SWM task compared to the linear track , raising the possibility that VTA-hippocampal coordination at SPW-R events may reflect task contingencies . However , this result should be considered with caution given the limited sample size acquired in each task . Although we did not detect the preferential coordination of RR cells with replayed reward locations immediately after or immediately prior to successful choice behavior , as compared with errors , it remains possible that RR unit coordination with replayed reward locations could reflect ( or predict ) choice behavior on longer timescales . Of note , the experimental design was not intended to dissect the relationship of other task dependent features , such as uncertainty , to hippocampal-VTA coordination , and this will be worth pursuing in future experiments . Interestingly , we found clear differences between RR neurons and nonRR neurons in their engagement with the hippocampus . A higher proportion of SPW-R modulated neurons were RR , RR neurons were more biased to fire in relation to replayed reward locations , and RR neurons demonstrated stronger phase-locking to the hippocampal theta rhythm . These results suggest that RR and nonRR neurons represent distinct functional classes of cells , perhaps associated with different cell types ( Cohen et al . , 2012; Lammel et al . , 2008; Margolis et al . , 2012; Hnasko et al . , 2012 ) , that differentially contribute to hippocampal-dependent spatial memory . Given the uncertainty in the confidence with which dopamine cells can be identified on the basis of electrophysiologic criteria ( Ungless and Grace , 2012; Cohen et al . , 2012; Margolis et al . , 2006 ) , however , we chose not to restrict our analysis to putative dopamine cells . Even so , over 80% of RR neurons had waveform properties that have been associated with dopamine cells , including a wide action potential and firing rates below 10 Hz . Models of reinforcement learning have suggested distinct contributions of dopamine to certain forms of learning ( Walsh and Anderson , 2014; Doll et al . , 2012 ) . The specific , wake-associated coordination of RR VTA neurons with hippocampal activity may mediate the capacity for task-associated replay content to predict future paths to goal locations ( Pfeiffer and Foster , 2013; de Lavilléon et al . , 2015 ) and may underlie dopamine’s stabilization of hippocampal replay ( Mcnamara et al . , 2014 ) . These specific VTA-hippocampal interactions are likely to play a critical role in context-dependent reward seeking behavior . In addition , the state-dependent coordination of VTA reinforcement activity with hippocampal spatial replay events directs attention to the differential processing of spatial memory in wakefulness and SWS . Five male Long-Evans rats ( 4–6 months old ) were implanted under anesthesia ( induction: ketamine ( 50 mg/kg ) and xylazine ( 6 mg/kg ) ; maintenance: isoflurane 0 . 5–3% , ) with 2 arrays of independently movable recording tetrodes ( for detailed methods , see Jones and Wilson , 2005 ) . One array of 6–10 tetrodes was directed to the dorsal CA1 pyramidal cell layer ( anterior-posterior ( AP ) -3 . 6 mm , lateral ( L ) + 2 . 4 mm; relative to Bregma ) . A reference electrode was placed in the white matter above the hippocampal cell layer for differential recordings . An additional array of 8–11 tetrodes was targeted to the VTA ( AP -5 . 3 mm , L + 1 . 0 mm ) . A tetrode without unit activity served as the local reference for VTA differential recordings . In one rat , stereotrodes as well as tetrodes were used for VTA recordings . Tetrodes were advanced to their target positions over several weeks . In 4 animals , an additional array of 4–8 electrodes was targeted to the prefrontal cortex for purposes unrelated to the present study . VTA tetrodes were lowered after each recording session and final electrode positions were confirmed with electrolytic lesions and histology ( Paxinos and Watson , 1998 ) after recording was completed ( Figure 8 ) . 10 . 7554/eLife . 05360 . 014Figure 8 . Histological location of tetrode tips targeting the VTA . For each rat , electrolytic lesions marked the tetrode tip locations , and these were mapped onto the stereotaxic atlas of Paxinos and Watson ( 1998 ) . Tetrode tips under-represent recording locations , which were acquired as electrodes were systematically lowered within the VTA along their tracks . SNR , substantia nigra reticulata . DOI: http://dx . doi . org/10 . 7554/eLife . 05360 . 014 For hippocampal recording , 1 ms windows around thresholded extracellular action potentials were acquired on-line at 31 kHz , 300-6 , 000 Hz filtering . In order to retain wave-shape information for VTA units , which often have long waveforms , VTA unit recordings were acquired continuously at 31 kHz , 300–6000 Hz filtering; and subsequently thresholded ( 60 μV ) offline to isolate extracellular action potentials . Local field potentials ( 2 kHz sampling , 1–475 Hz filtering ) were recorded from one electrode on each tetrode . Head position and direction were monitored using overhead camera tracking of two sets of infrared diodes that were mounted on the headstage and that alternated at 30 Hz each . Animals were trained over 2–4 weeks to run a spatial appetitive choice task ( Jones and Wilson , 2005 ) on an end-to-end T-maze ( Figure 1 ) . Each trial consisted of two phases . In the sample phase , rats were directed pseudorandomly to either the left ( or right ) reward site on the forced side of the maze , where a nosepoke-triggered grain pellet reward could be obtained ( MedAssociates ( Georgia , VT ) ; Bioserv ) . In the test phase of the trial , rats traversed the central arm to a choice point , where they chose to go right ( or left ) in order to win reward on the choice side of the maze . The reward contingency was set up such that if the rat had been forced to turn left in the sample phase , then the correct response in the test phase was to turn right . Individual trajectories between reward sites on forced and choice sides were 300 cm long . After training , the animals were implanted with tetrode arrays . Following recovery from surgery , animals were food deprived to 85% of their free-feeding weights . Animals relearned the task slowly , improving from 60 ± 3% ( mean ± s . e . m . ) performance in the first three days of behavior to 74 ± 2% in the final three behavioral sessions . Recordings on the spatial alternation task were acquired during the day over 22 days . Two animals ran only on a 200 cm linear track , to acquire food reward at both ends . Sleep sessions were acquired immediately after behavioral sessions in a sleep chamber with opaque walls within the recording room . Animals were housed in individual cages with a 12 hr light-12 hr dark standard light cycle . Established software was used for initial identification and characterization of unit activity . This includes identification of well-isolated clusters of spike waveforms and differentiation of putative hippocampal pyramidal cells , hippocampal interneurons , and VTA units ( Xclust , M . Wilson ) . Matlab ( MathWorks , Natick , Massachusetts ) was used for further data analysis ( https://github . com/stephengomperts/eLife_2015 ) . Unless otherwise stated , error bars reflect s . e . m . Reward responsive ( RR ) VTA units in the SWM task were identified as those with significantly different firing rates on correct versus error trials during approach to reward ( defined as the 2 s window prior to nosepoke ) or reward acquisition ( defined as the 3 s window following nosepoke; two-sided t-tests , p<0 . 05 for significance , n = 27 ) . Reward responsiveness on the linear maze was determined by comparing firing rates during reward acquisition to a 3 s window that ended 2 s before nosepoke . These two approaches were highly correlated for the SWM task ( r = 0 . 49 , p=2 . 4 × 10-4 , n = 51 ) . Differential VTA unit activity on correct versus error trials was common in our dataset and is consistent with prior reports in instrumental tasks ( Morris et al . , 2006; Roesch et al . , 2007; Totah et al . , 2013 ) . Such results have been interpreted as the representation of choice-associated reward prediction error , formalized for example in the Q-value associated temporal difference prediction error , where Q-value is the value of selecting a particular action at a given state ( Morris et al . , 2006 ) . Waveform duration ( time from peak to peak ) and trough to peak ratio ( trough/[peak + trough] ) were noted to distinguish two VTA unit populations , as shown previously ( Fujisawa and Buzsáki , 2011 ) ( Figure 1C ) . Waveform duration and the biphasic duration ( Ungless and Grace , 2012 ) were highly correlated ( r = 0 . 79 , p=8 . 53 × 10-20 , n = 145 ) . Of 145 VTA units recorded , 84 units ( 47 RR; 24 nonRR; 13 unclassified due to low firing rate [<0 . 3Hz] ) were recorded concurrent with wake-associated hippocampal activity , over 16 behavioral sessions in 5 rats , with 10 ( 2 , 3 , and 5 ) sessions on the end-to-end T maze; 6 ( 4 and 2 ) sessions on the linear track; and in subsequent slow wave sleep . Hippocampal activity in the 5th rat , acquired in 2 sessions on the linear track , was insufficient for position reconstruction and assessment of hippocampal frames , reducing the number of VTA units for replay and frame analyses to 66 , acquired over 14 behavioral sessions . Local field potentials were filtered to obtain hippocampal ripples ( Blackman filter; 100–300 Hz ) and theta oscillations ( 4–12 Hz ) ( Jones and Wilson , 2005 ) . Hippocampal action potentials that exceeded threshold ( 60 μV ) were aggregated as multiunit activity to measure SPW-R-associated bursts in hippocampal activity . Bursts with peak firing rate exceeding 4 standard deviations above the mean , behaviorally constrained to periods of speed < 10 cm/s , were identified as SPW-R multiunit events . The start and end of each event were defined as the times surrounding the event at which the multiunit firing rate fell back to its mean value . Although ripple power was not an explicit constraint for SPW-R multiunit events , 92 . 8% of SPW-R multiunit events had ripple power exceeding 2 Z scores above baseline ( 89 . 0% in replay; 93 . 5% in nonreplay ) . The majority of SPW-R events occurred at the force reward sites , where the animals paused longest . VTA single-unit activity during single trials was summed over repeated trials in a session to generate peri-event time histograms ( PETHs ) triggered on the start of SPW-R events . The PETH was smoothed with a Gaussian window ( σ = 50 ms; similar results were found over a range of σ ( 30–200 ms ) . SPW-R modulation amplitude was measured relative to a 300 ms baseline that ended 100 ms before the event , as the baseline-normalized difference between the PETH amplitude measured at the midpoint of the SPW-R event and the mean baseline amplitude . Results were similar using a 5 s baseline ending 1 s before the event . Units with low baseline firing rate ( <0 . 3 Hz ) were excluded from analysis to exclude undersampled PETHs . To compute significance of modulation , the SPW-R-aligned raster of each unit was bootstrapped to derive and compare confidence intervals , at the p<0 . 05 level , of a 50 ms bin at the midpoint of the SPW-R event and the average of the confidence intervals of the 300 ms baseline . The animal’s location was expressed as a linear distance along the track . For the end-to-end T-maze , the track was linearized by adjoining the 5 segments ( Figure 3—figure supplement 1A , B ) . To deal appropriately with the discrete jumps in the linearized representation , a distance look-up table was constructed for all pairs of densely sampled points along the track . We applied a Bayesian estimation algorithm ( Davidson et al . , 2009; Zhang et al . , 1998 ) to reconstruct position from hippocampal population activity . We expressed the relationship between neuronal activity and position in an encoding model that incorporated spike waveform amplitude features of unsorted spikes in run epochs with speed > 10 cm/s ( ‘clusterless decoding’; only putative pyramidal neuron spikes with peak amplitude > 100 μV and spike width > 300 μs were included ) . For direction reconstruction , we generated an independent encoding model that related running direction in run epochs to spike waveform amplitude features of unsorted spikes . Using the run velocity threshold of 10 cm/s , reward site arrival and departure were well represented . A non-informative uniform prior was used as we did not want to impose any spatial-temporal structure on the estimated positions in SPW-R events . To verify that position on the track could be accurately estimated from hippocampal population activity , we used a cross-validation procedure by dividing run epochs into alternating 1 s training and testing epochs . The rat’s position ( in 10 cm spatial bins ) was estimated in non-overlapping 500 ms time bins in the testing epochs and compared to the true location . Decoding performance was assessed by computing the median error and confusion matrices ( Figure 3—figure supplement 1C ) . We compared the clusterless decoding paradigm to the standard decoding of spike-sorted units . Spatial tuning curves were constructed for all manually sorted hippocampal place cells with peak place field firing rate > 3 Hz . Median error in clusterless position reconstruction was significantly lower than with spike sorting-based reconstruction ( clusterless: 8 . 3 ± 0 . 5 cm , spike sorting-based 15 . 2 ± 1 . 9 cm , p<1 . 22 × 10-4 , signed-rank test , n = 14 ) . Mean error in clusterless direction reconstruction was 0 . 26 ± 0 . 02 . Replay-detection was performed as described previously ( Davidson et al . , 2009 ) . We applied the clusterless Bayesian estimator to non-overlapping , 25 ms bins during SPW-R events occurring in non-run periods ( run speed < 10 cm/s; Figure 3—figure supplement 1D ) . Excluding running direction , four paths exist on the SWM task that connect the forced choice reward sites to the free choice reward sites ( two of which are correct and two incorrect ) . For each SPW-R event , a constant speed trajectory was fitted to the sequence of position estimates ( Davidson et al . , 2009; Kloosterman , 2012 ) for each of the four possible paths . The best fitting trajectory was selected based on a goodness-of-fit score ( ‘replay score’ ) , computed as the mean posterior probability within 15 cm of the fitted trajectory . To test if fitted trajectories could be expected by chance alignment of position estimates , the replay score for each candidate event was compared to replay score distributions derived with two shuffles of the data , using the approach described in ( Davidson et al . , 2009 ) . The first ‘column cycle shuffle’ controls for the linear alignment of consecutive position estimates by circularly shifting the estimated probability distribution over position ( PDF ) in each candidate event time bin by a random distance . The second ‘pseudo-event shuffle’ controls for bias towards particular locations , by constructing artificial candidate events generated by replacing each PDF in a candidate event with a PDF drawn at random from the complete set of candidate events in each session . Each shuffle was performed 1500 times for each candidate event to obtain sample distributions of the replay score . To increase detection sensitivity of possible replay events on the spatial working memory task , we considered replay events to be those with a Monte Carlo p value<0 . 05 for both shuffles on at least one path . We considered non-replay enriched events those with a p value>0 . 2 for both shuffles for all four trajectories . For replay content assessments ( below ) , we used a more stringent criterion for replay detection by performing the shuffles for each putative replay event on the one trajectory with the strongest replay score . We obtained similar results using the standard criterion . Replay/total ( R/T ) SPW-R events for each session ( s ) are as follows: R/T rat 1 , s1 173/798; s2 141/434; s3 92/425; s4 71/344; s5 55/203; rat2 , s1 120/377; s2 48/210; s3 68/219; rat 3 , s1 78/241; s2 30/93; rat 4 , s1 65/328; s2 53/275; s3 49/234; s4 64/464; rat 5 , s1 -/627; rat 5 , s2 -/904 . Distributions of replayed spatial locations were derived by accumulating the spatial posterior probability distribution function across all 25 ms time bins of all replay events , in each recording , for SPW-Rs that occurred while the rat paused at forced reward locations . We determined the temporal delay between hippocampal SPW-R events and VTA activity on the basis of the delay in the SPW-R event-triggered VTA local field potential ( 84 ± 13 ms ) . The distribution of VTA spike-associated replay content was derived by accumulating the spatial probability distribution functions for the subset of 25 ms replay time bins that preceded VTA unit spikes by this fixed delay . The probability of reward site content for each recording was measured as the fraction of the distribution of replayed spatial locations that was associated with reward sites . The probability of VTA units coordinating with reward site content ( reward site probability ) was measured similarly , as the fraction of the distribution of VTA spike-associated replay content that was associated with reward locations . To compare RR and nonRR VTA spike-associated replay content with overall replay content , we computed a reward site bias as follows: Each replay time bin was assigned a 1 ( or 0 ) when the average representation of reward site regions exceeded ( or did not exceed ) the average representation of nonreward regions . The same binary metric was applied to each RR and nonRR VTA spike-associated replay time bin . From these measures , we computed across the entire dataset the proportion of VTA spike-associated replay time bins that better represented reward site regions compared to nonreward site regions , and we compared this with the proportion of replay time bins that better represented reward site regions , accounting for differences in the number of spike-associated time bins across RR units and nonRR units across recordings , to derive the excess reward site bias . There were 5422 replay time bins , 2269 RR unit spike-associated bins , and 2455 nonRR unit spike-associated bins . On the SWM task , there were 3554 replay time bins and 1046 RR unit spike-associated bins . On the linear track , there were 1868 replay time bins and 818 RR unit spike-associated bins . The reward site bias was highly correlated with the reward site probability ( RR units , r = 0 . 80 , p=1 . 36 × 10-4 ) . To explore the sensitivity of the reward site bias to the temporal delay between replayed spatial content and VTA unit activity , we systematically varied the delay in 25 ms steps . Consistent with the latency of the SPW-R-associated VTA potential , the excess reward site bias of RR units was maximal at a 75 ms VTA lag relative to hippocampal activity ( data not shown ) . To evaluate whether RR unit coordination with replayed reward site representations correlated with choice behavior , we measured the excess reward site bias at force reward site locations immediately after correct and error trials; and separately , immediately prior to correct and error trials . There were 616 RR unit spike-associated bins following correct trials , 243 following error trials , 565 prior to correct trials , and 142 prior to error trials . For forward and reverse replay content analyses , we measured the reward site bias of replay event time bins that showed strong directional preference for outbound ( O ) or inbound ( I ) track direction ( direction index >0 . 5 , where the direction index is [O-I]/[O+I] ) ( Singer et al . , 2013 ) . For each replay event , we transformed the direction index of each time bin into an index of forward or reverse content , as follows . Since we restricted our analysis to replay events occurring while the rat paused at force reward sites , for centrifugal sequence replay away from the animal’s location , outbound content reflects forward replay , while inbound content reflects reverse replay . In contrast , for remotely initiated , centripetal replay towards the animal’s location , inbound content reflects forward replay , while outbound content reflects reverse replay . For centrifugal and centripetal replay content analyses , we aggregated forward and reverse replay event time bins together , defining centrifugal replay events as replay trajectories moving away from the current position of the animal and centripetal replay events as replay trajectories approaching the animal . We compared the reward site bias of centrifugal and centripetal replay event time bins and of forward and reverse replay event time bins to the reward site bias of VTA unit spikes occurring with 84 ms delay to those replay time bins . There were 1700 time bins with centrifugal replay content , of which 681 were associated with RR unit spikes and 770 with nonRR unit spikes . There were 1248 time bins with centripetal replay content , of which 585 were associated with RR unit spikes and 602 with nonRR unit spikes . In addition , there were 1561 time bins with forward replay content , of which 699 were associated with RR unit spikes and 720 with nonRR unit spikes . There were 1313 time bins with reverse replay content , of which 567 were associated with RR unit spikes and 652 with nonRR unit spikes . There were several cases in which we sought to determine whether the excess reward site bias of VTA units compared to hippocampal replay was greater across two comparisons ( B vs A compared to C vs A ) : 1 ) excess reward site bias of RR units versus nonRR units; 2 ) excess reward site bias of RR units on the SWM task versus on the linear track; 3 ) excess reward site bias of RR units after correct trials versus after error trials; 4 ) excess reward site bias of RR units before correct trials versus before error trials; and 5 ) excess reward site bias of RR units at forward and reverse replay . For each case , we ran a logistic regression in which we considered each element of the case separately , as well as their interaction . For example , in the first case , we computed a logistic regression to measure the interaction between RR unit-associated reward site bias and replay-associated reward site bias , and nonRR unit-associated reward site bias and replay associated reward site bias . We then compared the interaction term to a distribution of simulated interaction terms assuming the null hypothesis . We considered the reward site bias for replay ( A ) , the reward site bias for RR units ( B ) , and the reward site bias for nonRR units ( C ) . The logistic regression predicted reward site bias as a binary dependent variable ( present/absent ) with binary predictors of ( either B or C = 1 vs A = 0 ) , comparison ( B vs A = 1 , C vs A = 0 ) , and their interaction , the latter being the predictor pertinent to the research question . Since these hypotheses were in only one direction , we ran 1-tail tests . In order to determine the one tail p value in these nonlinear tests , we performed a nonparametric permutation test of 1 , 000 iterations of the logistic regression , assuming the null hypothesis ( i . e . , the coefficient for the interaction is centered at zero ) . In each iteration , the fraction of reward site bias ( A ) for each comparison was taken as the overall average of actual estimates separately incremented with a perturbation from a normal distribution . The mean of this distribution was set to 0 and the standard deviation chosen so as to produce simulated interaction regression coefficients with a standard deviation approximately equal to the standard error for the same predictor estimated from the logistic regression of the actual data . The proportion of simulated interaction coefficients greater than or equal to the actual interaction coefficient was taken as the estimated one-tail p value . To test for bias in the reconstruction algorithm towards reward sites , we first decoded the estimated position of nonreplay events . We did not detect a bias for reward sites in the distribution of spatial locations across nonreplay events ( reward site bias 0 . 424 ± 0 . 004; probability/spatial bin of content at reward locations 0 . 032 ± 0 . 007 bin-1; non-reward locations 0 . 023 ± 0 . 003 bin-1; p=0 . 1 , signed-rank test , n = 14 recordings ) . Because SPW-R events may encode hippocampal spatial representations other than replay sequences , and because we may have miscategorized some replay events as nonreplay events , we also assessed the output of the reconstruction algorithm in the absence of hippocampal activity . This approach did not detect a preference in the decoder toward reward sites ( probability/spatial bin of content at reward locations 0 . 029 ± 0 . 006 bin-1; non-reward locations 0 . 023 ± 0 . 003 bin-1; p=0 . 6 , signed-rank test , n = 14 recordings ) . The Hilbert transform of the theta-filtered LFP was used to assess theta phase relationships of VTA units . To evaluate for theta phase preference of VTA units , we computed Rayleigh’s test for uniformity of the circular theta phase distribution of each VTA neuron’s spikes against a unimodal alternative; and we computed the parameters mu and kappa of the von Mises distribution that best fit that distribution ( Jones and Wilson , 2005; Siapas et al . , 2005 ) , using custom Matlab code . The circular concentration coefficient kappa is inversely related to the variance of the distribution , such that in the limit of kappa = 0 , the circular distribution is uniform . Frames were identified as described previously ( Ji and Wilson , 2007 ) , within epochs of SWS defined on the basis of low hippocampal theta/delta ratio and clear sleep posture ( SWS median duration , range: 1226 s , 793–1998 s ) . Within SWS epochs , multiunit activity from all tetrodes were combined , counted in 10 ms bins , and smoothed with a Gaussian window , with σ = 30 ms . Frames were identified as periods of high activity between silent periods , with the spike count threshold defined as the spike count at which the spike count distribution reached its first minimum ( in 10 ms bins ) . Clusterless reconstruction was applied to frames of SWS to derive the spatial probability distribution function of each 25 ms time bin within each frame . Spatial content per frame was taken as the average of the maximum decoded probability of each time bin .	Whether it’s meeting a friend for dinner at a new restaurant or hiking through a forest for the first time , new experiences lead to new memories . In order to create these memories , our brains do a lot of work behind the scenes . A region of the brain called the hippocampus ‘records’ experiences as they happen . It then replays these recordings later on , both while we are awake and once we have fallen asleep . During active behaviour , dopamine cells in a different part of the brain , called the ventral tegmental area ( or VTA ) , recognize unexpected rewards , known as “reward prediction errors” , to guide learning . For example , was the chocolate mousse or conversation unusually good , or was there an unexpected and spectacular waterfall ? In order to reap rewards for our future selves , we need to remember such unexpected rewards in the context in which they occurred . However , despite the importance of this ability , research has not yet established how the brain combines our experiences as we navigate the world around us with information about reward prediction errors in our memories . Gomperts et al . have now investigated this process by giving rats a memory task . In the first phase of the task , the rats were placed in an arena and directed to turn either left or right to receive food . In the second phase , the rats were given a choice as to whether to turn left or right . If they turned in the opposite direction to that required in the first phase , they received food rewards . The rats performed multiple trials of this task each day , and over a number of days the rats learnt the layout of the arena and the rule to locate the food . Gomperts et al . recorded the rats’ brain activity across the multiple trials of the task and after the task as the rats slept . As expected , the findings showed that the hippocampus replayed the rats’ movements through the arena , both between the trials and after the task during sleep . Between trials , the activity of VTA neurons that encoded reward-related information coordinated with the hippocampus’ memories of paths through the arena and indicated where in the arena the rats found food rewards . Unexpectedly however , coordination of these VTA neurons with the hippocampus was greatly reduced when the rats were asleep . This observation suggests that waking memories and sleep-associated memories may serve distinct purposes , for example , to aid learning and planning or to stabilize the memories . A key challenge now is to determine these distinct roles .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2015	VTA neurons coordinate with the hippocampal reactivation of spatial experience
Activating mutations involving the PI3K pathway occur frequently in human cancers . However , PI3K inhibitors primarily induce cell cycle arrest , leaving a significant reservoir of tumor cells that may acquire or exhibit resistance . We searched for genes that are required for the survival of PI3K mutant cancer cells in the presence of PI3K inhibition by conducting a genome scale shRNA-based apoptosis screen in a PIK3CA mutant human breast cancer cell . We identified 5 genes ( PIM2 , ZAK , TACC1 , ZFR , ZNF565 ) whose suppression induced cell death upon PI3K inhibition . We showed that small molecule inhibitors of the PIM2 and ZAK kinases synergize with PI3K inhibition . In addition , using a microscale implementable device to deliver either siRNAs or small molecule inhibitors in vivo , we showed that suppressing these 5 genes with PI3K inhibition induced tumor regression . These observations identify targets whose inhibition synergizes with PI3K inhibitors and nominate potential combination therapies involving PI3K inhibition . The phosphatidylinositol 3-kinase ( PI3K ) pathway is frequently activated in breast cancers due to ( i ) amplification of ERBB2 , an oncogene that stimulates the PI3K pathway and ( ii ) activating mutations of the PI3K catalytic subunit , PIK3CA ( Koboldt et al . , 2012 ) . In addition , other genetic aberrations can lead to the activation of the PI3K pathway including PTEN deletion or loss-of-function mutations , PIK3CA amplification and activating AKT mutations . Constitutive PI3K pathway activation promotes cell proliferation and survival , and previous reports have demonstrated that tumors harboring mutations that activate the PI3K pathway require constitutive signaling of this pathway for tumor maintenance . Specifically , tumors that harbor mutant PIK3CA alleles exhibit significant dependence on PIK3CA expression and activity ( Cheung et al . , 2011; Liu et al . , 2011; Samuels et al . , 2005 ) . In addition , oncogenic activation of PIK3CA leads to intrinsic resistance of HER2-positive breast cancer cells to HER2 inhibition ( Berns et al . , 2007; Hanker et al . , 2013 ) , and is more frequently activated in patients that exhibit acquired resistance to HER2 inhibition ( Chandarlapaty et al . , 2012 ) . The prevalence of PI3K pathway activation in breast cancer and its importance to cancer cell proliferation and tumor survival make targeting this pathway an attractive therapeutic approach . However , inhibition of the PI3K pathway often leads to proliferative arrest rather than cell death ( Elkabets et al . , 2013; Klempner et al . , 2013; Serra et al . , 2008 ) and to date has shown limited clinical benefit . Specifically , PI3K/AKT/mTOR inhibitor therapy induced a partial response in 18–30% of patients whose tumors harbor PIK3CA and/or PTEN mutations ( Janku et al . , 2014 , 2013 , 2012 ) . Although this rate of partial responses was significantly higher than that achieved following treatment with therapies other than PI3K/AKT/mTOR inhibitors , this response was not associated with an improvement in either progression-free or overall survival of treated patients . Combination therapy consisting of Trastuzumab and Buparlisib , a PI3K inhibitor , resulted in a 17% partial response ( Saura et al . , 2014 ) , and mTOR inhibition combined with aromatase inhibitors in patients with hormone-receptor positive advanced breast cancer showed extended progression-free survival ( Baselga et al . , 2012 ) . Together , these studies suggest that targeting the PI3K pathway alone is only partially effective clinically . We hypothesized that identifying targets whose inhibition in the context of PI3K inhibition leads to cell death would provide a foundation to develop combination therapies . Here using a genome-scale loss of function screen , we identified genes whose suppression induces cell death only in the presence of PI3K inhibition both in vitro and in vivo . To identify genes whose suppression converts the cytostatic response to PI3K inhibition into a cytotoxic response , we performed a positive-selection genome scale shRNA screen ( Figure 1A ) using MDA-MB-453 breast cancer cells , which harbor a PIK3CA H1047R mutation and ERBB2 amplification . Treatment with the PI3K inhibitor GDC0941 leads to a complete proliferation arrest ( Figure 1—figure supplement 1A ) and suppression of AKT activity ( Figure 1—figure supplement 1B ) with minimal basal- and PI3Ki-induced cell death ( Figure 1—figure supplement 1C–D ) . 10 . 7554/eLife . 24523 . 003Figure 1 . Genome scale shRNA screen identifies genes whose suppression facilitates PI3Ki-induced cell death . ( A ) A schematic representation of the pooled shRNA screen design . ( B ) Z-scores for fold-change of proliferation of MDA-MB-453-eGFP cells infected with multiple shRNAs targeting the indicated genes and treated for 9 days with GDC0941 ( 0 . 625 μM; red ) , or vehicle ( DMSO; blue ) . Cells infected with five different control shRNAs ( shCTRLs ) were used to calculate Z-scores . Bars indicate standard deviation between the different shRNAs targeting each gene . Data shown are representative of three independent experiments . ( C–D ) MDA-MB-453 cells were infected with the indicated shRNAs , and then treated for 4 days with GDC0941 ( 0 . 625 μM ) ( C ) or left untreated ( D ) . Adherent and floating cells were collected and subjected to immunoblot analysis for induction of PARP cleavage . Cells infected with a shRNA targeting ZNF565 and treated with GDC0941 ( 0 . 625 μM ) for 4 days were used as positive control for PARP cleavage ( D ) . Data shown are representative of two independent experiments . ( E ) MDA-MB-453 cells were infected as in B and treated for 4 days with GDC0941 ( 0 . 625 μM ) . Adherent and floating cells were collected and analyzed for DNA content by flow cytometry . Bars indicate standard deviation between the different shRNAs targeting each gene . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 00310 . 7554/eLife . 24523 . 004Figure 1—figure supplement 1 . Supporting information for shRNA screen setup and scoring . ( A ) MDA-MB-453 cells were counted at the indicated time points after initiation of treatment with either DMSO or GDC0941 at the indicated concentrations . Experiments were performed in triplicates , with error bars representing standard deviation . Data shown are representative of two independent experiments . ( B ) MDA-MB-453 cells were seeded and treated as in A . Seventy-two hours after treatment initiation , cells were lysed and Western blotting analysis was performed using the indicated antibodies . Staurosporine ( STS , 1 μM , 6 hr ) was used as a positive control for cleaved-Caspase3 antibody . ( C–D ) MDA-MB-453 cells were seeded and treated as in A . Adherent and floating cells were collected and fixed at the indicated time-points . Cells were then stained with cleaved-cPARP antibody and analyzed by flow cytometry . At least 30 , 000 cells were acquired per sample . ( E ) Scoring of individual shRNAs according to a T-statistic . Each dot represents an individual shRNA T-statistic value . Red dots represent shRNAs that contributed to the nomination of genes by STARS . ( F ) Venn diagram summarizes the gene calling by either RIGER or STARS algorithms . ( G ) qPCR was used to measure the expression of the targeted genes in MDA-MB-453 cells infected with the indicated shRNAs relative to their mean expression in cells infected with five different control shRNA . Data shown are representative of two independent experiments . ( H ) MDA-MB-453 cells were infected with the indicated shRNAs . Five days post-infection cells were collected and subjected to Western blot analysis of AKT phosphorylation . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 004 We introduced a pooled lentivirally-delivered shRNA library in quadruplicate into MDA-MB-453 breast cancer cells . Cells that incorporated shRNAs were selected by resistance to puromycin and propagated in culture to deplete cells containing shRNAs targeting essential genes . Seven days post-infection , cells were treated with either the pan-PI3K inhibitor GDC0941 ( 0 . 625 μM ) or with DMSO ( vehicle control ) . We then isolated apoptotic cells using fluorescence activated cell sorting ( FACS ) with an antibody specific for cleaved PARP and a fluorescence-conjugated secondary antibody . We isolated genomic DNA and identified shRNAs present in this apoptotic population by massively parallel sequencing . To rank the shRNAs by their relative abundance in the PI3Ki-treated samples compared to DMSO-treated samples , we used a T-score to account for both the magnitude of the difference in means and the consistency of replicates . We then used the STARS ( Doench et al . , 2016 ) and RIGER ( Luo et al . , 2008 ) algorithms , which require that at least two shRNAs targeting a gene are ranked significantly higher than random , to rank genes targeted by the top scoring shRNAs . Based on these two algorithms , 57 genes met this significance threshold ( Figure 1—figure supplement 1E ) , of which 36 genes were called by both algorithms ( Figure 1—figure supplement 1F and Supplementary file 1 ) . To confirm the 57 candidates , we used a GFP-based cell proliferation assay in which shRNAs were introduced into GFP-expressing cells , and fluorescence was used as a surrogate for cell number both prior to and after 9 days of treatment . The fluorescent measurements were then used to calculate the fold-change of proliferation ( FC-proliferation ) , where a value of 1 corresponds to proliferative arrest , and cell death results in FC-proliferation <<1 . Cells infected with control shRNAs and treated with GDC0941 exhibited FC-proliferation≈1 and were used to calculate Z-scores for the FC-proliferation of experimental shRNAs , with a Z-score <<0 corresponding to cell death . We first retested the shRNAs that scored in the screen and then performed a second level validation by testing additional shRNAs targeting these genes for their ability to facilitate cell death upon PI3K inhibition ( Supplementary file 2 ) . In parallel , we also measured the knockdown efficiency of all shRNAs by qPCR . We used two criteria to select genes for further study: ( i ) their expression was suppressed by three or more targeting shRNAs ( Figure 1—figure supplement 1G ) and , ( ii ) their suppression by at least three targeting shRNAs modified the response to PI3K inhibition ( Figure 1B ) . Using these criteria , we confirmed five candidate genes: PIM2 , TACC1 , ZAK , ZFR , and ZNF565 . In addition , to determine whether suppression of these candidates affected PI3K signaling , we assessed the consequences of suppressing these genes on AKT phosphorylation ( Figure 1—figure supplement 1H ) and found that suppressing none of these candidates reduced phosphorylation at Ser473 of AKT , indicating that the observed interaction with PI3K inhibition was not due to further inhibition of the pathway . To further validate the interaction of the five identified genes with PI3K inhibition , we measured two features characteristic of apoptotic cells . First , we measured PARP cleavage as an indicator of the induction of apoptosis ( Figure 1C ) . We found that cells infected with control shRNAs ( EmptyT or Luciferase ) and treated with GDC0941 did not exhibit cleaved PARP . In contrast , every shRNAs targeting the identified genes induced PARP cleavage following treatment with GDC0941 , indicating the induction of apoptosis . Notably , although suppression of each of these genes resulted in reduction in the proliferation rate of untreated cells by 2–4 fold as compared to control shRNAs , we did not observe evidence of apoptosis as measured by cleaved PARP ( Figure 1D ) , indicating that the screen identified genes that only affect cell survival in the presence of the PI3K inhibitor . In addition , we used flow cytometry to examine the cell cycle profile to assess the proportion of cells with a sub-G1 DNA content , as a second measure of apoptosis ( Figure 1E ) . In consonance with the cleaved PARP observations , suppression of each of the five identified genes resulted in an increased percentage of cells with a sub-G1 DNA content in the presence of GDC0941 compared to control cells . Taken together , we identified and validated five genes that convert the cytostatic PI3K inhibitory response into apoptotic cell death . As indicated above , the observed effects were obtained using at least three different shRNAs containing different seed sequences , which ensured that these effects were the consequence of on-target activity of the shRNAs , rather than off-target suppression due to miRNA effects mediated by seed sequences ( Jackson et al . , 2006 ) . In addition to using multiple shRNAs per target genes , we used two additional strategies to validate on-target effects . First , we expressed the corresponding ORF to rescue the effect of each shRNAs . To prevent the suppression of exogenously expressed ORFs by shRNAs , we introduced at least three synonymous mutations into the shRNA-targeted sequence ( Figure 2—figure supplement 1A ) . MDA-MB-453 cells were infected with lentivirally-delivered V5-tagged ORFs , followed by infection with lentivirally-delivered shRNAs . Cells were then treated with DMSO or GDC0941 for 9 days . We used In-Cell-Western to detect expression of the V5 tag as a measure of the ORF-infected cell population . Over-expression of PIM2 , ZNF565 , and TACC1 rescued the effect of endogenous gene suppression by their corresponding shRNAs ( Figure 2A ) . We also found that over-expression of the kinase-inactive mutant of PIM2 ( K61A ) failed to rescue the suppression of endogenous PIM2 , suggesting that the kinase activity of PIM2 is essential for preventing cell death following PI3K inhibition . Overexpression of ZAK reduced cell viability in a kinase-dependent manner , but independent of shRNA suppression or PI3K inhibition ( Figure 2—figure supplement 1B ) , suggesting that ZAK overexpression itself reduces cell fitness . We found that overexpression of ZFR failed to rescue ZFR suppression ( Figure 2—figure supplement 1C ) . To investigate potential interactions between the five candidate genes , we utilized the rescue assay described here to test whether any of the genes can be rescued by another gene . These experiments revealed that PIM2 and TACC1 over-expression partially interfered in the interaction between PI3K inhibition and suppression of the other genes ( Figure 2—figure supplement 1D ) , suggesting that PIM2 or TACC1 generally contribute to cell survival in the presence of PI3K inhibition . Taken together , ORF expression-mediated rescue further validated three of the five identified genes . 10 . 7554/eLife . 24523 . 005Figure 2 . Validation of genes identified in the screen . ( A ) MDA-MB-453 cells were infected with the indicated ORFs and shRNAs . Cells were then treated with GDC0941 ( 0 . 625 μM ) for 9 days , followed by detection of V5-expressing cells by ICW . Cells infected with five different control shRNAs were used to calculate Z-scores . Data shown are representative of three independent experiments . ( B–G ) MDA-MB-453 cells were treated for 9 days with the indicated inhibitors ( 0 . 01–20 μM ) combined with GDC0941 ( 0 . 625 μM; red ) , or vehicle ( DMSO; blue ) . Plotted are mean and standard deviation of fold-change proliferation from quadruplicates . The expected fold-change of proliferation for treatment combination ( black ) was calculated according to the Bliss independence model using single treatment effects . The gray shading indicates compound concentrations for which significant synergy with GDC0941 was observed . Data shown are representative of two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 00510 . 7554/eLife . 24523 . 006Figure 2—figure supplement 1 . Supporting information for ORF rescue validation . ( A ) MDA-MB-453 cells were infected with the indicated ORFs and shRNAs . V5 expression was detected by ICW 5 days after shRNA infection . The average V5 stain value of 5 different control shRNAs ( shCTRL ) was used to normalize the stain of the targeting shRNAs . Data shown are representative of two independent experiments . ( B ) MDA-MB-453 cells were infected with the indicated ORF vectors . 5 days after infection , cell viability was determined by using DRAQ5 DNA stain . Bars indicate standard deviation from 4 replicates . ( C ) MDA-MB-453 cells were infected with the indicated ORF and shRNAs . Cells were then treated with GDC0941 ( 0 . 625 μM ) for 9 days , followed by detection of V5-expressing cells by ICW . Cells infected with five different control shRNAs were used to calculate Z-scores . Data shown are representative of two independent experiments . ( D ) MDA-MB-453 cells were infected with the indicated ORFs ( columns ) and shRNAs ( rows ) . Cells were then treated with GDC0941 ( 0 . 625 μM ) for 9 days , followed by detection of V5-expressing cells by ICW . Cells infected with five different control shRNAs along with each of the ORFs were used to calculate Z-scores . Presented are Z-scores color-coded as indicated by the colorbar . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 006 A second strategy to confirm on-target effects of shRNAs is to functionally mimic target suppression by using small molecule inhibitors . PIM2 and ZAK are serine/threonine kinases , therefore are potential targets for small molecule inhibitors . Searching for compounds that exhibit a selective and potent binding to either PIM2 or ZAK , we identified two compounds that target PIM2 and three compounds targeting ZAK ( Supplementary file 3 ) . Specifically , PIM2 scored high in a kinase selectivity assay for two compounds , HTH-01–091 , and HTH-02–096 , which also exhibited high potency for PIM2 inhibition in a biochemical assay ( Supplementary file 3 ) . Of note , the potency of both of these compounds as PIM2 inhibitors is similar to their potency toward PIM1 , in contrast to commercially available PIM inhibitors that have a strong preference for PIM1 inhibition . Similarly , ZAK scored high in a focused kinase selectivity assay for one compound ( TL-01–124 ) . In a biochemical assay to determine IC50 , three compounds exhibited high potency as ZAK inhibitors ( TL-01–124 , HYJ-2-129-6 and HYJ-2-34-2 ) . We used these small molecule inhibitors to further test the capacity of PIM2 or ZAK inhibition to modify the response to PI3K inhibition . We treated MDA-MB-453 cells with either DMSO or GDC0941 ( 0 . 625 μM ) in combination with each of the several compounds targeting PIM2 and ZAK , and GFP fluorescence was used to determine changes in proliferation . We then used the Bliss independence model to determine synergy based on the expected versus the observed effect for each combination . When significant synergy occurs , as indicated by gray shading ( Figure 2B–G; Bliss index <0 . 6 ) , the observed effect ( red ) is greater than the expected effect ( black ) . At higher concentrations , the calculated expected combined effect is limited by population death , which makes it impossible to determine whether synergy exists at these points . Combining inhibitors targeting ZAK ( HYJ-2-129-6 , HYJ-2-34-2 , and TL-01–124 ) and PI3K resulted in a synergistic response and conversion of the GDC0941-induced proliferation arrest into cell death ( Figure 2B–D ) . We also found that TL-02–082 , an analog of TL-01–124 that does not bind ZAK , but otherwise has a similar kinase selectivity profile , did not exhibit a synergistic interaction with GDC0941 ( Figure 2E ) . Similarly , co-inhibition of PIM2 ( HTH-01–091 , and HTH-02–096 ) together with PI3K also resulted in a synergistic response and conversion into cell death ( Figure 2F–G ) . These observations further validate that PIM2 and ZAK inhibition synergizes with PI3K inhibition to induce cell death . To confirm the relevance of the candidates in other cell lines , we suppressed these same five genes in three additional breast cancer cell lines , including HCC1954 , T47D , and HCC1937 , which growth arrest upon PI3K inhibition . Each of these cell lines show activation of the PI3K pathway through different mechanisms: ERBB2 amplification ( HCC1954 ) , PIK3CA amplification ( HCC1937 ) or activating mutation ( T47D , HCC1954 ) , and PTEN loss ( HCC1937 ) . We introduced shRNAs targeting the five genes into these cell lines and monitored their proliferation ( Figure 3A ) . Similar to the effects observed in MDA-MB-453 cells , we found that suppression of these five genes in each of these cell lines lead to apoptosis in the setting of PI3K inhibition . In contrast to the proliferation inhibitory effect of PI3K inhibition on breast cancer cell lines with activated PI3K pathway , inhibition of PI3K in breast cancer cell lines that have a wild-type PI3K pathway ( CAL-120 and MDA-MB-231 ) failed to affect cell proliferation . In these cell lines , suppression of the five genes had no effect on cell proliferation or apoptosis when combined with PI3K inhibition . We also found that although PI3K inhibition in cells with a wild-type PI3K pathway activated by growth factors can suppress proliferation , suppression of these five genes did not modify this response ( Figure 3A; MCF10A ) . In addition to breast cancer cell lines , we also tested four additional non-breast cancer cell lines . The ovarian cancer cell line EFO-21 ( PIK3CA amplified ) exhibited similar response to that observed in breast cancer cell lines with oncogenic PI3K activation ( Figure 3—figure supplement 1A ) . However , the interactions we observed between PI3K inhibition and suppression of the five candidate genes in breast cancer cell lines did not occur in the three tested GBM cell lines ( all that harbor PTEN inactivation ) ( Figure 3—figure supplement 1B ) . Taken together , these results suggest that the response to PI3K inhibition can be modified by the suppression of the five identified genes only in the context of oncogenic PI3K pathway activation . 10 . 7554/eLife . 24523 . 007Figure 3 . Effects of manipulating candidate gene expression in multiple cell lines . ( A ) T47D-eGFP , HCC1954-eGFP , HCC1937-eGFP , CAL-120-eGPF , MDA-MB-231-eGPF , and MCF10A-eGPF were infected with shRNAs targeting the indicated genes . Cells were then treated with GDC0941 ( 0 . 5 , 1 . 25 , 1 , 0 . 312 , 0 . 312 , and 0 . 156 μM , respectively ) or vehicle ( DMSO ) for 9 days . Z-scores for the fold-change of proliferation were calculated based on control shRNAs . Bars indicate standard deviation between the different shRNAs targeting each gene . Data shown are representative of at least two independent experiments . ( B ) T47D-eGFP , HCC1954-eGFP , HCC1937-eGFP , and SKBR3-eGFP were treated for 9 days with the indicated inhibitors ( 0 . 01–10 μM ) combined with GDC0941 ( 0 . 5 , 1 . 25 , 1 , and 0 . 625 μM , respectively; red ) , or vehicle ( DMSO; blue ) . Plotted are mean and standard deviation of fold-change of proliferation from 4 replicates . Expected effect of combination treatment ( gray ) was calculated as in Figure 2 . Significant synergistic combinations according to Bliss independence model are indicated by asterisks . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 00710 . 7554/eLife . 24523 . 008Figure 3—figure supplement 1 . Effects of manipulating candidate genes in non-breast cancer cell lines . ( A-B ) EFO-21-eGPF , LN443-eGFP , LN382-eGFP and SF295-eGFP were infected with shRNAs targeting the indicated genes . Cells were then treated with GDC0941 ( 0 . 625 μM ) or vehicle ( DMSO ) for 9 days . Z-scores for the fold-change of proliferation were calculated based on control shRNAs . Bars indicate standard deviation between the different shRNAs targeting each gene . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 008 In addition , we found that the combination of ZAK or PIM2 inhibition and PI3K inhibition also induced cell death at a synergistic manner in four additional cell lines: T47D , HCC1954 , HCC1937 , and SKBR3 ( bearing PTEN deletion and ERBB2 amplification ) ( Figure 3B ) . These observations confirm that suppression of these 5 genes converts PI3Ki-mediated proliferative arrest into a cytotoxic response in multiple cell lines with oncogenic PI3K pathway activation . To explore the pathway specificity of these observed interactions , we tested whether the identified genes also modify the response to up- or downstream inhibition of the PI3K pathway . Following the transduction of the shRNAs , MDA-MB-453 cells were treated with either a HER2 inhibitor ( BIBW2992 , 1 μM ) or an AKT inhibitor ( MK2206 , 0 . 5 μM ) at doses that induce significant attenuation of cell proliferation . Suppression of all five genes modified the response to HER2 or AKT inhibition , similar to our observation with PIK3CA inhibition ( Figure 4A ) . Similarly , suppression of the five identified genes combined with inhibition of PDK1 ( GSK2334470 , 2 . 5 μM ) or mTOR ( Sirolimus , 10 nM ) modified the response to PDK1 or mTOR inhibitors ( Figure 4—figure supplement 1A–B ) . These observations demonstrate that the observed synergy between these genes and PI3K pathway inhibition is not specific to the particular PI3K inhibitor ( GDC0941 ) used in the screen . To further explore the specificity of PI3K inhibition , we tested isoform specific inhibitors of PI3K , to inhibit either the alpha ( BYL-719 ) or the beta ( GSK2636771 ) isoforms . In addition , we also tested the beta-sparing PI3K inhibitor GDC0032 . As shown in Figure 4B , the only inhibitor that failed to interact with the suppression of the five identified genes is the beta isoform inhibitor , thus suggesting that the alpha isoform , PIK3CA , is the major isoform that mediates the observed interactions . 10 . 7554/eLife . 24523 . 009Figure 4 . PI3K pathway specificity of identified interactions . ( A ) Z-scores for fold-change of proliferation of MDA-MB-453-eGFP cells infected with shRNAs targeting the indicated genes and treated for 9 days with GDC0941 ( 0 . 625 μM; red ) , MK2206 ( 0 . 5 μM; green ) , BIBW2992 ( 1 μM; yellow ) or vehicle ( DMSO; blue ) . Cells infected with five different control shRNAs were used to calculate Z-scores . Bars indicate standard deviation between the different shRNAs targeting each gene . Data shown are representative of two independent experiments . ( B ) Z-scores for fold-change of proliferation of MDA-MB-453-eGFP cells infected with shRNAs targeting the indicated genes and treated for 9 days with GDC0941 ( 0 . 625 μM; red ) , BYL-719 ( 1 μM; green ) , GSK2636771 ( 2 . 5 μM; brown ) , GDC0032 ( 50 nM; purple ) or vehicle ( DMSO; blue ) . Cells infected with five different control shRNAs were used to calculate Z-scores . Bars indicate standard deviation between the different shRNAs targeting each gene . Data shown are representative of three independent experiments . ( C ) Z-scores for fold-change of proliferation of T47D-eGFP cells infected with shRNAs targeting the indicated genes and treated for 9 days with GDC0941 ( 0 . 5 μM; red ) , Palbociclib ( 0 . 3 μM; green ) , Fulvestrant ( 0 . 1 μM; yellow ) or vehicle ( DMSO; blue ) . Cells infected with five different control shRNAs were used to calculate Z-scores . Bars indicate standard deviation between the different shRNAs targeting each gene . Data shown are representative of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 00910 . 7554/eLife . 24523 . 010Figure 4—figure supplement 1 . Validation of pathway specificity with additional PI3K-pathway inhibitors . ( A–B ) Z-scores for fold-change of proliferation of MDA-MB-453-eGFP cells infected with shRNAs targeting the indicated genes and treated for 9 days with Sirolimus ( 10 nM; red ) ( A ) , GSK2334470 ( 2 . 5 μM; red ) ( B ) , or vehicle ( DMSO; blue ) . Cells infected with five different control shRNAs were used to calculate Z-scores . Bars indicate standard deviation between the different shRNAs targeting each gene . Data shown are representative of two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 010 Since PI3K inhibition attenuates proliferation in these cells , it is possible that the observed modified response to PI3K inhibitors by these five genes is the result of suppressing these genes in a state of impaired proliferation rather than being specifically related to PI3K inhibition . To determine whether this was the case , we combined shRNA-mediated suppression of these candidate genes with two PI3K-independent anti-proliferative drugs , Palbociclib , a CDK4/6 inhibitor , and Fulvestrant , an ER antagonist . In contrast to what we observed following PI3K inhibition , suppression of the five genes did not convert the cytostatic effect of either CDK4/6 inhibitor or ER antagonist into cytotoxicity ( Figure 4C ) . These observations indicate that the genes that we identified in this screen are specific modifiers of PI3K pathway inhibition and do not non-specifically affect cells whose proliferation is attenuated by other mechanisms . BAD , a BH3-only protein whose phosphorylation leads to its sequestration by 14-3-3 proteins , is a substrate of PIM2 ( Danial , 2008; Yan et al . , 2003 ) . We hypothesized that BAD phosphorylation mediates the effect of PIM2 on PI3K inhibition . To test this hypothesis , we first confirmed that BAD phosphorylation was reduced following PIM2 suppression with either shRNAs or a small molecule PIM2 inhibitor ( Figure 5A–B ) . To test the consequences of reducing BAD phosphorylation by suppressing PIM2 , we applied the BH3 profiling assay to assess whether PIM2 suppression primed mitochondria for apoptosis ( Certo et al . , 2006 ) . Mitochondrial priming level is determined by measuring the ratio between pro- and anti-apoptotic Bcl-2 proteins . BH3 profiling measures mitochondrial priming by determining whether synthetic peptides modeled after the BH3 domains of Bcl-2 proteins induce loss of cytochrome c from mitochondria , which indicates the likelihood of cells to undergo apoptosis and correlates with patient response to chemotherapy ( Ni Chonghaile et al . , 2011 ) . Suppression of PIM2 significantly increased overall mitochondrial priming ( Figure 5C ) , consistent with the observation that PIM2 suppression reduces the phosphorylation of BAD . In contrast , suppression of any of the other four genes did not lead to an increase in mitochondrial priming level ( Figure 5—figure supplement 1 ) , suggesting that the mechanism of interaction between PIM2 and PI3K inhibition is unique among the five identified genes . Supporting this mechanism of interaction , suppression of PIM2 resulted in caspase 9 cleavage ( Figure 5D ) when combined with PI3K inhibition , indicating the involvement of mitochondria depolarization in the observed cell death . 10 . 7554/eLife . 24523 . 011Figure 5 . PIM2 suppression increases mitochondrial priming . ( A ) MDA-MB-453 cells were infected with the indicated shRNAs . Five days after infection , cells were collected and phosphorylation of Bad was detected by Western blotting . ( B ) MDA-MB-453 cells were treated with PIM2 inhibitor HTH-02–096 for 24 hr , and Bad phosphorylation was detected as in A . ( C ) MDA-MB-453 cells were infected with the indicated shRNAs . Five days after infection , cells were collected and subjected to BH3 profiling . The percentage of cells exhibiting mitochondrial depolarization at 0 . 4 μM BIM peptide is plotted . Bars indicate standard deviation of 3 replicates . Data shown are representative of three independent experiments . ( D ) MDA-MB-453 cells were infected with the indicated shRNAs , and then treated for 30 hr with GDC0941 ( 0 . 625 μM ) or left untreated . Adherent and floating cells were collected and subjected to immunoblot analysis for caspase 9 . Triangle marks the product of cleaved caspase 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 01110 . 7554/eLife . 24523 . 012Figure 5—figure supplement 1 . MDA-MB-453 cells were infected with the indicated shRNAs . Five days after infection , cells were collected and subjected to BH3 profiling . Percent cells exhibiting mitochondrial depolarization at 0 . 8 μM BIM peptide is plotted . Bars indicate standard deviation from 3 replicates . Data shown are representative of two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 012 To determine whether the synergy between suppression or inhibition of the identified genes and PI3K inhibition was relevant for tumor maintenance in vivo , we used an implantable microscale device for simultaneous intratumoral delivery of several siRNAs or small molecules directly into xenografts of MDA-MB-453 cells ( Jonas et al . , 2015 ) . The major advantage of this device is that it allows more precise delivery of siRNA or small molecules to the tumor , allows for better quantification of responses and facilitates the testing of combinations of treatments . Each of 16 reservoirs on the device was loaded with ~1 μg of a unique siRNA or a small molecule . The device allows compounds to be released into confined regions within the tumor over 48 hr for siRNA and for 24 hr for small molecules , respectively . The device was then removed along with a cylindrical column of surrounding tumor tissue , which was stained by IHC and analyzed to determine cellular outcome . To facilitate efficient delivery , siRNAs were Accell-modified ( Dharmacon , GE Healthcare Life Sciences ) and labeled at the 5'-end with a fluorescent Cy3 dye to allow for detection of siRNA delivery into tissue ( Figure 6A ) . 10 . 7554/eLife . 24523 . 013Figure 6 . In vivo validation of the five identified genes . ( A ) A representative image showing siRNA ( red ) spreading in tumor tissue . ( B ) Devices loaded with the indicated siRNAs were implanted into MDA-MB-453 xenografts of un-dosed mice ( blue ) or mice that were pre-dosed with GDC0941 ( 75 mg/kg ) for 5 days ( red ) . Two days later , the devices were removed , xenograft tissue surrounding each reservoir was stained for Hematoxylin and Eosin , and the density of intact nuclei was quantified . Plotted are the proportional effects of the indicated siRNAs with or without PI3K inhibition as compared to the effect of siGFP in non-treated mice . Bars indicate SD from at least 4 replicates . The expected proportional effect for treatment combination ( grey ) was calculated according to the Bliss independence model using single treatment effects . ( C ) Devices were loaded with the indicated small molecules and were implanted into MDA-MB-453 xenografts of un-dosed mice ( blue ) or mice that were pre-dosed with GDC0941 ( 75 mg/kg ) for 5 days ( red ) . 24 hr later , the devices were removed , xenograft tissue surrounding each reservoir was stained for hematoxylin and eosin and for cleaved-CASP3 . Percentage of cells stained positive for cleaved caspase 3 was quantified . Mean and SD of at least 6 replicates is shown , asterisks indicate T-test p-value<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 01310 . 7554/eLife . 24523 . 014Figure 6—figure supplement 1 . Supporting information for in vivo validation . ( A ) Representative images of tumor tissue surrounding the microscale device loaded with the indicated siRNAs and implemented into xenografts of GDC0941-pre-dosed mice . Tissue stained with hematoxylin and eosin . ( B–C ) Devices loaded with the indicated siRNAs were implanted into MDA-MB-453 xenografts of un-dosed mice ( blue ) or mice that were pre-dosed with GDC0941 ( 75 mg/kg ) for 5 days ( red ) . Two days later , the devices were removed , xenograft tissue surrounding each reservoir was stained for Hematoxylin and Eosin . Density quantification of intact nuclei ( B ) and % reduction in intact nuclei upon PI3K inhibition ( C ) are presented . Shown is the mean and SD of at least 4 replicates , asterisks indicate T-test p-value<0 . 05 . ( D ) Representative images of tumor tissue surrounding the microscale device loaded with the indicated small molecules , and implanted into xenografts of GDC0941-pre-dosed mice . Tissue was stained with anti-cleaved-CASP3 antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 24523 . 014 To assess the effect of combining suppression of our five genes with PI3K inhibition , mice harboring xenograft tumors of MDA-MB-453 cells were systemically dosed with GDC0941 daily for 5 days prior to device implantation and during the period until the device was removed . Tissue surrounding the device was stained with hematoxylin and eosin and the density of intact nuclei around each reservoir was quantified as a measure of cell viability . Combining PI3K inhibition with control siGFP did not affect cell density compared to the effect of siGFP on tumor cells from non-treated mice . However , combining PI3K inhibition with any of the siRNAs targeting the five identified genes lead to a significant reduction in the density of intact nuclei as compared to non-treated control mice ( Figure 6—figure supplement 1B–C ) . For all siRNAs targeting any of the five candidate genes , this reduction in density of intact nuclei upon combination treatment was statistically significant ( p<0 . 05 ) when compared to either single treatments . We calculated the proportional effect from non-treated control siRNA ( siGFP ) for single treatments ( each of the siRNAs or PI3K inhibition alone ) and combined treatments , and compared it to the expected effect according to the Bliss independence model . All siRNAs targeting the five identified genes combined with PI3K inhibition exhibited greater effect then expected with a Bliss value significantly smaller than 1 ( Figure 6B ) . Thus , siRNA suppression of each of the five identified genes combined with PI3K inhibition lead to a significant increase in cell death ( Figure 6B and Figure 6—figure supplement 1A ) . Similarly , we also tested the combined effect of small molecule ZAK inhibitors and PI3K inhibition , and found that the combination resulted in a significant increase in the percentage of apoptotic cells as evidenced by staining for cleaved-CASP3 ( Figure 6C and Figure 6—figure supplement 1B ) . Notably , loading doxorubicin into a reservoir of the device lead to cleaved-CASP3 staining in ~20% of the exposed cells in tissue , which was not further affected by PI3K inhibition . This observation indicated that the observed increase in cleaved-CAPS3 with ZAK and PI3K inhibitors reflects a specific synergistic interaction . Taken together , these studies confirm that suppression or inhibition of the genes identified herein show synergistic effects with PI3K inhibition on tumor survival in vivo . Here we report the identification of genes that when suppressed in the setting of a PI3K inhibitor induce cell death . Specifically , we identified five genes , PIM2 , TACC1 , ZAK , ZFR , and ZNF565 whose suppression converts the cell cycle arrest induced by PI3K inhibition into apoptosis ( Figure 1 ) . Of these five genes , two are kinases for which we identified small molecule compounds that recapitulated the effects observed by shRNAs ( Figure 2 ) . In addition to using multiple distinct shRNAs targeting each of the five genes , we independently confirmed the initial screen using either small molecule inhibitors ( for PIM2 and ZAK ) or a rescue assay ( PIM2 , TACC1 , and ZNF565 ) ( Figure 2 ) . Although we were unable to independently confirm ZFR by either CRISPR-CAS9-mediated gene deletion or by a rescue assay , we included ZFR suppression in the experiments herein to provide information on this candidate for future studies . Our observations suggest that co-targeting PI3K and either PIM2 or ZAK is a synergistic combination with therapeutic potential . Furthermore , using a microscale implementable device , we found that inhibition or suppression of these genes also synergizes with PI3K inhibition in vivo ( Figure 6 ) . Together these studies identify potential targets whose perturbation augments the effects of PI3K inhibition . Although the primary screen described here was conducted in a single cell line , we verified these interactions in additional breast cancer cell lines in which the PI3K pathway is activated by different oncogenic aberrations ( Figure 3 ) . We also showed that PIK3CA is the major isoform that mediates this interaction ( Figure 4A–B ) . Importantly , these genes did not modulate the effect of drugs unrelated to the PI3K pathway , such as a CDK4/6 inhibitor and an ER antagonist ( Figure 4C ) . We also note that the five genes failed to interact with PI3K inhibition in cells that lack oncogenic activation of the pathway ( Figure 3 ) . These observations serve to further define the context in which these targets could be therapeutically combined . Although we identified and validated several candidates , we recognize that this screen was not saturating due to technical shortcomings of the shRNA library and inefficiencies of flow cytometry based assay that we employed . Thus , although we identified genes whose suppression in the presence of a PI3K inhibitor induced apoptotic cell death , extending this approach is likely to identify additional candidates that when suppressed or inhibited synergize with PI3K inhibition . A concern in drug-modifier screens is the potential for false positive results due to a combined effect on cell fitness rather than a specific functional interaction . In such cases , each perturbation independently reduces the cellular fitness and when combined , the effect on cellular fitness leads to cell death . One would expect that alternate combinations with other growth-arresting perturbations would have similar cytotoxic consequences . Here we demonstrated that the five identified genes from our screen caused cell death specifically with PI3K pathway inhibition and not with other cytostatic perturbations ( Figure 4 ) , suggesting that these interactions are specific for this genetic and pharmacologic context . We previously showed that over-expression of PIM1 can similarly rescue cells from lapatinib-induced apoptosis ( Moody et al . , 2015 ) , corroborating a role for PIM2 in promoting resistance to PI3K inhibition . In addition , it has been shown that PIM and AKT inhibitors synergize in acute myeloid leukemia ( Meja et al . , 2014 ) . PIM2 promotes tumor survival mainly through phosphorylation of the pro-apoptotic BH3-only protein BAD , as well as through regulation of MYC activity , and Cap-dependent protein translation ( Nawijn et al . , 2011 ) . Because BAD phosphorylation is a major pro-survival mechanism of PIM2 , we tested and confirmed that PIM2 suppression leads to mitochondrial priming ( Figure 5 ) . However , it is likely that additional substrates of PIM2 contribute to the interaction with PI3K inhibition . Of note , AKT and the PIM family of proteins share several common substrates , including BAD , while also having a number of unique targets ( Amaravadi and Thompson , 2005; Nawijn et al . , 2011 ) . Additional kinases such as PKA and p90RSK , also regulate BAD by phosphorylation ( Danial , 2008 ) . Accordingly , BAD phosphorylation was shown previously to integrate survival signals from the PI3K and MAPK pathways ( She et al . , 2005 ) . One of the other candidates identified in this screen , TACC1 , has been shown to enhance mammary tumor formation induced by heterozygous PTEN loss in vivo and was shown to mediate cell survival following PI3K pathway inhibition induced by PTEN over expression ( Cully et al . , 2005 ) . TACC1 interacts with Aurora A-chTOG complex at the spindle pole during metaphase ( Conte et al . , 2003 ) . At later stages of M phase , during anaphase and cytokinesis , TACC1 localizes to the midzone spindle and the midbody together with Aurora B ( Delaval et al . , 2004 ) . Suppression of TACC1 leads to formation of multipolar spindles and thus perturbs cell division ( Conte et al . , 2003 ) . Notably , PI3K activity contributes to genomic stability by regulating spindle orientation and chromosomal segregation ( Silió et al . , 2012 ) . These converging mechanisms suggest that combined suppression of TACC1 and PI3K may induce cell death by causing un-resolved spindle and chromosomal segregation defects . ZAK is a MAPK kinase kinase ( MAP3K ) that transduces signals through JNK and p38 , and is associated with activation of AP1 and NF-κB ( Liu et al . , 2014 ) . ZAK is up-regulated in several cancers including breast cancer ( Liu et al . , 2014 ) . ZAK over-expression has been shown to promote in vivo transformation ( Cho et al . , 2004 ) and cell migration ( Rey et al . , 2016 ) , and its suppression reduced proliferation ( Liu et al . , 2014 ) and β-catenin activity ( Firestein et al . , 2008 ) . Growth-factor stimulation , specifically EGF stimulation , activates ZAK ( Cho et al . , 2004; Rey et al . , 2016 ) , which then contributes to the activation of ERK signaling ( Vinayagam et al . , 2011 ) . In contrast , it was recently shown that ZAK inhibition by Sorafinib causes cutaneous squamous cell carcinoma due to suppression of JNK activity ( Vin et al . , 2014 ) , which might explain why over-expression of exogenous ZAK lead to reduced viability . Taken together , these observations suggest that ZAK influences cell state in a context-dependent manner , and that ZAK can either promote or suppress cell proliferation . Here we have demonstrated that ZAK suppression induced cell death upon PI3K inhibition , indicating that ZAK-mediated signaling is essential for cell survival in the context of PI3K inhibition in breast cancer cell lines . The role of ZFR and ZNF565 in cellular proliferation and survival , especially in cancer and in the context of PI3K pathway inhibition , are not yet known . Further studies will be necessary to investigate the interaction of ZFR and ZNF565 with PI3K inhibition . In consonance with the known pro-survival function of PIM2 and TACC1 in the context of PI3K inhibition , we show here that their over-expression partially rescued the effect of combining PI3K inhibition with the other four candidate genes ( Figure 2—figure supplement 1D ) , further emphasizing the importance of combining their inhibition with PI3K inhibitors . Although there is no reason to expect that these 5 genes would be altered in human cancers , analysis of breast cancer genome sequencing data collected by the TCGA project ( http://www . cbioportal . org ) reveals that the five genes are either amplified or their expression is up-regulated in 37% of breast cancer patients . In this dataset , we failed to find co-occurrence or mutual exclusivity between any of the five genes . Samples sub-classified as PI3K pathway active or inactive ( classification was based on PIK3CA mutations and amplification , ERBB2 amplification or over-expression , or PTEN deletion or mutation ) exhibited similar rates of these alterations , regardless of PIK3CA mutation status or breast cancer subtype . Moreover , there is no information regarding the response of the tumors analyzed by TCGA to PI3K pathway inhibition . As data from patients treated with PI3K inhibitors accumulates , it would be important to explore whether these genes are associated with response rate , or resistance . In summary , we report the identification of five genes whose suppression both in vitro and in vivo modulates the cellular response to PI3K inhibition , converting cytostatic effects into cytotoxicity . Importantly , we report that small molecule inhibition of two of these genes , PIM2 and ZAK , is synergistic with PI3K inhibition , a finding that has therapeutic implications for the treatment of breast cancer . MDA-MB-453 ( RRID:CVCL_0418 ) , SKBR3 ( RRID:CVCL_0033 ) , MDA-MB-231 ( RRID:CVCL_0062 ) , CAL-120 ( RRID:CVCL_1104 ) , LN-443 ( RRID:CVCL_3960 ) , LN-382 ( RRID:CVCL_3956 ) , and SF-295 ( RRID:CVCL_1690 ) were grown in DMEM supplemented with 10% FBS . T47D ( RRID:CVCL_0553 ) , HCC1954 ( RRID:CVCL_1259 ) , HCC1937 ( RRID:CVCL_0290 ) , and EFO-21 ( RRID:CVCL_0029 ) were grown in RPMI supplemented with 10% FBS . MCF10A ( RRID:CVCL_0598 ) cells were grown in MEGM supplemented with Bullet kit ( Lonza , USA ) . To generate eGFP-expressing cell lines , cells were infected with a lentivirally-delivered eGFP- and blasticidin-encoding plasmid . Cells were selected with blasticidin for at least 7 days . Unless specified , all media and supplements were from Gibco ( Thermo Fisher Scientific , USA ) . All drugs were from Selleck Chemicals ( USA ) . All cancer cell lines were obtained from the Cancer Cell Line Encyclopedia , which obtained them directly from original sources and confirmed their identity by SNP fingerprinting ( Barretina et al . , 2012 ) . LN-443 is included in the list of commonly misidentified cell lines ( International Cell Line Authentication Committee ) and is considered to be LN-444 . Both harbor the same PTEN inactivating mutation and therefore for the purpose of this report it is considered as PTEN inactive . MCF10A cells were obtained from ATCC ( ATCC , USA ) . All cells were tested for mycoplasma periodically as well as prior to screening . Lentivirus was produced in 293 T cells by co-transfection with VSV-G envelope encoding plasmid , psPAX2 packaging plasmid , and pLKO or pLEX viral vector . Virus-containing media were harvested 48 hr post transfection . For a single virus infection , cells were plated 24 hr prior to infection . Virus-containing media and polybrene ( 8 ug/ml final concentration ) were added to cells , after which cells were centrifuged at 1200 RPM for 45 min to promote infection . Virus-containing media were replaced 24 hr post infection . eGFP-expressing cells were seeded in phenol red-free media in 384 well plates using a Multidrop Combi reagent dispenser ( Thermo Scientific , USA ) . Infection and/or drug treatment were done in quadruplicates to obtain technical replicates . At least five different control shRNAs that do not target an endogenous transcript were used , and each control shRNA was infected into at least four quadruplicates across the plate , so that at least 20 quadruplicates were infected with control shRNAs per plate . Fluorescence intensity was measured by a SpectraMax M5 ( Molecular Devices , USA ) immediately after initial addition of GDC0941 ( day 0 ) and after 9 days of treatment ( day 9 ) . Media and drugs were replaced every 3–4 days . Background fluorescence from media-only wells was subtracted from fluorescent values of all cell-containing wells . Fold-change in proliferation was calculated by dividing the mean fluorescence on day nine by the mean fluorescence on day 0 for each quadruplicate . The mean FC-proliferation of all control shRNAs was used to calculate Z-scores for the FC-proliferation of experimental shRNAs . Floating and adherent cells were collected and fixed over-night in either 100% cold methanol ( for cell-cycle analysis ) , or in 70% cold ethanol ( for cPARP staining ) . For cell cycle analysis , cells were re-hydrated in PBS for 30 min , followed by staining with Propidium Iodide ( 25 ug/ml ) / RNAse A ( 50 ug/ml ) . Staining with anti-cPARP antibody ( Cell Signaling Technology , #5625 ) was conducted according to manufacturer's recommendations . A BDBiosciences LSR II flow cytometer was used to acquire samples . At least 30 , 000 cells were acquired per sample . FlowJo software was used for analysis . A BDBiosciences FACSAria instrument was used for cell sorting . In order to rescue shRNA knock down , at least three silent mutations were introduced in the shRNA-targeted sequence of each ORF . ORFs were cloned into a Gateway expression vector fused to a C-terminal V5 tag . For control ORF , HcRed was cloned into the same expression vector . Cells were seeded in 384 well plates using a Multidrop Combi reagent dispenser . Twenty-four hours later , cells were infected with lentivirally-delivered ORF vectors . 48 hr after ORF vector infection , cells were infected with lentivirally-delivered shRNAs . Infections were done in quadruplicates . Following a 48 hr recovery , cells were either treated with GDC0941 or DMSO for 9 days , with refreshing media and treatment every 3–4 days . At the end point , cells were fixed in 4% PFA , and stained for V5 ( Invitrogen #46–0705 ) expression as a surrogate reporter for ORF expression , and DARQ5 DNA stain to assess total cell viability . Staining was detected and quantified using a LiCor Odyssey scanner ( LiCor , USA ) and Image Studio software . Z-scores were calculated as described for cell proliferation assay . Cells were collected and lysed in cold 1% NP-40 lysis buffer ( 50 mM Tris pH7 . 5 , 150 mM NaCl , 2 mM EDTA , 1% NP-40 , 1 mg/ml NaF ) supplemented with phosphatase and protease inhibitors . Lysates were cleared by centrifugation , and protein concentration was determined by BCA assay . Samples were denaturated by adding NuPage LDS sample buffer ( Thermo Fisher Scientific , USA ) and reducing agent and boiling for 5 min . Equal amounts of protein were separated by electrophoresis on a 4–12% Bis-Tris gel , and then transferred to a nitrocellulose membrane . Membranes were incubated with primary antibody over-night at 4°C , and then incubated with a secondary antibody conjugated to either IRDye or HRP . Signals were detected by either a LiCor Odyssey scanner , or by using an ECL kit ( Perkin-Elmer , USA ) . All primary antibodies were from Cell Signaling Technology ( USA ) . Cellular RNA was purified with a kit ( PerfectPure; 5 Prime GmbH , Germany ) and complementary DNA ( cDNA ) was synthesized with the High Capacity RNA-to-cDNA kit ( ABI , Thermo Fisher Scientific , USA ) . Real-time-qPCR analysis was performed with Power-SYBR Green ( ABI , Thermo Fisher Scientific , USA ) in a QuantStudio 6 RT-PCR machine ( Thermo Fisher Scientific , USA ) . Three replicates were made for all qPCR assays . Probe-specific results were normalized to GAPDH RNA levels . Kinase selectivity assays for PIM2 inhibitors were done by the International Center for Kinase Profiling ( Dundee , UK ) . A KiNativ ( ActivX Biosciences ) ( Patricelli et al . , 2011 ) was used as a focused kinase selectivity assay for ZAK inhibitors . Potency of compounds was tested using a SelectScreen assay ( Thermo Fisher Scientific , USA ) . BH3 profiling was done as described previously ( Ryan and Letai , 2013 ) . In brief , cells were permeabilized and incubated with increasing concentrations of synthetic BIM peptide ( New England Peptide , USA ) . Cells were fixed and stained for endogenous cytochrome c ( Biolegend , USA ) . Cells were analyzed by FACS ( Fortessa BD Biosciences ) to determine the population rate of peptide induced loss of cytochrome c . Assay was conducted in triplicates . The initial screen was conducted by infecting 300 million cells with a pooled viral-shRNA library consisting of 98K shRNA plasmids , at a MOI of ~0 . 3 , such that each individual shRNA is represented by approximately 1000 cells . This infection was carried out in four biological replicates . Cells were then selected with puromycin for 2 days , after which cells were expanded for an additional 3 days . Cells were then divided into five separate samples , four of which were treated with GDC0941 ( 0 . 625 μM ) , which were then collected daily at 48 , 72 , 96 and 120 hr . One sample was treated with vehicle control ( DMSO ) and was collected at 72 hr . Cells were fixed , stained for cPARP and sorted twice to achieve >95% purity of cPARP-positive population . Genomic DNA was extracted from isolated cells and was used to amplify the integrated shRNA , which was then sequenced by Illumina HiSeq ( Cowley et al . , 2014 ) . Abundance of shRNAs in each sample was used to calculate a T-statistic per shRNA across all samples . Animals were maintained under conditions approved by the Institutional Animal Care and Use Committee at the Dana-Farber Cancer Institute and at the Massachusetts Institute of Technology . MDA-MB-453 cells were injected subcutaneously into both flanks of female SCID mice ( Taconic , USA ) at a concentration of 10^6 cells/flank in 50% Matrigel . Xenograft tumors were allowed to grow until reaching a diameter of ~5–8 mm . For GDC0941 dosing , mice were treated by a daily GDC0941 ( 75 mg/kg/day , in MCT ) delivered by oral gavage of for 5 days prior to device implantation . For siRNA-loaded devices , mice received an additional dose of GDC0941 24 hr after device implantation . siRNAs ( Dharmacon , USA ) were Accell-modified , Cy3-conjugated at the 5' end , and processed for in-vivo use . Microscale devices for intratumor delivery of siRNA and small molecule compounds were manufactured as described in Jonas et al . , 2015 , and were implanted directly into the mouse xenograft tumor . Devices containing the small molecule compounds remained in situ for 24 hr , while siRNA-loaded devices remained in situ for 48 hr . The flank tumor was excised and the tissue containing the device was fixed for 24 hr in 10% formalin and perfused with paraffin . The specimen was sectioned using a standard microtome and tissue sections were collected from each reservoir level . Sections were stained with Hematoxylin and Eosin , as well as antibody-stained by standard immunohistochemistry using cleaved-caspase-3 antibody ( Cell Signaling #9661 ) and scored using an ImageJ ( v1 . 48 ) image analysis algorithm in a blinded manner .	When cells become cancerous , they accumulate mutations in their DNA that switch on some genes at the wrong time and to higher levels than normal . These over-active genes help cancer cells to survive , grow and evade death . One of the genes that is often mutated and over-active in breast cancer encodes an enzyme called PIK3CA . There are several drugs that bind to and inhibit the mutant version of PIK3CA . Recent experiments show that inhibiting over-active forms of this enzyme can stop cancer cells from growing , but it does not cause them to die . This means that the cells have the opportunity to become resistant to the drug , which can subsequently lead to tumor relapse . Therefore , researchers have been looking for other drugs that , when combined with the PIK3CA-inhibiting drug , are able to kill the cancer cells . The first step to developing such a therapy is to identify genes that are essential for cancer cells to survive when they are exposed to the PIK3CA-inhibiting drug . One way to achieve this is to test what happens when you switch off individual genes one by one in these cells , an approach known as a functional genomics screen . Zwang et al . used this approach to identify genes in human breast cancer cells that have the potential to be useful drug targets . The screen identified five genes that can be individually switched off to kill the cancer cells . Two of these five genes encode enzymes known as PIM2 and ZAK . Zwang et al . went on to find drugs that inhibit PIM2 and ZAK . As expected , administering these drugs together with the PIK3CA inhibitor caused breast cancer cells to die . Further experiments will be necessary to find out what roles these five genes play in breast cancer cells . In the future , these findings may lead to the development of more effective therapies for human cancers in which PIK3CA is over-active .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cancer", "biology" ]	2017	Synergistic interactions with PI3K inhibition that induce apoptosis
Experience-dependent reorganisation of functional maps in the cerebral cortex is well described in the primary sensory cortices . However , there is relatively little evidence for such cortical reorganisation over the short-term . Using human somatosensory cortex as a model , we investigated the effects of a 24 hr gluing manipulation in which the right index and right middle fingers ( digits 2 and 3 ) were adjoined with surgical glue . Somatotopic representations , assessed with two 7 tesla fMRI protocols , revealed rapid off-target reorganisation in the non-manipulated fingers following gluing , with the representation of the ring finger ( digit 4 ) shifted towards the little finger ( digit 5 ) and away from the middle finger ( digit 3 ) . These shifts were also evident in two behavioural tasks conducted in an independent cohort , showing reduced sensitivity for discriminating the temporal order of stimuli to the ring and little fingers , and increased substitution errors across this pair on a speeded reaction time task . Evidence for experience-dependent plasticity in the adult mammalian brain exists across a variety of sensory modalities ( Fu and Zuo , 2011 ) . The ordered somatotopy of primary somatosensory cortex ( SI ) has long served as a model system for studies of cortical reorganisation , with a wealth of evidence from both the non-human primate and rodent literature linking both extreme and subtle peripheral changes in somatosensory inputs over months or years to alterations in cortical maps ( Buonomano and Merzenich , 1998; Feldman and Brecht , 2005 ) . In the human brain , there has also been evidence of experience-dependent remapping in SI . Considerable emphasis has been placed upon cross-sectional differences in the cortical organisation of SI observed in specialist populations , such as musicians , or patients with sensorimotor dysfunction , such as focal dystonia ( Elbert et al . , 1995; Bara-Jimenez et al . , 1998; Nelson et al . , 2009; Kalisch et al . , 2009 ) . However , only limited longitudinal evidence exists that directly demonstrates remapping at the level of human SI , either in response to altered hand usage patterns ( Stavrinou et al . , 2007 ) or more intensive Hebbian co-activation paradigms delivering specific patterns of tactile stimulation to the fingertips ( Pleger et al . , 2001 , 2003; Hodzic et al . , 2004; Vidyasagar et al . , 2014 ) . There remains a limited understanding of the speed of SI plasticity and how cortical changes relate to behaviour . Here we address this gap in the literature , investigating the propensity for rapid experience-dependent cortical reorganisation and the behavioural relevance thereof . Using a well-validated paradigm of single-subject fMRI mapping of human SI at 7 tesla ( Sanchez-Panchuelo et al . , 2010; Kolasinski et al . , 2016 ) , we asked two questions . First , can experience-dependent plastic remapping of SI somatotopy be elicited in the human cortex in just 24 hr ? Second , are any observed cortical changes reflected in altered tactile function ? A complementary combination of fMRI and behavioural psychophysics were used to investigate the effect of a 24 hr manipulation in which the right index ( D2 ) and right middle digit ( D3 ) were joined together using skin glue . Based on the primate literature ( Clark et al . , 1988; Allard et al . , 1991 ) , we hypothesised that forced co-use of the glued digit pair would result in increased tactile co-activation across the digits , and an increase in their shared cortical representation . An alternative hypothesis posited that compensatory behaviour during the 24 hr gluing manipulation would promote changes in the cortical representations of adjacent , but unaffected digits . In the fMRI cohort , SI digit somatotopy of the right hand was mapped at 7 tesla ( Kolasinski et al . , 2016 ) after two periods of normal hand use ( control 1 and control 2 ) , and after the gluing manipulation . In each session , fMRI data were acquired during a block design task and phase-encoding task . We first asked whether the amount of cortical overlap between digit representations in SI changed after the gluing manipulation in comparison with two control sessions using the phase-encoding fMRI data . Measures of inter-digit overlap were calculated for adjacent digit pairs ( D2-D3: index-middle , D3-D4: middle-ring , D4-D5: ring-little; Figure 1A ) . A two-way repeated measures ANOVA indicated a significant interaction between session and digit pair on the amount of cortical overlap ( F ( 4 , 32 ) = 13 . 412 , p<0 . 0005 , η2:0 . 626 ) ( Figure 1B ) . This was driven by a significant reduction in the cortical overlap between D3-D4 ( Simple main effect: F ( 2 , 16 ) = 23 . 379 , p<0 . 0005; Pairwise Sidak-corrected p<0 . 05 ) and a significant increase in the overlap of D4-D5 ( Simple main effect: F ( 2 , 16 ) = 13 . 384 , p<0 . 0005; Pairwise Sidak-corrected p<0 . 05 ) in the glued condition compared with both control sessions ( Figure 1B ) . No significant change was found in the overlap between D2 and D3 , the glued digits , where changes have been observed in similar but longer term studies in non-human primates ( Clark et al . , 1988; Allard et al . , 1991 ) . No shift in peak-to-peak distance between the digit representations ( Supplementary file 1 ) or the overall surface area of each digit representation was observed ( Figure 1—figure supplement 1 ) . No systematic difference in the fit between the phase-encoding models and fMRI signal were observed across sessions ( Supplementary file 2 ) . 10 . 7554/eLife . 17280 . 003Figure 1 . Patterns of rapid experience-dependent remapping in human SI . ( A ) An example digit map from an individual participant and session , shown on an inflated brain surface ( bottom right ) , and a zoomed panel ( Threshold FDR α = 0 . 01 ) ; sulci and gyri are shown in dark and light grey , respectively . The region of interest used for cortical overlap analysis is indicated by a white line on the inflated brain . ( B ) Cortical overlap between pairs of digit representations showed a significant reduction between D3 and D4 , and a significant increase between D4 and D5 after the gluing manipulation compared with the two control conditions . p<0 . 05 p<0 . 005 Sidak corrected . Data in ( B ) are presented normalised to control one session; all statistics were performed on raw un-normalised data . Dice: Dice coefficient; Error bars represent standard error of mean . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 00310 . 7554/eLife . 17280 . 004Figure 1—source data 1 . Data presented in Figure 1B . DICE coefficients representing digit overlap for three digit pairs ( D2–D3 , D3–D4 and D4–D5 ) across three conditions ( Control 1 , Control 2 and Glued ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 00410 . 7554/eLife . 17280 . 005Figure 1—figure supplement 1 . Cortical surface area of individual digit representations used in calculation of DICE coefficient at each time point . No significant change in cortical digit surface area was observed across the three time points under study ( Control 1 , Control 2 and Glue ) ; two-way repeated measures ANOVA indicated no significant interaction between session and digit on the surface area ( F ( 6 , 48 ) = 1 . 259 , p=0 . 294 , η2: . 136 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 00510 . 7554/eLife . 17280 . 006Figure 1—figure supplement 2 . Overview of the fMRI experiment and the behavioural experiment study design . The order of conditions was counterbalanced across participants , resulting in four participants experiencing the control period first ( Order A; top row ) , and five participants experiencing the gluing manipulation first ( Order B , bottom row ) . Control 0 was treated as an habituation session and was not included in the analysis . Beh: behavioural testing . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 00610 . 7554/eLife . 17280 . 007Figure 1—figure supplement 3 . Data presented in Figure 1B split according to the order of sessions . ( A ) Order A ( Control 1 separated by 4 weeks from Control 2 and Glued ) : n = 4; Order B ( Glued separated by 4 weeks from Control 1 and Control 2 ) , n = 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 007 To further explore the observed change in inter-digit overlap with no change in overall cortical surface area of each digit representation , the fMRI representations of digit 4 used to calculate the cortical overlap metrics ( Dice coefficients ) were visualised for each participant and session ( Figure 2 ) . This data revealed that the observed changes presented in Figure 1B were driven by an expansion in the representation of digit 4 adjacent to digit 5 and a corresponding contraction at the boundary with digit 3 . 10 . 7554/eLife . 17280 . 008Figure 2 . Pattern of shift in the cortical representation of digit 4 from inter-digit overlap analysis . Data displayed for all nine participants ( A–I ) showing the outline of the digit 4 representation mapped during the three sessions ( Control 1 , Control 2 , Glued ) overlaid on the individual participants’ cortical surface reconstruction ( sulci: dark grey , gyri: light grey; zoomed panel showing anatomical hand knob ) . The location of the peak vertex is shown with a coloured circle . These results demonstrated a consistent shift in the representation of digit 4 at the level of individual participants , such that while there is a minimal change in peak activation and area of the representation , the flank adjacent to the representation of digit 5 expands , and the flank adjacent to digit 3 contracts , consistent with the observed changes in cortical overlap ( Figure 1B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 008 Representational similarity analysis conducted on the block design fMRI data also implicated changes in the representation of digit 4 ( Figure 3 ) . These data demonstrated a shift in the representation of D4 away from D3 and towards D5 . 10 . 7554/eLife . 17280 . 009Figure 3 . Representational similarity analysis of block design data yields complementary evidence of shift in the S1 representation of digit 4 away from digit 3 and towards digit 5 . ( A ) Noise normalised parameter estimates from a standard GLM for each digit were used to construct representational dissimilarity matrices ( RDMs ) using Euclidean distance within a hand knob ROI derived individually for each subject from phase-encoding data from all sessions . Average raw distance values are shown for each session ( B ) Multidimensional scaling and Procrustes analysis of individual participants’ distance matrices at each time point demonstrate schematically the observed shift in the representation of digit 4 , away from digit 3 , and towards digit 5 , consistent with the observed pattern of cortical overlap and tactile discrimination changes . Two-way repeated measures ANOVA indicated a significant interaction between session and digit pair on the amount of cortical overlap ( F ( 2 . 0 , 16 . 2 ) = 4 . 430 , p=0 . 029 , η2:0 . 356 ) , driven by a shift in the representation of digit 4 away from digit 3 ( Simple Main Effect: F ( 2 , 16 ) = 16 . 076; pairwise comparisons glued vs . control 1 and control 2: p<0 . 01 ) . Multidimensional scaling yields the spatial relationship of representations in arbitrary units ( a . u . ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 00910 . 7554/eLife . 17280 . 010Figure 3—source data 1 . Data presented Figure 3A . Representational similarity analysis distance metrics for adjacent digit pairs ( D2–D3 , D3–D4 and D4–D5 ) across three conditions ( Control 1 , Control 2 and Glued ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 010 Given the fMRI data were strongly suggestive of changes in cortical overlap outside of the glued digit pair ( D2-D3 ) , we sought to more fully investigate the observed pattern of short-term reorganisation by assessing the behavioural correlates of the gluing manipulation . The same experimental design and gluing manipulation were applied in a separate cohort . Instead of an fMRI scan , nine participants undertook behavioural psychophysics tasks during each session . The first task involved temporal order judgment ( TOJ ) , where pairs of rapid vibrotactile stimuli were applied to adjacent digit pairs of the right hand at varying interstimulus intervals . Participants judged which digit was stimulated first . A psychometric function was fitted to the resulting accuracy data , from which a metric of tactile discrimination ability was calculated for each adjacent digit pair ( Just Noticeable Difference: JND ) . Greater values of JND are associated with poorer discrimination across an adjacent digit pair ( Figure 4A ) . We asked whether the ability to distinguish the order of two stimuli delivered in rapid succession , one each on two adjacent digits , differed in the session directly following the gluing manipulation in comparison with the control sessions . 10 . 7554/eLife . 17280 . 011Figure 4 . Patterns of rapid experience-dependent remapping in SI mirror peripheral changes in tactile discrimination performance . ( A ) An example of accuracy data from the temporal order judgment task . Accuracy scores ( black dots ) are illustrated for an individual participant and a single run assessing performance across D2-D3 . Data are fitted with a logistic function ( red line ) from which the just noticeable difference ( JND ) is calculated: a measure of temporal tactile acuity; greater JND means poorer tactile discrimination . ( B ) Tactile discrimination improved significantly between D3 and D4 , and worsened significantly between D4 and D5 after the gluing manipulation compared with the two control conditions . In summary , fMRI evidence of rapid cortical remapping ( Figure 3 ) concurs with behavioural changes in tactile function ( B ) , such that the digit pair with reduction in cortical overlap ( D3–D4 ) also shows increases in tactile discrimination , whereas the digit pair showing increase in cortical overlap ( D4–D5 ) demonstrates worsening of tactile discrimination p<0 . 05 *p<0 . 005 Sidak corrected . Data in ( B ) are presented normalised to time point control 1; all statistics were performed on raw un-normalised data . ISI: inter-stimulus interval; JND: Just Noticeable Difference . Error bars represent standard error of mean . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 01110 . 7554/eLife . 17280 . 012Figure 4—source data 1 . Data presented Figure 4B . Just Noticeable Difference ( JND ) values representing tactile discrimination ability for three digit pairs ( D2–D3 , D3–D4 and D4–D5 ) across three conditions ( Control 1 , Control 2 and Glued ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 01210 . 7554/eLife . 17280 . 013Figure 4—source data 2 . Data presented Figure 4—figure supplement 1B . Frequency of mis-localisations for motor confusion task across adjacent digit pairs ( D2–D3 , D3–D4 and D4–D5 ) across three conditions ( Control 1 , Control 2 and Glued ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 01310 . 7554/eLife . 17280 . 014Figure 4—figure supplement 1 . In a motor confusion task involving rapid button presses using the four digits under study ( D2 , D3 , D4 , D5 ) there is an increase in the number of mis-presses between digits 4 and 5 , consistent with the observed pattern of increased cortical overlap and representational similarity observed between digit 4 and 5 . ( A ) Distribution of presses and mis-presses for each target and each digit , averaged across Control 1 and Control 2 , demonstrating a pattern of mis-presses predominantly in digits adjacent to the target . ( B ) Plotting the average number of mis-presses across adjacent digit pairs during different sessions reveals an increased confusion in presses between digit 4 and digit 5 in the Glued condition compared with Control 1 and Control 2 , not seen in other digit pairs . Two-way repeated measures ANOVA indicated a significant interaction between session and digit pair on number of mis-presses ( F ( 4 , 32 ) = 3 . 828 , p=0 . 012 , η2: 0 . 324 ) , driven by an increase in confusion between digit 4 and 5 in the glued condition compared with controls; p<0 . 05 Sidak corrected . DOI: http://dx . doi . org/10 . 7554/eLife . 17280 . 014 We generated behavioural predictions based on our fMRI findings of altered cortical maps after the gluing manipulation ( Figure 3 ) . Specifically , we hypothesised that the gluing manipulation would result in improved tactile discrimination ( reduced JND ) across D3-D4 , which demonstrates reduced cortical overlap , and diminished tactile discrimination ( increased JND ) across D4-D5 , which demonstrates increased cortical overlap . Data from eight participants met the goodness of fit threshold for all sessions ( R2 > 0 . 4 ) . A two-way repeated measures ANOVA indicated a significant interaction between session and digit pair on tactile discrimination ( JND ) ( F ( 4 , 28 ) = 14 . 613 , p<0 . 0005 , η2:0 . 676 ) . This was driven by a significant reduction in JND across D3-D4 ( Simple main effect: F ( 2 , 14 ) = 14 . 631 , p<0 . 0005; Pairwise Sidak-corrected p<0 . 05 ) , and a significant increase in the JND across D4-D5 ( Simple main effect: F ( 2 , 14 ) = 10 . 578 , p=0 . 002; Pairwise Sidak-corrected p<0 . 05 ) in the gluing condition compared with both control sessions ( Figure 4B ) . No significant change in tactile discrimination was found for D2-D3 , the glued digits . In other words , after the gluing manipulation , tactile discrimination was improved across D3-D4 and was diminished across D4-D5 , with no change in the glued digits D2-D3 . This finding was supported by similar results from a second task , assessing motor confusion , which involved rapidly cued button presses using individual digits . There was an increase in the level of confusion between digits 4 and 5 after the gluing manipulation compared with the control conditions ( Figure 4—figure supplement 1 ) . The fMRI and psychophysics results present complementary evidence for functionally relevant reorganisation in human SI following just a 24 hr peripheral change in hand use . Previous longitudinal studies of detailed somatotopic remapping in humans using MEG or standard resolution fMRI have provided evidence for a general change in the distance between digit representations ( Stavrinou et al . , 2007; Mogilner et al . , 1993 ) , or a synchronisation in activity pattern of non-adjacent digits ( Vidyasagar et al . , 2014 ) after coupled stimulation across digits . Conversely , an intervention involving hand and arm immobilisation drove diminished tactile acuity and a corresponding reduction in BOLD activity in contralateral SI ( Lissek et al . , 2009 ) . Our results build on this work , demonstrating that usage-dependent changes in the overlap of SI digit representations are associated with a corresponding change in the ability to differentiate tactile inputs to the digits . Specifically , an increase in the cortical overlap between D4 and D5 was accompanied by a reduction in the tactile discrimination ability across these two digits . A consistent feature of our findings was that cortical reorganisation and behavioural change was not observed for the glued digit pair ( D2-D3 ) . Instead , we saw changes in cortical overlap , and corresponding change in behavioural performance , for the other digit pairs . Specifically , we saw a shift of D4 away from D3 and towards D5 . These findings reject our hypothesis of increased tactile synchronisation driving an increase in cortical overlap . Instead , these data support the alternative hypothesis that compensatory behaviour in other digits during the 24 hr manipulation drives off-target effects . In light of the considerable dexterous abilities specific to the human hand ( Young , 2003 ) and the short duration of the gluing manipulation , it is feasible that the observed remapping reflects compensatory behavioural changes in the pattern of hand use during the manipulation . Rather than learning to coordinate and co-use the glued digits , participants may have adapted synergies with the non-manipulated digits . In this case , the increased physical separation of D3 and D4 during the gluing reduces the usually high degree of anatomical enslavement seen peripherally between these two digits ( Kim et al . , 2008 ) , freeing D4 to function even more synergistically with D5 . This argument is supported by the observed pattern of increased motor confusion between digits 4 and 5 after the gluing manipulation ( Figure 4—figure supplement 1 ) , demonstrating potential changes in movement synergies , that may explain the somatosensory changes observed . This finding complements recent observations that patterns of synergistic digit usage are strongly reflected in the generalisation of tactile learning from trained to adjacent untrained digits ( Dempsey-Jones et al . , 2016 ) . The two fMRI datasets showed independent evidence of a shift in the digit 4 representation during the gluing manipulation: away from digit 3 , and towards digit 5 ( Figures 1 and 3 ) , raising questions regarding the underlying mechanism driving this reorganisation . Similar studies undertaken over a period of months in non-human primates have reported the emergence of overlapping receptive fields in response to the fusion of two adjacent digits ( Clark et al . , 1988; Allard et al . , 1991 ) . The spatial resolution of BOLD fMRI is , of course , poorer than invasive mapping techniques . Nonetheless , it is possible that the observed changes in cortical activation could result from shifts in population receptive fields at the boundary of cortical digit representations ( Besle et al . , 2014 ) , or even changes in multidigit receptive fields ( Thakur et al . , 2012 ) . As well as reorganisation in the cortex , strong evidence also suggests considerable plastic potential in subcortical structures , which is mirrored in the cortex ( Jones , 1996; Rausell et al . , 1998; Kambi et al . , 2014 ) . The remapping observed in SI here may therefore result from reorganisation in lower level nuclei of the ascending somatosensory pathway . A shift in the cortical representation of digit 4 ( Figure 2 ) , rather than an overall enlargement of the representation , distinguishes the observed pattern of cortical change from previous studies of artificial tactile co-activation , wherein a generalised cortical magnification of the manipulated digit is observed ( Pleger et al . , 2001 , 2003; Hodzic et al . , 2004 ) . The cortical shift perhaps reflects how naturalistic behavioural changes in hand use are reflected in the cortex , with remapping based on functional need; in this case , leading to perceptual improvements in certain digit pairs ( D3/D4 ) , and worsening in others ( D4/D5 ) . What mechanism drives the observed change in cortical overlap ? In vivo studies in rodent barrel cortex demonstrate that a 24 hr period of altered sensory input can induce increases in synaptic density in the corresponding cortical representation ( Knott et al . , 2002 ) . This could unmask or potentiate pre-existing divergent or silenced connections between adjacent digit representations ( DeFelipe et al . , 1986; Recanzone et al . , 1992; Huntley , 1997 ) . The question remains: how could this cortical shift occur with no associated change in the surface area or location of peak activation of the digit 4 representation ? Patterns of local horizontal connectivity in the cortex are established during development and moulded throughout life ( Sur and Rubenstein , 2005 ) . Homeostatic mechanisms also exist to maintain balance in cortical activity ( Turrigiano and Nelson , 2000 ) , and to preserve the scaling and pattern of cortical topography ( Sharma et al . , 2000; Vanderhaeghen et al . , 2000 ) . It is plausible that the gluing manipulation results in the potentiation of excitatory horizontal connectivity between the representations of digits 4 and 5 , and the concurrent weakening of connectivity between the representations of digits 3 and 4 . After 24 hr , we observe the resulting pattern of cortical reorganisation ( Figures 2 and 3 ) , wherein the relative positions of the representations have changed , but features of the maps firmly established in development , remain in place . A comparable process of reduced inhibition prior to remapping has been reported in studies of sensory deprivation in rodent barrel cortex . Reduced inputs after whisker clipping produce a local pattern of cortical disinhibition and broadening excitation , followed by subsequent cortical contraction of the representation ( Albieri et al . , 2015 ) . In this case , changes in the relative inputs across different digits may prompt a transient disinhibitory cortical milieu , and the subsequent re-establishment of modified excitatory networks within the cortex , wherein the cortical representations have shifted to reflect new usage patterns . The unmasking of latent excitatory connections prior to any homeostatic increase in lateral inhibition to rebalance cortical excitability could explain the observation of more marked changes in tactile perception and representational surface area after very short and intense tactile training interventions , or rTMS protocols ( Pleger et al . , 2003; Tegenthoff et al . , 2005; Ragert et al . , 2008; Dinse and Tegenthoff , 2015 ) . The consistent cortical changes at the flanks of the digit 4 representation ( Figure 2 ) in the absence of changes in the location of peak activation suggest that the latter may not be the most informative feature of rapid map reorganisation . The existence of such hard-wired elements in cortical somatotopy is supported by evidence of SI constraining the extent of reorganisation within the somatosensory system: acting as a blueprint for lower level nuclei ( Zembrzycki et al . , 2013 ) , enforcing some limits on plasticity , and potentially explaining the persistence of SI somatotopic features long after marked peripheral changes , such as amputation ( Kikkert et al . , 2016 ) . Given the increasingly recognised contribution of motor efference information to activity patterns in somatosensory cortex ( Lee et al . , 2008 , 2013 ) , it is also possible that changes in motor outputs over the course of 24 hr may shape patterns of SI activity , promoting a top-down form of SI experience-dependent plasticity . The rapidity of cortical changes shown herein supports further investigation of rapid and targeted therapeutic interventions in conditions such as focal dystonia , where maladaptive changes in somatotopy have previously been reported ( Bara-Jimenez et al . , 1998 ) . The mechanistic link between short-term plasticity and long-term circuit changes remains unclear ( Holtmaat and Svoboda , 2009; Johansen-Berg et al . , 2012 ) . Therefore , further questions remain as to how short-term peripheral manipulations could be used to induce sustained modification or refinement of fine grain functional cortical organisation , as a means to enhance or rehabilitate tactile function .	The areas of the brain that receive inputs from our senses have a map-like structure . In an area called the visual cortex this map represents our field of vision; in the auditory cortex , it represents the range of different tones we can hear . The sense of touch is processed in the somatosensory cortex: an area of the brain that is organised around a map of the body , with adjacent regions of the cortex representing adjacent regions of the body . The clear structure of these brain regions makes them ideal for exploring how the organisation of the brain changes over time . How quickly can changes to the touch inputs that the brain receives cause the map in the somatosensory cortex to reorganise ? Can these effects be produced in just 24 hours ? And would this remapping affect how we perceive touch ? To investigate these questions , Kolasinski et al . glued together the right index and right middle fingers of healthy human volunteers . This separated the middle and ring fingers: a pair that usually move together due to the anatomical structure of the hand . Functional magnetic resonance imaging of the brain’s activity revealed that within 24 hours of the gluing , the brain’s representation of the ring finger moved away from that of the middle finger , and towards the representation of the little finger . A perceptual judgment task mirrored this finding: after 24 hours of gluing , the participants became better at distinguishing between the middle and ring fingers and worse at distinguishing between the ring and little fingers . This is a powerful demonstration of the human brain’s potential to adapt and reorganise rapidly to changes to sensory inputs . The sense of touch declines gradually with age and may also be reduced as a result of disease such as stroke . A long-term challenge is to understand how the sensory regions of the brain change during this loss of sensation . Further research could then investigate how to maintain the structure of the cortical map to prolong or restore high quality touch sensation .	[ "Abstract", "Introduction" ]	[ "short", "report", "neuroscience" ]	2016	Perceptually relevant remapping of human somatotopy in 24 hours
Cell proliferation and quiescence are intimately coordinated during metazoan development . Here , we adapt a cyclin-dependent kinase ( CDK ) sensor to uncouple these key events of the cell cycle in Caenorhabditis elegans and zebrafish through live-cell imaging . The CDK sensor consists of a fluorescently tagged CDK substrate that steadily translocates from the nucleus to the cytoplasm in response to increasing CDK activity and consequent sensor phosphorylation . We show that the CDK sensor can distinguish cycling cells in G1 from quiescent cells in G0 , revealing a possible commitment point and a cryptic stochasticity in an otherwise invariant C . elegans cell lineage . Finally , we derive a predictive model of future proliferation behavior in C . elegans based on a snapshot of CDK activity in newly born cells . Thus , we introduce a live-cell imaging tool to facilitate in vivo studies of cell-cycle control in a wide-range of developmental contexts . Organismal development requires a delicate balance between cell proliferation and cell cycle exit . In early embryos , the emphasis is placed on rapid cell proliferation , which is achieved by omitting gap phases ( G1 and G2 ) and establishing a biphasic cell cycle that rapidly alternates between DNA synthesis ( S phase ) and mitosis ( M phase ) ( Edgar and O'Farrell , 1989; Newport and Kirschner , 1982 ) . After several rounds of embryonic cell division , the gap phases are introduced , coincident in many organisms with cell fate decisions and the execution of morphogenetic cell behaviors ( Foe , 1989; Grosshans and Wieschaus , 2000 ) . These gap phases are believed to function as commitment points for cell-cycle progression decisions . The earliest point of commitment occurs during G1 , which is the focus of this study . Cells either engage in cell-cycle progression and enter S phase , or they exit the cell cycle altogether and enter a cell-cycle phase referred to as G0 and undergo quiescence or terminal differentiation ( Sun and Buttitta , 2017 ) . Although the location of the G1 commitment point in yeast ( Start ) and cultured mammalian cells ( Restriction Point ) has in large part been spatiotemporally mapped and molecularly characterized ( Hartwell et al . , 1974; Pardee , 1974; Spencer et al . , 2013 ) , when cells make this decision in living organisms while integrating intrinsic and the extrinsic cues of their local microenvironment during development remains poorly understood . A cell-cycle sensor that is amenable to such in vivo studies can shed new light on this four-decade-old biological phenomenon . In 2008 , Sakaue-Sawano and colleagues engineered a multicolor fluorescent ubiquitination-based cell-cycle indicator ( FUCCI ) for mammalian cell culture ( Sakaue-Sawano et al . , 2008 ) . FUCCI has since been adapted for many research organisms ( Özpolat et al . , 2017; Zielke and Edgar , 2015 ) . However , FUCCI on its own cannot distinguish between a cell residing in G1 that will cycle again upon completing mitosis and a cell that is poised to enter G0 ( Oki et al . , 2015 ) . Separating G1 from G0 is an essential first step to understanding mechanisms controlling cell cycle exit during quiescence or terminal differentiation . To distinguish G1 from G0 in mammalian cell culture , Hanh , Spencer and colleagues developed and implemented a single-color ratiometric sensor of cell-cycle state composed of a fragment of human DNA helicase B ( DHB ) fused to a fluorescent protein that is phosphorylated by CDKs ( Hahn et al . , 2009; Schwarz et al . , 2018; Spencer et al . , 2013 ) . Notably , through quantitative measurements of CDK activity , this sensor provided new insights into the proliferation-quiescence decision in cultured mammalian cells by identifying cycling cells that exit mitosis in a CDK-increasing ( CDKinc ) state and quiescent cells that exit mitosis in a CDK-low ( CDKlow ) state ( Spencer et al . , 2013 ) . Nonetheless , a DHB-based CDK sensor has not been utilized to evaluate the proliferation-quiescence decision in vivo . In this study , we investigate the proliferation-quiescence decision in Caenorhabditis elegans and zebrafish , two powerful in vivo systems with radically different modes of development . We generate transgenic CDK sensor lines in each organism to examine this decision live at mitotic exit . By quantifying CDK activity , or DHB ratios , at mitotic exit , we are able to predict future cell behavior across several embryonic and post-embryonic lineages . Despite cells generally exiting mitosis with decreased CDK-activity levels , we reliably distinguish cycling cells that exit mitosis into G1 , in a CDKinc state , from quiescent cells that exit mitosis into G0 , in a CDKlow state . To gain insights into cell-cycle progression commitment , we examine the activity of C . elegans cki-1 , a cyclin-dependent kinase inhibitor ( CKI ) of the Cip/Kip family , demonstrating that endogenous CKI-1 levels are anti-correlated with CDK activity during the proliferation-quiescence decision . We propose that integration of CKI-1 levels in the mother cell and high CKI-1 , low CDK activity at mitotic exit mediate this decision . By utilizing the CDK sensor to predict future cell behavior , we uncover a cryptic stochasticity that occurs in a temperature-dependent fashion in the C . elegans vulva , an otherwise invariant and well-characterized lineage . Finally , we reveal cell-cycle dynamics in zebrafish , an organism that lacks a defined cell lineage , demonstrating that quiescent embryonic tissues display DHB ratios that correlate with those observed in G0 cells in C . elegans . Together , we present a tool for visualizing G1/G0 dynamics in vivo during metazoan development that can be used to study the interplay between cell proliferation and quiescence . We synthesized a C . elegans codon-optimized fragment of human DHB composed of amino acids 994–1087 ( Hahn et al . , 2009; Spencer et al . , 2013 ) . The fragment contains four serine residues that are consensus CDK phosphorylation sites ( Moser et al . , 2018; Spencer et al . , 2013 ) . These serines flank a nuclear localization signal ( NLS ) that is adjacent to a nuclear export signal ( NES ) ( Figure 1A ) . When CDK activity is low , the NLS is strong , the NES is weak and DHB localizes to the nucleus . However , when CDK activity increases during cell-cycle entry , the NLS is masked and DHB re-localizes to the cytoplasm ( Figure 1B ) . Using this DHB fragment , we generated two CDK sensors by fusing green fluorescent protein ( GFP ) or two copies of a red fluorescent protein , mKate2 ( 2xmKate2 ) , to the DHB C-terminus ( Figure 1A ) . To visualize the nucleus , we co-expressed his-58/histone H2B fused to 2xmKate2 or GFP , respectively , which is separated from DHB by a P2A self-cleaving viral peptide ( Ahier and Jarriault , 2014 ) . We drove the expression of each CDK sensor via a ubiquitous rps-27 promoter ( Ruijtenberg and van den Heuvel , 2015 ) . To test both the GFP ( Figure 1—figure supplement 1A and B ) and 2xmKate2 ( Figure 1C , Figure 1—figure supplement 1B and C ) versions of our CDK sensor , we began by examining cell divisions in the C . elegans embryo and germline ( Figure 1—video 1 ) . First , we visualized cells in the embryonic intestine , which is clonally derived from the E blastomere , as these are the first cells in the embryo to have gap phases ( Edgar and McGhee , 1988 ) . The E blastomere goes through four rounds of divisions to give rise to 16 descendants ( E16 cells ) about 4 hr after first cleavage . While 12 of the E16 cells have completed their embryonic divisions at this stage ( Leung et al . , 1999 ) , four cells called E16* star cells divide once more to generate the 20-celled intestine ( E20 ) ( Rasmussen et al . , 2013; Yang and Feldman , 2015 ) . Thus , we wondered whether our CDK sensor could be used to distinguish between cycling E16* star cells and quiescent E16 cells . To accomplish this , we tracked E16* star cell division from the E16–E20 stage and observed that DHB::GFP localizes in a cell-cycle-dependent fashion during these divisions , with DHB::GFP translocating from the nucleus to the cytoplasm and then re-locating to the nucleus at the completion of E16* star cell division ( Figure 1—figure supplement 1A ) . Consistent with our observations using the GFP version of our CDK sensor in mid-embryogenesis , DHB::2xmKate2 also dynamically translocates from the nucleus to the cytoplasm during cell divisions in the early embryo ( Figure 1C ) . Second , we examined the localization of DHB::GFP and DHB::2xmKate2 in the adult C . elegans germline ( Figure 1—figure supplement 1B and C ) . Here we detected a gradient of live CDK activity , from high in the distal mitotic progenitor zone to low in the proximal meiotic regions , as described with EdU incorporation and phospho-histone H3 staining ( Kocsisova et al . , 2018 ) . Together , these results demonstrate that our CDK sensor is dynamic during cell-cycle progression in the C . elegans embryo and germline . The ability to distinguish cycling cells from quiescent cells in the embryo made us wonder whether we could also distinguish these cellular states post-embryogenesis . Therefore , we examined our CDK sensor in several post-embryonic somatic lineages that undergo proliferation followed by cell cycle exit ( Sulston and Horvitz , 1977 ) . Specifically , we selected the sex myoblasts ( SM ) , the somatic sheath ( SS ) and ventral uterine ( VU ) cells of the somatic gonad , and the vulval precursor cells ( VPCs ) ( Figure 1D and D’ ) . To define each phase of the cell cycle while these lineages are proliferating , we combined static and time-lapse imaging approaches to measure cytoplasmic:nuclear DHB ratios for G1 , S , and G2 ( Figure 1E–J , Figure 1—figure supplement 1D–M ) . First , we quantified DHB ratios following time-lapse of VPC ( Figure 1F–H ) , SM ( Figure 1—figure supplement 1D , F and G ) and uterine ( Figure 1—figure supplement 1E , F and H ) divisions to determine peak values of G2 CDK activity . All lineages exhibited the same CDK sensor localization pattern during peak G2 ( i . e . maximal nuclear exclusion ) . We then RNAi depleted the sole C . elegans CDK1 homolog , cdk-1 , to induce a penetrant G2 phase arrest in the SM cells to corroborate these results . Quantification of DHB ratios following cdk-1 RNAi treatment showed a mean ratio of 1 . 00 ± 0 . 28 and 2 . 36 ± 0 . 70 in the GFP and 2xmKate2 versions of our CDK sensor , respectively ( Figure 1—figure supplement 1I and J ) . Next , for each lineage ( Figure 1F , Figure 1—figure supplement 1D and E ) , we quantified DHB ratios 25 min after anaphase from our time-lapses to determine a threshold for G1 phase CDK activity . In G1 , DHB::GFP and DHB::2xmKate2 were nuclear localized after mitotic exit with mean ratios of 0 . 35 ± 0 . 14 and 0 . 58 ± 0 . 32 in VPCs , 0 . 59 ± 0 . 11 and 0 . 97 ± 0 . 20 in SMs , and 0 . 67 ± 0 . 10 and 1 . 13 ± 0 . 17 in uterine cells ( Figure 1G and H , Figure 1—figure supplement 1F–H ) . Finally , we paired DHB::2xmKate2 with a reporter for S phase , fusing GFP to the sole C . elegans proliferating cell nuclear antigen ( PCNA ) homolog , pcn-1 , expressed under its own endogenous promoter at single copy . Although nuclear localized throughout the cell cycle , PCNA forms sub-nuclear puncta only in S phase ( Brauchle et al . , 2003; Dwivedi et al . , 2019; Strzyz et al . , 2015; Zerjatke et al . , 2017 ) . Analysis of time-lapse data found that punctate expression of PCN-1::GFP correlated with mean DHB::2xmKate2 ratios of 1 . 02 ± 0 . 22 in VPC ( Figure 1I and J ) , 0 . 89 ± 0 . 16 in SM ( Figure 1—figure supplement 1K and M; Figure 1—video 2 ) , and 1 . 00 ± 0 . 10 in uterine ( Figure 1—figure supplement 1L and M ) lineages . Despite individual lineages showing differences in CDK activity ( Figure 1—figure supplement 1F and M–O ) , primarily in G1 , we can establish DHB ratios for each interphase state ( G1/S/G2 ) across several post-embryonic somatic lineages using our CDK sensor paired with a PCNA reporter . We next wondered if we could distinguish G1 from G0 as these somatic lineages exit their final cell division; therefore , allowing us to visibly and quantitatively detect cellular quiescence in vivo . We mainly chose the DHB::GFP version of our CDK sensor to conduct the following experiments as it was more photostable . In asynchronously dividing MCF10A epithelial cell lines , cells that exited mitosis into a CDK2low state had a high probability of staying in G0 compared to cells that exited at a CDK2inc state ( Spencer et al . , 2013 ) . We therefore wanted to determine whether the cytoplasmic:nuclear ratio of DHB::GFP following an in vivo cell division could be used to predict if a cell will enter G1 and divide again or enter G0 and undergo quiescence . Taking advantage of the predictable cell lineage pattern of C . elegans , we quantitatively correlated DHB::GFP ratios with the decision to proliferate or exit the cell cycle . We first quantified DHB::GFP ratios from time-lapse acquisitions of SM cell divisions . The SM cells undergo three rounds of cell division during the L3 and L4 larval stages before exiting the cell cycle and differentiating into uterine muscle ( um ) and vulval muscle ( vm ) ( Figure 2A; Sulston and Horvitz , 1977 ) . Quantification of DHB::GFP in this lineage revealed that shortly after the first and second divisions , CDK activity increases immediately after mitotic exit from an intermediate level , which we designate as a CDKinc state ( Figure 2B and C; Figure 2—video 1 ) , Conversely , CDK activity following the third and terminal division remains low , which we designate as a CDKlow state . Bootstrap analyses support a significant difference in DHB::GFP ratios between pre-terminal ( CDKinc ) and terminal divisions ( CDKlow ) , but not among pre-terminal divisions ( Figure 2—figure supplement 1A–C ) . We then quantified DHB::GFP ratios during the division of two somatic gonad lineages , the VU and SS cells . VU and SS cells undergo several rounds of division during the L3 larval stage and exit the cell cycle in the early L4 stage ( Figure 2D; Sulston and Horvitz , 1977 ) . We quantified a pre-terminal division and the subsequent division that leads to quiescence . Similar to the SM lineage , both somatic gonad lineages exit the round of cell division prior to their final division into a CDKinc state and then exit into a CDKlow state following their terminal division ( Figure 2E and F , Figure 2—figure supplement 1D–F; Figure 2—video 2 ) . Bootstrap analyses also support a significant difference between DHB::GFP ratios in pre-terminal versus terminal divisions in the developing somatic gonad ( Figure 2—figure supplement 1D ) . Next , we sought to determine how the CDK sensor behaves under conditions in which cells are experimentally forced into G0 . To accomplish this , we generated a single copy transgenic line of mTagBFP2-tagged CKI-1 , the C . elegans homolog of p21Cip1/p27Kip1 , under an inducible heat shock promoter ( hsp ) , paired with a rps-0>DHB::mKate2 variant of the CDK sensor . Induced expression of CKI-1 is expected to result in G0 arrest ( Hong et al . , 1998; Matus et al . , 2014; van der Horst et al . , 2019 ) . Indeed , in the SM and uterine lineages , induced expression of CKI-1 resulted in cells entering a CDKlow G0 state , with mean DHB ratios of 0 . 10 ± 0 . 05 and 0 . 12 ± 0 . 05 , respectively ( Figure 2G ) , as compared to control animals that lacked heat shock-induced expression ( SM: 0 . 99 ± 0 . 82 , uterine: 0 . 71 ± 0 . 35 ) or lacked the inducible transgene ( SM: 0 . 96 ± 0 . 77 , uterine: 1 . 00 ± 0 . 37 ) . Thus , induced G0 arrest by ectopic expression of CKI-1 is functionally equivalent , by CDK-activity levels , to the G0 arrest that occurs following mitotic exit in an unperturbed cell destined to undergo quiescence . We next examined the divisions of the 1°- and 2°-fated VPC lineage . The C . elegans vulva is derived from three cells ( P5 . p–P7 . p ) , which undergo three rounds of cell division during the L3 and early L4 larval stages ( Figure 3A and B; Katz et al . , 1995; Sternberg and Horvitz , 1986; Sulston and Horvitz , 1977 ) . Rather than giving rise to 24 cells , the two D cells , the innermost granddaughters of the 2°-fated P5 . p and P7 . p , exit the cell cycle one round early . This results in a total of 22 cells , which comprise the adult vulva ( Katz et al . , 1995; Sulston and Horvitz , 1977 ) . Quantification of DHB::GFP ratios during VPC divisions yielded the expected pattern . The daughters of P5 . p–P7 . p all exited their first division into a CDKinc state ( Figure 3C and D ) . After the next division , the 12 granddaughters of P5 . p–P7 . p ( named A–F symmetrically ) are born , including the D cell ( Katz et al . , 1995; Sulston and Horvitz , 1977 ) . At this division , the strong nuclear localization of DHB::GFP in the D cell was in stark contrast to the remaining proliferating VPCs . The D cell exited into and remained in a CDKlow state , while the remaining VPCs exited into a CDKinc state and continued to progress through the cell cycle ( Figure 3C and D; Figure 3—video 1 ) . All remaining VPCs exited into a CDKlow state at their terminal division . Consistent with these results , bootstrap analyses ( Figure 3—figure supplement 1A–G ) support our qualitative results , such that we can quantitatively distinguish between a cell that has completed mitosis and will continue to cycle ( CDKinc ) from a cell that exits mitosis and enters a G0 state ( CDKlow ) . In mammalian cell culture , endogenous levels of p21Cip1 during G2 are predictive of whether a cell will go on to divide or enter quiescence , senescence , or terminal differentiation ( Hsu et al . , 2019; Moser et al . , 2018; Overton et al . , 2014; Spencer et al . , 2013 ) . This raises the intriguing possibility that endogenous levels of CKI-1 in C . elegans correlate with CDKlow or CDKinc activity . To co-visualize CKI-1 dynamics with our CDK sensor , we inserted a N-terminal GFP tag into the endogenous locus of cki-1 via CRISPR/Cas9 and introduced a DHB::2xmKate2 variant of the sensor ( devoid of histone H2B ) into this genetic background . Since endogenous levels of GFP::CKI-1 were too dim for time-lapse microscopy , likely due to its short half-life ( Yang et al . , 2017 ) , we collected a developmental time series of static images over the L3 and L4 larval stages to characterize GFP::CKI-1 levels during pre-terminal and terminal divisions in the VPC lineage . We detected generally low levels of GFP::CKI-1 at the Pn . p 2 cell stage ( Figure 4A–C , Figure 4—figure supplement 1A–C ) . In their daughter cells , at the Pn . p 4 cell stage , we detected an increase in GFP::CKI-1 levels in cycling cells prior to their next cell division , peaking in G2 ( Figure 4A , B and D , Figure 4—figure supplement 1A–C ) . Notably , the D cell , which becomes post-mitotic after this cell division , exits mitosis with higher levels of GFP::CKI-1 than its CD mother ( Figure 4B , Figure 4—figure supplement 1B ) . This trend holds true for the remaining VPCs at the Pn . p 6 cell and 8 cell stage , which show high levels of GFP::CKI-1 that peak immediately after mitotic exit and remain high during the post-mitotic L4 stage ( Figure 4A , B , E and F , Figure 4—figure supplement 1A–C ) . We also observed increasing levels of GFP::CKI-1 in the G2 phase of mother cells that peak in their quiescent daughter cells in the uterine ( Figure 4—figure supplement 1D ) and SM cell lineages ( Figure 4—figure supplement 1E ) . Thus , levels of GFP::CKI-1 increase in mother cells and remain high upon mitotic exit in daughter cells with CDKlow activity . These results suggest that the proliferation-quiescence decision is already underway in the G2 phase of the previous cell cycle and correlates with CKI-1 levels in the mother cell . During our collection of static images of GFP::CKI-1 animals , we observed significant deviations in the expected VPC lineage pattern in the early L4 larval stage . In particular , we noted that many cells appeared to bypass their final division and undergo early cell-cycle quiescence with coincident high levels of GFP::CKI-1 and low DHB ratios . We hypothesized that the line we generated could be behaving as a gain-of-function mutant , as GFP insertions at the N-terminus could interfere with proteasome-mediated protein degradation of CKI-1 ( Bloom et al . , 2003 ) . The penetrance of this early cell-cycle quiescence defect varied across VPC lineages . While the A ( 2% of cases observed ) and E ( 3% of cases observed ) lineages showed a low penetrance of this defect , the B ( 26% of cases observed ) and F ( 58% of cases observed ) lineages showed a moderate penetrance ( Figure 4G and H ) . We speculate that the A and E lineages are largely insensitive to the gain-of-function mutant because CKI-2 , an understudied paralog of CKI-1 , may be the dominant CKI in these cells . The C cell , sister to the D cell , had a highly penetrant early cell-cycle quiescence defect ( 98% of cases observed; Figure 4G and H ) . Consistent with our finding that high levels of endogenous GFP::CKI-1 can lead to early cell-cycle quiescence , heat shock-induced CKI-1 expression uniformly drove VPCs at the Pn . p 2 cell stage into a CDKlow G0 state with mean DHB ratios of 0 . 11 ± 0 . 05 ( Figure 4I ) , as compared to control animals that lacked heat shock-induced expression ( 0 . 46 ± 0 . 87 ) or lacked the inducible cki-1 transgene ( 0 . 47 ± 0 . 42 ) . Strikingly , most of the VPCs of heat shocked larvae that were allowed to recover for 5 hr remained quiescent ( 0 . 35 ± 0 . 29 ) ( Figure 4I ) , as VPCs that received an effective pulse of CKI-1 failed to divide again . We observed an average of 6 . 86 VPCs present hours later at the L4 stage as opposed to the wild type vulva composed of 22 total cells ( Figure 4—figure supplement 1F and G ) . Together , these results demonstrate that cycling cells are highly sensitive to levels of CKIs and that increased expression can induce a G0 state . A strength of C . elegans is the organism’s robust ability to buffer external and internal perturbations to maintain its invariant cell lineage . However , not all cell divisions that give rise to the 959 somatic cells are completely invariant . Studies have identified several lineages , including the vulva , where environmental stressors , genetic mutations and/or genetic divergence of wild isolates leads to stochastic changes in a highly invariant cell fate pattern ( Braendle and Félix , 2008; Hintze et al . , 2020; Katsanos et al . , 2017 ) . Thus , we wondered if the CDK sensor generated here could be utilized to visualize and predict stochastic lineage decisions during C . elegans development . The VPC lineage that gives rise to the adult vulva is invariant under most conditions ( Figure 5A , Figure 5—figure supplement 1A; Sulston and Horvitz , 1977 ) . However , at high temperatures it has been observed that the D cell , the inner-most granddaughter of P5 . p or P7 . p , will go on to divide ( Figure 5A; Sternberg , 1984; Sternberg and Horvitz , 1986 ) . Unexpectedly , we noticed a rare occurrence of D cells expressing elevated DHB ratios during the course of time-lapse analysis of VPC divisions captured under standard laboratory conditions . To determine the penetrance of the cycling D cell phenotype , we inspected each of our CDK sensor lines grown at 25°C , a high temperature that is still within normal range for C . elegans . In both strains we observed a cycling D cell with a 4–6% penetrance ( Figure 5B , Figure 5—figure supplement 1B ) . To test whether this cycling D cell phenotype resulted from the presence of the DHB transgene or environmental stressors , such as temperature fluctuation , we examined the VPC lineage in animals lacking the CDK sensor at 25°C and 28°C . At 25°C , we observed a low penetrance ( 2% ) of cycling D cells in a strain expressing an endogenously tagged DNA licensing factor , CDT-1::GFP ( Figure 5B , Figure 5—figure supplement 1B ) , which is cytosolic in cycling cells ( Matus et al . , 2014; Matus et al . , 2015 ) . From lineage analysis , L2 larvae , expressing a seam cell reporter ( scm>GFP ) , that were temperature shifted from 20°C to 28°C displayed approximately a 30% occurrence of extra D cell divisions ( Figure 5—figure supplement 1C–E ) . Lastly , we wanted to determine whether D cells that show CDKinc activity divide . To accomplish this , we collected time-lapses of DHB::GFP animals grown at 25°C . These time-lapses revealed 10 occurrences of D cells born into a CDKinc rather than a CDKlow state ( Figure 5C and D; Figure 5—video 1 ) . In all 10 cases , the CDKinc D cell goes on to divide ( Figure 5—figure supplement 1A ) . Thus , we find that CDK activity shortly after mitosis is a predictor of future cell behavior , even in rare stochastic cases of extra cell divisions in C . elegans , an organism with a well-defined cell lineage . To investigate the predictive capability of DHB ratios in zebrafish , we generated two CDK sensor lines with different fluorescent protein combinations , DHB-mNeonGreen ( DHB-mNG ) and DHB-mScarlet ( DHB-mSc ) with H2B-mSc and H2B-miRFP670 , respectively , to allow for flexibility with imaging and experimental design ( Figure 6A ) . Both transgenes are under the control of the hsp70l heat shock-inducible promoter , which produces robust ubiquitous expression after shifting the temperature from 28 . 5 to 40°C for 30 min ( Figure 6B; Halloran et al . , 2000; Shoji et al . , 1998 ) . We also generated a transgenic line , Tg ( ubb:Lck-mNG ) , that ubiquitously labels the plasma membrane with mNG , which we crossed into the HS:DHB-mSc-2A-H2B-miRFP670 line to simultaneously visualize CDK activity ( DHB-mSc ) , segment nuclei ( H2B-miRFP670 ) and segment the plasma membrane ( LCK-mNG ) ( Figure 6A ) . To verify that DHB localizes in a cell-cycle-dependent manner in both CDK sensor lines , we first used time-lapse microscopy and quantified DHB ratios across cell divisions in the tailbud of bud or 22 somite-stage embryos ( Figure 6C and D , Figure 6—figure supplement 1A ) . We observed the expected localization pattern for both CDK sensor lines , with maximal nuclear exclusion of the sensor shortly before mitosis in G2 ( 3 . 42 ± 0 . 56 ( DHB-mNG ) and 6 . 57 ± 2 . 00 ( DHB-mSc ) ) and low ratios ( 0 . 69 ± 0 . 17 ( DHB-mNG ) and 0 . 51 ± 0 . 21 ( DHB-mSC ) ) representing nuclear accumulation of the sensor shortly after mitosis in G1 ( Figure 6F , Figure 6—figure supplement 1A ) . To establish the DHB ratio for S phase we visualized PCNA-GFP in the tailbud of DHB-mSC embryos as PCNA forms puncta in the nucleus at S phase entry and returns to a uniform nuclear distribution in G2 ( Figure 6E; Leonhardt et al . , 2000; Leung et al . , 2011 ) . Approximately 38 . 5 min after puncta formation , corresponding to mid-S phase , the DHB ratio is 1 . 36 ± 0 . 36 , which is significantly higher than the G1 DHB value ( 0 . 51 ± 0 . 21; Figure 6F ) . Thus , we conclude that both CDK sensor lines localize in a cell-cycle-dependent fashion , and that quantitative measurements can be used to determine interphase states . Next , using both DHB transgenic lines , we examined CDK activity in a number of defined embryonic tissues . Imaging of the developing tailbud revealed cells in all phases of the cell cycle with mean DHB ratios of 1 . 95 ± 1 . 74 ( mNG ) and 1 . 67 ± 2 . 05 ( mSC ) ( Figure 7A and B , Figure 7—figure supplement 1A ) . The tailbud of vertebrate embryos contain neuromesodermal progenitors ( NMPs ) ( Martin , 2016 ) , which in zebrafish have been reported to be predominantly arrested in the G2 phase of the cell cycle ( Bouldin et al . , 2014 ) . Consistent with this , we observed cells with high CDK activity in the tailbud ( orange arrows; Figure 7A , Figure 7—figure supplement 1A ) . This enrichment is eliminated when embryos are treated with the CDK4/6 inhibitor palbociclib , leading to a significant increase of cells in the tailbud with low CDK activity ( 0 . 58 ± 0 . 3 ) , similar in range to the G1/G0 values we measured during time-lapse ( 0 . 69 ± 0 . 17; Figure 7C–D ) . We also made the surprising observation that primitive red blood cells in the intermediate cell mass of 24 hr post-fertilization ( hpf ) embryos , which are nucleated in zebrafish , display high CDK activity ( 3 . 00 ± 0 . 97 ) indicating that they are likely in the G2 phase of the cell cycle ( Figure 7—figure supplement 1E and F ) , suggesting that cell-cycle regulation may be important for hematopoiesis ( Brönnimann et al . , 2018; De La Garza et al . , 2019 ) . To examine differences between proliferating and quiescent cells , we examined CDK activity in the somites , which are segmental mesodermal structures that give rise to skeletal muscle cells and other cell types ( Martin , 2016 ) , and adaxial cells , cells positioned at the medial edge of the somite next to the axial mesoderm ( Figure 7F and G ) . The adaxial cells are the slow muscle precursors and are considered to be in a quiescent state through the cooperative action of Cdkn1ca ( p57 ) and MyoD ( Osborn et al . , 2011 ) . In the most recently formed somites at 24 hpf , cells can be observed in all phases of the cell cycle ( Figure 7F and K ) . Consistent with what we observed in the tailbud , treatment with palbociclib also caused somite cells to arrest with low CDK activity in G1/G0 ( 0 . 33 ± 0 . 42; Figure 7—figure supplement 1B–D ) . As opposed to the majority of cells in the lateral regions of recently formed somites , adaxial cells possess low CDK activity ( 0 . 13 ± 0 . 04; Figure 7K ) . At later stages , the majority of cells in the lateral regions of the somite will differentiate into fast skeletal muscles fibers , which are also considered to be in a quiescent state ( Halevy et al . , 1995 ) . Examination of DHB ratios in 72 hpf skeletal muscle fibers revealed they have low CDK activity ( 0 . 14 ± 0 . 04 ( mNG ) and 0 . 13 ± 0 . 04 ( mSc ) ) , similar to the adaxial cells , but significantly different than the mean DHB ratios of undifferentiated cells at 24 hr ( 0 . 82 ± 0 . 70 ( mNG ) and 0 . 99 ± 0 . 084 ( mSc ) ; Figure 7H and K , Figure 7—figure supplement 1G–I ) . Thus , from our static imaging , we can identify cell types with low CDK activity that are thought to be quiescent . We next sought to determine if we can differentiate between the G1 and G0 state based on ratiometric quantification of DHB . We compared adaxial cells to notochord progenitor cells , which are held transiently in G1/G0 before re-entering the cell cycle upon joining the notochord ( Figure 7I; Sugiyama et al . , 2014; Sugiyama et al . , 2009 ) . Notably , the mean DHB-mNG ratio of the notochord progenitors ( 0 . 32 ± 0 . 08 ) is significantly higher than the DHB-mNG ratio of the quiescent adaxial cells ( 0 . 13 ± 0 . 04; Figure 7F and J ) . This elevated DHB ratio is consistent in notochord progenitors at two other earlier developmental stages , 90% epiboly ( 0 . 28 ± 0 . 08 ) and 18 somites ( 0 . 27 ± 0 . 09; Figure 7K ) . The mean DHB-mSc ratio in the notochord progenitors ( 0 . 33 ± 0 . 08 ) is also significantly different than the differentiated epidermis ( 0 . 13 ± 0 . 03; Figure 7J , Figure 7—figure supplement 1K–L ) . Based on this difference in DHB ratios between notochord progenitors and differentiated cell types , including muscle and epidermis ( Figure 7K ) , and our knowledge of the normal biology of these cells , we conclude that the CDK sensor can infer cell cycle state in the zebrafish , as it can distinguish between a cycling G1 state and a quiescent G0 state . We next investigated whether zebrafish cells separate into G1/CDKinc and G0/CDKlow populations as they do in the nematode C . elegans and whether these CDK-activity states are a general predictor of future cell behavior in both animals . First , we plotted all of the time-lapse CDK sensor data we collected in C . elegans ( Figure 8A and B ) and zebrafish ( Figure 8C ) . For C . elegans , plotting of all CDK sensor trace data , irrespective of lineage , demonstrated that cells entering a CDKlow state after mitosis corresponded to quiescent cells , while cells that exited mitosis into a CDKinc state corresponded to cells from pre-terminal divisions . For zebrafish , in a lineage agnostic manner , we plotted all the traces from the tailbud . We classified cells as CDKlow that remained below 0 . 19 , the upper bound of the DHB ratio for quiescent adaxial cells at this stage of development ( Figure 7K ) , for three or more frames post-anaphase . Indeed , we found that these traces could also be classified into CDKlow and CDKinc populations ( Figure 8C ) . In addition to a fast cycling population of cells at this developmental stage , we also identified cells that maintain a CDKinc DHB ratio but appear to stay in a prolonged G1 phase , potentially representing a slow cycling population of cells . As we were able to detect a rare stochastic lineage change in the C . elegans vulval lineage ( Figure 5 ) , we selected all CDK sensor trace data from the C . elegans VPCs ( Figure 8—figure supplement 1A ) and used this data to build a classifier to predict proliferative ( G1 ) versus quiescent ( G0 ) cell fates based on CDK activity after anaphase ( Figure 8—figure supplement 1A–C ) . Cross-examining our modeling with the known VPC lineage demonstrated that at 20 min after anaphase we had 85% accuracy in predictions with near-perfect prediction 60 min after anaphase ( Figure 8—figure supplement 1B ) . To test the predictive power of the classifier , we analyzed CDK trace data from the births of C and D cells , where some D cells stochastically divide ( Figure 5 ) . Our classifier correctly predicted cell fate 92% ( n = 24/26 single-cell traces ) of the time , including the two occurrences of a stochastic mitotic D cell in the data set ( Figure 8D ) . Finally , to determine whether we could predict future cell behavior independent of cell type , we trained a new classifier using 75% of all collected C . elegans time-lapse trace data from the SM , uterine , and VPC lineages . We used the remaining 25% of traces as test data . When cross-referenced with the known C . elegans lineage , our cell-type agnostic classifier correctly predicted the difference between a CDKinc proliferative cell and a CDKlow quiescent cell 93% ( 62/67 ) of the time . Together , these results demonstrate that during development , cycling cells encounter a bifurcation in CDK activity following mitosis where they either: ( 1 ) increase in CDK activity and become poised to cycle , or ( 2 ) exit into a CDKlow state and undergo cell-cycle quiescence ( Figure 8E ) . Thus , we suggest a model where cells from developing tissue in C . elegans and zebrafish must cross an early commitment point in the cell cycle where these cells must make the decision to divide or enter G0 . The decision to undergo quiescence is crucial to tissue integration and organization and is in part likely controlled by the activity of evolutionarily conserved CKI ( s ) in the mother cell that control daughter cell CDK activity ( Figure 8E ) . We introduce here a CDK-activity sensor to visually monitor interphase and the proliferation-quiescence decision in real-time and in vivo in two widely used research organisms , C . elegans and zebrafish . This sensor , which reads out the phosphorylation of a DHB peptide by CDKs ( Hahn et al . , 2009; Spencer et al . , 2013 ) , allows for quantitative assessment of cell cycle state , including G0 . The use of FUCCI in zebrafish ( Bouldin and Kimelman , 2014; Sugiyama et al . , 2009 ) and past iterations of a CDK sensor in C . elegans ( Deng et al . , 2020; Dwivedi et al . , 2019; van Rijnberk et al . , 2017 ) and Drosophila ( Hur et al . , 2020 ) have been informative in improving our understanding of cell-cycle regulation of development , but have not addressed the proliferation-quiescence decision . The DHB transgenic lines generated in this study will allow researchers to distinguish G1 from G0 shortly after a cell has divided and directly study G0-related cell behaviors , such as quiescence , terminal differentiation , and senescence , in living organisms . Previously , CDK sensors have been used to distinguish between proliferative and quiescent cells in asynchronous mammalian cell culture populations ( Arora et al . , 2017; Cappell et al . , 2016; Gast et al . , 2018; Gookin et al . , 2017; Miller et al . , 2018; Moser et al . , 2018; Overton et al . , 2014; Spencer et al . , 2013; Yang et al . , 2015 ) . As mammalian cells complete mitosis , they are born into either a CDK2inc state in which they are more likely to divide again or a CDK2low state in which they remain quiescent . Here we have examined the CDK activity state of cells in an invertebrate with a well-defined and invariant lineage , C . elegans , and a vertebrate that lacks a defined cell lineage , the zebrafish . In both contexts , we can visually and quantitatively differentiate between cells that are in a CDKinc state following cell division and cells that are in a CDKlow state . Strikingly , in C . elegans these states precisely correlate with the lineage pattern of the three post-embryonic tissues we examined: the SM cells , uterine cells , and VPCs . Cells born into a CDKinc state represented pre-terminal divisions , whereas cells born into a CDKlow state were quiescent and represented cells that had undergone their terminal division . By distinguishing these two states in CDK activity , we were able to accurately identify shortly after cell birth a rare stochasticity that was first described through careful end-point lineage analysis nearly 36 years ago in the C . elegans vulval lineage ( Sternberg , 1984; Sternberg and Horvitz , 1986 ) . Further , statistical modeling demonstrated that we could predict future cell behavior with >85% accuracy in C . elegans just 20 min post-anaphase . From static imaging in zebrafish , we found that we could readily distinguish between CDKinc cells in G1 , such as notochord progenitors , which re-enter the cell cycle after joining the notochord , and quiescent tissues that contain CDKlow cells in G0 , such as skeletal muscle and epidermis . While analysis of time-lapse data did lead to the identification of cells born into either CDKinc or CDKlow states , we were unable to follow and quantify enough cell births to determine whether CDK activity at mitotic exit is also predictive of future proliferation behavior during zebrafish development . We attribute our inability to capture an adequate number of cell births largely to a combination of conventional confocal microscopy and manual cell tracking and quantification . Nonetheless , in both organisms the CDK sensor can be easily used to separate G1 from G0 without the need for multiple fluorescent reporters ( Bajar et al . , 2016; Oki et al . , 2015 ) or fixation followed by antibody staining for FACS analysis ( Tomura et al . , 2013 ) . The classic model of the Restriction Point , the point in G1 at which cells in culture decide to commit to the cell cycle and no longer require growth factors ( e . g . mitogens ) , is that mammalian cells are born uncommitted and that the cell-cycle progression decision is not made until several hours after mitosis ( Jones and Kazlauskas , 2001; Pardee , 1974; Zetterberg and Larsson , 1985; Zwang et al . , 2011 ) . An alternative model has been proposed in studies using single-cell measurements of CDK2 activity in asynchronous populations of MCF10A cells ( Spencer et al . , 2013 ) and other nontumorigenic as well as tumorigenic cell lines ( Moser et al . , 2018 ) . This model extends the classic Restriction Point model for cell cycle commitment . During the G2 phase of the cell cycle , the mother cell is influenced by levels of p21 and cyclin D and these levels affect the phosphorylation status of Rb in CDKlow and CDKinc daughter cells , respectively ( Min et al . , 2020; Moser et al . , 2018 ) . In CDKlow daughter cells , phospho-Rb is low , and these cells are still sensitive to mitogens . Whether cells in vivo coordinate cell-cycle commitment with levels of CKI and CDK over this extended Restriction Point was poorly understood . By first quantifying the cytoplasmic:nuclear ratio of the CDK sensor in time-lapse recordings of cell divisions in C . elegans somatic lineages , we were able to use DHB ratios as a proxy for CDK levels to distinguish two populations of daughter cells: the first being actively cycling cells in a CDKinc state ( G1 ) and the second being quiescent cells in a CDKlow state ( G0 ) . We then quantified cytoplasmic:nuclear ratio of the CDK sensor in time-lapse recordings of cell divisions in zebrafish and we were also able to distinguish two populations of daughter cells . As data from asynchronous cell culture studies suggest that the decision to commit to the cell cycle is made by the mother cell as early as G2 ( Moser et al . , 2018; Spencer et al . , 2013 ) , we wanted to determine if this same phenomenon occurred in vivo . To accomplish this , we endogenously tagged one of two CKIs in the C . elegans genome , cki-1 , with GFP using CRISPR/Cas9 . We paired static live-cell imaging of GFP::CKI-1 with DHB::2xmKate2 during vulval development . Similar to in vitro experiments ( Moser et al . , 2018; Spencer et al . , 2013 ) , we found that mother cells whose daughters are born into a CDKinc G1 state will divide again , expressing low levels of GFP::CKI-1 . In contrast , mother cells of daughters that will exit the cell cycle express a peak of GFP::CKI-1 in G2 which increases as daughter cells are born into a CDKlow G0 state . Thus , our data demonstrate that an extended Restriction Point exists in the cell cycle of intact Metazoa . Furthermore , the in vivo proliferation-quiescence decision can be predicted in C . elegans by CDK activity shortly after mitotic exit and , based on our gain-of-function studies , is highly sensitive to levels of CKI-1 shortly before and after the mother cell divides . We demonstrate here that the CDK sensor functions in both C . elegans and zebrafish to read out cell cycle state dynamically , and unlike other in vivo cell-cycle sensors , can distinguish between proliferative and quiescent cells within an hour of cell birth . As nematodes and vertebrates last shared a common ancestor over 500 million years ago , this suggests that the CDK sensor is likely to function in a similar fashion across Metazoa . With advances in time-lapse in vivo 4D imaging and machine learning methods that facilitate the collection and analyses of CDK sensor activity in 4D , we envision an increased demand for this tool to study cell-cycle-regulated biology in other animals . The broad functionality of the sensor will offer researchers a unique opportunity to dissect the relationship between cell cycle state and cell fate during normal development , cellular reprogramming , and tissue regeneration . Finally , as an increasing body of evidence suggests that cell cycle state impinges on morphogenetic events ranging from gastrulation ( Grosshans and Wieschaus , 2000; Murakami et al . , 2004 ) , convergent extension ( Leise and Mueller , 2004 ) and cellular invasion ( Kohrman and Matus , 2017; Matus et al . , 2015; Medwig-Kinney et al . , 2020 ) , this CDK sensor will provide the means to increase our understanding of the relationship between interphase states and morphogenesis during normal development and diseases arising from cell-cycle defects , such as cancer . This manuscript is accompanied by Supplementary file 1 , a spreadsheet of all reported p-values from statistical tests performed . Transgene insertion was performed via CRISPR/Cas9 genome engineering to generate single copy knock-ins to a known neutral locus on chromosome I or II using a self-excising cassette ( SEC ) -based method ( de la Cova et al . , 2017; Dickinson et al . , 2015 ) . Homologous repair templates and guide plasmids were graciously provided by Bob Goldstein , targeting the MosSCI integration sites ttTi4348 and ttTi5605 on chromosome I and II , respectively . CRISPR microinjection products were prepared using the PureLink HQ Mini Plasmid DNA Purification Kit from Invitrogen ( K210001 ) . An additional wash step was included prior to the final ethanol wash , using 650 µL of 60% 4 M guanidine hydrochloride ( Fisher Scientific , BP178-500; pH 6 . 5 , 40% isopropanol ) yielding a marked increase in knock-in efficiency . All purified microinjection products were stored at 4 °C . Injection mixes were freshly made before each round of injection . These mixes contain Cas9-sgRNA plasmids ( 50 ng/µL ) , homologous repair templates ( 50 ng/µL ) , and a co-injection marker ( pCFJ90 , 2 . 5 ng/µL ) . Injection mixes were injected into the gonads of young adult C . elegans N2 hermaphrodites . Successful integrants were identified in the F3 offspring of injected worms ( Dickinson et al . , 2015 ) . Injected young adult hermaphrodites of the relevant parent strain were then each individually transferred to a fresh OP50 plate and allowed to lay eggs for three days at 25 °C . On day 3 , 400 µL of a 5 mg/mL stock of hygromycin B ( Millipore , 400052 ) was added to the plates to a final plate concentration of 0 . 25 mg/mL . After five days of hygromycin B exposure , surviving dominant sqt-1 roller ( Rol ) worms were singled out onto fresh OP50 plates , checked for expression of the desired transgene/genomic edit and the presence of extrachromosomal array markers on a fluorescence dissecting microscope ( frame and automation: Zeiss Axio Zoom . V16 , light source: Lumencor SOLA light engine ) . The Rol phenotype was assessed for Mendelian inheritance , and if possible , the genomic edit was homozygosed . Once homozygosed , selectable markers ( hygromycin B resistance and dominant sqt-1 Rol phenotype ) were removed from the genome using heat shock-inducible Cre-Lox recombination via either a 3–4 hr heat shock at 34 °C or overnight ( 8–12 hr ) heat shock of large numbers of L1 and L2 stage animals at 26 °C in an air incubator . After two days , wild type worms were singled out one to a plate and progeny assessed for expression and homozygosity of the desired genomic insertion . Three transgenic lines were generated , including Tg ( ubb:Lck . mNeonGreen ) sbu107 , Tg ( hsp70l:DHB . mNeonGreen-p2a-H2B . mScarlet ) sbu108 , and Tg ( hsp70l:DHB . mScarlet-p2a-H2B . miRFP670 ) sbu109 . These lines were created using the Tol2 transposable element system ( Kawakami , 2004 ) . Zebrafish plasmids for generating transgenic lines were created using a tol2 plasmid vector containing the hsp70l promoter based on previous plasmids constructs ( Row et al . , 2016 ) . For the hsp70l:DHB . mNeonGreen-p2a-H2B . mScarlet plasmid , Gibson cloning was used to insert DNA encoding amino acids 994–1087 of human DHB fused to the N-terminus of mNeonGreen , followed by the P2A viral peptide sequence and human H2B with a C-terminal mScarlet fusion . The same method was used to generate hsp70l:DHB . mScarlet-p2a-H2B . miRFP670 , except mScarlet and miRFP670 were used instead of mNeonGreen and mScarlet , respectively . The tol2 hsp70l vector was also used to create the ubb:Lck . mNeonGreen plasmid . The hsp70l promoter was replaced with the ubb promoter ( Mosimann et al . , 2011 ) , followed by mNeonGreen with an N-terminal membrane targeting sequence from Mus musculus LCK ( amino acids MGCVCSSNPE ) . Each plasmid was co-injected with in vitro transcribed tol2 transposase mRNA . One nanoliter of injection mix containing 25 pg/nl of plasmid and 25 pg/nl of tol2 mRNA were injected into wild type zebrafish embryos at the 1 cell stage . Injected embryos were raised to adults and screened for germline transmission . One nanoliter of injection mix containing 25 pg/nl of HS-PCNA-GFP and 25 pg/nl of tol2 mRNA were injected into Tg ( hsp70l:DHB . mScarlet-p2a-H2B . miRFP670 ) sbu109 zebrafish embryos at the 1 cell stage . Synthetic DNAs were ordered as gBlocks from Integrated DNA Technologies ( IDT ) or gene fragments from Twist BioScience ( see Key Resources Table ) . The nucleotide sequence of DHB ( index 1 . 0 ) was codon-optimized for C . elegans somatic expression and the P2A sequence used in pWZ193 ( index 0 . 2; see KRT ) de-optimized to increase the efficiency of ribosome stalling ( Lo et al . , 2019; Redemann et al . , 2011 ) . The C . elegans rps-0 and rps-27 promoters and the pcn-1 promoter and coding sequence were all amplified from N2 genomic DNA . Sequences of all primers and synthetic DNAs are provided in the KRT . Synthetic gene fragments and amplified DNAs were cloned via Gibson Assembly ( Barnes , 1994; Gibson et al . , 2010; Gibson et al . , 2009 ) or NEBuilder HiFi into target plasmids . Constructs used for zebrafish transgenes were made from PCR products amplified from synthetic Twist BioScience gene fragment sequences followed by NEBuilder HiFi cloning . Human DHB and H2B sequences were used for making the DHB transgenes , and the human membrane targeting Lck sequence was used for the ubb:Lck . mNeonGreen transgene . All primers and synthetic gene fragment sequences are available in the KRT . Microinjections for C . elegans transgenesis were performed on an injection setup combining a Zeiss Axio Observer A1 inverted compound frame , EC Plan-Neofluar 40x/0 . 75 NA DIC objective and floating stage , with a Narashige manual micromanipulator and a picoliter injection system from Warner for fine control of delivered volume . Microinjection needles ( Sutter ) were pulled on a Sutter P-97 reconditioned and calibrated by Sutter . Zebrafish microinjections were performed on either a Leica S6e or a Zeiss Stemi 508 dissecting microscope using a Narishige manual micromanipulator and a Warner picoliter injecting system . Glass needles were pulled on a P-1000 puller from Sutter Instruments . RNAi was delivered by feeding E . coli strain HT115 ( DE3 ) expressing double-stranded RNA ( dsRNA ) to synchronized L1 stage strains . Transcription of dsRNA was induced with 1 mM isopropyl-b-D-1-thiogalactopyranoside ( Thermo Scientific , R0393 ) in bacterial cultures for 1 hr at 37 °C . After an hour , cultures were plated on NGM plates topically treated with 2 . 5 µl each of 30 mg/mL carbenicillin ( Alfa Aesar , J61949 ) and 10 µl of 1 M IPTG . The RNAi vector targeting cdk-1 was obtained from the Vidal RNAi library ( Rual et al . , 2004 ) . The empty vector L4440 was used as a negative control . RNAi vectors were verified by Sanger sequencing . For heat shock CKI-1 experiments , the following strains were used DQM406 ( hsp>CKI-1::BFP; rps-0>DHB::mKate2 ) and DQM394 ( rps-0>DHB::mKate2 ) . Synchronized L1 animals were plated on OP50 and allowed to develop to mid-L3 . Plates were then placed at 30 °C in an air incubator for 3 hr . Animals were then placed at 20 °C and allowed to recover from heat shock for 20–40 min before being mounted for static imaging . For assessing endogenous CKI-1 levels in Figure 4 , strain DQM586 ( GFP::CKI-1; rps-27>DHB::2xmKate2 ) was utilized . Briefly , L1 animals were synchronized via sodium hypochlorite treatment and plated on OP50 at 25 °C and analyzed at the P6 . p 2 cell , 4 cell , and 8 cell stages . DQM586 was superficially wild type , but several phenotypes , revealed by confocal microscopy and/or analyzed in this study ( e . g . the presence of larger somatic cells than normal in the L3 and L4 stages , including the anchor cell ) , led to the conclusion that the N-terminal GFP fusion ( which lacks a flexible linker ) resulted in animals displaying a gain-of-function effect of GFP::CKI-1 . Early cell cycle exit in the VPCs was determined by lineage analysis of each image and comparing the size of individual VPCs in GFP::CKI-1 animals to wild type . A VPC was considered to have undergone early cell cycle exit if it failed to divide ( larger nucleus than normal ) and showed strong nuclear localization of DHB::2xmKate2 consistent with a CDKlow state . Palbociclib ( PD-0332991 ) , a selective inhibitor of CDK4/6 , was purchased from MedChemExpress ( HY-A0065 ) . A 5 mM stock solution in embryo media was prepared and stored at −80 °C for up to six months . Prior to each experiment , palbociclib was thawed and diluted in embryo media to a final concentration of 50 µM . Control experiments were performed by treating zebrafish embryos with embryo media only . Embryos were placed in palbociclib at 16 somites for 5 hr at 22 °C . DHB measurements were performed blinded to avoid bias . All live-cell imaging of C . elegans and zebrafish , unless indicated otherwise , was performed on a custom-assembled spinning disk confocal microscope consisting of a Zeiss Axio Imager A2 frame , a Borealis modified Yokogawa CSU10 spinning disc , an ASI 150-micron piezo stage controlled by a MS2000 , an ASI filter wheel and a Hamamatsu ImagEM X2 EM-CCD camera . The imaging objective used for C . elegans imaging was a Plan Apochromat 100x/1 . 4 NA DIC objective ( Carl Zeiss ) . For zebrafish imaging , a Plan Apochromat 63x/1 . 0 NA water dipping objective ( Carl Zeiss ) was used . L3 stage C . elegans larvae shown in Figure 1D and D’ were imaged on a separate custom-assembled spinning disk confocal microscope consisting of an automated Zeiss frame , a Yokogawa CSU10 spinning disc , a Ludl stage controlled by a Ludl MAC6000 and an ASI filter turret attached to a Photometrics Prime 95B camera . The imaging objective used was a Plan Apochromat 63x/1 . 4 NA DIC objective ( Carl Zeiss ) . For both aforementioned microscopes , laser illumination was provided by a six-line , 405/442/488/514/561/640 nm Vortran laser merge driven by a custom Measurement Computing Microcontroller integrated by Nobska Imaging , Inc Both microscopes were controlled with Metamorph software ( version: 7 . 10 . 2 . 240 ) and laser power levels were set with Vortran’s Stradus VersaLase eight software . In Figure 1—figure supplement 1A , live imaging of C . elegans embryos was performed on a Nikon Ti-E inverted microscope using a Plan Apochromat 60x/1 . 4 NA oil immersion objective and controlled by Nikon’s NIS-Elements software ( version: 4 . 30 ) . Images were acquired with an Andor Ixon Ultra back thinned EM-CCD camera using 488 nm or 561 nm imaging lasers and a Yokogawa X1 confocal spinning disk head equipped with a 1 . 5 Å magnifying lens . For time-lapse imaging of the C . elegans germline and embryos in Figure 1C and Figure 1—figure supplement 1C and Figure 1—video 1 , recordings were acquired using a Yokogawa CSUW1 SoRa spinning disk confocal in SoRa disk mode with 1 . 0x relay lens , a 60x/1 . 27 NA water immersion objective and a Prime 95B sCMOS camera mounted on a Nikon Ti-2 stand . Nikon’s NIS-Elements software ( version: 4 . 3 ) was used for image acquisition . For static imaging experiments , worms were anesthetized by mounting on a 7 . 5% noble agar pad containing sodium azide ( Sigma-Aldrich , S2002 ) ( Martinez and Matus , 2020; Matus et al . , 2015 ) . Time-lapse imaging of C . elegans was performed using a modified version of a previously published protocol ( Kelley et al . , 2017 ) . We substituted in a 24 mm square coverslip #1 . 5 ( Fisher Scientific , 12–541-B ) and divided the imaging agar pad into two asymmetric smaller portions ( each 2–3 mm squares ) , filling the void space under the coverslip with 5 mM levamisole in M9 buffer or M9 buffer alone . These modifications allowed for much longer imaging durations and substantially reduced sample Z-drift over the course of the imaging session on both upright and inverted microscope systems . Anesthesia was performed in a spot dish in ~50 µl of a 0 . 1% tricaine ( Sigma-Aldrich , E10521 ) /0 . 01% levamisole hydrochloride ( Sigma-Aldrich , L9756 ) anesthetic ( Kirby et al . , 1990; Maddox and Maddox , 2012; Wong et al . , 2011 ) . For some experiments , this tricaine-levamisole solution was substituted for 5 mM levamisole in M9 buffer . When levamisole was used alone , to maintain animals in an anesthetized state for long-duration time-lapse imaging , imaging chambers were flooded with 5 mM levamisole in M9 instead of M9 . Embryos for imaging ( Figure 1C , Figure 1—figure supplement 1A ) were collected by dissection from gravid hermaphrodites and incubated for 4–4 . 5 hr in M9 at room temperature ( Figure 1—figure supplement 1A ) or imaged immediately ( Figure 1C , Figure 1—video 1 ) . For live imaging , images were taken at a sampling rate of 0 . 5 µm . For time-lapse , z-stacks were collected every four ( Figure 1—figure supplement 1A ) or three min ( Figure 1C , Figure 1—video 1 ) . For time-lapse of the germline ( Figure 1—figure supplement 1C , Figure 1—video 1 ) , young adult worms were lightly immobilized using 0 . 1 mM levamisole in M9 buffer and mounted on 5% agarose pads . Zebrafish were mounted in a 35 mm glass bottom dish with uncoated #1 . 5 coverslip and 20 mm glass diameter ( MatTek ) . A thin layer of 1% agarose dissolved in embryo media ( Westerfield , 2007 ) , was added to the dish covering the glass bottom . Once solidified , a P10 pipette tip was used to punch holes in the agarose . Embryos were added to 1% low melting point agarose dissolved in embryo media containing 1x tricaine ( 24x stock 0 . 4 g/l; Pentair , TRS1 ) , and then one embryo was added to each of the punched holes . Embryos were manipulated gently with an eyelash while the agarose solidified to ensure proper orientation . For 72 hpf embryos , animals were anesthetized in 1x tricaine prior to mounting in 1% low melt agarose with 1x tricaine . In all cases imaging dishes were filled with embryo media containing 1x tricaine . Hand quantification of images was performed in Fiji ( version: 2 . 0 . 0-rc-69/1 . 52 p ) ( Schindelin et al . , 2012 ) . Due to the high level of amplifier noise in EM-CCD images , and to remove any remaining out-of-focus fluorescence in these confocal micrographs , a rolling ball background subtraction was used ( size = 50 ) ( Sternberg , 1983 ) . After a recording was qualified for inclusion , ratiometric measurements were obtained . First , the Z plane containing the center of the cell of interest was located . Using the freehand tool , a conservative toroid was drawn around the nucleus and excluding the nucleolus if present , which does not localize the CDK sensor . The fluorescent histone and corresponding DIC and DHB images were used to assess the accuracy of this toroid . A measurement of mean gray value was obtained . Then , a region of perinuclear cytoplasm was chosen , avoiding pixels belonging to the cytoplasm of neighboring cells . The mean gray value of the cytoplasmic patch was then measured . These values were recorded and a cytoplasmic: nuclear ratio was calculated . If there were multiple cells of interest in the image , the procedure was repeated for each cell . For time-lapse recordings , this procedure was repeated at each time point . To evaluate the predictability of CDK activity ( readout as the ratio of cytoplasmic-to-nuclear intensity of DHB ) on proliferative versus quiescent cell fate in different cell-cycle phases , we created a receiver operating characteristic ( ROC ) curve for CDK activity at each time point relative to anaphase . Using the perfcurve function in MATLAB , we calculated the area under the curves ( AUC ) as the indicator of predictability . We then built a classifier to predict proliferative vs . quiescent cell fates based on CDK activity after anaphase . For each time point , we chose the CDK-activity threshold for classification that maximizes the geometric mean of specificity ( 1 – false positive rate ) and sensitivity ( true positive rate ) . We tested the classifier in a second dataset , the stochastic division of the vulval D cell ( see Figure 5 ) . To predict the cell fate of each trace , we made independent classifications on each relevant time point based on CDK activity and use the majority class of all relevant time points as the classification for the trace . For traces recorded beyond 60 min after anaphase , we used all time points after 60 min post-anaphase , since these time points allow near-perfect prediction ( AUC>0 . 9 ) . For traces recorded beyond 20 min but within 60 min after anaphase , we used their last three time points , since these time points show good and increasing prediction power with AUC>0 . 8 . Bootstrapping was performed in MATLAB R2019A . The code used is available at GitHub ( https://github . com/abraham-kohrman/matus-dhb-stats; Kohrman , 2020; copy archived at swh:1:rev:9c88bc74fa1ca0793b2ee9598d1842a482581400 ) . Custom code for statistical testing may not be compatible with MATLAB releases older than R2019A and may require the use of MATLAB Toolboxes . Briefly , when single timepoint samples did not exhibit normal distributions , empirical statistics were calculated . For single timepoint experiments , a bootstrapped distribution of the difference between mean groups was calculated for each comparison ( Equation 1 ) . ( 1 ) \|x´1−x´2\| 108 statistical simulations were performed by random sampling without replacement in MATLAB . A p-value was calculated by determining the proportion of simulated differences with values greater than the true difference . For comparisons of time course data , a mean trend line was calculated for each dataset to be compared . The area between the mean trend lines was calculated . In MATLAB , this was performed as the sum of the absolute value of the difference at each time point . Where x1 corresponds to the first trend line and x2 corresponds to the second trend line . ( 2 ) ‖x1−x2‖1 Statistical simulations were performed by random partitioning of the data without replacement into two groups with the same sizes as the original groups . Mean trend lines were then calculated for these randomly assigned groups , and as before the statistic was calculated . 108 simulated replicates were performed to estimate the distribution of the difference statistic . In a manner analogous to bootstrapping , p-value was calculated by determining the proportion of simulations with more extreme statistical values than the observed statistic . See Figure 8—figure supplement 1 for a detailed schematic of the procedure . The α value for this study was nominally 0 . 05 , however exact p-values and n ( number of cells ) are reported in all cases . When no simulation produced a more extreme result than the true data configuration , p-values are reported as p<1×10−7 , rather as the true probability value is so small , as to be outside the range of accurately calculable probability values . For every comparison performed , plots of distributions of empirically calculated statistics are available upon request . To interpret p-values as presented , it is important to note our null hypothesis which can be formulated as: The categorization ( e . g . into C lineage vs . D lineage cells or treated vs untreated cells ) is not better than random . In short , the p-values we have corresponded to the probability that the difference between the mean or mean trend lines arose by chance . Another formulation would be the odds that the categorization of the data is meaningless . Throughout the study , an α value of 0 . 05 is used for significance . A p-value of 0 . 05 corresponds to the statement that 95% of random reassortments of the data yielded a difference between the means/mean trend lines less extreme than the true , observed difference . In the course of data collection for this manuscript , many animals were recorded that were not included in this manuscript . In order to be considered for analysis , recordings had to satisfy the following criteria: ( 1 ) a cell of interest had to have been present in the recording , ( 2 ) the cell of interest must have exhibited at least one anaphase during the recording , and ( 3 ) the animal must have appeared phenotypically normal at the beginning and end of the recording . Additional criteria for exclusion were the presence of a stalled metaphase plate at any point in the Video or unexpected developmental arrest . For data analysis , two workstation computers were used . Both systems boot into Windows 10 ( Microsoft ) off a 1 TB M . 2 drive ( Samsung 970 EVO Plus ) . The first system consists of an I9-9900X processor ( Intel ) , a GeForce GTX 1070 Ti GPU ( Nvidia ) and 128 GB of DDR4 RAM ( Corsair ) . The second system has an I9-9900K processor ( Intel ) , a GeForce RTX 2070 GPU ( Nvidia ) and 64 GB of DDR4 RAM ( G . Skill Ripjaws ) . Data were stored on a 4 TB RAID0 array consisting of two 2 TB drives ( Samsung ) and a 2 TB RAID0 array consisting of two 1 TB Drives ( Samsung ) , respectively . System integration , support and maintenance performed by Nobska Imaging , Inc . Data for figures were plotted in GraphPad Prism ( version: 8 . 1 . 2 ) . Micrographs in all figures were reviewed and selected in Fiji . Figure micrographs were contrast and brightness adjusted for ease of display in Adobe Photoshop CC ( version: 20 . 0 . 6 ) or Fiji . Figures were assembled in Adobe Illustrator CC ( version: 23 . 0 . 26 ) . Supplemental Videos were selected in Fiji and clipped to the desired length . The plane of interest was selected , and a time-lapse montage of channels was created . Time-lapse Videos were rotated to standard orientation , cropped to the relevant region and timestamps and scale bars annotations were added . Brightness and contrast were adjusted for ease of viewing . Videos showing more than one channel were assembled using the multi-stack montage plugin ( https://github . com/BIOP/ijp-multi-stack-montage ) .	All living things are made up of cells that form the different tissues , organs and structures of an organism . The human body , for example , is thought to consist of some 37 trillion cells and harbor over 200 cell types . To maintain a working organism , cells divide to create new cells and replace the ones that have died . Cell division is a tightly controlled process consisting of several steps , and cells continuously face a Shakespearean dilemma of deciding whether to continue dividing ( also known as cell proliferation ) or to halt the process ( known as quiescence ) . This difficult balancing act is critical during all stages of life , from embryonic development to tissue growth in an adult . Problems in the underlying pathways can result in diseases such as cancer . Cell division is driven by proteins called CDKs , which help cells to complete their cell cycle in the correct sequence . To gain more insight into this complex process , scientists have developed tools for monitoring CDKs . One such tool is a fluorescent biosensor , a molecule that can be inserted into cells that glows and moves in response to CDK activity . The biosensor can be studied and measured in each cell using a microscope . Adikes , Kohrman , Martinez et al . adapted and optimized an existing CDK biosensor to help study cell division and the switch between proliferation and quiescence in two common research organisms , the nematode Caenorhabditis elegans and the zebrafish . Analysis of this biosensor showed that CDK activity at the end of cell division is higher if the cells will divide again but is low if the cells are going to become quiescent . This could suggest that the decision of a cell between proliferation and quiescence may happen earlier than expected . The optimized biosensor is sensitive enough to detect these differences and can even measure variations that influence proliferation in a region on C . elegans that was once thought to be unchanging . The development of this biosensor provides a useful research tool that could be used in other living organisms . Many research questions relate to cell division and so the applications of this tool are wide ranging .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology", "tools", "and", "resources" ]	2020	Visualizing the metazoan proliferation-quiescence decision in vivo
Odorant binding proteins ( Obps ) are remarkable in their number , diversity , and abundance , yet their role in olfactory coding remains unclear . They are widely believed to be required for transporting hydrophobic odorants through an aqueous lymph to odorant receptors . We construct a map of the Drosophila antenna , in which the abundant Obps are mapped to olfactory sensilla with defined functions . The results lay a foundation for an incisive analysis of Obp function . The map identifies a sensillum type that contains a single abundant Obp , Obp28a . Surprisingly , deletion of the sole abundant Obp in these sensilla does not reduce the magnitude of their olfactory responses . The results suggest that this Obp is not required for odorant transport and that this sensillum does not require an abundant Obp . The results further suggest a novel role for this Obp in buffering changes in the odor environment , perhaps providing a molecular form of gain control . Like many animals , insects rely on their sense of smell to navigate through the environment towards food sources and mates . The Drosophila antenna is covered with olfactory sensilla that fall into three main morphological classes: basiconic , trichoid and coeloconic sensilla ( Figure 1A , B ) ( Shanbhag et al . , 1999 ) . Basiconic sensilla detect many fruit odors ( de Bruyne et al . , 2001; Hallem and Carlson , 2006; Hallem et al . , 2004 ) ; trichoid sensilla sense pheromones ( Clyne et al . , 1997; Dweck et al . , 2015; Ha and Smith , 2006; van der Goes van Naters and Carlson , 2007 ) ; coeloconic sensilla detect organic acids and amines ( Abuin et al . , 2011; Ai et al . , 2010; Benton et al . , 2009; Silbering et al . , 2011; Yao et al . , 2005 ) . Each class can in turn be divided into functional types . For example , 10 types of basiconic sensilla , designated ab1 ( antennal basiconic 1 ) through ab10 , detect different subsets of odorants ( Couto et al . , 2005; de Bruyne et al . , 2001; Elmore et al . , 2003; Hallem et al . , 2004; Marshall et al . , 2010 ) . Olfactory sensilla are perforated by pores or channels through which odorants can pass , and they contain an aqueous lymph in which the dendrites of up to four olfactory receptor neurons ( ORNs ) are bathed ( Figure 1C ) . Many sensilla contain two ORNs , designated A and B based on their spike amplitudes . Sensilla also contain a thecogen ( sheath ) cell , a trichogen ( shaft ) cell , and one or two tormogen ( socket ) cells . These cells produce the lymph and wrap around the ORNs ( Shanbhag et al . , 2000 ) . 10 . 7554/eLife . 20242 . 003Figure 1 . Organization of the antenna and expression of Obps . ( A ) Scanning electron micrograph of antennae ( arrowheads ) on a Drosophila head . Adapted from http://www . sdbonline . org/sites/fly/aimain/images . htm , ( Menuz et al . , 2014 ) . ( B ) Higher magnification image of antennal surface showing basiconic ( B ) , trichoid ( T ) , and coeloconic ( C ) sensilla surrounded by non-innervated spinules ( Sp ) . Adapted from ( Menuz et al . , 2014; Shanbhag et al . , 1999 ) . ( C ) A generic sensillum containing two ORNs and thecogen ( Th ) , trichogen ( Tr ) , and tormogen ( To ) cells , separated from neighboring sensilla by epidermal cells ( E ) . Adapted from ( Menuz et al . , 2014; Steinbrecht et al . , 1992 ) . ( D ) Members of the Obp gene family detected in the third antennal segment , where olfactory sensilla are located , at >1 read per million mapped reads in each of three samples . Genes are listed in decreasing order of expression level , indicated in terms of reads per million mapped reads per kilobase of gene length ( RPKM ) . Obp76a is also known as LUSH . Adapted from ( Menuz et al . , 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 00310 . 7554/eLife . 20242 . 004Figure 1—figure supplement 1 . Obp expression levels . Members of the Obp gene family detected with at least one read per million mapped reads ( RPM ) in each of three samples in the third antennal segment of Canton-S ( CS ) flies . The 27 genes detected at this level are listed in decreasing order of reads per million mapped reads per kilobase of gene length ( RPKM ) . Adapted from ( Menuz et al . , 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 004 Two broad classes of proteins are believed essential for the responses of sensilla to odorants . One class , the odor receptors , have been mapped to specific types of sensilla and to individual neurons within those sensilla , and their functional specificities have been characterized ( Ai et al . , 2010; Benton et al . , 2009; Couto et al . , 2005; Hallem and Carlson , 2006; Hallem et al . , 2004; Silbering et al . , 2011 ) . The other class , called odorant binding proteins ( Obps ) , have been the subject of much investigation in many insects but remain poorly understood ( Leal , 2013; Pelosi et al . , 2006; Vogt and Riddiford , 1981 ) . Obps are remarkable in three ways: they are numerous , being encoded by a family of 52 genes in Drosophila; they are abundant , with some encoded by the most abundant mRNAs in the antenna; they are diverse , with members sharing only 20% amino acid identity on average ( Hekmat-Scafe et al . , 2002; Menuz et al . , 2014 ) . They are small proteins , on the order of 14 kDa , and despite their high sequence divergence they are believed to have a common structure ( Graham and Davies , 2002 ) . Many are found in the lymph of olfactory sensilla , where ORN dendrites are located ( Shanbhag et al . , 2001a ) . They bind odorants , with different degrees of affinity and selectivity reported for different Obps ( Gong et al . , 2010; Leal et al . , 2005 ) . Within a species , different Obps are expressed in different antennal sensilla ( McKenna et al . , 1994; Pikielny et al . , 1994; Schultze et al . , 2013 ) , and some are expressed in the taste system ( Galindo and Smith , 2001; Jeong et al . , 2013; Pikielny et al . , 1994; Shanbhag et al . , 2001b ) or in larval chemosensory organs ( Galindo and Smith , 2001; Park et al . , 2000 ) . Despite numerous studies of Obps , much remains to be learned about their function . They are widely believed to bind , solubilize , and transport hydrophobic odorants across the aqueous sensillum lymph to receptors in the dendrites ( Gomez-Diaz et al . , 2013; Sandler et al . , 2000; Vogt et al . , 1985; Wojtasek and Leal , 1999; Xu et al . , 2005 ) . Obps have also been proposed to accelerate the termination of odor response , by removing odorants from receptors or from the sensillar lymph ( Vogt and Riddiford , 1981; Ziegelberger , 1995 ) . A variety of studies support a role for Obps in olfactory perception in vivo ( Biessmann et al . , 2010; Pelletier et al . , 2010; Swarup et al . , 2011 ) . However , to date , the physiological role of only one Obp , Obp76a , has been thoroughly investigated in the olfactory system of Drosophila ( Gomez-Diaz et al . , 2013; Laughlin et al . , 2008; Xu et al . , 2005 ) . Obp76a , also called LUSH , is required in trichoid sensilla for normal response of the odor receptor Or67d to the pheromone cis-vaccenyl acetate ( cVA ) , although responses of Or67d to cVA have been detected in the absence of Obp76a ( Benton et al . , 2007; Gomez-Diaz et al . , 2013; Li et al . , 2014; van der Goes van Naters and Carlson , 2007 ) . LUSH has been found to bind cVA in vitro ( Kruse et al . , 2003; Laughlin et al . , 2008 ) , but also binds other insect pheromones ( Katti et al . , 2013 ) , short-chain alcohols ( Bucci et al . , 2006; Thode et al . , 2008 ) , and phthalates ( Zhou et al . , 2004 ) . One reason the role of Obps has remained unclear is that their molecular organization has not been established . Previous work has not defined which Obps are expressed in individual sensilla , nor in which sensilla individual Obps are expressed . Thus it has been difficult to carry out well-defined manipulations of Obp content . Here we take advantage of a recent RNA-Seq analysis that quantified Obp expression in the antenna ( Menuz et al . , 2014 ) . We examine the expression of the abundant Obps at the level of sensillum morphology , sensillum functional type , and cell type . The results illustrate basic principles of olfactory system organization , and they allow construction of an Obp-to-sensillum map . The map identifies a sensillum type that contains only one highly expressed Obp , Obp28a . We delete this Obp gene and analyze the effects physiologically . Surprisingly , we find that deletion of the only abundant Obp in this sensillum type does not reduce the magnitude of its ORN responses . Additionally , we find evidence suggesting that this Obp plays a role not previously demonstrated for Obps , in buffering sudden increases in odorant levels . This Obp could provide a molecular mechanism of gain control that acts prior to receptor activation . Together , this work provides a foundation for incisive studies of Obp function , suggests that some sensilla do not require an abundant Obp for odorant transport , and encourages a broader view of the functions performed by the large and diverse family of Obps . We systematically analyzed the expression patterns of the 10 Obps that are expressed most abundantly in the antenna . The remaining Obps are all expressed at much lower levels ( Figure 1D , Figure 1—figure supplement 1 ) . In situ hybridization was used to identify the regions of the antenna and the morphological classes of sensilla in which the abundant Obps are expressed . These Obps are expressed in a wide diversity of spatial patterns ( Figure 2A ) . Expression of some of the Obps was distributed broadly ( e . g . Obp19d ) , expression of another was concentrated narrowly ( Obp59a ) , and others showed intermediate patterns ( Obp76a ) . The expression levels of these 10 Obps are striking not only in their magnitude but also in their wide range , from ~30 , 000 RPKM to ~2000 RPKM in the third antennal segment ( Figure 1D ) ( Menuz et al . , 2014 ) . Consistent with the breadth of their spatial patterns , Obp19d is the most abundant transcript , and Obp59a is the least abundant of the 10 Obps . 10 . 7554/eLife . 20242 . 005Figure 2 . Diverse expression patterns of abundant Obps in the antenna . ( A ) In situ hybridization of Obp antisense RNA probes to antennal sections . Scale bar = 20 μm . Male and female antennae were examined with each probe and no sexual dimorphism was observed except that females appeared to show stronger labeling than males with Obp56d in the region where trichoid sensilla are found . ( B ) Higher magnification images of sensilla labeled with Obp probes: Obp19a in a basiconic sensillum ( B ) , Obp76a in a trichoid sensillum ( T ) , Obp84a in a coeloconic sensillum ( C ) , and Obp19d in cells that are located between sensilla and that are associated with uninnervated spinules ( Sp ) . Scale bar = 2 μm . ( C ) Summary of Obp expression patterns . +a indicates expression in the arista as well as in cells between sensilla of the third antennal segment . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 005 Expression was observed in each of the three major morphological classes of sensilla ( Figure 2B ) . Four , such as Obp19a , are expressed in basiconic sensilla; four , including Obp76a , are detected in trichoid sensilla; two , including Obp84a , are expressed in coeloconic sensilla ( Figure 2B , C ) . Some Obps are expressed only in basiconic sensilla , some only in trichoid sensilla , and some only in coeloconic sensilla , but two Obps were expressed in both basiconic and trichoid sensilla , consistent with an earlier report ( Hekmat-Scafe et al . , 1997 ) . Of the two Obps in coeloconic sensilla , Obp84a is expressed in most if not all coeloconic sensilla on the antennal surface as well as those of the sacculus , a three-chambered cavity of the antenna . Obp59a was detected only in the sacculus . The localization of Obp84a and Obp59a to coeloconic sensilla is thus consistent with their absence in the antennae of atonal , a mutant lacking coeloconic sensilla , as revealed by an RNA-Seq analysis ( Menuz et al . , 2014 ) . Obp19d and Obp56d appear to be expressed not in olfactory sensilla but rather in epidermal cells , some of which flank olfactory sensilla and some of which are associated with uninnervated spinules ( Figure 2B ) . Obp56d is also expressed in the arista , a feathery structure associated with thermosensation and mechanosensation ( Figure 2A ) ( Foelix et al . , 1989; Ni et al . , 2013 ) . Having mapped the abundant Obps to morphological classes of sensilla , we next mapped them at higher resolution , to functional types of sensilla . We focused on the basiconic sensilla , whose function has been analyzed in particular detail ( de Bruyne et al . , 2001; Hallem et al . , 2004 ) . Our goal was to determine which of the abundant Obps are expressed in each of 10 individual functional types of basiconic sensilla . We used 10 Or-GAL4 drivers , each chosen to label a particular type of basiconic sensillum . For example , Or42b-GAL4 labels an ORN located in the ab1 type , and Or59b-GAL4 labels an ORN in the ab2 type . We systematically carried out a double-label analysis with 10 Or-GAL4 drivers and in situ hybridization probes for each of the four Obps that are expressed in basiconic sensilla ( Figure 3 ) . 10 . 7554/eLife . 20242 . 006Figure 3 . Obps are differentially expressed within basiconic sensillum types . ( A ) Confocal images of in situ hybridization to antennal sections labeled with antisense probes for the Obps ( red ) and an antibody against GFP ( green ) driven by Or42b-GAL4 ( ab1 ) , Or59b-GAL4 ( ab2 ) , Or22a-GAL4 ( ab3 ) , Or56a-GAL4 ( ab4 ) , Or82a-GAL4 ( ab5 ) , Or49b-GAL4 ( ab6 ) , Or67c-GAL4; Or98a-GAL4 ( used together to label ab7 ) , Or43b-GAL4 ( ab8 ) , Or 67b-GAL4 ( ab9 ) , and Or49a-GAL4 ( ab10 ) . Many panels show more than one sensillum . Thus in some panels , such as Obp83a/ab2 , two sensilla are labeled by both the Or-GAL4 driver and the Obp probe . In some other panels , such as Obp28a/ab1 , Obp probes label multiple neighboring sensilla of which some are unlabeled by the GAL4 driver . Scale bar = 5 μm . ( B ) Summary of Obp expression in ten basiconic types . An Obp is considered to be expressed in a sensillum type if it labeled a cell that wraps around the dendrites of ORNs in the majority of labeled sensilla examined . Obp expression was more difficult to identify with confidence in ab9 because of its proximity to other sensilla with strong Obp expression . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 006 In this manner we found that ab1 , one of whose ORNs is labeled green in the left column of Figure 3A , contained a cell labeled with Obp83a ( red ) . Likewise , ab2 , ab3 , ab7 , and ab10 all express Obp83a; in each case the dendrite of the labelled ORN in the sensillum appears to be surrounded by a cell expressing Obp83a . By contrast , the other five types of basiconic sensilla ( ab4 , ab5 , ab6 , ab8 , and ab9 ) did not express detectable levels of Obp83a . The Obp83b gene , which lies less than 1 kb from Obp83a and encodes a protein with 68% amino acid identity to Obp83a ( Hekmat-Scafe et al . , 1997 ) , maps to the same set of basiconic sensilla . By contrast , Obp19a maps to a different subset of sensilla . There are sensilla that express Obp83a and b but not Obp19a , sensilla that express Obp19a but not Obp83a or b , sensilla that express all , and sensilla that express none ( Figure 3B ) . Interestingly , Obp28a is expressed in all 10 types of basiconic sensilla , suggesting that it could play a broad role in odor coding . In summary , different Obps map to different functional subsets of basiconic sensilla , and basiconic sensilla express distinct subsets of Obps . A conclusion of particular interest is that ab8 , which has been well characterized ( Elmore et al . , 2003 ) , expresses a single abundant Obp , as considered further below . We next examined Obp expression at still higher resolution: at the level of individual cell types . Olfactory sensilla contain not only neurons , but also thecogen , tormogen , and trichogen cells ( Shanbhag et al . , 2000 ) . We found no evidence for expression of any Obps in neurons: ( i ) we did not observe axons or dendrites in the cells labeled by any of the 10 Obp probes ( Figures 3 and 4 ) ; ( ii ) none of the Obp in situ hybridization probes co-labeled the neurons labeled by any of the 10 Or-GAL4 drivers in the double-labeling analysis ( Figure 3 ) ; ( iii ) we carried out additional double-label experiments with Obps and GAL4 drivers that label neurons in basiconic , trichoid , and coeloconic sensilla , and found no co-labeling ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 20242 . 007Figure 4 . Obps are expressed in different cell types . ( A ) Diagram of a generic sensillum containing ORNs , thecogen cells labeled with nompA-GAL4 , trichogen cells ( Tr ) , and tormogen cells labeled with ASE5-GAL4 , separated from neighboring sensilla by epidermal ( E ) cells . Adapted from ( Steinbrecht et al . , 1992 ) . ( B–D ) , ( F–J ) Confocal images of antennal sections labeled with Obp antisense probes ( red ) and an antibody against GFP ( green ) driven by the thecogen cell driver nompA-GAL4 ( B , D , F ) , and the tormogen cell driver ASE5-GAL4 ( C–D , G–J ) . Yellow indicates coexpression . ( E ) Summary of coexpression experiments . The dark , solid red rectangle indicates that Obp84a was co-expressed with the thecogen cell type marker sensilla consistently in those sensilla that express Obp84a . In many cases , indicated by light , stippled red rectangles , an Obp was co-expressed in a specific cell type in some but not all sensilla . An empty rectangle indicates that the Obp did not co-localize with that cell type marker in any sensilla examined . F , G Images of whole antennal sections . Obp83a is expressed in many cells that are not labeled by the thecogen ( F ) or tormogen cell driver ( G ) . ( H ) Two trichoid sensilla that each house one tormogen cell ( green ) and two Obp83b-expressing cells ( red ) . ( I ) Obp19a is coexpressed with the tormogen cell driver in one sensillum . The image shows the same cell as in ( C ) but in a different focal plane . ( J ) Obp19a is not coexpressed with the tormogen cell driver in another sensillum . Scale bars = 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 00710 . 7554/eLife . 20242 . 008Figure 4—figure supplement 1 . Lack of Obp expression in ORNs . Confocal images of antennal sections labeled with antisense probes for the indicated Obps ( red ) and an antibody against GFP ( green ) driven by neuronal GAL4 drivers . ( A ) The Obp28a probe labels all basiconic sensilla . Orco-GAL4 labels all ORNs in basiconic sensilla , as well as all ORNS in trichoid sensilla ( Larsson et al . , 2004 ) . Coexpression is not observed . ( B ) The Obp76a probe labels most if not all trichoid sensilla . Or67d-GAL4 labels an ORN in at1 sensilla ( Couto et al . , 2005 ) . Coexpression is not observed . ( C ) The Obp84a probe and IR8a-GAL4 label most coeloconic sensilla ( Abuin et al . , 2011 ) . Coexpression is not observed . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 008 We observed expression of Obps in thecogen and tormogen cells , as judged by double-label analysis using Obp probes and GAL4 drivers that have previously been used as markers of these cell types in other tissues ( Barolo et al . , 2000; Chung et al . , 2001; Jeong et al . , 2013 ) . For example , an in situ hybridization probe for Obp84a labels cells that express a marker of thecogen cells , nompA-GAL4 ( Figure 4A , B ) , and an Obp19a probe labels cells that express a marker of tormogen cells , ASE5-GAL4 ( Figure 4A , C ) . A systematic double-label analysis was carried out with all 10 Obps and the thecogen and tormogen cell markers ( Figure 4D ) . Besides Obp84a , only one other Obp , Obp28a , was coexpressed with the thecogen marker , and this Obp28a-thecogen coexpression was observed in a very limited number of cells per antenna . Six of the Obps were coexpressed with the tormogen marker ( Figure 4D , E ) . Obp84a , which showed strong labeling of thecogen cells , did not label tormogen cells . Obp76a did not label cells with either marker , suggesting that it is expressed in trichogen cells , a suggestion that we have been unable to confirm directly for lack of a suitable marker specific for trichogen cells in olfactory sensilla . Moreover , most of the Obp probes ( including Obp83a , Obp83b , Obp28a , Obp19a , Obp69a , Obp76a ) seem likely to be expressed in trichogen cells , based on the substantial number of cells they label in antenna sensilla that are not labeled by thecogen or tormogen markers; for example , Obp83a is not co-expressed with the thecogen marker ( Figure 4F ) and labels many cells that are not labeled by the tormogen marker ( Figure 4G ) . Many individual sensilla contain more than one Obp-labeled cell , as can be seen in each of the two neighboring sensilla shown in Figure 4H . This finding is consistent with the interpretation that some Obps label more than one cell type ( Shanbhag et al . , 2001a; Steinbrecht et al . , 1992 ) . Obp19d and Obp56d , which label cells located between sensilla , do not co-localize with either marker ( Figure 4D , E ) , consistent with the interpretation that they are expressed in epidermal cells . Finally , we note with interest that Obp19a labeled the same cell as the tormogen marker in some sensilla ( Figure 4I ) but a different cell in other sensilla ( Figure 4J ) . A simple interpretation of this result is that Obp19a is expressed in different cell types in different basiconic sensilla . The construction of an Obp-to-sensillum map provided an unprecedented opportunity to investigate Obp function in an incisive way . Of particular interest , the map shows that some sensilla express a single abundant Obp , Obp28a ( Figure 3B ) . We sought to remove Obp28a genetically , with the goal of examining olfactory physiology in a sensillum whose Obp content had been drastically reduced if not eliminated . Accordingly , we created a CRISPR/Cas9-mediated deletion of Obp28a ( Figure 5—figure supplement 1 and Supplementary file 1 ) and then outcrossed the deletion five times to a control genetic background . The mutation was verified by PCR analysis both before and after outcrossing , and we further confirmed the loss of Obp28a RNA by qPCR ( not shown ) . We then analyzed the olfactory response of mutant ab8 sensilla via single-unit electrophysiology to assess the effect of removing its only abundant Obp . Since Obps have been proposed to be required for the transport of hydrophobic odorants through the aqueous sensillum lymph , we first examined the effect of removing Obp28a on the detection of 1-octanol , an odorant that is highly hydrophobic ( logP = 3 . 07 ) . This odorant is found in citrus fruits and other plants , and it elicits modest responses from the receptors of both ab8A and ab8B neurons . Due to the difficulty of sorting spikes from the two ab8 neurons , we quantified the total number of spikes following stimulation . We were surprised to find that the mutant sensillum responded robustly to the odorant across a broad concentration range , despite the lack of its single abundant Obp ( red line in Figure 5A ) . The simplest interpretation of this result is that this sensillum can maintain a strong olfactory response with little if any Obp . Moreover , we were surprised to find that not only was the response robust , but that it was greater in the Obp mutant than in the control , over a broad concentration range . To investigate the effect of Obp depletion in more detail , we examined the dynamics of olfactory response . Rather than focusing on the total number of spikes in a single 0 . 5 s interval , as in Figure 5A , we examined the numbers of spikes in 50 ms intervals and plotted the results as a peri-stimulus time histogram ( PSTH ) ( Figure 5B ) . Again we observed a robust response in the mutant . The shape of the response was affected by Obp depletion , with the greatest effect occurring during the initial phase of the response . The peak response of the mutant was greater than that of the control across a broad range of 1-octanol concentrations . We note that the baseline firing rate was the same in the Obp mutant and the control ( Figure 5B , 500–1000 ms ) . 10 . 7554/eLife . 20242 . 009Figure 5 . Robust and increased response of ab8 to 1-octanol in an Obp28a mutant . ( A ) Dose-response curves of control ( black ) and Obp28a- ( red ) ab8 neuronal responses to a 0 . 5s pulse of 1-octanol . Responses were quantified by subtracting the spontaneous firing rate from the rate during the stimulus . The spike rates of both neurons were summed . ( B ) Peri-stimulus time histograms ( PSTHs ) of ab8 responses to increasing doses of 1-octanol . Gray boxes denote 0 . 5s stimulus presentations . The baseline firing rates of the mutant and control are comparable . Graphs display 2s time windows in 50 ms bins . Shaded areas surrounding each curve indicate SEM . * p<0 . 05 , p<0 . 01 , * p<0 . 001 , **** p<0 . 0001 , n = 12 . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 00910 . 7554/eLife . 20242 . 010Figure 5—figure supplement 1 . CRISPR mutant cloning and verification primers . Primers #1–3 were used for creating CRISPR Guide chiRNA , primers #4–7 were used for constructing the CRISPR donor plasmid , and primers #8–11 were used to verify Obp28a deletion in transgenic flies . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 010 The preceding analysis concerned a brief odor stimulus: 0 . 5 s . In nature , flies also experience sustained olfactory stimuli . For example , flies spend prolonged periods of time in direct contact with food sources , which are intense sources of odor . We were interested in the possibility that Obp28a might be essential for response to such prolonged stimuli or for the recovery therefrom . The extremely high levels of Obp28a expression might have evolved to enhance odor coding under extreme conditions of olfactory stimulation . We delivered a strong 1-octanol stimulus for 30 s . Consistent with our earlier results , the initial response of the mutant was greater than that of the control immediately following odor onset ( Figure 6A ) . During the ensuing long stimulus period , the response of the mutant was comparable to that of the control , supporting the notion that the ab8 sensillum is capable of a robust olfactory response to a prolonged stimulus in the absence of an abundant Obp . 10 . 7554/eLife . 20242 . 011Figure 6 . Altered responses of Obp28a in prolonged stimulus paradigms . ( A ) PSTH of ab8 responses to 30s presentations of a 10−1 . 5 dilution of 1-octanol in control ( black ) and Obp28a- ( red ) . Gray box indicates stimulus presentation . Shaded areas display SEM . The graph displays a 60 s time window in 1 s bins . • p<0 . 05 , n = 18; we note that for five of the bins , p<0 . 0001 . We also asked whether the increased initial response of the mutant , indicated by * , was significant when the responses were examined in 50 ms bins , as in Figure 5 , and found that p<0 . 05 . ( B ) Background firing rates in response to prolonged ( >5 min ) stimulation of a 10−2 . 5 dilution of 1-octanol . Each bar represents data prior to the administration of a short pulse of the indicated dose . ns = not significant , n = 16 . C Dose-response curves of ab8 neurons to increasing doses of 1-octanol superimposed on the 10−2 . 5 1-octanol background . Spike rates are calculated from the number of spikes during the stimulus period , without subtracting the background firing rate . **** p<0 . 0001 , n = 16 . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 011 We then terminated the odor stimulus and measured the response . If Obp28a played an essential role in clearing high levels of odorant from the sensillum lymph , one might expect the Obp28a+ control to show a faster decline in response than its Obp28a− counterpart . The opposite result was observed ( Figure 6A ) . These findings argue against the possibility that Obp28a plays an essential role in clearing 1-octanol from the sensillum lymph after intense stimulation . Having thus analyzed responses to both short and long odor stimuli , we then tested the response to short stimuli superimposed upon long stimuli . In nature , the olfactory system is often faced with the challenge of detecting an olfactory signal against a high background of odorant , for example in assessing local variation in the quality of a food source . Accordingly , we delivered a sustained background of 1-octanol and measured the response to a pulse of 1-octanol that was superimposed upon this background . More specifically , we kept the background at a constant level , and measured the responses to a series of superimposed 1-octanol pulses of increasing intensity . Three results were obtained . First , prior to the odor pulses , the background firing rate was the same in the mutant as in the control ( Figure 6B ) , consistent with the equivalent firing rates of mutant and control during the prolonged stimulus applied in Figure 6A . Second , the mutant responded robustly to superimposed pulses of odor , with a threshold between 10−3 and 10−2 dilutions ( Figure 6C ) . Third , the magnitude of the mutant firing level in response to a pulse was greater than that of the control at all doses above the threshold , consistent with our earlier findings with pulses delivered in the absence of a background stimulus ( Figure 5 ) . The preceding analysis used a highly hydrophobic odor that elicited modest responses ( <50 spikes/s ) even at high doses . We wanted to determine whether the depletion of an Obp would affect: ( i ) responses to a more hydrophilic odorant; ( ii ) strong responses; ( iii ) responses of both the ab8A and ab8B neurons . Although extensive screening ( n>170 odorants ) did not identify a highly hydrophobic odorant that strongly activated ab8A or ab8B ( not shown ) , the more hydrophilic odorants ethyl acetate ( logP = 0 . 73 ) and butyric acid ( logP = 0 . 79 ) elicit strong responses ( n>100 spikes/s ) from the receptors of ab8A and ab8B , respectively ( Hallem and Carlson , 2006 ) . We tested the response of the Obp28a mutant to 0 . 5 s pulses of butyric acid and ethyl acetate and found robust responses over a broad concentration range for both odorants ( red lines in Figure 7A and B ) . The mutant responded more strongly than the control to each odorant at one or more concentrations . Butyric acid elicited a stronger response from the mutant over a wide range of concentrations ( Figure 7A ) , and ethyl acetate elicited a stronger response from the mutant at one dose in the middle of its dynamic range ( Figure 7B ) . 10 . 7554/eLife . 20242 . 012Figure 7 . Obp28a mutants show robust and increased responses to other odorants that activate ab8A or ab8B Dose-response curves of control ( black ) and Obp28a- ( red ) ab8 neuronal responses to a 0 . 5s pulse of butyric acid ( A ) and ethyl acetate ( B ) . Responses were quantified by subtracting the spontaneous firing rate from the rate during the stimulus . * p<0 . 05 , p<0 . 01 , * p<0 . 001 . PSTHs of butyric acid ( C ) and ethyl acetate ( D ) responses are shown , with shaded areas indicating SEM . Gray boxes indicate 0 . 5 s stimulus presentations . Graphs depict 2s time windows with spikes summed in 50 ms bins . • indicates p<0 . 05 in all cases and p<0 . 01 in 80% of these cases . n = 12 ( A , C ) and 13 ( B , D ) . ( E ) Responses of control ( black ) and Obp28a- ( red ) ab8 neurons to 0 . 5s pulses of odorants of different chemical classes: 3-methylthio-1-propanol ( sulfur compound ) , γ-hexalactone ( lactone ) , linalool oxide ( terpene ) , methyl benzoate ( aromatic ) , furfural ( aldehyde ) , 2-pentanone ( ketone ) , 2-pentanol ( alcohol ) , and ethyl-3-hydroxybutyrate ( ester ) . Responses were quantified by subtracting the spontaneous firing rate from the rate during the stimulus . * p<0 . 05 , p<0 . 01 . n = 12 . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 01210 . 7554/eLife . 20242 . 013Figure 7—figure supplement 1 . Robust responses from ab4 sensilla of Obp28a mutants to odorants that activate ab4A or ab4B . Responses of control ( black ) and Obp28a- ( red ) ab4 neurons to 0 . 5s pulses of the odorants representing different chemical classes that were tested against ab8 as well as additional odorants that elicit strong responses from ab4A ( E2-hexenal ) and ab4B ( geosmin ) . Responses were quantified by subtracting the spontaneous firing rate from the rate during the stimulus . ** p<0 . 0001 n = 12 . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 013 PSTH analysis revealed higher peak responses in the mutant for both butyric acid and ethyl acetate across a wide range of doses ( Figure 7C , D ) . Following the termination of these short , 0 . 5s stimuli , decay dynamics appeared comparable in the two genotypes for butyric acid , but were faster in the mutant following stimulation with the highest ethyl acetate doses , arguing against a model in which the Obp is essential for clearing these odorants after such pulses . We then tested a diverse panel of other odorants to ask whether the loss of an abundant Obp affected the response profile of the ORNs in ab8 . A priori , Obp28a could differentially expedite the transport of a subset of odorants , or perhaps selectively filter some odorants so as to reduce their access to the dendrites in the sensillum . We selected a panel of eight odorants representing eight chemical classes—a sulfur compound , a lactone , a terpene , an aromatic , an aldehyde , a ketone , an alcohol and an ester—that elicit responses from the odor receptors of ab8 neurons ( Hallem and Carlson , 2006 ) . The odorants were tested at 10−2 dilutions , and also at 10−3 dilutions in the cases of the odorants that elicited the strongest responses . Each was delivered as a 0 . 5s pulse and the responses of both neurons were summed . The response profiles appeared very similar in Obp28a− and Obp28a+ ( Figure 7E ) . The mean responses varied over a broad range , from 50 spikes/s to >200 spikes/s , and for both genotypes the lowest mean response was to 3-methylthio-1-propanol and the greatest response was to ethyl-3-hydroxybutyrate . In between these extremes , the rank order of stimuli was similar . The Obp28a mutant showed greater responses than the control to a subset of the compounds tested . In no case was the response of the mutant lower than that of the control . Our expression analysis highlights the great diversity of Drosophila Obps . The 10 abundant Obps are expressed in diverse antennal regions and in different morphological classes of sensilla . Within a morphological class , they are expressed in distinct functional types , and within different cell types . The diversity of these genes in expression pattern and amino acid sequence , together with their remarkably high abundance , provokes questions about their roles in olfactory function . The Obp-to-sensillum map lays a foundation for addressing these questions . Most of these Obps are expressed in only one morphological class of sensillum . It seems likely that many Obps have evolved to fill specialized needs of particular sensilla . For example , coeloconic sensilla express a subset of Obps that does not overlap with those expressed by other sensilla . This pattern of mutually exclusive expression may reflect the ancient origin of the coeloconic sensilla , their unique double-walled architecture ( Shanbhag et al . , 1999 ) , their response to many polar odorants ( Silbering et al . , 2011 ) , or their expression of a different family of receptors , IRs ( Abuin et al . , 2011; Benton et al . , 2009; Croset et al . , 2010 ) . The Obp-to-sensillum map now provides a foundation for designing ectopic expression experiments in which the function of an individual Obp can be examined in different sensillum types . All 10 of the basiconic sensillum types express at least one of the abundant Obps . Most if not all of the coeloconic and trichoid sensilla also express at least one abundant Obp ( Hekmat-Scafe et al . , 1997; Pikielny et al . , 1994; Shanbhag et al . , 2001 ) . These results are consistent with the concept that Obps are essential to the coding of olfactory information within sensilla . The map reveals that some functionally distinct basiconic sensilla , such as ab1 and ab2 , contain the same subset of abundant Obps . This finding supports the notion that Obps do not dictate the response profile of ORNs; rather , it is consistent with the conclusion from ‘empty neuron’ and heterologous expression analysis that the odor response profile of an ORN is conferred by the odor receptor that it expresses ( Abuin et al . , 2011; Benton et al . , 2009; Dobritsa et al . , 2003; Hallem et al . , 2004; Silbering et al . , 2011 ) . When Ors from eight functional types of basiconic sensilla , ab1-ab8 , were individually expressed in the empty neuron , all yielded response profiles that agreed well with those observed in their endogenous neurons ( Hallem et al . , 2004 ) . The map reveals that the sensillum used in the empty neuron expression system , ab3 , expresses all four of the abundant Obps expressed in basiconic sensilla . Thus any abundant Obp that might be essential to the response of one of the 10 basiconic sensilla would be present in the ab3 test system . We note that the Obp-to-sensillum map also invites analysis of the regulatory mechanisms by which it is established . A receptor-to-neuron map allowed incisive investigation of mechanisms by which individual ORNs select , from among 60 Or genes , which to express ( Barish and Volkan , 2015; Ray et al . , 2007 , 2008 ) . Likewise , it should now be possible to elucidate mechanisms by which individual sensilla select which Obps to express , for example by comparing regulatory regions of co-expressed Obps . We were surprised to find that elimination of the sole abundant Obp from a sensillum did not reduce the magnitude of its response to the tested odorants . We tested odorants of widely varying hydrophobicity and chemical class , including odorants that activate each of the ORNs in the sensillum . The results do not support the widespread belief that Obps are essential for the transport of odorants to receptors within all sensilla ( Figure 8A ) ; rather , the simplest interpretation of our results is that ab8 does not require an Obp to transport odorants to receptors . 10 . 7554/eLife . 20242 . 014Figure 8 . Models of odorant transport via Obps or pore tubules . ( A ) Odorant transport via Obps in a sensillum . An odorant molecule contacts the membrane , diffuses in the surface of the cuticle ( dark gray ) until it reaches a pore , enters the sensillum lymph through the pore , binds to an Obp ( red hexagon ) , and is transported to an olfactory receptor ( green square ) on an ORN dendrite ( D ) . ( B ) Odorant transport via pore tubules . An odorant molecule contacts and diffuses on the cuticle surface , enters a pore , and is transported along a pore tubule to an olfactory receptor on a dendrite ( D ) . An Obp could bind to the odorant and affect dynamics at any point after the odorant reaches the pore . ( C ) Transmission electron micrograph of the pore ( P ) and pore tubules ( Pt ) of a trichoid sensillum of Bombyx mori . Adapted from ( Steinbrecht , 1973 ) . Pore tubules can be seen to contact an ORN dendrite ( D ) in two locations ( arrowheads ) . Scale bar = 0 . 1 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20242 . 014 How could a hydrophobic odorant traverse the aqueous sensillum lymph of ab8 , if not via an Obp ? Many sensilla of widely diverse insects contain tubular structures called pore tubules ( Steinbrecht , 1997 ) ( Figure 8B , C ) . These structures , which can be up to 1 μm long and 15–20 nm in diameter , extend from the wall pores of the sensilla into the interior of the sensillum , often making contact with dendritic membranes . The number of pore tubules per pore ranges up to 20 , and the estimated number of pore tubules per sensillum ranges as high as 83 , 000 ( Steinbrecht , 1997 ) . In Drosophila , pore tubules have been observed in basiconic sensilla , but not in trichoid sensilla ( Shanbhag et al . , 1999 ) , where Obp76a is expressed . Pore tubules have been proposed as a conduit for the transport of hydrophobic odorants from the wall pores to the dendritic membranes . Thus an odorant would follow a three-dimensional trajectory through the air to the antenna , a two-dimensional path in the hydrophobic antennal surface cuticle to a wall pore , and then a one-dimensional trajectory toward the dendritic membrane via a pore tubule ( Figure 8B ) ( Adam and Delbruck , 1968; Kaissling et al . , 1987; Steinbrecht , 1997 ) . There is evidence that pore tubules consist largely of lipids and proteins ( Keil , 1982 ) , likely allowing for diffusion of a hydrophobic odorant . After the discovery of Obps , the view that Obps transport odorants to receptors was proposed and has prevailed in the field for over 30 years . Our results support the possibility that in some sensilla , pore tubules provide an alternative mechanism . We acknowledge that the following formal possibilities can not be excluded: A second surprise was that the response to an odor pulse was greater in the mutant , for a number of odorants . The mutant also showed an accelerated decline in the firing level following the termination of a prolonged odor stimulus . The overall response profile to a panel of diverse odorants , however , was similar between the mutant and control . These results are interesting because they argue against some alternative models of Obp28a function . Specifically , in addition to transporting odorants , Obps have been proposed to protect odorants from degradative enzymes , to inactivate odorants following stimulus termination , or to filter them ( Kaissling , 2001; Leal , 2013; Pelosi and Maida , 1995 ) . If Obp28a protected odorants in ab8 we would expect a greater initial response in the control than in Obp28a mutants . If Obp28a inactivated odorants at the end of the odor response , we would expect a faster decline in the control at the end of odor stimulation . If Obp28a selectively filtered certain odorants , we would expect differences in the response profiles of mutant and control . We note moreover that the similarity in response profiles is consistent with results from the empty neuron system . If Obp28a selectively filtered certain odorants , we would expect the Or35a receptor to confer a different odor response profile in basiconic sensilla , which express Obp28a , and in coeloconic sensilla , which do not ( Figure 2C ) . By contrast , the response profiles of Or35a in the two sensillum types are very similar ( Hallem et al . , 2004 ) . We do not claim that the role of Obp28a in ab8 can be extrapolated to all Obps and all olfactory sensilla . Obp28a is expressed in all basiconic sensilla , and could play a broader role than Obps expressed in only a subset of sensilla . One of the most intriguing aspects of Obps and sensilla is their diversity . Different Obps are likely to have different functions: there is evidence for reduction in olfactory function following reduction in the levels of certain Obps ( Biessmann et al . , 2010; Pelletier et al . , 2010; Swarup et al . , 2011; Xu et al . , 2005 ) , and another Obp has been found to contribute to the inhibition of neurons in the taste system ( Jeong et al . , 2013 ) . In this regard , we found that the most abundant Obp transcript in the antenna , Obp19d , is expressed in epidermal cells and not accessory cells , consistent with the findings of ( Park et al . , 2000 ) , and is thus unlikely to be secreted into sensillar lymph . It seems plausible that Obp19a binds ligands other than odorants , as may other Obps that are expressed outside the olfactory system ( Arya et al . , 2010; Li et al . , 2005; Pitts et al . , 2014 ) . Not only are Obps highly diverse , but sensilla also are diverse in structure and function , as illustrated by the lack of visible pore tubules in trichoid sensilla of Drosophila ( Shanbhag et al . , 1999 ) . We have focused on a single Obp and a single sensillum type because the map constructed in this study allowed us to manipulate them with unparalleled definition . How does the presence of Obp28a reduce the initial response to odorant and prolong the response following termination of a long , intense stimulus ? It is possible that following the sudden influx of an odor stimulus , some of the odorant binds to Obp28a , thereby reducing the amount available for receptor activation . After the termination of a prolonged stimulus , odorant released from the Obp-odorant complex might increase the level available for receptor binding . In both cases the Obp would buffer the system against sudden changes in odor level . The olfactory circuit has evolved a cellular mechanism of gain control , in which local interneurons inhibit ORNs ( Olsen and Wilson , 2008; Root et al . , 2008 ) . This gain control is believed to play a critical role in preventing network saturation and in improving odorant discrimination . It is possible that Obp28a provides an additional mechanism of gain control , which acts at the molecular level as opposed to the cellular level . This molecular form of gain control would precede the cellular form , occurring even before the odor receptor or the neuron were activated . We note finally that even small effects on ORN firing rate can have large effects on odor perception and behavior . ORNs form synapses with second-order neurons called projection neurons ( PNs ) , and the relationship between the firing rates of ORNs and PNs is non-linear ( Olsen and Wilson , 2008 ) . Moreover , PNs also appear to respond most vigorously to odor onset ( Olsen and Wilson , 2008; Wilson , 2013 ) . Thus , even a modest alteration in the firing rate during the initial phase of response could have an important effect on the way olfactory information is represented in the CNS and on the behavior that it drives . Or-GAL4 drivers were from the Bloomington Drosophila Stock Center ( NIH P40OD018537 ) . nompA-GAL4 and ASE5-GAL4 were provided by C . Montell and described previously ( Barolo et al . , 2000; Chung et al . , 2001 ) . GAL4 drivers were crossed to a UAS-mCD8:GFP line ( Lee and Luo , 1999 ) . white Canton-S ( wCS ) and Obp28a- flies ( described below ) were used for electrophysiology experiments . Obp28a- was outcrossed for five generations to wCS to reduce the risk of background effects . The coding region of each Obp was amplified from CS antennal cDNA and cloned into the pGEM-T Easy vector ( Invitrogen , Waltham , MA ) for transcription . For genes with multiple transcripts , the primers were designed to encompass the region present in all versions . Plasmids were linearized with SpeI , NotI , or AatII ( New England BioLabs ) . Digoxigenin ( DIG ) and Fluorescein ( FITC ) labeled probes were created using DIG RNA Labeling Kit SP6/T7 and Fluorescein-labeled UTP ( Roche , Branford , CT ) , and purified with the RNEasy Cleanup Kit ( QIAGEN , Germantown , MD ) . We used seven day-old flies in all staining experiments except that we used flies less than 1-day-old , many within a few hours of eclosion , when labeling with ASE5-Gal4 and nompA-Gal4 because these markers are developmentally regulated and staining was weak in older flies . Male and female flies were anesthetized , placed in a collar , covered with OCT ( Tissue-Tek , VWR , Radnor , PA ) , and frozen on dry ice . 14 µm antennal cryosections were collected on slides and stained as previously described ( Menuz et al . , 2014 ) . In brief , sections were fixed and acetylated at room temperature , pre-hybridized , and then incubated with DIG and/or FITC probes at 65°C overnight . Detection of probes was carried out with anti-DIG-POD or anti-FITC-POD in 1% Blocking Reagent ( Roche ) and amplified with Cy3 or Cy5 TSA ( Perkin Elmer , Waltham , MA ) . GFP was detected with mouse-anti-GFP ( Roche ) and Alexa-fluor-488 donkey-anti-mouse antibodies ( Invitrogen ) . All microscopy was performed using a Carl Zeiss LSM 510 Laser Scanning Confocal Microscope and images were processed with ImageJ software . We used 6–8 day old female flies for single-sensillum recordings essentially as described previously ( Dobritsa et al . , 2003 ) . Filtered AC signals ( 50–2000 Hz ) were recorded and digitized with a Digidata 1440 digitizer and Axoscope 10 . 5 software ( Molecular Devices ) . Action potentials were detected and counted using custom Matlab ( MathWorks ) scripts written by Carlotta Martelli ( Martelli et al . , 2013 ) , with few if any modifications . All spikes were counted due to difficulty in reliably sorting A and B neurons in ab8 sensilla . Impulse responses are defined as the firing rate during the stimulus period minus the baseline firing rate . Odors typically take 100 ms to reach the antenna , due to the length of the odor delivery tube , so the stimulus period is considered to start 100 ms after the opening of the odor delivery valve and persists for the length of the pulse , which in nearly all cases was 500 ms . The baseline firing rate was calculated by counting the number of spikes during the 500 ms period prior to odorant release . To generate dose response curves , each sensillum was tested with all concentrations of odor in ascending order of dose , and no more than 3 sensilla were analyzed per fly . ab8 sensilla were identified by their location and physical attributes and confirmed by their strong response to 10−3 2 , 3-butanediol ( Hallem and Carlson , 2006 ) . Spikes for peri-stimulus time histograms ( PSTHs ) were binned in 50 ms intervals except that in the 30s stimulus paradigm they were binned in 1s intervals . Statistical significance was assessed using Prism’s ( GraphPad ) two-way repeated measures ANOVA followed by Bonferroni’s post-hoc test . Values shown are the mean +/− SEM . Odor stimuli were prepared and delivered essentially as described previously ( Hallem et al . , 2004 ) . 1-octanol , butyric acid , 3-methylthio-1-propanol , γ-hexalactone , furfural , 2-pentanone , 2-pentanol , ethyl-3-hydroxybutyrate , and were purchased from Sigma Aldrich ( St . Louis , MO ) , methyl benzoate and linalool oxide were purchased from Fluka ( St . Louis , MO ) , and ethyl acetate was purchased from J . T . Baker ( Center Valley , PA ) . Odorants of the highest grade available were used and were then diluted in paraffin oil ( Fluka ) . For 500 ms pulses , 50 μl of diluted odorant was applied to a 13 mm filter paper disc inside a Pasteur pipette and capped with a 1 ml pipette tip to create an odor cartridge . Cartridges were allowed to equilibrate for 20 min before use . Each cartridge was used no more than three times and with more than 10 min between uses , to allow the odor to re-equilibrate . Mutant and control flies were tested in parallel , on the same day , and in the same manner . Stimuli were presented by placing the tip of the cartridge into a glass tube delivering a humidified air stream ( ~2000 ml/min ) to the fly , and administering a 500 ms pulse of air ( ~200 ml/min ) through the cartridge . For experiments with a 30 s odor stimulus , 25 ml pipettes were used in place of Pasteur pipettes , with three filter papers , each 55 mm in diameter , and 1 ml of odor dilution . These large cartridges were used up to eight times , since independent PID measurements showed no substantial decrease in odor concentration after more than 10 uses . For experiments with a background odor , a 125 ml flask with 5 ml of odor dilution was inserted between the main airstream and the odor delivery tube . Flies were exposed to the background odor for more than 5 min before beginning the experiment .	Insects use their sense of smell to find mates , to find food and – in the case of insects that transmit diseases such as malaria and Zika – to find us . If we can understand how insect scent detection works at the molecular and cellular level , we may be able to devise new ways of manipulating the insects’ sense of smell and prevent them from finding us . Insects contain a family of proteins called odorant binding proteins that are intriguing in several ways . They are numerous ( there are 52 kinds in the fruit fly Drosophila ) , they are diverse and some are made in remarkably large amounts in the antennae . Fine hair-like structures known as olfactory sensilla protrude from the surface of the antennae . Odorant binding proteins are widely believed to carry odorant molecules through the fluid inside the sensilla to olfactory neurons , which then send signals that trigger the insect’s response to the scent . Larter et al . have now mapped the most abundant odorant binding proteins to the various olfactory sensilla of Drosophila . This revealed that a type of sensillum known as ab8 contained only one abundant odorant binding protein , called Obp28a . Unexpectedly , Larter et al . found that ab8 sensilla that are deprived of this protein respond strongly to odorant molecules . This result suggests that Obp28a is not required to transport odorants to the neurons in ab8; indeed , it appears that these neurons do not require an abundant odorant binding protein in order to respond to a scent . Instead , Obp28a helps to moderate the effects of sudden changes in the level of an odorant in the environment , so that concentrated odors do not trigger too large a response from the olfactory neurons . The details of the role that Obp28a plays in olfactory sensilla remain to be investigated in future studies , and the map created by Larter et al . also lays a foundation for studying the roles of other odorant binding proteins . The discovery that Obp28a is not needed to transport odorant molecules also raises questions about how insects are able to detect smells . Many odorant molecules repel water , so how do these molecules travel through the fluid in the sensilla if odorant binding proteins are not needed to transport them ?	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2016	Organization and function of Drosophila odorant binding proteins
Bacterial chemotaxis is a paradigm for how environmental signals modulate cellular behavior . Although the network underlying this process has been studied extensively , we do not yet have an end-to-end understanding of chemotaxis . Specifically , how the rotational states of a cell’s flagella cooperatively determine whether the cell ‘runs’ or ‘tumbles’ remains poorly characterized . Here , we measure the swimming behavior of individual E . coli cells while simultaneously detecting the rotational states of each flagellum . We find that a simple mathematical expression relates the cell’s run/tumble bias to the number and average rotational state of its flagella . However , due to inter-flagellar correlations , an ‘effective number’ of flagella—smaller than the actual number—enters into this relation . Data from a chemotaxis mutant and stochastic modeling suggest that fluctuations of the regulator CheY-P are the source of flagellar correlations . A consequence of inter-flagellar correlations is that run/tumble behavior is only weakly dependent on number of flagella . Many species of bacteria swim by rotating helical filaments called flagella ( Berg , 2004 ) . A typical Escherichia coli cell is propelled by a bundle composed of multiple flagella . Each flagellum is controlled by a rotary motor that can switch between clockwise ( CW ) and counter-clockwise ( CCW ) rotation . When flagella on a cell rotate CCW , the cell swims along an approximately straight path called a ‘run’ . When some of the flagella rotate CW , the bundle is disrupted causing an abrupt change in direction called a ‘tumble’ ( Macnab and Ornston , 1977 ) . E . coli modulates the probability of being in one of these two swimming states in response to its environment , allowing it to navigate chemical , temperature , and light gradients ( Berg and Brown , 1972; Berg , 2004 ) . At any point in time , the probability that a flagellar motor rotates CW is determined by the concentration of phosphorylated signaling protein CheY ( CheY-P ) . Coupling CheY phosphorylation to chemicals from the environment allows the cell to bias its random walk and migrate towards more favorable conditions . This biased random walk is called chemotaxis , and serves as a model for understanding how living organisms process information ( Berg and Brown , 1972; Wadhams and Armitage , 2004; Shimizu et al . , 2010 ) . Tremendous progress has been made towards elucidating the mechanism of bacterial chemotaxis . The relationship between the chemotaxis signaling network and the CCW/CW rotational bias of the individual flagellar motor is now well mapped ( [Block et al . , 1982; Cluzel et al . , 2000; Sourjik and Berg , 2002; Yuan et al . , 2012]; for a review see Berg , 2004 ) , and has also been described using detailed mathematical models ( Emonet et al . , 2005; Jiang et al . , 2010; Shimizu et al . , 2010 ) . Despite this wealth of knowledge , how the CCW/CW states of individual motors collectively determine the run/tumble swimming behavior of the whole , multi-flagellated cell remains poorly understood . The number of flagella on an individual swimming cell can vary greatly , from one to more than ten ( Cohen-Ben-Lulu et al . , 2008 ) ( Figure 1—figure supplement 1 ) , yet very few studies are available to indicate how flagellar number affects swimming behavior . The only direct measurements of flagellar dynamics in swimming cells have been limited to short durations ( ∼1 s ) ( Turner et al . , 2000; Darnton et al . , 2007 ) . The absence of long-term observations has precluded the development of a detailed mapping between flagellar state and cell swimming behavior . As a result , most theoretical models of bacterial chemotaxis are limited to treating an individual motor , or simply assume that all cells have a single flagellum ( Bray et al . , 2007; Kalinin et al . , 2009; Matthaus et al . , 2009; Jiang et al . , 2010; Flores et al . , 2012 ) . Quantifying the mapping from single-flagellum state to whole-cell swimming behavior thus remains a missing link to developing an end-to-end picture of bacterial chemotaxis . A number of theoretical models have been put forward in an attempt to describe this mapping . One such model invokes a ‘voting’ mechanism , in which cells tumble only if a majority of flagella rotate CW ( Ishihara et al . , 1983; Spiro et al . , 1997; Andrews et al . , 2006; Vladimirov et al . , 2008; Jiang et al . , 2010 ) . However , by observing fluorescently labeled flagella during individual tumbles , Turner et al . established that CW rotation of a single flagellum is sufficient to ‘veto’ a run ( Turner et al . , 2000 ) ( Figure 1A ) . Refined versions of this ‘veto model’ were recently developed ( Vladimirov et al . , 2010; Sneddon et al . , 2012 ) , based on careful , slow-motion observations of tumbles ( Darnton et al . , 2007 ) . However , the extent to which these details are relevant for modeling swimming behavior is unknown , because no measurements have directly correlated long-term swimming behavior with flagellar activity in the same cell . 10 . 7554/eLife . 01916 . 003Figure 1 . Wild-type E . coli cells deviate from the ‘veto’ model . ( A ) The mapping relating the run/tumble state of the cell to the CCW/CW state of its flagella according to the veto model . Schematic time trace from a cell with 3 flagella , showing CW ( purple ) and CCW ( white ) intervals for each flagellar motor and the resulting tumbles ( black ) and runs ( white ) of the cell . The veto model corresponds to an AND gate , by which cell runs only occur when all flagella rotate CCW ( where CCW = 1 , CW = 0 , run = 1 , and tumble = 0 ) . ( B ) Schematic of a cell held by two optical traps ( red cones ) in the fluorescence excitation volume ( green ) within the sample chamber . ( C ) Representative data trace from a trapped cell with three flagella . Still images of fluorescently labeled flagella at 0 . 5-s intervals ( top ) . The approximate location of the unlabeled cell body is indicated by a dashed yellow line . Flagella rotating CW ( purple ) and CCW ( white ) are numbered in frames in which they appear distinct . Corresponding cell-body rotation signal for the same cell ( red line , bottom ) as detected from deflections of the trapping laser . Tumbles ( shaded area ) were determined from the erratic cell-body rotation signal . ( D ) Long-term time trace of CCW/CW flagellar rotation state and run/tumble cell swimming state . CW intervals ( purple , top ) for each flagellum were determined from the fluorescence images . Tumbles ( black , bottom ) were determined from the cell-body rotation signal . ( E ) Mean deviation η from the veto model vs number of flagella per cell . Wild-type cells ( solid black circles ) with multiple flagella deviate significantly from the model ( p=0 . 0003 , N = 69 cells ) . CheY* cells ( open gray circles; N = 46 cells ) match the model ( p=0 . 77 ) . Error bars denote SEM . See ‘Materials and methods’ for more details . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 00310 . 7554/eLife . 01916 . 004Figure 1—figure supplement 1 . Distribution of flagellar number . Top , fluorescently-labeled flagella on trapped E . coli cells . Images show cells with 1 , 2 , 3 , 4 , and 6 flagella ( left to right ) . The image contrast was adjusted to make flagella more visible ( see ‘Materials and methods’ for details on the labeling , imaging and image processing protocols ) . Middle , transmission electron microscopy ( TEM ) images of cells possessing 1 , 2 , 3 , 4 and 7 flagella ( left to right ) . The images were enhanced to make flagella more visible ( see ‘Materials and methods’ for details of the imaging and image processing ) . Bottom , distributions of the number of flagella per cell . Cells with zero flagella were excluded because non-swimming cells were not trapped . The distributions from trapped cells ( blue , N = 86 cells ) and TEM ( red , N = 56 cells ) have very similar means . Error bars denote SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 00410 . 7554/eLife . 01916 . 005Figure 1—figure supplement 2 . Instrument layout . Layout of the combined optical trap/epi-fluorescence microscope , showing the 1064-nm trapping laser beam ( red ) , 532-nm fluorescence excitation laser beam ( green ) , fluorescence emission path ( yellow ) , objectives ( 01 and 02 ) , quadrant photodiode ( QPD ) , and fluorescence imaging , charge-coupled device camera ( EMCCD ) . To the right is a schematic of a cell held by two optical traps ( red cones ) in the fluorescence excitation volume ( green ) within the sample chamber . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 00510 . 7554/eLife . 01916 . 006Figure 1—figure supplement 3 . Laser temporal interlacing scheme . Schematic of the temporal interlacing of the trapping laser ( red ) , excitation laser ( green ) and camera exposure ( black ) . The 1064-nm trapping laser alternates with the 532-nm fluorescence excitation laser , such that they are never on at the same time . The camera exposure ( 30-μs exposure ) is synchronized with the fluorescence excitation laser for stroboscopic imaging . Lower panel shows the interlacing over a longer time scale . The fluorescence excitation pulse and camera exposure occur once every 10 ms to create movies at 100 fps . See ‘Materials and methods’ for additional details . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 00610 . 7554/eLife . 01916 . 007Figure 1—figure supplement 4 . Sample data from representative cells . Typical data from trapped cells with different numbers of flagella . A few still images ( out of hundreds for each sample ) of the fluorescently labeled flagella at different time points are shown . In the panels below , colors indicate the swimming state as runs/tumbles ( white/black ) and flagella waveforms as ‘normal’/‘semi-coiled’/‘curly-1’/‘semi-coiled’ or ‘curly-1’ ( white/red/blue/purple ) . Light colors indicate periods when the flagellum was transitioning between two different waveforms . Light blue indicates a transition between normal and semi-coiled , light red indicates a transition between normal and curly-1 . Samples A and B are wild-type cells , samples C–E are CheY* cells . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 00710 . 7554/eLife . 01916 . 008Figure 1—figure supplement 5 . Cells with flagella in the curly-1 state rarely run . Top , the fraction of time that cells spent running vs the number of flagella that were in the curly-1 waveform ( while all other flagella were in the normal waveform ) . When all flagella were in the normal waveform , cells ran 91% of the time . When the waveform of a single flagellum was curly-1 , and the rest were normal , cells ran 18% of the time . Bottom , mean deviation from veto model , η , vs number of flagella per cell . Same as Figure 1E , with an additional theoretical model in which cells run 18% of the time when the waveform of a single flagellum is curly-1 ( magenta ) . Error bars denote SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 008 In this study , we present simultaneous , prolonged observations of individual flagella and whole-cell swimming . Using an optical trap to hold a swimming cell ( Min et al . , 2009 ) while simultaneously imaging its fluorescently labeled flagella , we relate directly the number and state of each flagellum on a cell to its swimming behavior . Our measurements establish a simple mapping between CCW/CW flagellar rotation and run/tumble bacterial swimming state for cells with arbitrary numbers of flagella . Surprisingly , we find that E . coli cells wild-type for chemotaxis do not strictly comply with the veto model , because the states of individual flagella in the same cell are strongly coupled . Flagella do not switch independently , and as a result , tumbles typically involve many of the flagella on the cell . The behavior of multi-flagellated cells can still be mapped to the simple veto model by renormalizing the flagellar number to a lower effective number of independent flagella . As for the cause of inter-flagella coupling , our data strongly favor a mechanism involving fluctuations of the signaling network rather than direct , physical interactions between flagella . This model is supported by the observation that mutant cells in which flagellar switching is decoupled from the chemotaxis network do obey the simple veto model , as well as by stochastic simulations of the swimming behavior . According to the veto model , a cell tumbles whenever any one flagellum rotates CW ( Figure 1A ) . Thus , the probability that a cell runs equals the probability that all of its flagella remain CCW . As a consequence , cells with more flagella are expected to tumble more , since there is a higher chance that at least one flagellum will deviate from the consensus and ‘veto’ the run . These predictions can be stated mathematically , under the assumption that the rotational direction of each flagellum is independent of the other flagella . In a cell with Nflag flagella , the average tumble bias TB—the fraction of time a cell spends tumbling—will be given by ( 1 ) TB =1− ( 1− CB ) Nflagwhere CB is the average clockwise bias—the fraction of time the cell’s flagella rotate CW ( ‘Materials and methods’ ) . To test this prediction , we quantified the swimming behavior of individual E . coli cells wild-type for chemotaxis ( strain HCB1660 [Turner et al . , 2010]; see ‘Materials and methods’ , Table 1 ) using an instrument combining optical tweezers and epi-fluorescence imaging ( Figure 1B and Figure 1—figure supplement 2 ) . The instrument allowed us to measure simultaneously run/tumble behavior and flagellar dynamics in the same cell . The optical trap was used to hold each end of a single cell in place and the light scattered by the cell was utilized to monitor its swimming behavior , as described previously ( Min et al . , 2009 , 2012 ) . As shown in Figure 1C ( Video 1 ) , cell runs were identified from oscillatory time signals due to cell body rotation at a frequency of ∼10 Hz . Cell tumbles were identified as periods of erratic motion during which the flagellar bundle was disrupted ( Min et al . , 2009 ) . Flagella were fluorescently labeled using the method of Turner et al . ( 2010 ) . High speed , epi-fluorescent , stroboscopic imaging ( Turner et al . , 2000; Figure 1—figure supplement 3 ) was used to resolve individual flagella ( Figure 1C ) . Since the trapped cell remained in the field of view for a prolonged period , flagella were observed through multiple runs and tumbles ( typically ∼5 events ) , limited by the time until flagella became too dim to discern due to photobleaching ( ∼8 to 40 s ) . The rotational direction of each flagellum was determined by observing its shape during 100-ms time windows . As shown by Darnton et al . ( 2007 ) , flagella may take on different helical waveforms depending on their rotational state . These waveforms , termed ‘normal’ , ‘semi-coiled’ , ‘curly-1’ , and ‘curly-2’ , can be visually identified based on their pitch and wavelength ( Figure 1C , Figure 1—figure supplement 4 ) . CCW rotating flagella were identified based on the normal conformation , which they have been shown to adopt exclusively ( Darnton et al . , 2007 ) , while CW rotating flagella were identified by their curly-1 or semi-coiled shape . From the identification of CCW and CW intervals , the cell’s mean CW bias was determined by averaging the fraction of time that all the flagella on the cell spent CW ( ‘Materials and methods’ ) . 10 . 7554/eLife . 01916 . 009Table 1 . Strains and plasmids used in this workDOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 009StrainGenotypeCommentsSourceHCB1660fliC::Tn5 ( KanR ) ‘wild type’ Contains plasmid pBAD33-fliCS219C ( Turner et al . , 2010 ) Gift of H BergPM87cheBYZ::FRT , fliC::Tn5 ( KanR ) ‘CheY’ Contains plasmids pMS164 and pPM5This studyRP437Wild-type for chemotaxis ( Parkinson and Houts , 1982 ) SK109cheBYZ::CmThis studySK110cheBYZ::FRTThis studySK112cheBYZ::FRT , fliC::Tn5 ( KanR ) This studyPlasmids pBAD33 fliCS219CfliCS219C under ParaBAD promoter , CmR , p15a originExpresses mutant version of FliC for fluorescent labeling ( Turner et al . , 2010 ) Gift of H Berg pPM5fliCS219C under ParaBAD promoter , AmpR , colE1 originExpresses mutant version of FliC for fluorescent labelingThis study pMS164cheYD13K under PlacOP promoter , CmR , pSC101 originExpresses constitutively active version of CheY ( Alon et al . , 1998 ) Gift of P Cluzel pDK46Helper plasmid ( Datsenko and Wanner , 2000 ) pKD3Template for CmR cassette ( Datsenko and Wanner , 2000 ) pCP20Helper plasmid ( Cherepanov and Wackernagel , 1995 ) 10 . 7554/eLife . 01916 . 010Video 1 . Video of trapped wild-type cell with three labeled flagella as it runs and tumbles . Slow motion video of the wild-type cell in Figure 1C with three long , fluorescently labeled flagella . The approximate location of the unlabeled cell body is indicated by the dotted line . The trap signal used to determine runs and tumbles ( bottom , scrolling blue curve ) measures the position of the cell body in the trap as it rotates . At the beginning of the video ( time stamp = 2 . 4 s ) , all three flagella are in a bundle and the cell is running . One by one , the flagella switch to CW rotation ( 2 . 7–3 . 0 s ) , which disrupts the bundle and causes the cell to tumble . Flagella can be observed in all three waveforms , normal , semi-coiled , and curly-1 . Near the end of the video , the flagella all return to CCW rotation and coalesce into a bundle , causing the cell to resume running ( 3 . 4 s ) . In addition to the three long flagella , a short flagellar stub is visible . The stub does not affect the swimming behavior and is not analyzed . Scale bar in bottom left corner is 1 µm . Frames were recorded at 400 frames per second , video shows every other frame at 20 frames per second . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 010 Our assay allowed us to determine all the parameters in Equation ( 1 ) directly . For each cell , we measured the tumble bias ( using the optical trap ) , flagellar number Nflag , and CW bias ( using fluorescence imaging ) . We used these values to compare our experimental data to the prediction of the veto model . Reorganizing Equation ( 1 ) , we define the parameter η: ( 2 ) η ≡log ( 1− TB ) log ( 1−CB ) −Nflagwhich quantifies the deviation of the data from the veto model . Comparing Equation ( 2 ) to Equation ( 1 ) , η may also be interpreted as the difference between two terms: the number of flagella estimated from the veto model based on the cell’s swimming behavior , and the number of flagella on the cell as determined by counting directly . Figure 1E ( black circles ) shows η against the flagellar number Nflag . η was calculated for each individual cell and then averaged over all cells with a given number of flagella . Unexpectedly , we found that wild-type cells with multiple flagella systematically deviated from the predicted behavior . Specifically , η was consistently negative for cells with Nflag > 1 ( 35/48 cells ) , indicating that cells with multiple flagella tumbled less than expected from the model . In the context of the veto model , the cells behaved as if they had a smaller number of flagella than what they actually had . We first considered the possibility that a more detailed version of the veto model might explain the observed behavior and reconcile this discrepancy . A recent study by Sneddon et al . ( 2012 ) used the observations of Darnton et al . ( 2007 ) to refine the veto model . Specifically , the Sneddon model states that a cell with a minimum of X CCW flagella will run rather than tumble , provided the remaining ( Nflag–X ) CW flagella are in the curly-1 conformation only ( X is a parameter in the model , with possible values in the range [1 , Nflag − 1] ) . Thus , the simple veto model considered above corresponds to X = Nflag , and the least perturbative refinement to the model corresponds to X = Nflag − 1 , in which a cell with a single curly-1 flagellum still runs . However , in our measurements we observed that cells with a single CW flagellum in the curly-1 state still tumbled 82% of the time ( 44 s of cumulative time in which one flagellum was in the curly-1 state; Figure 1—figure supplement 5 ) . Modifying the Sneddon model to allow runs 18% of the time was not sufficient to reproduce the trend observed in Figure 1D , Figure 1—figure supplement 5 . To investigate the discrepancy between our data and the veto model , we next examined individual tumble events in greater detail . In agreement with the original observation of Turner et al . ( 2000 ) , we found that CW rotation of a single flagellum was indeed sufficient to cause a tumble in multi-flagellated cells ( Figure 1—figure supplement 4 , samples A , D and E ) . However , we also observed that more than half of tumbles in multi-flagellated cells ( 56% , 117/210 events ) actually involved multiple CW flagella . Figure 2A shows a representative trace from a wild-type cell with three flagella . There are times during each tumble in the trace where all three flagella are in a CW state . As shown in Figure 2C , the number of CW flagella ‘participating’ in a tumble ( black circles ) was significantly larger than would be expected if flagella were independently switching ( gray dashed line , obtained from simulations of a cell with independent flagella; see ‘Materials and methods’ ) . Our results thus suggest that while a single CW flagellum is sufficient to induce a tumble ( in agreement with a simple veto model ) , flagella are also coupled and may thus switch in groups , in a correlated fashion . Further evidence for inter-flagella coupling was obtained by calculating the cross-correlation between pairs of flagella on a given cell . We found a significant correlation between the rotational directions of pairs of flagella on the same cell ( Figure 2D , black data points ) . This correlation persisted for ∼1 s , the average duration of a tumble . Our findings are consistent with previous observations by Terasawa et al . on surface-immobilized cells ( Terasawa et al . , 2011 ) . There , correlations between individual motors on the same cell were detected by monitoring beads attached to flagellar stubs , as opposed to complete flagella on swimming cells in our present work . 10 . 7554/eLife . 01916 . 011Figure 2 . Tumbles in wild-type cells involve multiple CW flagella . ( A ) A typical time trace from a wild-type cell with 3 flagella . Colors indicate runs/tumbles ( white/black , top ) and CCW/CW ( white/purple , middle ) . The blue line ( bottom ) shows the corresponding number of CW flagella at each time point . ( B ) Same as ( A ) for a typical CheY cell with 3 flagella . ( C ) Mean of the maximum number of CW flagella during a tumble vs number of flagella per cell . Consistently more flagella are CW during tumbles in the wild-type ( black circles; N = 61 cells ) compared to the CheY* strain ( open gray circles; N = 24 cells ) . Simulations incorporating fluctuations in CheY-P ( black line ) and without fluctuations ( gray dashed lines ) reproduce the observed trends ( simulations detailed in the text and Figure 4 ) . ( D ) Cross-correlation between flagella pairs , averaged over all pairs and all cells . Wild-type ( black circles ) match simulations with fluctuations in CheY-P ( black line ) . CheY* strain ( open gray circles ) matches simulations without fluctuations in CheY-P ( gray dashed line ) , which exhibit almost no correlation . Error bars denote SEM . See ‘Materials and methods’ for more details . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 01110 . 7554/eLife . 01916 . 012Figure 2—figure supplement 1 . Flagellar transition rates vs number of flagella per cell . Flagella motor transition rates , separated into two groups corresponding to cells with a single flagellum ( red ) and cells with multiple flagella ( blue ) . All data from wild-type cells ( N = 52 cells ) . Error bars denote SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 01210 . 7554/eLife . 01916 . 013Figure 2—figure supplement 2 . Flagellar transition rates vs number of flagella per cell . Transitions between the three observed flagella waveforms ( normal , semi-coiled and curly-1 ) , separated into two groups corresponding to cells with a single flagellum ( red ) and cells with multiple flagella ( blue ) . All data from wild-type cells ( N = 52 cells ) . Error bars denote SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 013 The source of inter-flagellar correlation remains under debate . Terasawa et al . observed that mutant cells , in which the concentration of CheY-P was decoupled from the chemotaxis network , displayed no correlations ( Terasawa et al . , 2011 ) . This led us to likewise investigate the behavior of a strain , PM87 , expressing a constitutively-active CheY ( CheYD13K [Alon et al . , 1998] , denoted CheY* , see Table 1 and 2 ) . The protein was exogenously expressed , with the expression level chosen such that the population-averaged tumble bias matched that of wild-type cells ( ‘Materials and methods’ ) . A representative trace from a CheY* cell with three flagella is shown in Figure 2B ( see also Figure 1—figure supplement 4 , and Video 2 ) . Upon inspection , flagellar switching appears far less correlated than in wild-type cells ( Figure 2D , compare black and gray data points ) . Comparing 54 wild-type and 24 CheY* cells with the same mean CW biases ( 0 . 11 ± 0 . 07 vs 0 . 11 ± 0 . 07 , mean ± SD ) we found that , on average , fewer CW rotating flagella participated in tumbles in the CheY* strain ( Figure 2C , open circles ) . Moreover , the number of participating flagella in the CheY* strain closely matched the expectation for cells with independently switching flagella ( Figure 2C , dashed line ) . ( This number deviates from unity and trends upwards with number of flagella simply because of the finite probability that two tumbles overlap by chance . ) Our results indicate that when the signal for flagellar motors to switch their rotational state is decoupled from the chemotaxis network , the motors switch independently . Based on our interpretation of the wild-type data , we thus expect CheY* cells to adhere to the simple veto model . As shown in the plot of η in Figure 1D ( open circles ) , CheY* cells indeed match the veto model closely ( η = −0 . 08 ± 0 . 15 , mean ± SEM ) . The existence of correlations between flagella states in wild-type cells may thus explain why cells with multiple flagella deviate from the veto model . 10 . 7554/eLife . 01916 . 014Table 2 . Primers used in this workDOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 014PrimerSequenceCommentsSK140FTGCGTGGTCAGACGGTGTATGCGCTAAGTAAGGATTAACG GTGTAGGCTGGAGCTGCTTCcheBYZ deletion forwardSK140RGCCTGATATGACGTGGTCACGCCACATCAGGCAATACAAA CATATGAATATCCTCCTTAGcheBYZ deletion reverseSK141FCCTTAAACCCGACGGATTGCcheBYZ deletion check forwardSK141RTTGCTGCCACACATCAAGCcheBYZ deletion check reverseSK163FAGGGTTATTGTCTCATGAGCpZE11 sequencing forwardSK163RGTTTTATTTGATGCCTCTAGpZE11 sequencing reversePM7FGGG GACGTC ATCGATGCATAATGTGCCTGamplify ParaBAD fliCS219C forwardPM7RGGG GTCGAC TTAACCCTGCAGCamplify ParaBAD fliCS219C reverse10 . 7554/eLife . 01916 . 015Video 2 . Video of trapped CheY* cell with two labeled flagella as it runs and tumbles . Slow motion video , similar to Video 1 , of a trapped CheY* cell with two fluorescently labeled flagella . Still images from this cell are shown in Figure 1—figure supplement 4 , sample D . The approximate location of the unlabeled cell body is indicated by the dotted line . At the beginning of the video ( time stamp = 7 . 4 s ) , both flagella are in a bundle , rotating CCW in the normal waveform , and the cell is running . At 7 . 6 s , one flagellum switches to CW rotation and transitions to the semi-coiled waveform , which disrupts the bundle and causes the cell to tumble . At the 8 . 5 s , the semi-coiled flagellum returns to CCW rotation and both flagella coalesce into a bundle , causing the cell to resume running . Scale bar in bottom left corner is 1 µm . Frames were recorded at 100 frames per second , video plays at 20 frames per second . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 015 Our results so far suggest that , while wild-type cells obey the fundamental premise of the veto model—that is , a single CW flagellum is sufficient to induce a tumble—the presence of inter-flagella correlations leads to the failure of Equation ( 1 ) in relating the observed CW bias and tumble bias . To describe the relation between single flagellar state and whole-cell behavior successfully , this expression must then be modified to account for flagellar correlations . To this end , we examined the relation between CW bias and tumble bias in all individual cells having a given flagellar number ( Figure 3A , B ) . Equation ( 1 ) defines a single curve , along which the CW bias and tumble bias of all cells with Nflag flagella should lie ( dashed line in Figure 3A , B ) . As expected , the CheY* cells follow these predicted curves closely for all Nflag values ( Figure 3B ) ( R2 = 0 . 89 ) . In contrast , wild-type cells with multiple flagella consistently fall below the predicted curves ( Figure 3A ) ( 35/48 cells ) . 10 . 7554/eLife . 01916 . 016Figure 3 . Wild-type behavior matches the veto model for cells with a lower effective number of flagella . ( A ) Tumble bias vs CW bias for individual wild-type cells ( N = 69 ) , plotted separately for different numbers of flagella per cell ( Nflag = 1 , purple; 2 , blue; 3 , green; 4 , red; 5 , cyan ) . The prediction from the veto model in Equation ( 1 ) ( dashed lines ) does not match the data for cells with multiple flagella ( R2 = 0 . 88 , 0 . 60 , 0 . 41 , 0 . 39 for Nflag = 2 , 3 , 4 , 5 ) . The data were fit ( solid lines ) to Equation ( 1 ) , while allowing the number of flagella to be used as a fitting parameter , Neff . Error bars denote SD . ( B ) Same as ( A ) for CheY* ( open circles , same color code as [A] N = 46 cells ) . The veto model prediction ( dashed lines ) matches the data well ( R2 = 0 . 91 , 0 . 97 , 0 . 93 , 0 . 67 , 0 . 98 for Nflag = 1 , 2 , 3 , 4 , 5 ) . Fits ( solid lines ) yield Neff values almost identical to Nflag . ( C ) Fitted Neff values vs number of flagella per cell for wild-type ( black circles ) and CheY* ( open gray circles ) cells . Simulations ( described in the text ) reproduce the observed trends . ( D ) Data points from individual wild-type ( solid circles ) and CheY* ( open circles ) cells all collapse onto a single line when using Neff from fits to wild-type data in ( A ) and the actual flagellar number Nflag for CheY* cells in ( B ) . Error bars denote SEM . See ‘Materials and methods’ for more details . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 01610 . 7554/eLife . 01916 . 017Figure 3—figure supplement 1 . Fit to Neff vs Nflag . Data points show fitted Neff values vs number of flagella per cell for wild-type cells as in Figure 3C . The data was fit empirically to a power law ( red line ) given by Neff = 1 . 27 × Nflag0 . 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 017 Based on our observation that wild-type cells exhibited η values consistent with cells with a lower number of flagella than the actual value , we hypothesized that wild-type behavior may be described within the framework of the veto model by allowing the parameter Nflag in Equation ( 1 ) to deviate from the actual flagellar number . As shown in Figure 3A ( solid lines ) , using the flagellar number as a fitting parameter ( now denoted Neff ) indeed allows for a good match for the wild-type data ( R2 = 0 . 85 ) . In this revised equation , Neff can be thought of as the ‘effective’ number of independent flagella on a cell , which captures the fact that flagella in wild-type cells switch in a correlated manner . Consistent with this picture , the effective number of flagella Neff is consistently smaller than the actual flagellar number Nflag ( Figure 3C , black circles; Neff can be approximated by Neff = 1 . 27 × Nflag0 . 5 , see Figure 3—figure supplement 1 ) . As a control , estimating Neff for CheY* cells produces values very close to the original flagellar number Nflag ( Figure 3B , solid line and Figure 3C , open circles ) . The introduction of the parameter Neff allows us to formulate a generalized veto model , which describes the mapping between the CW bias and tumble bias for both wild-type and CheY* cells . The generalized model now defines a universal curve , ( 3 ) 1−TB= ( 1−CB ) Neffalong which all individual cells of both genotypes should lie ( using Neff = Nflag for CheY* strain and the best fit value of Neff for wild-type ) . As seen in Figure 3D , this expression successfully collapses all single-cell data from both strains and all flagellar numbers . Our results show that E . coli cells adhere to the veto model , but that inter-flagellar correlations lead to a renormalization of the effect of flagellar number . What is the source of these flagellar correlations ? The absence of correlations in the CheY* strain and its adherence to a simple veto model provide an important clue to understanding the mechanism of inter-flagellar coupling . In wild-type cells , CheY-P levels are subject to phosphorylation and de-phosphorylation reactions by chemotaxis network components and are believed to fluctuate in time ( Korobkova et al . , 2006; Sneddon et al . , 2012 ) . In contrast , in the mutant strain , CheY* levels are decoupled from the network and are thus expected to be constant over the timescales of interest ( Korobkova et al . , 2006; Min et al . , 2009 ) . Terasawa et al . proposed that fluctuations in CheY-P levels may thus provide a mechanism by which the CW biases of multiple flagella on a cell can be coupled ( Terasawa et al . , 2011 ) . To test whether such a mechanism could account for the different features of bacterial swimming observed in our study , we performed simulations of whole-cell swimming driven by the chemotaxis network . In particular , we investigated how fluctuations in CheY-P concentration could produce differences in inter-flagellar correlations between wild-type and CheY* cells and consequent differences in their respective mappings of CW bias to tumble bias . We performed stochastic simulations of the chemotaxis network and resulting flagellar motor activity and then applied the veto rule that CW rotation of a single flagellum leads to cell tumbling ( Figure 4A , B; ‘Materials and methods’ ) . For wild-type cells , we incorporated fluctuations in CheY-P concentration using the approach of Sneddon et al . ( 2012 ) . The model was constrained using transition rates between flagella waveforms that were directly extracted from our experiments ( Figure 4—figure supplement 1 and Tables 3 and 4 ) . We found that , consistently , simulations with CheY-P fluctuations were able to duplicate the observed wild-type behavior , while simulations without fluctuations matched CheY* data and the predictions of the ‘naïve’ veto model . Specifically , using only two parameters—the amplitude and the characteristic timescale for CheY-P fluctuations ( Table 3 ) —as fitting parameters , our simulations simultaneously reproduced ( i ) the relation between flagellar number and flagellar participation in tumbles ( Figure 2C ) ; ( ii ) temporal correlations between flagella ( Figure 2D ) ; ( iii ) the effective number of flagella in multi-flagellated wild-type cells ( Figure 3C ) ; and ( iv ) the degree of deviation from the veto model ( Figure 4C ) . 10 . 7554/eLife . 01916 . 018Figure 4 . A theoretical model incorporating CheY-P fluctuations reproduces wild-type data . ( A ) Simulated time trace for a wild-type cell . Representative simulated time trace of CheY-P concentration ( gray line , top ) , CW bias ( red ) , run/tumble state ( white/black ) , CCW/CW flagellar rotational direction ( white/purple ) and number of CW flagella ( blue line , bottom ) for a cell with 3 flagella . CheY-P simulation parameters are described in the text and in Table 3 . ( B ) Same as ( A ) from a simulated CheY* cell , in which CheY-P concentration ( gray line ) does not fluctuate . ( C ) Deviation from veto model . The theoretical model that includes CheY-P concentration fluctuations ( black line ) reproduces the wild-type data ( black circles ) . A simple veto model with constant CheY-P concentration ( gray dashed line ) reproduces the CheY* data ( open gray circles ) . Error bars denote SEM . See ‘Materials and methods’ for more details . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 01810 . 7554/eLife . 01916 . 019Figure 4—figure supplement 1 . Flagellar waveform transition rates . Diagram of the transitions rates between different flagellar waveforms: normal ( CCW ) , semi-coiled and curly-1 ( both CW ) . Data from wild-type cells ( N = 52 cells , 203 tumbles ) . Values are mean ± SEM . Arrow thickness is proportional to the transition rate . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 01910 . 7554/eLife . 01916 . 020Figure 4—figure supplement 2 . Flagellar waveform transition sequences . Histogram of the sequence of flagellar waveforms when flagella motor rotation switches from CW to CCW ( N = normal , S = semi-coiled , C = curly-1 ) . Note that prior to CW rotation all flagella had the normal waveform . The majority of CCW-to-CW switches caused flagella to transition to the semi-coiled waveform ( 77% , 164/213 of all flagellar switching events ) . Of those flagella , most remained semi-coiled until the motor returned to CCW ( 71% , 117/164 ) , while some instead transitioned to the curly-1 waveform ( 29% , 47/164 ) . Some fraction of flagella ( 22% , 47/213 ) also skipped the semi-coiled waveform altogether and transitioned immediately from CCW to curly-1 . Those flagella never switch to the semi-coiled state . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 02010 . 7554/eLife . 01916 . 021Figure 4—figure supplement 3 . Flagellar waveform transition sequences . Decision tree showing the probability of each sequence of flagellar waveforms when flagella switch from CCW to CW rotation ( N = normal , S = semi-coiled , C = curly-1 ) . Arrow thickness is proportional to the probability of a specific transition . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 02110 . 7554/eLife . 01916 . 022Figure 4—figure supplement 4 . Tumble bias vs number of flagella per cell . Mean tumble bias of wild-type cells vs number of flagella per cell . Data were filtered to exclude outliers with a CW bias outside the range of 0–0 . 25 ( N = 61 cells remaining ) . Gray line shows the mean tumble bias from all cells . Error bars denote SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 02210 . 7554/eLife . 01916 . 023Table 3 . Model parametersDOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 023ParameterDescriptionValueSourcekccw->cwMotor switching rate from CCW→CW0 . 26 s−1Our datakcw->ccwMotor switching rate from CW→CCW1 . 7 s−1Our dataCBAverage clock-wise bias of wild-type motors0 . 13Our data<Y>Mean concentration of CheY-P2 . 59 µMFit to our dataωCharacteristic motor switching time0 . 5 sOur dataλ−1Transition rate from semi-coiled to curly-1 state0 . 68 s−1Our dataxFrom the model in Sneddon et al . ; number of flagella that must be normal for a run to occur ( while other flagella are curly-1 ) ( Sneddon et al . , 2012 ) Nflag ( variable in Figure 1—figure supplement 5 ) Our dataσ2Variance in [CheY-P]1 . 0 µM2Fit to our dataτCharacteristic time-scale of fluctuations in [CheY-P]0 . 2 sFit to our dataKdMidpoint of CW bias vs CheY-P response curve3 . 1 µM ( Cluzel et al . , 2000 ) HHill coefficient for CW bias vs CheY-P response curve10 . 3 ( Cluzel et al . , 2000 ) dtSimulation time steps0 . 001 s–Note: ω≡CBkccw→cw10 . 7554/eLife . 01916 . 024Table 4 . Flagella waveform transition ratesDOI: http://dx . doi . org/10 . 7554/eLife . 01916 . 024InitialFinalNormalSemi-coiledCurly-1Normal0 . 28 ± 0 . 03 s−10 . 08 ± 0 . 01 s−1Semi-coiled1 . 6 ± 0 . 2 s−10 . 54 ± 0 . 08 s−1Curly-11 . 8 ± 0 . 2 s−10 . 04 ± 0 . 02 s−1Transition rates between different flagellar waveforms: normal ( CCW ) , semi-coiled and curly-1 ( both CW ) . Data from wild-type cells ( N = 52 cells , 203 tumbles ) . Values are mean ± SEM . The experimental approach described here allows the simultaneous , long-term observation of flagellar activity and swimming behavior in a single cell . By imaging many individual tumbles ( N = 203 in wild-type cells ) , we are able to describe in great detail the underlying structure of a tumbling event . For instance , we can follow the sequence of flagellar waveforms that occurs when motors switch from CCW to CW rotation and back to CCW . Figure 4—figure supplement 2 shows the distribution of possible sequences of flagellar waveforms during tumbles . In particular , the sequence of states from normal to semi-coiled to curly-1 that we observed was described by Darnton et al . ( 2007 ) as a ‘canonical tumble’ . Although we cannot rule out that runs and tumbles in the optical trap are different in some ways than those in free swimming cells ( Min et al . , 2009 ) , our results are in qualitative agreement with these previous observations . Our measurements also reveal the relationship between the cell’s run/tumble state and the CCW/CW rotational state of its flagella . In a multi-flagellated wild-type cell , a single CW flagellum ( either semi-coiled or curly-1 ) is sufficient to induce a tumble , in agreement with the simple veto model ( Turner et al . , 2000; Darnton et al . , 2007 ) . However , the number of CW flagella during a tumble typically exceeds that expected from a cell with independently switching flagella ( Figure 2C ) . The high fraction of CW flagella during tumbles in wild-type cells is in qualitative agreement with previous measurements by Turner et al . ( 2000 ) , who observed that a majority of tumbles involved multiple flagella leaving the bundle . Our measurements using the CheY* strain provide an important piece of evidence linking inter-flagellar coupling to the chemotaxis network . We propose that fluctuations in the concentration of CheY-P are at the heart of wild-type E . coli behavior . In our theoretical analysis , the existence of temporal fluctuations was sufficient to explain all of our data . Stochastic simulations with and without CheY-P fluctuations ( representing wild-type and CheY* cells , respectively ) reproduced all of the observed differences between our two strains . Figure 4A summarizes how CheY-P fluctuations could lead to correlated flagellar switching . A well-known feature of the chemotaxis network is the sigmoidal relation between CW bias and CheY-P concentration ( Cluzel et al . , 2000; Yuan and Berg , 2013 ) . A consequence of this non-linearity is that the probability of CW rotation is highly sensitive , and can respond dramatically to fluctuations in CheY-P levels , provided their amplitude is sufficiently large . As shown in Figure 4A , when CheY-P concentration is high , the cell experiences a near-100% probability of CW rotation and multiple motors switch CW at approximately the same time . In contrast , when CheY-P concentration is low , the probability of any motor rotating CW is essentially zero . This mechanism can explain the elevated number of CW flagella involved in tumbles ( Figure 2C ) and correlation between flagella states ( Figure 2D ) . By contrast , in simulations where CheY-P level was held constant , flagellar switching was not as correlated and the majority of tumbles involved only a single CW flagellum ( Figure 4B ) . Despite the success of this model in reproducing our data , we must acknowledge that there is no direct experimental evidence to-date for the CheY-P fluctuations depicted in Figure 4A . Fluctuations in CheY-P have been inferred from experimental observations of CW bias in tethered-bead assays ( Korobkova et al . , 2004 ) . However , the fluctuations described in that study are different in their time scale and amplitude from what we found required to produce the observed correlations in flagellar rotational direction ( ‘Materials and methods’ ) . Future investigation will be essential to resolving this issue and will likely have to involve direct measurements of CheY-P temporal dynamics in individual cells . Such measurements are challenging , but the development of intra-cellular fluorescence sensors for kinase activity in the network ( Sourjik and Berg , 2002 ) provides a promising approach . As an alternative mechanism for the inter-flagellar correlations observed by Terasawa et al . ( 2011 ) ( and in the present work ) , Hu and Tu ( 2013 ) proposed that hydrodynamic interactions between nearby flagella could also generate correlations in their rotational direction . One consequence of their model is that the flagellar switching rates in a cell with a single flagellum will be different than those in multi-flagellated cells . However , this prediction is not borne out by our data . In our experiments , the number of flagella per cell did not have a significant effect on the switching rates between CCW and CW states , nor on the switching rates between different flagellar waveforms ( Figure 2—figure supplements 1 and 2 ) . While we cannot rule out the presence of hydrodynamic interactions between flagella , these must satisfy the strict requirement that switching rates remain independent of flagellar number . In light of these constraints , we believe a mechanism in which chemotaxis network fluctuations engender inter-flagella correlations to be more plausible . For cells with Nflag > 4 , we note that both strains appear to deviate from the generalized veto model ( Figure 3C ) . It is possible that hydrodynamic effects must be taken in consideration in cells with many flagella . Hydrodynamic coupling in the Hu model leads to a lower Neff , in the direction of the deviation . Alternatively , a mechanism as described by Sneddon et al . ( 2012 ) , in which cells with many flagella can run while some of its flagella rotate CW in the curly-1 state , could lead to a similar deviation ( Figure 1—figure supplement 5 ) . Finally , we must consider the possibility that the apparent deviation is due to systematic experimental error , since it is more difficult to determine visually the state of each flagellum on cells with many flagella . While a large number of studies have elucidated mathematical relationships between many of the components of the chemotaxis network ( Block et al . , 1982; Cluzel et al . , 2000; Sourjik and Berg , 2002; Shimizu et al . , 2010; Yuan et al . , 2012 ) , there have been few experimental studies devoted to the relationship between individual flagella and whole-cell swimming ( Turner et al . , 2000; Darnton et al . , 2007 ) . As a result , existing models of bacterial chemotaxis have made drastically different assumptions in order to describe the swimming behavior of the whole cell ( Bray et al . , 2007; Jiang et al . , 2010; Vladimirov et al . , 2010; Sneddon et al . , 2012 ) . To the best of our knowledge , the current study provides the first experimentally-derived mathematical relation between flagellar and whole-cell-swimming states . We propose that the details of this mapping are crucial for fully understanding bacterial chemotaxis . Results from recent theoretical models suggest that the details of flagellar mechanics can have significant effects on chemotactic drift . Turner et al . observed that , on average , the angular change in swimming direction upon tumbling increases as a function of the number of flagella that leave the bundle ( Turner et al . , 2000 ) . Vladimirov et al . showed that when this effect is incorporated into a theoretical model of bacterial chemotaxis , the chemotactic drift is nearly doubled ( Vladimirov et al . , 2010 ) . Our observations that multi-flagellated wild-type cells tumble significantly less than expected also implies that the cell’s swimming behavior ( and presumably its chemotactic response ) is robust against variations in the number of flagella ( Figure 4—figure supplement 4 ) . We hypothesize that this phenomenon may confer evolutionary advantages , in light of the large fluctuations in flagellar numbers within a cell population ( Figure 1—figure supplement 1 ) . If cells with many flagella did not behave like cells with fewer flagella , then they would spend the majority of their time tumbling , a behavior that would inhibit chemotaxis . E . coli thus appears to have developed a mechanism to achieve similar tumble biases with a wide range of flagellar number . TEM images were recorded using the JEOL 2100 cryo-Transmission Electron Microscope ( TEM ) at the Frederick Seitz Materials Research Laboratory Central Facilities at the University of Illinois at Urbana–Champaign , following the protocol of Saini et al . ( 2010 ) . Briefly , cells were grown as described above , and used without fluorescent labeling . Cells were fixed with glutaraldehyde and then placed on 200 Mesh Carbon Coated Copper grids ( Cat . # 182; Canemco , Lakefield , Canada ) , which were used as supports for sample loading and imaging . Images were taken at 200 kV with a camera exposure lasting 1 s . Finally , images were contrast adjusted and an image dilation was performed in Matlab to make flagella more visible . The distribution of flagella per cell is shown in Figure 1—figure supplement 1 .	Escherichia coli is a rod-shaped bacterium commonly found in the lower intestines of humans and other warm-blooded animals . While most strains of E . coli are harmless , including most of those found in the human gut , some can cause diseases such as food poisoning . Due to its close association with humans and the fact that it is easy to grow and work with in the laboratory , E . coli has been intensively studied for over 60 years . Many bacteria are capable of ‘swimming’ by using one or more flagella . These rotating whip-like structures are each driven by a reversible motor , and they act a bit like a propeller on a boat . While some bacteria have only a single flagellum , others , such as E . coli , have multiple flagella distributed over the cell surface . Rotating all their flagella in a counterclockwise direction allows the bacterium to swim—and it has been proposed that the clockwise movement of at least one flagellum will cause the bacterium cell to stop swimming and start tumbling . E . coli is able to control the time it spends swimming or tumbling to move towards a nutrient , such as glucose , or away from certain harmful chemicals . However , the details of how the number of flagella and the direction of rotation ( clockwise or counterclockwise ) influence the motion of the bacterium are not fully understood . Now , Mears et al . have used ‘optical tweezers’ to immobilize individual E . coli cells under a microscope , and then track both their swimming behavior and the movements of their flagella . This revealed that the individual flagella on the same cell tend to move in a coordinated way . Therefore , whilst tumbling could be caused by a single flagellum stopping swimming behavior , it often involved a concerted effort by many of the cell’s flagella . After observing that E . coli cells with more flagella spent less time tumbling than would be predicted if a single flagella always ‘vetoed’ swimming , Mears et al . propose a new mathematical relationship between the number of flagella on the cell , the direction of rotation , and the resulting probability that the cell will tumble . This work shows that swimming behavior in bacteria is less affected by variations in the number of flagella than previously thought—and this phenomenon may provide evolutionary advantages to E . coli . The next step is to explore the mechanism by which bacteria coordinate their flagella .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease" ]	2014	Escherichia coli swimming is robust against variations in flagellar number
We have examined the role of Fam60a , a gene highly expressed in embryonic stem cells , in mouse development . Fam60a interacts with components of the Sin3a-Hdac transcriptional corepressor complex , and most Fam60a–/– embryos manifest hypoplasia of visceral organs and die in utero . Fam60a is recruited to the promoter regions of a subset of genes , with the expression of these genes being either up- or down-regulated in Fam60a–/– embryos . The DNA methylation level of the Fam60a target gene Adhfe1 is maintained at embryonic day ( E ) 7 . 5 but markedly reduced at E9 . 5 in Fam60a–/– embryos , suggesting that DNA demethylation is enhanced in the mutant . Examination of genome-wide DNA methylation identified several differentially methylated regions , which were preferentially hypomethylated , in Fam60a–/– embryos . Our data suggest that Fam60a is required for proper embryogenesis , at least in part as a result of its regulation of DNA methylation at specific gene promoters . The Sin3a protein is a core component of a mammalian transcriptional corepressor complex that includes histone deacetylases ( Hdacs ) ( Hassig et al . , 1997; Zhang et al . , 1997 ) . Although Sin3a does not bind DNA on its own , it provides a scaffold for several transcription factors with specific DNA binding activities and thereby promotes the recruitment of Hdacs to and consequent repression of specific target genes . Mice lacking Sin3a die during embryogenesis around the time of implantation ( McDonel et al . , 2012 ) , suggesting that the Sin3a-Hdac complex is essential for early embryonic development . Although this complex was initially thought only to repress gene expression , it can also stimulate transcription in a manner dependent on cellular context ( Icardi et al . , 2012 ) . Gene expression is also regulated by DNA methylation . In mammals , DNA methylation occurs predominantly at CpG sequences , with ~70% of gene promoters in mammalian genomes containing CpG islands . In general , CpG islands of transcriptionally active promoters are not methylated , whereas methylation of CpG in a promoter is associated with transcriptional silencing . During mouse development , the methylation pattern of genomic DNA is established at the peri-implantation stage by the de novo methyltransferases Dnmt3a and Dnmt3b . Once established , this methylation pattern is faithfully maintained by Dnmt1 during DNA replication . The precise formation and maintenance of the DNA methylation pattern are essential for mouse embryogenesis , given that embryos lacking Dnmt enzymes develop pronounced morphological defects and die in utero ( Li et al . , 1992; Okano et al . , 1999 ) . Methylated DNA can undergo demethylation , a process mediated by the Tet family of 5-methylcytosine dioxygenases that catalyze the conversion of 5-methylcytosine ( 5mC ) to 5-hydroxymethylcytosine ( 5hmC ) ( Tahiliani et al . , 2009 ) . Demethylation of DNA by Tet proteins serves to activate gene promoters , but these proteins are also able to regulate gene expression via histone modification ( Wu et al . , 2011; Wu and Zhang , 2017 ) . Strict regulation of Tet proteins is also required for proper development , given that mouse embryos lacking Tet1 and Tet2 as well as chimeric embryos that include cells deficient in Tet1 , Tet2 , and Tet3 become malformed ( Dawlaty et al . , 2014; 2013 ) . The mechanisms responsible for such Tet regulation have remained unknown , however . We have previously identified Fam60a ( Sinhcaf ) as a gene of unknown function ( gene 226 reported in [Saijoh et al . , 1996] ) that is highly expressed in mouse embryonic stem ( ES ) cells and whose expression in these cells is down-regulated on their differentiation . We have now examined the role of Fam60a in mouse development . Our data show that Fam60a is an embryonic component of the Sin3a-Hdac corepressor complex and regulates gene expression at least in part by regulating DNA methylation at a subset of gene promoters . To examine the biochemical function of Fam60a , we generated mice harboring a Fam60a::Venus BAC ( bacterial artificial chromosome ) transgene ( Figure 1—figure supplement 1A; as described below , the Fam60a-Venus fusion protein encoded by this transgene is functional ) . Immunostaining revealed that the Fam60a-Venus protein was present in nuclei of embryonic day ( E ) 9 . 5 embryos harboring the transgene , and that Fam60a was localized to the nucleus of undifferentiated P19 ( mouse embryonic carcinoma ) cells ( Figure 1—figure supplement 2 ) . To identify proteins that might interact with Fam60a , we prepared nuclear extracts from E10 . 5 embryos harboring the Fam60a::Venus transgene under three different conditions , subjected the extracts to immunoprecipitation with antibodies to green fluorescent protein ( GFP ) , and analyzed the precipitated proteins by mass spectrometry . The major proteins identified were Arid4a , Arid4b , Sin3a , Sap130 , Hdac1 , Hdac2 , Suds3 , and Brms1l ( Figure 1A ) , all of which are components of the Sin3a-Hdac corepressor complex ( Cunliffe , 2008; Fleischer et al . , 2003; Grzenda et al . , 2009; Nikolaev et al . , 2004; Shiio et al . , 2006; Silverstein and Ekwall , 2005 ) . Arid4a and Arid4b were not detected if nuclear extracts were prepared with radioimmunoprecipitation assay ( RIPA ) buffer ( Figure 1A ) , the most stringent of the three conditions used , suggesting that these proteins interact weakly with the other components of the Sin3a-Hdac complex ( Lai et al . , 2001 ) . Further co-immunoprecipitation analysis confirmed that Fam60a interacts with components of the Sin3a-Hdac complex . Immunoprecipitates prepared from nuclear extracts of E10 . 5 wild-type ( WT ) embryos with antibodies to Fam60a were thus found to contain Sin3a , Hdac1 , and Hdac2 ( Figure 1B ) . In addition , these three proteins were detected in immunoprecipitates prepared from nuclear extracts of Fam60a::Venus transgenic embryos with antibodies to GFP ( Figure 1B ) . Immunoprecipitates prepared from undifferentiated P19 cells with antibodies to Fam60a also contained Sin3a and Hdac1 but not Hdac2 ( Figure 1B ) . Reciprocal co-immunoprecipitation analysis with nuclear extracts of E10 . 5 WT embryos revealed that Fam60a was present in immunoprecipitates prepared with antibodies to Sin3a or to Hdac1 but not in those prepared with antibodies to Hdac2 ( Figure 1B ) , suggesting that the association between Hdac2 and Fam60a is relatively weak . Together , these data indicated that Fam60a is a component of the Sin3a-Hdac corepressor complex in developing mouse embryos and in undifferentiated P19 cells . This is consistent with recent findings that Fam60a is a core subunit of a variant Sin3a complex in ES cells ( Streubel et al . , 2017 ) . Formation of the Sin3a-Hdac complex was not affected by the absence of Fam60a , however , given that Hdac1 , Hdac2 , and RbAp46/48 were co-immunoprecipitated with Sin3a from Fam60a–/– ES cells ( Figure 1—figure supplement 3 ) . To shed light on the physiological function of Fam60a , we first examined the pattern of Fam60a expression during mouse embryogenesis . Expression of Fam60a was ubiquitous at E9 . 5 , but it gradually became restricted to a subset of cells as development proceeded ( Figure 2—figure supplement 1 ) . At E12 . 5 , Fam60a expression was thus apparent in the neural tube , neural crest cells , lung , pancreas , and intestine , but not in liver . Epithelial cells of the intestinal tract showed a high level of Fam60a expression at E15 . 5 ( Figure 2A ) , and intervilli of the intestinal tract continued to express Fam60a at E17 . 5 ( Figure 2B and C ) . In adult mice , Fam60a expression was maintained in crypts of the duodenum ( Figure 2D–F ) . Given that intestinal stem and progenitor cells reside in crypts , we examined the fate of Fam60a+ cells in crypts by administering tamoxifen to adult mice harboring a Fam60a-CreERT2 transgene and lacZ reporter gene . Examination of the mice at 1 , 3 , and 5 days after tamoxifen injection revealed that LacZ+ cells were present at the base of intestinal villi at 1 day and that they subsequently migrated toward the tip of the villi during the next 4 days ( Figure 2G–I ) . These data thus suggested that Fam60a is expressed in a subset of cells including somatic stem cells in the intestine . We next generated mice lacking Fam60a . Two types of mutant allele were generated: Fam60a– and Fam60aβgeo ( Figure 3—figure supplement 1A and B ) . Fam60a–/– and Fam60aβgeo/βgeo mice showed indistinguishable phenotypes , suggesting that both alleles are functionally null , with subsequent analyses being performed with Fam60a–/– mice unless indicated otherwise . Both types of heterozygote also appeared indistinguishable from WT mice . We confirmed that Fam60a mRNA and Fam60a protein were absent in Fam60a–/– embryos ( Figure 3—figure supplement 1C and D , Figure 3—source data 1 ) . Fam60a–/– mice were born at a frequency much lower than that expected . They were detected at the expected frequency at E9 . 5 and E10 . 5 , but their number started to decline thereafter and was greatly decreased at E18 . 5 ( Supplementary file 1 ) . Examination of Fam60a–/– embryos at E13 . 5 revealed that many visceral organs including the heart , lungs , liver , and gut were markedly smaller than those of WT embryos ( Figure 3A–C ) . In particular , hypoplasia of the right ventricle of the heart was apparent , and a ventricular septum defect was also frequently observed , in Fam60a–/– embryos ( Figure 3D ) . Fam60a was expressed in the developing heart , predominantly in the right ventricle and outflow tract , of WT embryos at E13 . 5 ( Figure 3G ) . Many of the Fam60a–/– embryos that survived to E18 . 5 manifested transposition of the great arteries , double-outlet right ventricle , and ventricular septum defects as well as spleen hypoplasia , incomplete lobulation of the lungs , and abnormal rotation of the gut ( Figure 3—figure supplements 2 and 3 ) . Although these abnormalities appeared reminiscent of laterality defects , left-right asymmetric expression of Pitx2 was maintained at E8 . 0 ( data not shown ) , suggesting that the abnormalities are not directly due to impaired left-right patterning . The Fam60a–/– embryos already showed morphological abnormalities including growth retardation as well as cardiac ( shortening of the outflow tract ) and neural tube defects at E9 . 5 ( Figure 3H and I ) . Given that most of the mutant embryos manifested growth retardation , we examined the rate of cell proliferation in various tissues of embryos at E9 . 0 to E9 . 5 by labeling with bromodeoxyuridine ( BrdU ) and counting of BrdU-positive cells ( Figure 3—figure supplement 4 , Figure 3—source data 2 ) . The extent of cell proliferation was significantly reduced in the septum transversum , secondary heart field ( SHF ) , and proepicardium , whereas it was unaffected in the heart ventricle and slightly increased in the neural tube , of Fam60a–/– embryos compared with control embryos . Given that the outflow tract is derived from SHF cells ( Buckingham et al . , 2005 ) and that Fam60a is expressed in SHF-derived regions of WT embryos ( Figure 3G ) , the reduced proliferation rate of SHF cells may give rise to the shortening of the outflow tract and subsequent right ventricle hypoplasia apparent in the mutant embryos . These results thus suggested that Fam60a is required for cell proliferation and organogenesis in mouse embryos . Given that the Sin3a-Hdac complex is thought to repress gene expression by binding to promoter regions , we examined the global gene expression pattern in Fam60a–/– embryos by RNA-sequencing ( RNA-seq ) analysis . Comparison of Fam60a–/– and WT embryos at E9 . 5 revealed that the expression of 558 genes was up-regulated and that of 172 genes was down-regulated in the mutant embryos ( Figure 4A and B , Figure 4—source datas 1 and 2 ) . Gene ontology analysis revealed that the expression of genes related to the response to nutrients or to extracellular matrix organization was increased , whereas that of those related to lipid biosynthesis was decreased , in the mutant embryos ( Figure 4—figure supplement 1 ) . These data suggested that Fam60a regulates gene expression in both a negative and positive manner , but predominantly in a negative manner , in E9 . 5 embryos . We also performed chromatin immunoprecipitation followed by deep sequencing ( ChIP-seq ) analysis with E9 . 5 Fam60a::Venus transgenic embryos and antibodies to GFP to identify Fam60a binding sites in the genome . The Fam60a-Venus fusion protein encoded by this transgene was able to rescue the defects of Fam60a mutant mice ( Figure 1—figure supplement 1B and C ) , suggesting that it is fully functional . Approximately 17 , 000 and 14 , 000 peaks were detected in two independent experiments ( ChIP-seq1 and ChIP-seq2 , respectively ) , with ~80% of the peaks being localized at gene loci , in particular in the vicinity of transcription start sites ( TSSs ) ( Figure 5A and B; Figure 5—figure supplement 1A and B , Figure 5—source data 1 ) . This distribution pattern was highly similar to that previously determined for Sin3a ( Bowman et al . , 2014 ) . Co-immunoprecipitation analysis of E10 . 5 transgenic embryos revealed that Fam60a-Venus interacts with Ing2 ( Figure 5—figure supplement 1C ) , a protein that binds to Lys4-trimethylated histone H3 ( H3K4me3 ) , suggesting that Fam60a is recruited predominantly to the promoters of transcribed genes . Examination of the TSS region ( between –3 kb and +3 kb relative to the TSS ) of all genes resulted in the identification of 7989 genes that reproducibly showed at least one Fam60a binding site in this region ( Figure 5C ) , suggesting that these genes may be directly regulated by Fam60a . Among the 558 up-regulated and 172 down-regulated genes identified in Fam60a–/– embryos , 245 and 45 genes , respectively , had at least one Fam60a binding peak in the TSS region ( Figure 4A ) . Given that 74% ( 127/172 ) of the down-regulated genes lacked a Fam60a binding site in this region , the change in expression of most of the down-regulated genes was likely due to a secondary effect of Fam60a loss . We selected for further analysis 18 genes from the 290 ( 245 + 45 ) identified genes on the basis of their large fold change in expression in the mutant embryos as revealed by RNA-seq ( Figure 4B ) . Reverse transcription and quantitative polymerase chain reaction ( RT-qPCR ) analysis confirmed significant differences in expression level for at least six of these putative Fam60a target genes between WT and Fam60a–/– embryos at E9 . 5 , with the expression of Leng9 , Adhfe1 , Mxd3 , Dchs1 , and Nagk being up-regulated and that of Gt ( ROSA ) 26Sor being down-regulated in the mutant ( Figure 4C , Figure 4—source data 3 ) . The expression of some of these up-regulated genes ( such as Leng9 , Dchs1 , and Nagk ) was also increased in Fam60a–/– ES cells compared with control ES cells ( Figure 4—figure supplement 2A and B , Figure 4—source data 4 ) . ChIP-qPCR analysis for three of the up-regulated genes ( Adhfe1 , Nagk , Dchs1 ) also revealed the association of their promoter regions with Fam60a-Venus and Sin3a in E9 . 5 transgenic and WT embryos , respectively ( Figure 5D , Figure 5—source data 2 ) . The Fam60a binding peaks identified by ChIP-seq analysis in the TSS regions of Adhfe1 , Nagk , and Dchs1 are shown in Figure 5E . Although the Sin3a-Hdac complex possesses histone-deacetylating activity , the level of Lys9-acetylated histone H3 ( AcH3K9 ) at the promoter regions of Fam60a target genes ( Adhfe1 , Nagk , Dchs1 ) did not differ between WT and Fam60a–/– embryos ( Figure 5—figure supplement 2 , Figure 5—source data 3 ) . However , similar analysis with Fam60a–/– ES cells revealed that the level of AcH3K9 at the promoter regions of three such genes ( Leng9 , Dchs1 , Nagk ) was increased ( Figure 4—figure supplement 2C , Figure 4—source data 5 ) . We examined the molecular phylogeny of Fam60a with a sequence data set containing invertebrate homologs as well as a paralog , designated Fam60b ( Figure 6A ) . The phylogenetic tree revealed the gene duplication event that gave rise to Fam60a and Fam60b in the early vertebrate lineage before the radiation of jawed vertebrates , likely during the well-studied genome expansion ( 2R-WGD , two-round whole genome duplication ) that occurred in this period . It also highlighted the origin of the preduplication ortholog Fam60 in the early metazoan era . Analysis of the families of genes encoding Sin3 , Tet , and Dnmt proteins as well as the presence or absence of DNA methylation in individual species suggested an association of Fam60a with DNA methylation , Tet , and Sin3 ( Figure 6B ) . Fam60a proteins of ~220 amino acid residues were thus found in all vertebrates examined , and Fam60a orthologs were also detected in insects but not in nematodes or yeasts ( Figure 6—figure supplement 1 ) ( Smith et al . , 2012 ) . DNA methylation and Tet proteins are also conserved from humans to insects but not in nematodes or yeasts , whereas Sin3 is more widely conserved from yeasts to humans . The association of Fam60a with DNA methylation and Tet , together with the fact that the Sin3a-Hdac complex interacts with methylation-regulating proteins such as methylated CpG binding protein2 ( MeCP2 ) , Dnmt1 , and Tet1 ( Nan et al . , 1998; Williams et al . , 2011 ) , suggested that Fam60a might regulate Tet-mediated DNA demethylation . We tested this possibility in NIH3T3 cells transfected with a doxycycline-inducible expression vector for FLAG epitope–tagged Tet1 and with either an expression vector for both Fam60a and Venus or the corresponding empty vector . Exposure of the transfected cells to doxycycline thus induced the expression of Tet1 in the absence or presence of that of Fam60a ( Figure 6—figure supplement 2 ) . In the absence of Fam60a , 83% of FLAG-Tet1+ cells were positive for 5hmC ( that is , only 17% of FLAG-Tet1+ cells remained negative for 5hmC ) at 24 hr after the administration of doxycycline , suggestive of the efficient conversion of 5mC to 5hmC by FLAG-Tet1 . In the presence of Fam60a , however , 55% of FLAG-Tet1+ cells remained negative for 5hmC ( Figure 6C–F , Figure 6—source data 1 and 2 ) , suggesting that Fam60a might inhibit Tet1 activity . Recruitment of Tet1 to the promoter regions of Fam60a target genes ( Leng9 , Dchs1 , Nagk ) was not affected in Fam60a–/– ES cells ( Figure 4—figure supplement 2D , Figure 4—source data 6 ) , suggesting that Fam60a negatively regulates Tet1 activity without affecting its recruitment to promoter regions . Given that our results suggested that Fam60a inhibits Tet1 activity in cultured cells , we next determined whether DNA methylation is affected in Fam60a–/– mouse embryos . Bisulfite sequencing of the promoter regions of Nagk and Leng9 revealed little or no DNA methylation in WT or Fam60a–/– embryos at E9 . 5 ( Figure 7—figure supplement 1 ) , even though our ChIP analyses showed that Fam60a-Venus was recruited to these promoter regions in transgenic embryos . In contrast , the promoter region of Adhfe1 was found to be hypomethylated in Fam60a–/– embryos , with a methylation level of 4 to 10% compared with a value of ~20% in WT embryos at E9 . 5 ( Figure 7 , Figure 7—source data 1 ) . This hypomethylation might have been due to reduced de novo DNA methylation or increased demethylation mediated by Tet . To distinguish between these possibilities , we examined methylation of the Adhfe1 promoter at earlier developmental stages , given that de novo DNA methylation occurs predominantly before implantation . No significant difference in methylation was observed between WT and Fam60a–/– embryos at E7 . 5 , after which the methylation level of this promoter gradually decreased in the mutant embryos ( Figure 7 , Figure 7—source data 1 ) . These results suggested that impaired maintenance of methylation or increased demethylation is responsible for the hypomethylation of the Adhfe1 promoter in Fam60a–/– embryos , consistent with our observation that Fam60a inhibited Tet1 activity in cultured cells . Providing further support for this notion , hydroxymethyl DNA immunoprecipitation ( hMeDIP ) analysis revealed 5hmC deposition at almost all Fam60a target gene promoters examined in WT embryos ( Figure 7—figure supplement 2 , Figure 7—source data 2 ) . Hypomethylation was not detected at the imprinting control regions of Kcnq1ot1 or Peg3 in Fam60a–/– embryos ( Figure 7—figure supplement 3 ) . Together , these findings suggested that Fam60a regulates Tet-mediated demethylation at a subset of gene promoters . To verify the role of Fam60a in regulation of DNA methylation , we examined the methylation status of promoters , CpG islands , and CpG shores in the genome of Fam60a–/– and WT embryos at E9 . 5 . These target regions were captured , subjected to bisulfite conversion , and sequenced with a next-generation sequencer . The overall methylation level of CpG sites in the captured DNA was around 45% and showed a similar distribution pattern in both Fam60a–/– and WT embryos ( Supplementary file 2 , Figure 8—figure supplement 1 ) . Given that genome-wide DNA methylation level did not appear to be affected by the absence of Fam60a , we first examined DNA methylation levels over Fam60a-bound promoters ( ~8000 promoters ) in the wild-type and Fam60a–/– embryos . Hypomethylation was commonly observed at the Fam60a-binding regions , but there was no obvious difference in the profile between the wild-type and Fam60a–/– embryos ( Figure 8—figure supplement 2 ) . We next examined if the DNA methylation level was affected in a subset of gene promoters , by focusing on differentially methylated regions ( DMRs ) . 7245 DMRs were detected with average changes of DNA methylation 11 . 87 and 10 . 99% for hyper- and hypomethylated DMRs , respectively ( Figure 8—figure supplement 3 ) . Among the 7245 DMRs detected , 3049 and 4196 regions were hyper- and hypomethylated , respectively , in Fam60a–/– embryos , with 388 hypermethylated DMRs ( 12 . 7% ) and 1257 hypomethylated DMRs ( 30 . 0% ) being found to overlap with Fam60a binding regions ( Table 1 ) . Among the top 500 hyper- and hypomethylated DMRs showing the largest differences in methylation level between mutant and WT embryos , 83 of the hypermethylated DMRs ( 16 . 6% ) and 254 of the hypomethylated DMRs ( 50 . 8% ) contained Fam60a binding sites ( Table 1 ) , suggestive of a preferential association of Fam60a binding sites with hypomethylated DMRs . The promoter of Adhfe1 , which was found to be hypomethylated in Fam60a–/– embryos ( Figure 7 ) , was included in the top 500 hypomethylated DMRs ( Figure 8—source data 1 ) . We next examined the positions of the top 500 hypermethylated and top 500 hypomethylated DMRs in the genome . The distributions of these two types of region differed , with hypermethylated DMRs being preferentially located in exonic regions of genes at 5 to 50 kb downstream of the TSS ( Figure 8A and C ) , whereas most hypomethylated DMRs were located in intronic regions at 0 to 5 kb downstream of the TSS ( Figure 8B and D ) . The distribution pattern of hypomethylated DMRs ( Figure 8B ) was similar to that of Fam60a binding sites ( Figure 5B and Figure 5—figure supplement 1B ) . These data thus suggested that Fam60a is associated with DNA methylation status in mouse embryos . Fam60a is expressed ubiquitously during mouse embryonic development until at least E9 . 5 , after which its expression gradually becomes restricted to a subset of cells , including those engaged in proliferation . In the adult mouse , Fam60a is expressed in stem cells located in intestinal crypts , suggesting that its expression may be associated with differentiation potential . Consistent with this notion , Fam60a knockout mice manifest growth retardation in visceral organs . Gene ontology analysis revealed that genes whose expression is dysregulated in Fam60a–/– embryos include those related to the response to nutrients , extracellular matrix organization , and lipid biosynthesis , suggesting that disruption of these processes may contribute to the retardation of organ growth apparent in the mutant embryos . A search for Fam60a-interacting proteins identified the Sin3a-Hdac transcriptional corepressor complex . The stoichiometry of Fam60a and components of this complex recovered in immunoprecipitates ( Figure 1 ) suggested that most Fam60a in a given cell is associated with the complex . Fam60a may therefore function in association with the Sin3a-Hdac complex . Whereas Sin3a knockout mice die during embryogenesis around the time of implantation ( McDonel et al . , 2012 ) , Fam60a–/– embryos develop until later stages . It is thus possible that Sin3a has functions independent of Fam60a , including functions in multiple protein complexes , or that the earlier defects of Sin3a knockout mice are due to the lack of this protein in oocytes . Fam60a was recently shown to be a core subunit of a variant Sin3a complex in ES cells that includes Tet1 and Ogt ( Streubel et al . , 2017 ) . In general , the Sin3a-Hdac complex is thought to repress gene expression via histone deacetylation . However , this complex can also facilitate transcriptional activation in a manner dependent on cellular context ( Suganuma and Workman , 2013; Icardi et al . , 2012 ) . Indeed , we found that the expression of many genes was either up-regulated or down-regulated in Fam60a–/– embryos . Fam60a may therefore contribute not only to the transcriptional corepressor activity of the Sin3a-Hdac complex but also to its promotion of transcriptional activation . Fam60a likely does not serve as a simple regulator of Hdac activity , given that the level of histone acetylation at Fam60a target gene promoters did not differ between WT and Fam60a–/– embryos . Phylogenetic analysis revealed a wide taxonomic distribution of the ancestral Fam60 gene in eumetazoans and a duplication of this gene during early vertebrate evolution that gave rise to Fam60a and Fam60b paralogs . The absence of Fam60 and Tet genes as well as of DNA methylation in both Caenorhabditis elegans and yeasts suggests that Fam60a may contribute to Sin3a function related to DNA methylation and Tet . Consistent with this possibility , Sin3a is known to interact with MeCP2 , Dnmt1 , and Tet1 ( Nan et al . , 1998; Williams et al . , 2011 ) . Of note , Tet proteins play a role in demethylation of evolutionarily conserved gene enhancers during the phylotypic period of early development ( Bogdanović et al . , 2016 ) . Adhfe1 , whose promoter was found to be hypomethylated in Fam60a–/– embryos , appears to be a typical gene regulated by Fam60a and Tet activity . Expression of Adhfe1 is thus normally repressed because of the methylation of its promoter that results from Fam60a-mediated inhibition of Tet activity , but it is up-regulated in Fam60a–/– embryos because of the promoter hypomethylation that results from the absence of Fam60a . Other genes whose expression was up-regulated in Fam60a–/– embryos ( such as Leng9 and Nagk ) showed almost no DNA methylation in their promoter regions in either WT or mutant embryos , even though Fam60a-Venus was efficiently recruited to these promoters in transgenic embryos . In addition to functioning as DNA demethylases , Tet proteins associate with Sin3a-Hdac and act as transcriptional repressors in a manner independent of their demethylating activity ( Williams et al . , 2011; Zhang et al . , 2015 ) . Up-regulation of genes such as Leng9 and Nagk in Fam60a–/– embryos may thus be due to the lack of the latter function of Tet proteins . A genome-wide search for Fam60a binding sites revealed that Fam60a is recruited to gene promoter regions that overlap with CpG islands . In general , such CpG island promoters of transcriptionally active genes are enriched in H3K4me3 . Consistent with the genomic localization of Fam60a , we found that Fam60a interacts with Ing2 , which is known to bind to H3K4me3 ( Goeman et al . , 2008 ) . Tet proteins interact with Sin3a and are thought to localize to CpG island promoters in order to maintain the CpG islands unmethylated , with such promoters often being marked with H3K4me3 . These observations suggest that Fam60a is localized mostly to transcribed gene promoters , where it regulates the level of gene expression both negatively and positively via Sin3a and Tet . How might Fam60a regulate Tet activity ? It may inhibit dioxygenase enzymatic activity or impair recruitment of Tet to DNA . In this regard , PGC7 ( also known as Stella ) protects the female pronucleus from Tet3-dependent conversion of 5mC to 5hmC in mouse zygotes as well as inhibits the binding of Tet3 to chromatin in mouse ES cells ( Nakamura et al . , 2012 ) . Fam60a may similarly affect the binding of Tet to chromatin , although this is unlikely given that recruitment of Tet1 to Fam60a target genes was not affected in Fam60a–/– ES cells . Alternatively , Fam60a may physically interact with Tet proteins and inhibit their activity . However , given that Tet proteins were not identified in our search for Fam60a-interacting proteins , it is unlikely that Fam60a directly interacts with Tet . Further characterization of the mechanisms by which Fam60a affects the function of Sin3a and Tet should provide new insight into gene regulation during embryogenesis . Fam60aβgeo , a mutant allele of Fam60a in which an internal ribosome entry site ( IRES ) –βgeo cassette and a loxP site are inserted in intron 4 and intron 1 , respectively , was generated by gene targeting in mouse ES cells ( Figure 2—figure supplement 1A ) . A Fam60aflox allele was subsequently generated with the use of the CAG-Flpe transgene ( Kanki et al . , 2006 ) , and a Fam60a– allele lacking exons 2 to 4 was generated with the use of the CAG-Cre transgene ( Sakai and Miyazaki , 1997 ) . Both Fam60aβgeo and Fam60a– alleles are functionally null . Mutant mice were maintained on the 129/C57B6 mixed background . PCR primers for genotyping were Fam60a-5A ( 5′-ATATGCTGCTAGGTGCCACAG-3′ ) , Fam60a-3A ( 5′-TTCTCTACTCCATAGCACAGG-3′ ) , and Fam60a-3C ( 5′-CTACTGTGGTCACAAGCAGAC-3′ ) . A BAC transgene ( Fam60a::Venus ) encoding a Fam60a-Venus fusion protein was constructed from mouse BAC clone RP23-100A22 with the use of a BAC recombination system ( Figure 3—figure supplement 1A ) ( Copeland et al . , 2001 ) . The Fam60a-Venus protein , in which Venus is fused to the COOH-terminus of Fam60a , is functional , given that the transgene is able to rescue the phenotype of Fam60a mutant mice ( Figure 3—figure supplement 1C ) . A BAC transgene ( Fam60a-CreERT2 ) was constructed by inserting CreERT2 into the Fam60a BAC clone . P19 embryonal carcinoma cell line ( McBurney and Rogers , 1982 ) is a gift from Michael McBurney ( University of Ottawa ) . NIH3T3 Tet-On 3 G cell line ( 631197 , Clontech ) was purchased from Clontech , Takara-bio ( Kyoto , Japan ) . E10 . 5 embryos harboring the Fam60a::Venus transgene were recovered in PBS for the preparation of nuclear extracts . The embryos were passed through a 70 µm cell strainer with a plunger , and the cells were allowed to swell by incubation in buffer A ( 10 mM Hepes-KOH ( pH 7 . 9 ) , 10 mM KCl , 1 . 5 mM MgCl2 , 0 . 1 mM EGTA , 1 mM dithiothreitol , and Roche complete protease inhibitor cocktail ) for 15 min on ice before homogenization with 20 strokes of a loose-fitting pestle in a Dounce homogenizer . Nonidet P-40 was then added to the homogenate at a final concentration of 0 . 1% , and another 20 strokes of the pestle were applied . The homogenate was centrifuged at 960 × g for 5 min at 4°C , and the resulting nuclear pellet was suspended and incubated for 3 hr at 4°C either in RIPA buffer ( 50 mM Tris-HCl ( pH 8 . 0 ) , 150 mM NaCl , 2 mM EDTA , 1% Nonidet P-40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , 1 mM dithiothreitol , and Roche complete protease inhibitor cocktail ) , in buffer C ( 20 mM Hepes-KOH ( pH 7 . 9 ) , 400 mM NaCl , 0 . 1 mM EDTA , 0 . 1 mM EGTA , 0 . 1% Nonidet P-40 , 1 mM dithiothreitol , and Roche complete protease inhibitor cocktail ) , or in nondenaturing lysis buffer containing Benzonase nuclease ( 20 mM Tris-HCl ( pH 8 . 0 ) , 137 mM NaCl , 2 mM EGTA , 1 . 5 mM MgCl2 , 10% glycerol , 1 mM dithiothreitol , Benzonase nuclease ( 125 U; 70 , 446–3 , Novagen ) , and Roche complete protease inhibitor cocktail ) . The samples were centrifuged at 18 , 000 × g for 10 min at 4°C , and the resulting supernatants ( nuclear extracts ) were incubated with Dynal Protein G beads ( Invitrogen ) for 3 hr at 4°C . After removal of the beads , the extracts were divided into two halves . One half was incubated for 3 hr at 4°C with Dynal Protein G beads conjugated with antibodies to GFP , whereas the other half was incubated with identical antibody-conjugated beads that had been previously exposed to recombinant GFP ( ab84191 , Abcam ) to mask the antigen binding site . Proteins that bound to the beads were eluted by incubation for 30 min at 37°C with 1 × SDS sample buffer not containing dithiothreitol . They were then fractionated by SDS-polyacrylamide gel electrophoresis and silver-stained . Target proteins were identified by liquid chromatography and tandem mass spectrometry with a nano-UPLC Q-TOF MS/MS system ( SYNAPT G2 , Waters ) . Nuclear extracts prepared from E10 . 5 embryos or undifferentiated P19 cells with RIPA buffer as described above were incubated for 3 hr at 4°C first with Dynal Protein G beads alone and then with antibody-conjugated beads . Proteins that bound to the antibody-conjugated beads were eluted by incubation for 30 min at 37°C with 1 × SDS sample buffer not containing dithiothreitol , fractionated by SDS-polyacrylamide gel electrophoresis , and transferred to a polyvinylidene difluoride membrane . The membrane was then subjected to immunoblot analysis with primary antibodies , horseradish peroxidase–conjugated secondary antibodies , and ECL Plus reagents ( RPN2133 , Amersham ) . Tamoxifen ( 6 mg; T5648 , Sigma-Aldrich ) in 1 ml of corn oil ( C8267 , Sigma-Aldrich ) was administered orally to Fam60a-CreERT2::ROSA26RlacZ mice at the age of 8 weeks age . One , 3 , or 5 days after tamoxifen administration , mice were killed and the duodenum was removed and then fixed overnight at 4°C in phosphate-buffered saline ( PBS ) containing 1% paraformaldehyde , 0 . 2% glutaraldehyde , and 0 . 02% Nonidet P-40 . Expression of the lacZ transgene was detected by staining with X-gal as described previously ( Saijoh et al . , 1999 ) . Embryos were dissected in PBS and fixed with 4% paraformaldehyde . In situ hybridization was performed with whole-mount preparations ( Sakai et al . , 2001 ) or sections ( Yashiro et al . , 2000 ) . The 3′untranslated region of Fam60a was used as a probe for in situ hybridization . For histological analysis , embryos were fixed with 4% paraformaldehyde , dehydrated , and embedded in paraffin . Serial sections ( thickness , 7 µm ) were stained with hematoxylin-eosin according to standard procedures . Antibodies to Fam60a ( α-E15W ) were generated in rabbits by injection of a synthetic peptide corresponding to the COOH-terminal region of the mouse protein ( EEQGPAPLPISTQEW ) and were affinity-purified . Additional antibodies included control rabbit IgG ( Kamiya Biomedical or Thermo Fisher Scientific ) , conformation-specific mouse monoclonal antibodies to rabbit IgG ( #3678 , Cell Signaling ) that can avoid detection of denatured rabbit IgG used for immunoprecipitation , as well as rabbit polyclonal antibodies to GFP ( 598 , MBL International ) , to Hdac1 ( ab31263 , Abcam ) , to FLAG ( F3165 , Sigma-Aldrich ) , to Hdac2 ( ab7029 , Abcam ) , to Sin3a ( sc-994 , Santa Cruz Biotechnology ) , to Ing2 ( ab109504 , Abcam ) , to RbAp46/48 ( 39199 , Active Motif ) , to AcH3K9 ( 39917 , Active Motif ) , to BrdU ( 347580 , BD Biosciences ) , and to 5hmC ( 39769 , Active Motif ) . Mouse monoclonal antibodies to FLAG for immunostaining were obtained from Sigma . Cryosections were incubated overnight at 4°C with primary antibodies . Immune complexes were detected with horseradish peroxidase–conjugated secondary antibodies ( ImmPRESS reagent , Vector labs ) or Alexa Fluor 488–conjugated secondary antibodies ( Molecular Probes ) . Nuclei were counterstained with 4' , 6-diamidino-2-phenylindole ( DAPI ) . Confocal images were acquired with a confocal microscope ( Olympus FV1000D or Zeiss LSM510META ) . Pregnant mice were injected intraperitoneally with undiluted BrdU labeling reagent ( RPN20LR , Amersham ) at a dose of 1 ml per 100 g of body weight . Embryos were dissected 30 min after BrdU injection and were subjected to immunostaining of BrdU as previously described ( Santarelli et al . , 2003 ) with the use of a Vectastain ABC Kit ( Vector labs ) and diaminobenzidine . The proliferation index was calculated as the percentage of cells positive for BrdU incorporation . Total RNA was isolated from E9 . 5 embryos with the use of TRIzol reagent ( 15596026 , Invitrogen ) , and portions of the RNA ( 1 in 20 µl ) were subjected to RT with the use of a PrimeScript RT Reagent Kit with gDNA Eraser ( RR047A , Takara ) . The resulting cDNA ( corresponding to an RNA amount of 15 , or 0 . 92 ng for β-actin qPCR ) was subjected to real-time PCR analysis with the use of Power SYBR Green PCR Master Mix ( 4367659 , Applied Biosystems ) . For quantitation of mRNAs , we established standard curves with serial dilutions of RNA of known concentrations . Data were normalized byβ-actin mRNA abundance . PCR primers ( forward and reverse , respectively ) were as follows: 5′-GGTCATCACTATTGGCAACG-3′ and 5′-ACGGATGTCAACGTCACACT-3′ for Actb ( β-actin ) ; 5′-TACCAGGGTAGCAACCCAAT-3′ and 5′-GGTTTCTGACAGCCCTCTTC-3′ for Adhfe1; 5′-CTGATTGAGGAGTTGAGGC-3′ and 5′-AGCCTACAGTTGGAGCCTG-3′ for Nagk; 5′-CTGACTTTTCGGTGGGCTACA-3′ and 5′-GGCGCAGAATGGCTCTTC-3′ for Leng9; 5′-GGCCTGCCTCCTTTAGTCTC-3′ and 5′-TGTCAGCATCTGTGGCTGTT-3′ for Dchs1; 5′-GCTCAGACTCAGACCAAGAG-3′ and 5′-TGCTGTGTGAGTAGCTGTGC-3′ for Mxd3; 5′-GGGGGAATGAGTGCTTGAAG-3′ and 5′-TCACCTGGACCTCCAAATGTC-3′ for AA465934; 5′-AAAGAGAAACTGCCAACGC-3′ and 5′-TATTCATACCTGGGCCGAAG-3′ for 2610020C07Rik; 5′-GTAGGGGATCGGGACTCTGG-3′ and 5′-TCCTCAAGGAATGATCCGGC-3′ for Gt ( ROSA ) 26Sor; and 5′-CAATTACAGAAGTGTGGGACT-3′ and 5′-CACCTTCCTCCCAGTTCTTT-3′ for Acsl3 . E9 . 5 embryos harboring the Fam60::Venus transgene were recovered in PBS for ChIP with antibodies to GFP performed as previously described ( Hayakawa et al . , 2007 ) . The isolated DNA was applied to ChIP-seq library construction with the use of a SOLiD Fragment Library Core Kit ( PN 4464412 , Life Technologies ) . Sequencing was performed with a SOLiD four instrument ( Life Technologies ) . Sequenced reads were aligned to the mouse genome ( mm9 ) with the use of LifeScope software ( Applied Biosystems ) . Aligned peaks were called and BED and Wig files were generated with MACS version 1 . 4 . 1 ( Zhang et al . , 2008 ) , and the files were visualized in the UCSC genome browser as custom tracks . The called peaks were filtered with the following criteria: false discovery rate ( FDR ) of ≤1% and fold enrichment of ≥2 . 0 . To obtain a peak distribution and averaged peak profile around genes , we analyzed the filtered peaks with CEAS version 1 . 0 . 2 . Genes with filtered peaks within ±3 kb of the TSS in UCSC RefGene were defined as Fam60a target genes . For RNA-seq , E9 . 5 embryos were collected in PBS and stored in RNAlater ( AM7020 , Ambion ) at –80°C . After genotyping with yolk sac DNA , RNA was isolated from WT and Fam60a–/– embryos with the TRIzol reagent and mRNA was extracted twice with the use of a MicroPoly ( A ) Purist Kit ( AM1922 , Life Technologies ) . Library preparation was performed with the use of a SOLiD Total RNA-Seq Kit ( 4445374 , Life Technologies ) . Three biological replicates were analyzed for each genotype . Libraries were labeled with distinct barcoding adapters . Sequencing was performed with a SOLiD4 instrument , and sequencing data were mapped to the mouse genome ( mm9 ) with the use of LifeScope software . Differentially expressed genes were identified with the edgeR Bioconductor package . Transcripts with an FDR of <0 . 01 were considered to be significantly up- or down-regulated . Genomic DNA was isolated from WT and Fam60a–/– embryos according to standard procedures , and its concentration was determined by spectrophotometry . The DNA ( 500 to 1000 ng ) was treated with bisulfite and purified with the use of an EpiTect Bisulfite Kit ( 59104 , Qiagen ) and was then subjected to PCR amplification with the following primer sets: 5′- ATTTAGTGGGGTTTTTGTTATTG-3′ ( Adhfe1 Bis F1 ) and 5′-TATTTCTACACATAAACCCATAC-3′ ( Adhfe1 Bis R1 ) for initial PCR and Adhfe1 Bis F1 and 5′-ACTAAACCACATTACACCATCC-3′ ( Adhfe1 Bis R2 ) for seminested PCR; 5′-TGGAAGGAGGTTAAAGGATTAG-3′ ( Leng9 Bis F1 ) and 5′-AAATTATCTAAACCCTACCCCC-3′ ( Leng9 Bis R1 ) ; 5′-ATTTTTTTAGGAGTTTTAGTTGGGGTG-3′ ( Nagk Bis F1 ) and 5′-CAACTCTACACAACTCTCCAAATTAAC-3′ ( Nagk Bis R1 ) ; 5′-AGAGGGTGTATGTTGTAGAGTAGTTAGGTG-3′ ( Peg3 Met11 ) and 5′-CATCCCATCCCCCTTTTCCAAACTCTAC-3′ ( Peg3 Met12 . 1 ) ; and 5′-GTATTTAGTTTATTATGAGGAAGAGTTT-3′ ( Kcnq1ot1 1F ) and 5′-CAAAAACAACTCCAAAAAAACTATAAA-3′ ( Kcnq1ot1 1R ) . The amplified fragments were separated by agarose gel electrophoresis and the target bands excised . DNA was recovered from the excised gel pieces with the use of a QIAquick Gel Extraction Kit ( 28706 , Qiagen ) and was then cloned into the pCRII vector with the use of a Dual Promoter TA Cloning Kit ( K207020 , Invitrogen ) . Sequenced fragments were analyzed with the QUMA tool ( quantification tool for methylation analysis; http://quma . cdb . riken . jp ) . NIH3T3 Tet-On 3G fibroblasts ( 631197 , Clontech ) were seeded at ~80% confluence on 15-mm-diameter cover slips coated with 0 . 1% gelatin and placed in 24-well plates . The cells were cultured for at least 2 hr at 37°C in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and were then transfected for 24 hr with 125 ng of pTRE3G-FLAG-Tet1 ( encoding FLAG-tagged mouse Tet1 ) with or without 250 ng of pEF-BOS-Fam60a-IRES-Venus ( encoding mouse Fam60a and Venus ) with the use of the Lipofectamine LTX reagent ( 15338500 , Invitrogen ) . The cells transfected without or with pEF-BOS-Fam60a-IRES-Venus were also transfected with 375 or 125 ng , respectively , of the pEF-BOS empty vector . Expression of FLAG-Tet1 was induced by exposure of the cells to doxycycline ( 1 µg/ml ) for 24 hr , after which the cells were fixed for 15 min with 4% paraformaldehyde in PBS , permeabilized for 15 min with 0 . 2% Triton X-100 in PBS , treated for 20 min with 2 M HCl , neutralized for 10 min with 100 mM Tris-HCl ( pH 8 . 0 ) , washed with PBS , and exposed for 1 hr to blocking buffer ( 1% bovine serum albumin and 0 . 1% Tween 20 in PBS ) , all at room temperature . The cells were then incubated overnight at 4°C with mouse monoclonal antibodies to FLAG ( 1:2000 dilution ) and rabbit polyclonal antibodies to 5hmC ( 1:2000 dilution ) in blocking buffer . Immune complexes were detected with Alexa Fluor 568– or Alexa Fluor 647–conjugated secondary antibodies ( Molecular Probes ) , respectively , and nuclei were stained with DAPI ( 250 ng/ml ) . The cells were mounted in ProLong Gold antifade reagent ( P36930 , Invitrogen ) , and images were acquired with a confocal microscope ( Olympus FV1000D ) . The fluorescence intensity of 5hmC was plotted against that of FLAG . If FLAG fluorescence intensity was >40 , the cell was considered as FLAG-Tet1 positive; if 5hmC fluorescence intensity was >30 , the cell was considered as 5hmC positive . ChIP was performed as described above , and the precipitated DNA was subjected to qPCR analysis with the following primers ( forward and reverse , respectively ) : 5′-CTAGCCACGAGAGAGCGAAG-3′ and 5′-AGCTTCTTTGCAGCTCCTTC-3′ for Actb; 5′-GACCGGATTGGCTGTTAGTG-3′ and 5′-TAGGTGCCTCAGCAAGTGTG-3′ for Adhfe1; 5′-CTAGGAAGAAGCGGCAGACC-3′ and 5′-GGCGTCACAGTTGGAGATCA-3′ for Leng9; 5′-CTGAGATTCATGCACAAGGG-3′ and 5′-TATAGGAACCAAGGGCGTTC-3′ for Nagk; 5′-GCGAGGACACTCACTGACTC-3′ and 5′-AGTGTGTGGTGGTGCTTGAG-3′ for Dchs1; 5′-GTGACGACAACTCGCGTAC-3′ and 5′-AATGGCCCTAATGAGAGACG-3′ for Mxd3; 5′-TTGGGAATCCAGTGGAAACT-3′ and 5′-AGCCATGCACAAAGTTCTTG-3′ for Acsl3; 5′-CTGGAGTTGCAGATCACGAG-3′ and 5′-CCTTTCTGGGAGTTCTCTGC-3′ for Gt ( ROSA ) 26Sor; 5′-TAAAGAGAAACTGCCAACGC-3′ and 5′ CTCATAGGACGTTCTGGCG 3′ for 2610020C07Rik; and 5′-CTGTCCAAGACTGCGGAATG-3′ and 5′-CCTGAAGCCATCCTTGGTAG-3′ for AA465934 . E9 . 5 embryos were recovered in PBS and stored at –80°C . After genotyping , embryos were lysed overnight at 55°C in a solution containing 20 mM Tris-HCl ( pH 8 . 0 ) , 4 mM EDTA , 20 mM NaCl , 1% SDS , and proteinase K ( 0 . 4 mg/ml , Nacalai ) . They were then exposed for 30 min at 37°C to RNase A ( 5 mg/ml , Sigma ) before purification of genomic DNA first by phenol-chloroform treatment and ethanol precipitation and then with the use of a QIAamp DNA Micro Kit ( 56304 , Qiagen ) . The DNA was sheared with the use of a Bioruptor UCD-250 ( Diagenode ) ( 15 s on and 15 s off for 10 min at low power ) . Portions ( 500 ng ) of the sheared DNA were denatured for 10 min at 98°C , placed on ice , and then incubated overnight at 4°C with rotation in 100 µl of hMeDIP buffer containing 20 mM Tris-HCl ( pH 8 . 0 ) , 2 mM EDTA , 150 mM NaCl , 1% Triton X-100 , 4 µg of antibodies to 5hmC , and 1% bovine serum albumin . Dynal Protein G beads were then added to the samples to precipitate the antibody-DNA complexes , after which the beads were washed three times with hMeDIP wash buffer ( 20 mM Tris-HCl ( pH 8 . 0 ) , 2 mM EDTA , 300 mM NaCl , 1% Triton X-100 , 0 . 1% SDS ) and then treated overnight at 55°C with proteinase K in hMeDIP elution buffer ( 20 mM Tris-HCl ( pH 8 . 0 ) , 8 mM EDTA , 300 mM NaCl , 0 . 5% SDS ) . The eluted DNA was purified with the use of a QIAquick PCR Purification Kit ( 28106 , Qiagen ) and subjected to qPCR analysis with the primers described above for ChIP-qPCR . Libraries compatible with the Illumina platform were prepared from 3 µg of genomic DNA with the use of a SureSelect Methyl-Seq Target Enrichment System ( Agilent Technologies ) . Genomic DNA was sheared at 4°C by focused ultrasonic disruption with a Focused-ultrasonicator E220 ( Covaris ) ( duty factor , 10%; PIP , 175; cycles per burst , 200; time , 360 s ) . The fragmented DNA was end-repaired , adenylated at the 3′ end , and ligated to a methylated adapter . The prepared libraries were subjected to hybridization with the biotinylated SureSelect Methyl-Seq Capture Library ( Agilent Technologies ) , which covers genomic regions of 109 Mb in total including GENCODE promoters; CpG islands , shores , and shelves; DNase I–hypersensitive sites; and RefGenes . Library molecules that overlapped the targeted regions were collected with streptavidin-conjugated beads and converted with bisulfite with the use of an EZ Methylation-Gold Kit ( Zymo Research ) before amplification by PCR . Further amplification was performed with the use of the SureSelect Methyl-Seq Indexing Primer ( Agilent Technologies ) to allow multiplexed sequencing on the Illumina platform . The amplified libraries supplemented with 20% of a phiX sequencing control library were sequenced with an Illumina HiSeq 1500 instrument with 2 × 127 cycles in the Rapid Run Mode . Sequence reads were obtained with HiSeq Control Software ( HCS ) version 2 . 2 . 58 and Real-Time Analysis ( RTA ) version 1 . 18 . 64 . 0 . The obtained paired-end reads were subjected to quality control with FastQC version 0 . 11 . 5 ( https://www . bioinformatics . babraham . ac . uk/projects/fastqc/ ) , and adapter sequences and low-quality reads were removed using Trim Galore ! version 0 . 4 . 2 and with the parameters ‘-e 0 . 1 -q 30’ ( http://www . bioinformatics . babraham . ac . uk/projects/trim_galore ) . After the removal of phiX-derived reads with Bowtie2 version 2 . 3 . 0 ( Langmead and Salzberg , 2012 ) , the valid reads were mapped to the UCSC mm9 reference genome sequence using Bismark version 0 . 17 . 0 ( Krueger and Andrews , 2011 ) and with the parameters ‘--bowtie2 -N 1 L 22 --score_min L , -0 . 6 , -0 . 6 . ’ Before methylation calling at each CpG site , potential PCR duplicates were removed and only read-pairs from the expected strand ( the original bottom strand of the reference genome sequence ) were extracted with the use of Bismark and Samtools version 1 . 3 . 1 ( Li et al . , 2009 ) , respectively . The on-bait coverage of mapped reads was calculated with CollectHsMetrics of the Picard package version 2 . 8 . 1 ( http://broadinstitute . github . io/picard ) . Methylated CpG was identified using the bismark_methylation_extractor function of Bismark and with the parameter ‘--cutoff 5 . ’ To compare methylation profiles among libraries , we performed a hierarchical clustering analysis according to Ward’s method with the use of the methylKit program version 1 . 0 . 0 ( Akalin et al . , 2012 ) in the Bioconductor package . For detection of DMRs in three mutant embryos compared with three WT embryos , we used BSseq version 1 . 10 . 0 ( Hansen et al . , 2012 ) in the Bioconductor package . After importation of the CpG report files of the Bismark output , the BSseq data were processed with the BSmooth algorithm for computation of smoothed methylation levels . The smoothed methylation data were selected for regions with a read coverage of at least five reads at the CpG sites in at least two of the three samples in both comparison groups . Comparison of the mutant and WT samples was then performed with t-statistics . DMRs were detected on the basis of the threshold ‘qcutoff ( low = 0 . 025 , high = 0 . 975 ) ’ and were further narrowed down to those with a minimum of three CpG sites and mean methylation difference of ≥0 . 05 . For examination of the relation between DMRs and Fam60a ChIP-seq peak regions , the peaks of the two ChIP-seq analyses were merged on the basis of their genomic locations and the merged peaks were then compared with DMRs with the use of bedtools version 2 . 26 . 0 ( http://bedtools . readthedocs . io ) . Regions of overlap were characterized by statistical evaluation of peak enrichment at genome features such as promoters , exons , introns , untranslated regions ( UTRs ) , and distal intergenic regions with the use of CEAS version 0 . 9 . 9 . 7 ( Shin et al . , 2009 ) , and plots of average profiles near TSSs were constructed with GREAT version 3 . 0 . 0 ( McLean et al . , 2010 ) . Methylation levels at imprinted genes and DMRs were visualized with the UCSC Integrative Genomics Viewer ( IGV ) version 2 . 3 . 72 ( Thorvaldsdóttir et al . , 2013 ) . Amino acid sequences similar to that of human Fam60a were collected by aLeaves ( Kuraku et al . , 2013 ) , and the resultant sequence set was then modified to remove redundant sequences . The modified sequence set was subjected first to multiple alignment with the use of the program MAFFT v7 . 299b ( Katoh and Standley , 2013 ) and with the option ‘-linsi’ and then to trimming of unaligned and gapped sites with the program trimAl v1 . 4 . rev15 ( Capella-Gutiérrez et al . , 2009 ) with the options ‘-automated1’ and ‘-nogaps’ in order . The obtained sequence file was used to infer the maximum-likelihood tree with the program RAxML v8 . 2 . 8 ( Stamatakis , 2014 ) according to the PROTCATWAG model and with 1000 bootstrap resamplings . RNA-seq , ChIP-seq and Methyl-seq data have been deposited in DNA Data Bank of Japan ( DDBJ ) with the accession numbers DRA004841 , DRA004842 and DRA006579 , respectively . Quantitative data are presented as means ± s . d . and were analyzed with the unpaired Student’s t test . A p value of < 0 . 05 was considered statistically significant .	As an embryo develops , its cells continue to divide and transform from unspecialized embryonic stem cells into the specialized cells that form the tissues and organs of the adult body . This complex process is controlled by a network of genes . Although most adult cells carry the same genes , different cell types each activate specific sets of genes , which ultimately gives them their unique properties . Likewise , developing cells also have unique patterns of gene expression that guide the cell’s development , behavior and its interaction with neighboring cells . For example , the gene Fam60a is highly active in embryonic stem cells , but until now , it was not known what role this gene had . To investigate this further , Nabeshima et al . studied mice that either had normal levels of Fam60a or reduced levels of Fam60a . The results showed that at a normal level , Fam60a was responsible for the intestines to develop properly . The guts of mice with reduced levels , however , grew very slowly . Moreover , Farm60a appears to regulate several other genes , and their activity was no longer controlled properly in these mice . Nabeshima et al . discovered that this was because Fam60a could interact with protein complexes responsible for repressing or activating genes . By changing the activity of these complexes , Fam60a could affect the activity of many other genes . A next step will be to find out how exactly Fam60a interacts with the protein complexes that affect the activity of genes . A better knowledge of how genes contribute to the development of an embryo may help understand the causes of miscarriage and find ways to prevent it .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine", "developmental", "biology" ]	2018	Loss of Fam60a, a Sin3a subunit, results in embryonic lethality and is associated with aberrant methylation at a subset of gene promoters
The majority of adult hippocampal newborn cells die during early differentiation from intermediate progenitors ( IPCs ) to immature neurons . Neural stem cells in vivo are located in a relative hypoxic environment , and hypoxia enhances their survival , proliferation and stemness in vitro . Thus , we hypothesized that migration of IPCs away from hypoxic zones within the SGZ might result in oxidative damage , thus triggering cell death . Hypoxic niches were observed along the SGZ , composed of adult NSCs and early IPCs , and oxidative byproducts were present in adjacent late IPCs and neuroblasts . Stabilizing hypoxia inducible factor-1α with dimethyloxallyl glycine increased early survival , but not proliferation or differentiation , in neurospheres in vitro and in newly born SGZ cells in vivo . Rescue experiments in Baxfl/fl mutants supported these results . We propose that localized hypoxia within the SGZ contributes to the neurogenic microenvironment and determines the early , activity-independent survival of adult hippocampal newborn cells . In the subgranular zone ( SGZ ) of the adult hippocampal dentate gyrus ( DG ) , one of the two neurogenic niches in the adult mammalian brain , new neurons are continuously generated throughout adulthood ( Taupin and Gage , 2002; van Praag et al . , 2002 ) . Adult hippocampal neurogenesis is a highly dynamic and regulated process that evolves slowly over several weeks ( Zhao et al . , 2008 ) . Adult neural stem/ radial glia-like cells ( Type I cells ) reside at the interface between the SGZ and the hilus , and give rise to rapidly dividing intermediate progenitors ( Type 2 cells ) . These cells migrate a short distance into the SGZ and progressively ( via Type 2a and Type 2b intermediate progenitors ) differentiate into neuroblasts ( Type 3 cells ) ( Zhao et al . , 2008; Fuentealba et al . , 2012; Bonaguidi et al . , 2012 ) . Neuroblasts then exit the cell cycle and become immature granule cell neurons , which in turn migrate into the granule cell layer and incorporate into the pre-existing functional hippocampal circuits ( Toni et al . , 2008 ) . In the young adult mouse DG as many as 4000 new cells are born daily , but only a subset ( ~30% ) survive at 4 weeks post-mitosis to become mature granule cell neurons ( Dayer et al . , 2003; Kempermann , 2003; Kempermann et al . , 2006; Sierra et al . , 2010 ) . The survival of adult-generated granule cells exhibits two critical periods; an early one during the transition from transient amplifying progenitors to neuroblasts and a later one during the integration of the immature neurons ( Dayer et al . , 2003; Kempermann , 2003; Sierra et al . , 2010; Mandyam et al . , 2007 ) . The early phase is associated with phagocytosis of apoptotic cells by microglia ( Sierra et al . , 2010 ) . GABA-mediated-depolarization and NMDA-receptor-mediated neuronal activity regulate the later phase of survival , at two and three weeks after mitosis respectively ( Jagasia et al . , 2009; Tashiro et al . , 2006 ) . However , less is known about the mechanisms responsible for the early death of newborn granule cells which occurs during the first days post mitosis , as cells exit the specialized microenvironment ( NSC niche ) of the adult SGZ . There is increasing evidence that differential oxygen tensions may be an important component of the NSC niche and that oxygen sensing can regulate the proliferation , survival and differentiation of neural stem cells and progenitors ( Panchision , 2009; de Filippis and Delia , 2011; Mazumdar et al . , 2010 ) . Thus , we investigated hypoxia within the SGZ , and whether stabilizing Hypoxia Inducible Factor-1 α ( HIF-1α ) , the master regulator of oxygen homeostasis , influences the early survival of the adult hippocampal newborn cells . We assessed for hypoxia within the SGZ of the adult DG following intraperitoneal injection of the hypoxia marker pimonidazole hydrochloride ( Hypoxyprobe , Figure 1B ) ( Varia et al . , 1998 ) . Pimonidazole ( PH ) is reductively activated and binds only to cells that have oxygen concentrations less than 14 μM , equivalent to a pO2 of 10 mm Hg ( 1 . 3% O2 ) ( Raleigh et al . , 1998 ) . Pimonidazole labeled hypoxic cells or groups of cells along the inner border of the SGZ ( Figure 1B ) . Double labeling with selective markers revealed that the majority ( 71 . 7 ± 10 . 4% ) of the hypoxic cells were radial glia ( GFAP+ ) and a smaller percentage ( 28 . 3 ± 9 . 5% ) were early intermediate progenitor cells ( Tbr2+/Dcx- ) ( Figure 1C , D ) in close proximity to the SGZ . In contrast , none of the hypoxic cells ( 0% ) expressed the neuroblast marker doublecortin ( DCX , Figure 1C , D ) . 10 . 7554/eLife . 08722 . 003Figure 1 . Detection of hypoxia and oxidative stress in SGZ niches of the adult DG . ( A ) Schematic diagram of experimental design . ( B ) Immunostaining with pimonidazole hydrochloride marks hypoxic areas ( white arrows ) within the adult SGZ ( scale bar: 20 μm ) . ( C ) Pimonidazole-positive cells colocalized with stem cell marker GFAP ( red , top row ) , intermediate progenitor marker Tbr2 ( blue , middle row ) but not with neuroblast marker DCX ( blue , bottom row ) . The enlargement in the rightmost panel in the top row indicates a hypoxic neural stem cell ( green ) that colocalizes with a GFAP+ radial glial cell ( yellow ) . A smaller fraction of early progenitors ( Tbr2+ ) also were pimonidazole-positive ( middle row , rightmost panel ) . Scale bar: 15μm left panel , 5 μm right panel . ( D ) Quantification of SGZ cells types expressing pimonidazole . ( E ) Oxidized 8-deoxyguanosine ( 8OHdG ) , a marker of oxidized nucleic acids , was not detected in GFAP positive neural stem cells , but in Tbr2- and DCX-expressing intermediate progenitors and neuroblasts ( scale bar: 25 μm ) . ( F ) Quantification of SGZ cell types expressing oxidized 8-deoxyguanosine . For all quantifications data are plotted as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 08722 . 003 Shortly after their generation , proliferating intermediate precursors translocate away from the proximal domain and into the intermediate domain ( Fuentealba et al . , 2012 ) . Thus , we hypothesized that migration away from hypoxic zones might result in oxidative damage as diagrammed in Figure 1A . Oxidative byproducts were highly localized in cells within the SGZ as assayed by 8-oxo-7 , 8-dihydro-2'-deoxyguanosine ( 8-OHdG ) labeling , an established biomarker of nucleic acid oxidative damage ( Figure 1E ) . Interestingly , phenotypic analysis revealed that 74 . 6 ± 5 . 8% of the 8-OHdG positive cells were late intermediate progenitors ( Tbr2+/DCX+ ) and 29 . 2 ± 12 . 2% were neuroblasts ( DCX+ only ) ( Figure 1E , F ) . Thus , a subset of newborn cells in the SGZ undergoes oxidative damage during their early migration and differentiation . The most critical survival period for adult-generated dentate granule cells occurs during the first few days post mitosis , as they differentiate from late intermediate progenitors to neuroblasts ( Dayer et al . , 2003; Kempermann , 2003; Sierra et al . , 2010; Mandyam et al . , 2007 ) . In order to test the role of oxidative damage in this early phase of apoptosis , we used the hypoxia mimetic agent , Dimethyloxallyl glycine ( DMOG ) ( Harten et al . , 2010 ) . Hypoxia Inducible Factor-1 α ( HIF-1α ) , the master regulator of oxygen homeostasis , is enzymatically degraded under normoxia by propyl hydroxylases ( Semenza , 2001 ) . DMOG suppresses propyl hydroxylase , thus stabilizing/enhancing HIF-1α activity under normoxic conditions ( Harten et al . , 2010 ) . We treated adult mice with DMOG ( 50 mg/kg once daily ) or vehicle for 3 days . Consistent with stabilization of HIF-1α by DMOG , mRNA levels of HIF-1α as well as its downstream targets VEGF , EPO and LEF-1 , were significantly higher after 3 days of DMOG treatment ( Figure 2A ) . As expected given the effect on downstream targets , DMOG treatment also increased HIF-1α protein levels ( Figure 2B , p = 0 . 03 ) . The increase in HIF1α mRNA levels occurs because of autoregulation of transcription by hypoxia response elements in the HIF1α promoter ( Iyer et al . , 1998 ) . Hypoxia also increases phosphorylation of Akt ( Ser473 ) , a serine/threonine kinase that promotes cell survival and reduces apoptosis ( Beitner-Johnson et al . , 2001 ) . Double immunohistochemistry with an antibody that specifically recognizes phosphorylated Akt together with an antibody against the neuroblast/immature neuron marker doublecortin ( DCX ) ( Figure 2C left panels ) , revealed that DMOG treatment induced a two-fold increase in the number of phospho-AKT+/DCX+ cells relative to vehicle treated animals ( p = 0 . 01 , n = 3 animals , Figure 2C right panel ) , also consistent with the hypoxia mimetic action of DMOG . 10 . 7554/eLife . 08722 . 004Figure 2 . DMOG stabilized and activated Hif-1α signaling in vivo . ( A ) DMOG treatment elevated mRNA levels of HIF1-α and its downstream targets . HIF1-α , VEGF , EPO and Lef1 in the microdissected DG of animals treated with DMOG for 3 days . ( HIF1-α , p = 0 . 04 , n = 3; VEGF , p = 0 . 03 , n = 3 , EPO = 0 . 02 , n = 3 , Lef1 , p = 0 . 002 , n = 3 ) . Data are mean ± SD . ( B ) Representative western blots of DG protein extracts probed with antibody against HIF1-α , from animals treated for 3 days with vehicle or DMOG ( n = 3 each group , left panel ) . Semiquantitative densitometry for HIF1-α protein normalized to β-tubulin levels ( right panel ) . HIF1-α DG protein levels were significantly elevated in DMOG treated animals . Data are mean ± SD , p = 0 . 03 . ( C ) DMOG increases phosphorylation of Akt ( Ser473 ) in DG newborn cells . Representative images of adult DG sections stained with anti-phospho-Akt ( green ) and anti-DCX ( blue ) after 3 days treatment with vehicle or DMOG . Note the higher density of phospho-Akt positive cells in the SGZ of DMOG treated animals ( below ) compared to vehicle treated controls ( above ) ( scale bar: 8 μm ) . Data are mean ± SD , p = 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 08722 . 004 Hypoxia in vitro influences neural precursors’ proliferation , differentiation and survival ( Panchision , 2009; de Filippis and Delia , 2011 ) . There was no significant effect of DMOG on the net proliferation of newborn cells at 3 dpi . ( control group , 12901 ± 1870 Ki67+ cells/mm3 , n = 5 animals , DMOG group 11859 ± 2953 Ki67+ cells/mm3 , n = 5 animals , p = 0 . 5 , Figure 3A , B , E ) . Similarly , there was not a detectable effect on the total density of Tbr2+ cells ( control group: 5776 ± 681 cells/mm3 , DMOG: 7331 ± 1381 cells/mm3 , p = 0 . 07 , n = 5 animals ) . At 3 dpi . 43 . 9 ± 5 . 8% of the proliferating cells were neural stem cells , nestin-only expressing cells , and 55 . 3 ± 7 . 1% were intermediate progenitors expressing both nestin and DCX ( n = 4 animals ) . DMOG did not cause a shift in the proliferative populations of SGZ progenitor at 3 dpi . ( 41 . 9 ± 9 . 53% Ki67+/Nestin+ , 58 . 1 ± 8 . 33 Ki67+/Nestin+/DCX+ , 2-way ANOVA , n = 4 animals , p = 0 . 56 , Figure 3C , D , F ) . The composition of the progenitor subtypes , measured as a percentage of BrdU+ cells was unaffected by DMOG treatment at 3 , 7 , 14 , 21 and 28 days post-injection ( 2-way ANOVA , no interaction between vehicle and DMOG groups , p >0 . 99 ) . This data is summarized in Figure 3G . Specifically , at 3 dpi . BrdU+ cells consisted mainly of late intermediate progenitors ( Tbr2+/DCX+ ) and neuroblasts ( Tbr2-/DCX+ ) , and a small number of early intermediate progenitors ( Tbr2+/DCX- ) . At 7 dpi . , the proportion of BrdU+ cells colabeled with Tbr2+/DCX+ decreased , whereas the proportion expressing only DCX+ increased , reflecting a shift towards immature neurons . By 14 dpi . Tbr2+ cells were not detected in either group with the majority of the BrdU+ cells expressing DCX . In the subsequent two weeks there was a significant decrease in the number of BrdU+/DCX+ cells consistent with lineage progression to mature neurons . DMOG had no effect on the total volume of the dentate gyrus ( Control: 0 . 61 mm3 ± 0 . 02 , n = 3 animals , DMOG: 0 . 59 ± 0 . 04 , n = animals , p = 0 . 89 ) , indicating that there were no macroscopic changes in the tissue . 10 . 7554/eLife . 08722 . 005Figure 3 . DMOG does not affect the proliferation and differentiation of 3 day old cells in the adult DG . ( A , B ) Representative images of proliferating cells Ki67+ cells in the SGZ ( A , B , scale bar: 100 μm ) . ( C , D ) Triple labeling with Ki67/ Nestin/DCX ( C , D , scale bar: 10 μm ) . ( E ) The density of proliferating cells ( Ki67+ ) was comparable between vehicle controls and DMOG treated animals . ( F ) The proportion of the DG proliferative progenitors remained unaltered following DMOG administration . ( G ) To analyze the phenotype of Brdu+ cells , brains were collected 3 , 7 , 14 , 21 and 28 days after two pulses of BrdU ( 300 mg/kg with a 4 hr interval between doses ) and triple labeled with BrdU/Tbr2/DCX . DMOG treatment did not affect the composition of the SGZ progenitor subtypes at any of the examined time-points . For all quantifications data are plotted as mean ± SD . The percentages at each time point are as follows: 3 dpi . : Control: 8 . 8 ± 3 . 2% Tbr2+/DCX- , 16 . 78 ± 3 . 6% Tbr2+/DCX+ , 68 . 4 ± 2 . 7 Tbr2-/DCX+; DMOG: 7 . 8 ± 1 . 1% Tbr2+/DCX- , 18 . 8 ± 1 . 9% Tbr2+/DCX+ , 70 . 7 ± 2 . 9% Tbr2-/DCX+ . 7 dpi . : Control: 8 . 4 ± 3 . 2% Tbr2+/DCX+ , 91 ± 3 . 1% Tbr2-/DCX+; DMOG: 6 . 8 ± 1 . 8% Tbr2+/DCX+ , 93 . 2 ± 1 . 7% . 14 dpi . : Control , 94 ± 5 . 6% Tbr2-/DCX+; DMOG: 95 ± 2 . 5% Tbr2-/DCX+ . 21 and 28 dpi . : Control: 21 dpi . 33 ± 3 . 2% Tbr2-/DCX+ , 28 dpi . 7 . 6 ± 2% Tbr2-/DCX+; DMOG: 21 dpi 37 . 2 ± 6 . 2% Tbr2-/DCX+ , 28 dpi . 7 . 9 ± 8 . 4% Tbr2-/DCX+ . DOI: http://dx . doi . org/10 . 7554/eLife . 08722 . 005 To assay the effect of DMOG on early cell survival , dividing cells were pulse-labeled with BrdU in adult mice just prior to DMOG administration , and the number of BrdU-labeled cells in the dentate gyrus was counted at several timepoints post-BrdU injection ( dpi ) ( Figure 4A ) . DMOG resulted in an increase of BrdU labeled nuclei in the SGZ ( Figure 4B ) . Quantitative analysis of the BrdU+ cells showed a significant effect of DMOG on survival of newborn cells in the SGZ ( 2-way ANOVA , p = 0 . 0002 , Figure 4C ) . One day of DMOG treatment did not alter the number of BrdU+ cells , consistent with no effect on proliferation ( p = 0 . 49 , n = 6 animals , Figure 4C ) . However , DMOG significantly increased the survival of BrdU+ cells by 3 days post-mitosis ( p <0 . 0001 , n = 12 animals ) . Interestingly , the relative increase in BrdU-labelled cells in DMOG persisted at later time points ( 7 dpi . , p = 0 . 0017 , n = 8 animals; 14 dpi . , p = 0 . 003 , n = 6; 21 dpi . , p = 0 . 007 , n = 6; 28 dpi . , p = 0 . 01 , n = 6 ) , as assessed by the rate of BrdU+ cell loss between 3 and 28 dpi . ( Comparison of fits , p = 0 . 5 , F = 0 . 7 ) . To delineate the effective time window for the action on adult newborn granule cell survival by 28 dpi . , dividing cells were labeled with BrdU at day zero , then exposed to DMOG from 0–3 , 0–7 or 7–14 days ( Figure 4D ) . Exposure to DMOG for the first three day post-mitosis resulted in a 33% increase in survival ( p = 0 . 023 , n = 3 ) , whereas exposure for 7 days did not yield a significant further increase ( p = 0 . 23 , n = 3 ) . However , DMOG administration from day 7–14 , a critical period during which the newborn cells begin to integrate into the hippocampal network , had no effect on their survival ( p = 0 . 76 , n = 3 ) . The differentiation of BrdU+ cells into mature neurons ( NeuN+ ) by 28 dpi . was not affected by exposure to DMOG for either the first 3 or 7 days post mitosis ( BrdU+NeuN+/BrdU+: 0–3 days , Vehicle: 96 ± 4% , DMOG: 95 ± 2% , n = 3; 0–7 days , Vehicle: 98 ± 3% , DMOG: 96 ± 2% , n = 3 , 2-way ANOVA p = 0 . 8 ) . The above results indicate that the hypoxia mimetic agent DMOG does not alter the proliferation or differentiation of newborn cells in the adult DG , but rather specifically promotes their survival during the first few days post-mitosis , during their transition from intermediate progenitors to neuroblasts . 10 . 7554/eLife . 08722 . 006Figure 4 . Hypoxia mimetic agent DMOG increases early survival of newborn cells in the adult SGZ . ( A ) The schema shows the experimental design for BrdU pulse labeling of newborn cells at day 0 followed by 3 days treatment with DMOG or vehicle and the different timepoints studied thereafter . ( B ) Representative immunofluorescence in sections of adult DG treated either with vehicle or with DMOG for 3 days double labeled with anti-BrdU and anti-Tbr2 , which increased BrdU+ cells following DMOG treatment ( scale bar: 100 μm ) . ( C ) Quantification of the BrdU positive cells along the time course of 28 days post injection ( dpi ) following 3-day treatment with vehicle or DMOG . ( D ) Quantification of the survival of BrdU positive cells at 28 dpi . following different DMOG treatment periods . For all quantifications data are plotted as mean ± SD ( p <0 . 05;p <0 . 01;*p <0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08722 . 006 Apoptotic newborn cells in the DG are rapidly phagocytosed and cleared by unchallenged microglia ( Sierra et al . , 2010 ) ; rendering apoptosis hard to detect and evaluate with apoptotic markers such as TUNEL and activated-Caspase-3 . Because the pro-apoptotic gene Bax mediates programmed cell death of adult-generated hippocampal cells ( Sun , 2004; Sahay et al . , 2011 ) , we hypothesized that DMOG treatment should not affect the survival of cells lacking Bax . To test this idea , the dentate gyrus of wildtype and Baxfl/fl mice was co-injected with a nuclear-Cre-GFP encoding retrovirus together with a control retrovirus encoding mCherry ( Figure 5A ) . These MMLV-based retroviral constructs selectively target the same population of adult-born granule cells ( van Praag et al . , 2002 ) , and could be used to count the numbers of labeled adult-born cells and selectively inhibit Bax expression from Cre-retrovirus infected cells . The ratio of nuclear-Cre-GFP expressing cells to the mCherry-expressing cells in Baxfl/fl animals as early as 7 dpi . , provides a measure of survival of newborn cells , independent of possible variation in injection sites and viral titers . The ratio of Cre-GFP/ mCherry cells was increased in vehicle treated Baxfl/fl compared to WT animals at 7 dpi . ( Figure 5B , C , 31 . 7 ± 10 . 3% , p = 0 . 002 ) , indicating that loss of Bax augments the survival of adult-born cells at one week post-mitosis . Treatment with DMOG for 7 days post injection decreased the Cre-GFP/ mCherry ratio compared to Baxfl/fl vehicle treated animals ( Figure 5B , C , 30 ± 9 . 4% , p = 0 . 03 ) , consistent with a DMOG-mediated increase in survival in mCherry+ cells . Together our results strongly suggest activation of HIF-1α signaling enhances the early phase of survival of the newborn cells in the adult DG . 10 . 7554/eLife . 08722 . 007Figure 5 . DMOG mimics the ablation of pro-apoptotic gene Bax in promoting survival of SGZ newborn cells . ( A ) A fixed ratio mixture of two retoviruses ( encoding nuclear Cre-GFP and mCherry ) was injected into the adult DG of WT and Baxfl/fl mice . Survival of the newborn cells was analyzed at 7 days post injection following treatment with either vehicle or DMOG . ( B ) Representative images of virus-labeled cells from co-injection experiments in WT mice treated with vehicle , Baxfl/fl mice treated with vehicle and Baxfl/fl mice treated with DMOG ( scale bar: left 100 μm , right 20 μm ) . ( C ) The ratio of Cre-positive to mCherry-positive cells at 7 dpi was used as a measure cell survival ( see text ) . Data are plotted as mean ± SD . : WT+ Vehicle vs Baxfl/fl + Vehicle , p = 0 . 002; #: Baxfl/fl + Vehicle vs Baxfl/fl + DMOG , p = 0 . 03; WT+ Vehicle vs Baxfl/fl vs DMOG: not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 08722 . 007 Neuronal survival can be activity-dependent ( Jagasia et al . , 2009; Tashiro et al . , 2006 ) , thus we investigated whether newborn granule cells have received synaptic innervation by three days post-mitosis , the time point at which DMOG maximally increased survival . Acute slices from wildtype animals were prepared 3 days after retroviral infection and GFP-expressing 3 day old cells were identified using fluorescence microscopy and recorded using whole-cell voltage clamp techniques . Labeled cells were predominantly located in the subgranular zone as expected , with passive membrane properties consistent with immature neuronal precursors ( Rinput = 13 . 2 ± 1 . 9 GΩ , Cm = 4 . 8 ± 0 . 5 pF , n = 11 cells ) , including a fast inward current in response to a depolarizing voltage step consistent with a Na+ spike in 9 of 11 cells ( Figure 6A ) . Strikingly , none of these cells in wildtype animals had any spontaneous post-synaptic currents ( PSCs ) in >5 min of continuous recording per cell ( n = 11 ) , compared to an sPSC frequency of 0 . 55 ± 0 . 15 Hz in adjacent mature cells ( n = 4 ) . Furthermore , none of the 3 day-old cells had any evoked synaptic responses in the middle molecular layer ( >ten 10-second 8 Hz trains per cell , n = 11; Figure 6A ) , despite robust extracellular field potentials in these same slices at the same stimulation intensity ( n = 7 , data not shown ) or large PSCs in neighboring mature cells ( n = 4; Figure 6B ) . As expected for this feedforward circuit , the same middle molecular layer stimuli also produced repetitive spiking in hilar neurons ( data now shown ) , indicating that hilar mossy cells or GABAergic interneurons were activated by our stimulation but were not functionally connected to 3 day-old cells . Together , these results indicate that there is minimal , if any , synaptic innervation at this maturational stage . 10 . 7554/eLife . 08722 . 008Figure 6 . Adult-born granule cells lack synaptic responses 3 days post-mitosis . ( A ) At 3-days post-mitosis , granule cell lack responses to afferent stimulation . Five consecutive sweeps from a retrovirus-labeled 3-day-old cell , recorded in whole-cell voltage clamp mode while stimulating afferent inputs in the middle molecular layer . A -10 mV voltage step at the beginning of each sweep demonstrated the high input resistance of these cells , which lacked any postsynaptic responses to stimulation ( dots ) . ( B ) A mature granule cell immediately adjacent to the cell in A had robust post-synaptic currents to the same stimulation . Right panels: Voltage-dependent sodium current evoked in a 3-day-old granule cell during a step from -70 to -10 mV , demonstrating the neuronal identity of the cell ( right upper panel ) . The voltage-dependent sodium current recorded from a mature granule cell was much larger ( right lower panel ) . ( C ) Representative confocal images of p-CREB+ and c-Fos+ cells in the DG of animals treated with vehicle or DMOG for 3 days ( scale bar: 10 μm ) . ( D ) Quantification of the normalized number of p-CREB+ and c-Fos+ cells in the DG of DMOG treated animals relative to vehicle treated ones . DMOG treatment did not alter the number of p-CREB+ or c-Fos+ cells . Bar charts are mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 08722 . 008 Likewise DMOG did not affect neuronal activity in immature or mature DG granule cells as measured by immunohistochemistry for the immediate early genes pCREB or c-Fos ( Figure 6C , D ) , which are widely used as surrogate markers for neural activity ( Jagasia et al . , 2009; Lonze and Ginty , 2002 ) . Thus the action of DMOG in our experiments cannot be attributed to a non-cell autonomous effect on neuronal activity of granule cells . Labeled cells were quantified as follows: pCREB - ( control group , 122535 ± 19273 p-CREB+ cells/mm3 , n = 4 animals , DMOG group 134051 ± 9733 p-CREB+ cells/mm3 , n = 4 animals , p = 0 . 35 ) ; c-Fos - ( control group , 20175 ± 1025 c-Fos+ cells/mm3 , n = 4 animals; DMOG group 20646 ± 2766 c-fos+ cells/mm3 , n = 4 animals; p = 0 . 9 ) . DMOG treatment also did not affect neuronal activity in the hilus as assayed by double immunohistochemistry with c-Fos and the GABAergic neuronal marker glutamic acid decarboxylase-64 ( Gad-67 ) . We detected only sparse activation of hilar mossy cells ( Fos+ only ) or hilar interneurons ( Fos+/GAD67+ ) under basal conditions , which was unaffected by DMOG ( Vehicle , Fos+: 144 ± 55 cells/mm3 , Fos+/GAD67+: 14 ± 16 cells/mm3 , n = 4; DMOG , Fos only+: 116 ± 14 cells/mm3 , n = 4 , Fos+/GAD67+: 13 ± 14 cells/mm3 , n = 4 , p = 0 . 4 ) These results indicate that stabilization and activation of HIF-1a signaling by DMOG rescues early survival of newborn cells in an activity-independent manner . To further test the role of hypoxia , we used the neurosphere assay to examine survival and proliferation of adult DG progenitors ( Azari et al . , 2010; Deleyrolle and Reynolds , 2009; Louis et al . , 2013 ) . Adult DG neurospheres were generated and cultured for 7 days at ambient oxygen tension ( 21% O2 ) , in the presence of DMOG ( 100 μM ) or in hypoxic conditions ( 2 . 5% O2 ) . Neurospheres under all three conditions exhibited normal morphology , each containing committed progenitors that co-expressed nestin and doublecortin ( Figure 7A ) . Treatment with DMOG or reducing O2 levels to 2 . 5% increased the number of neurospheres , a measure of cell survival ( p = 0 . 01 and p = 0 . 002 respectively ) , but had no effect on the size of the neurospheres , a measure of cells’ proliferation ( p = 0 . 98 , Figure 7B , C ) . Consistent with enhanced survival DMOG or hypoxia reduced the percentage of apoptotic cells in neurospheres as measured by caspase-3 immunoreactivity ( normoxia: 28 . 1 ± 1 . 7%; DMOG: 17 . 7 ± 2 . 6% , p = 0 . 01; hypoxia: 11 . 4 ± 2 . 3% , p <0 . 0001 , Figure 7D–G ) . The neurosphere assay also enabled us to test the sensitivity of the hypoxic marker pimonidazole hydrochloride ( PH ) , which labeled groups of cells in the SGZ ( Figure 1B ) . No pimonidazole binding was detected in neurospheres cultured in standard atmospheric culture conditions or in the presence of DMOG , whereas only a small percentage of PH+ cells ( 10 . 8 ± 4 . 4% ) , were detected in neurospheres cultured in 2 . 5% O2 ( Figure 7H–K ) , indicating that PH labeling occurs at O2 ≤2 . 5% . Consistent with our in vivo experiments , our neurosphere data results strongly suggest that hypoxia is an essential factor for the survival of DG-derived intermediate progenitors in vitro . 10 . 7554/eLife . 08722 . 009Figure 7 . DMOG and Hypoxia increase survival of adult DG- derived neurospheres . ( A ) Representative immunofluorescence image of an adult DG-derived neurosphere , composed of nestin+ and DCX+ progenitors . Quantification of neurosphere diameter ( B ) and number ( C ) in cultures grown at normoxia , in the presence of DMOG or under hypoxia ( 2 . 5% O2 ) for 7 days . ( D–F ) Representative immunocytochemistry for activated caspase-3 in neurospheres cultured under the 3 different experimental conditions . ( G ) Quantification of apoptotic ( activated Caspase-3+ ) cells among the total number of cells in neurospheres exposed to normoxia , DMOG or hypoxia for 7 days . ( H–J ) . Immunostaining with pimonidazole hydrochloride detected some hypoxic cells only in neurospheres cultured in 2 . 5% O2 for 7 days . ( Scale bar: 25 μm ) . ( K ) Quantification of pimonidazole hydrochloride-positive cells among the total cell number in neurospheres exposed to normoxia , DMOG or hypoxia . Bar charts are mean ± SD ( p <0 . 05; p <0 . 01; **p <0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08722 . 009 The partial oxygen pressure ( pO2 ) and concentration in brain tissue is well below the 40 mm Hg in venous blood exiting the brain and is also heterogeneous ( Erecińska and Silver , 2001 ) . In rat hippocampus the pO2 has been estimated to be 20 . 3 mmHg ( Erecińska and Silver , 2001 ) , whereas even more hypoxic regions have been detected in the SGZ of the adult mouse DG ( Mazumdar et al . , 2010 ) . Using the hypoxia marker , pimonidazole hydrochloride , which labels cells with oxygen levels <10 mm Hg ( 1 . 3% ) , we found that SGZ stem-cell like-radial glia and early intermediate progenitors lie within hypoxic zones . Interestingly , not all cells in the SGZ were labeled , but were localized to narrow zones of a few cell diameters wide along the hilar border of the SGZ . Only a fraction of GFAP+ or Tbr2+ cells were pimonidazole-positive whereas the hypoxic cell did not label with the proliferating marker Ki67 ( data not shown ) . Thus heterogeneous zones of tissue pO2 may reflect some degree of cell-state specificity as well as local metabolic demand that changes constantly even within a small specific area/volume of the brain . The two major determinants for tissue oxygen concentration are blood flow ( oxygen supply ) and oxygen consumption rate ( cellularity ) . The latter may predominate in the SGZ with its high ratio of nuclei to blood vessels compared to other brain regions ( Mazumdar et al . , 2010 ) . Although some neural stem cells are perivascular ( Ottone et al . , 2014; Tavazoie et al . , 2008; Palmer et al . , 2000 ) , it remains to be determined whether these vessels carry highly oxygenated blood or how easily oxygen is distributed to the cell . Interestingly , a recent paper ( Sun et al . , 2015 ) revealed that within the SGZ zone only late amplifying progenitors ( Tbr2+/DCX+ ) and horizontal early neuroblasts ( Tbr2-/DCX+ ) , among the entire population of SGZ progenitors are in direct contact with blood vessels . Numerous adult stem cells reside in hypoxic niches , where they maintain a quiescent state depending predominantly on anaerobic glycolysis ( de Filippis and Delia , 2011; Mohyeldin et al . , 2010; Platero-Luengo et al . , 2014 ) . Stem cells exhibit different metabolic properties than their differentiated progeny that may promote ‘stemness’ . However , whether anaerobic metabolism is an adaptation to low oxygen levels in the specific niches in vivo or is an intrinsic stem cell property is still unclear . In either case , our results are consistent with two-photon phosphorescence direct in vivo measurements of local oxygen tension in the bone marrow , which revealed a hypoxic adult stem cell niche with high heterogeneity in local pO2 ( Spencer et al . , 2014 ) . Thus our data supports the existence of steep pO2 gradients between the neurogenic niche and the granule cell layer . Although a microenvironment with low oxygen concentration confers resistance to reactive oxygen species and cytotoxic stressors , stem cells and their progeny are susceptible to changes in redox status upon migration away from hypoxic zones ( Mohyeldin et al . , 2010; Platero-Luengo et al . , 2014; Suda et al . , 2011 ) . Interestingly , we detected oxidative byproducts in intermediate progenitors and neuroblasts located adjacent to the hypoxic niches . This correlation suggests that migration of intermediate progenitors away from hypoxic zones leads to oxidative damage , and thus triggers an early phase of apoptosis . In vitro , hypoxic conditions ( defined as <5% O2 ) reduces apoptosis , promotes proliferation , and increases cultured embryonic , adult neural stem cells and neuronal progenitors ( Sierra et al . , 2010; Shingo et al . , 2001; Studer et al . , 2000; Felling , 2006; Gustafsson et al . , 2005; Clarke and van der Kooy , 2009; Chen et al . , 2007; Bürgers et al . , 2008 ) . Deletion of hypoxia-inducible factor-1 alpha subunit ( HIF-1α ) in vivo , a transcriptional activator mediating adaptive cellular responses to hypoxia , dramatically decreased adult hippocampal neurogenesis ( Mazumdar et al . , 2010 ) . Additionally , conditional ablation of HIF1-α in adult mouse brain resulted in hydrocephalus , decreased neurogenesis ( partly due to an increase in apoptosis ) and deficits in spatial memory ( Tomita et al . , 2003 ) . In our experiments , DMOG , an agent that stabilizes HIF1-α under normoxic conditions , reversed the early phase of cell death of newborn cells in the adult dentate gyrus . These results strongly support the idea that transition of progenitors from a hypoxic niche to normal tissue oxygen contributes to cell death . Prior studies using mitotic markers such as BrdU found that only 30–50% of adult-born newborn neurons survive by 30 days post-mitosis ( Dayer et al . , 2003; Kempermann , 2003; Mandyam et al . , 2007; Jagasia et al . , 2009; Tashiro et al . , 2006 ) , but these studies often did not investigate survival during the first few days post-mitosis . Thus the prevailing idea , until recently , had been that adult-generated hippocampal granule cell survival is regulated mostly at the immature neuron stage ( 2–4 weeks post-mitosis ) , when is affected by neuronal activity and behavior ( Jagasia et al . , 2009; Tashiro et al . , 2006; Gage et al . , 1999 ) . Our results support two critical periods for survival of newborn cells in the adult SGZ , early ( 1–4 days post-mitosis ) and late ( 1–3 weeks post-mitosis ) with most ( 2/3rds ) of cell death occurring in the first week ( Dayer et al . , 2003; Kempermann , 2003; Sierra et al . , 2010; Mandyam et al . , 2007; Jagasia et al . , 2009; Tashiro et al . , 2006; Mazumdar et al . , 2010 ) . The early apoptosis is likely executed by members of the p53 and Bcl-2 protein families that are key players regulating the apoptotic machinery of adult DG neural precursors and newborn neurons ( Sun , 2004; Sahay et al . , 2011; Cancino et al . , 2013; Fatt et al . , 2014 ) . Although programmed neuronal death is often activity-regulated , whether there is synaptic innervation during the first few days post-mitosis has been unclear ( Song et al . , 2013; Esposito , 2005 ) . Our results show that at 3 days post-mitosis , adult-born cells demonstrate no detectable neural activity . Rather , we propose that the metabolic milieu of the adult SGZ , a previously underexplored variable of this highly specialized neurogenic niche , plays a critical role in the early survival of adult generated hippocampal granule cells . All procedures were performed according to the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and were in compliance with approved IACUC protocols at Oregon Health & Science University . Subjects were young adult ( six weeks old ) C57BL/6J ( wildtype ) and B6;129-Baxtm2Sjk Bak1tm1Thsn/J ( Baxfl/fl ) transgenic mice ( Takeuchi et al . , 2005 ) . Homozygotic Baxfl/fl mutants generated on a Bak1 null background have exons 2–4 of Bax deleted following Cre-mediated recombination . The conditional deletion of Bax combined with the Bak1 null allele greatly reduces apoptotic cell death and thus makes these mice useful in studies of apoptosis regulation ( Takeuchi et al . , 2005 ) . Hypoxic zones in the adult hippocampus were detected using pimonidazole hydrochloride kit ( Hypoxyprobe ) according to manufacturer’s protocol . Pimonidazole hydrichloride in water was injected intraperitoneally at a dosage of 60 mg/kg . Animals were sacrificed at 1 hr and the brain removed following cardiac perfusion . Given that the half-life of pimonidazole is 22 min in mice , >99% of free drug had been cleared by the time of sacrifice , thus the pimonidazole imaged in our experiments was already bound at time of perfusion ( Walton et al . , 1987; Williams et al . , 1982 ) . To examine survival of newborn cells in the adult hippocampus , mice were injected intraperitoneally with Bromodeoxyuridine ( BrdU , Sigma-Aldrich , St . Louis , MO ) at 300 mg/kg twice with a 4 hr interval between doses , and sacrificed at different time points . This pulse-chase protocol was chosen to saturate mitotic cell labeling within a single cell cycle as determined previously ( Cameron and Mckay , 2001 ) . To stabilize hypoxia-inducible factor 1-α , mice were treated with dimethyloxallyl glycine ( DMOG , Cayman Chemicals , Ann Arbor , MI ) at 50 mg/kg intraperitoneally daily for 3 or 7 days and animals were sacrificed at different time points . The comparison group received vehicle ( 30% DMSO ) . The primary antibodies used were: anti-pimonidazole hydrochrolide ( 1:100 , Hypoxyprobe ) , anti-glial fibrillary acidic protein ( GFAP; 1:1000 , Dako , Denmark ) , anti-Tbr2 ( Heffner lab ) , anti-doublecortin ( DCX , 1:500 , Millipore , Billerica , MA ) , Anti-8-Hydroxyguanosine ( 1:100 , Calbiochem , Billerica , MA ) , anti-BrdU ( 1:500 , Abcam ) , anti-phospho-Akt ( 1:100 , Cell Signalling , Danvers , MA ) , anti-phospho-CREB ( 1:300 , Santa Cruz ) , anti-c-fos ( 1:300 , Santa Cruz , Dalla , TX ) , anti-NeuN ( 1:500 , Sigma ) and anti-GAD67 ( 1:500 , Sigma ) .	The hippocampus is a region in the mammalian brain that has been implicated in the formation of new memories . This process involves the birth of new neurons , which are created at a rate of ∼4000 a day in the hippocampus of a young adult mouse . Yet only a fraction of these cells survive to form mature neurons . These cells die in two main waves – the first occurs days after they form , and the second several weeks later when as immature neurons they integrate into the brain . During this later wave , new neurons become active and survive if they connect with other nerve cells and die if they don’t . But little is known about what causes the earlier wave of cell death . The tissues that contain the precursors of new neurons often have lower oxygen levels compared to other tissues . This means that when these cells start to become neurons and leave these sites , they have to face higher levels of oxygen and may undergo “oxidative” damage . This led Chatzi et al . to ask whether such oxidative damage might cause the early loss of new neurons in the hippocampus . First , the part of the hippocampus that contains the precursor cells ( called the subgranular zone or SGZ ) was found to have patchy areas of low oxygen . Further experiments then revealed that chemicals that may cause oxidative damage were present in the nearby cells that had already started on the path to become new neurons . Chatzi et al . then tested whether chemically stabilizing a protein called Hypoxia Inducible Factor-1α ( or HIF1α for short ) , which naturally helps cells to adapt to low oxygen environments , might increase the survival of the cells in the SGZ . Higher levels of HIF1α did indeed increase the survival of these cells . These findings suggest that newborn cells in the SGZ walk a tightrope between a low oxygen environment that supports the early precursors and the surrounding higher oxygen levels that can be toxic to those cells that start to become neurons . Further studies of the proteins and molecules that act downstream of HIF1α could shed light on ways to enhance the survival of these newly-generated neurons .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2015	Localized hypoxia within the subgranular zone determines the early survival of newborn hippocampal granule cells
Heterochromatin formed by the SUV39 histone methyltransferases represses transcription from repetitive DNA sequences and ensures genomic stability . How SUV39 enzymes localize to their target genomic loci remains unclear . Here , we demonstrate that chromatin-associated RNA contributes to the stable association of SUV39H1 with constitutive heterochromatin in human cells . We find that RNA associated with mitotic chromosomes is concentrated at pericentric heterochromatin , and is encoded , in part , by repetitive α-satellite sequences , which are retained in cis at their transcription sites . Purified SUV39H1 directly binds nucleic acids through its chromodomain; and in cells , SUV39H1 associates with α-satellite RNA transcripts . Furthermore , nucleic acid binding mutants destabilize the association of SUV39H1 with chromatin in mitotic and interphase cells – effects that can be recapitulated by RNase treatment or RNA polymerase inhibition – and cause defects in heterochromatin function . Collectively , our findings uncover a previously unrealized function for chromatin-associated RNA in regulating constitutive heterochromatin in human cells . The histone methyltransferase Su ( var ) 3–9 was first identified in forward genetic screens as a suppressor of position effect variegation in Drosophila melanogaster ( Tschiersch et al . , 1994 ) . Previous studies identified important functions for the evolutionarily conserved SUV39 proteins in the silencing of heterochromatin , as well as in chromosome segregation and cell division ( Ekwall et al . , 1996; Melcher et al . , 2000; Peters et al . , 2001 ) . This family of chromatin-modifying enzymes includes Clr4 in fission yeast ( Nakayama et al . , 2001 ) , as well as SUV39H1 and SUV39H2 in humans ( Rea et al . , 2000 ) . SUV39 proteins catalyze the di- and tri-methylation of lysine 9 of histone H3 ( H3K9me2/3 ) , and these histone modifications are bound by chromodomain-containing proteins , including the SUV39 enzymes themselves and the HP1 family of proteins ( Al-Sady et al . , 2013; Bannister et al . , 2001; Lachner et al . , 2001; Müller et al . , 2016; Wang et al . , 2012 ) . HP1 protein binding to H3K9me2/3 chromatin is then thought to drive chromatin compaction and transcriptional repression through oligomerization ( Canzio et al . , 2011; Fan et al . , 2004; Grewal and Jia , 2007 ) . SUV39H1 and H3K9me3 are predominately associated with constitutive heterochromatin , which represses ‘selfish’ genetic elements and repetitive DNA to promote genomic stability ( Bulut-Karslioglu et al . , 2014; Peters et al . , 2001 ) . In many eukaryotes , constitutive heterochromatin is concentrated at the repetitive sequences flanking centromeres , and is termed pericentric heterochromatin . In fission yeast , disruption of pericentric heterochromatin causes chromosome cohesion defects and chromosome missegregation ( Bernard et al . , 2001 ) ; and in mammals , defective pericentric heterochromatin and aberrant transcription of pericentric repeats are associated with genomic instability and cancer ( Peters et al . , 2001; Ting et al . , 2011; Zhu et al . , 2011 ) . These defects in constitutive heterochromatin are most evident in SUV39H1 and SUV39H2 double knockout mice , which exhibit reduced embryonic viability , small stature , chromosome instability , an increased risk of tumor formation , and male infertility owing to defective spermatogenesis ( Peters et al . , 2001 ) . Human SUV39H1 has been implicated in a variety of complex biological processes such as DNA damage repair ( Alagoz et al . , 2015; Ayrapetov et al . , 2014; Zheng et al . , 2014 ) , telomere maintenance ( García-Cao et al . , 2004; Porro et al . , 2014 ) , cell differentiation ( Allan et al . , 2012; Scarola et al . , 2015 ) , and aging ( Zhang et al . , 2015 ) . Despite the fundamental role of SUV39H1 and SUV39H2 in heterochromatin formation , it is largely unclear how these enzymes are localized at specific genomic sites to generate heterochromatin . Other chromatin modifiers – in addition to binding DNA , post-translationally modified histones , and other chromatin-associated proteins – depend on interactions with noncoding RNAs for their proper localization ( Margueron and Reinberg , 2011; Rinn and Chang , 2012 ) . In fission yeast , the localization of pericentric heterochromatin proteins , including the SUV39 homolog Clr4 , relies on the RNAi machinery ( Bühler and Moazed , 2007; Grewal and Jia , 2007; Moazed , 2011 ) , and RNAi has also been implicated in heterochromatin formation in other eukaryotic systems as well ( Fukagawa et al . , 2004; Pal-Bhadra et al . , 2004 ) . Recent studies reported that RNA is involved in targeting SUV39H1 to telomeres and to the Oct4 locus ( Porro et al . , 2014; Scarola et al . , 2015 ) ; however , it is unclear whether RNA plays a broader role in SUV39H1-dependent heterochromatin formation , and if direct RNA binding regulates the association of SUV39H1 with pericentric heterochromatin . In this study , we establish that chromatin-associated RNA contributes to the localization of SUV39H1 at constitutive heterochromatin in humans . We find that RNA associates with the pericentric heterochromatin of human mitotic chromosomes in primary and immortalized cell lines , and that a portion of this RNA is encoded by pericentric α-satellite sequences . We show that SUV39H1 binds without any observed sequence preference to both DNA and RNA in vitro , and that SUV39H1 binds RNA transcribed from pericentromeric repeats in human cells . Mutations that disrupt the nucleic acid binding function of SUV39H1 cause defects in its localization to pericentric heterochromatin , destabilize SUV39H1’s association with chromatin , and result in heterochromatin silencing defects . We propose a model in which the direct binding of SUV39H1 to RNA and to methylated histones ensures proper constitutive heterochromatin function in humans . Chromatin-associated RNA has a well-studied role in the formation of pericentric heterochromatin in fission yeast ( Bühler and Moazed , 2007; Grewal and Jia , 2007; Moazed , 2011 ) , but the role of RNA at human pericentric heterochromatin remains largely unexplored . To test if RNA is associated with pericentric heterochromatin in human cells , we used fluorescent pulse labeling of RNA to observe its localization on mitotic chromosomes . Because transcription is largely repressed in mitosis ( Gottesfeld and Forbes , 1997 ) , RNAs bound to mitotic chromosomes may be more likely to play regulatory roles than RNAs associated with chromatin during interphase . The morphology of condensed mitotic chromosomes also provides landmarks such as the primary centromere constriction and the telomere to facilitate the localization of RNAs . We visualized chromosome-associated RNA by treating HeLa cells with the modified nucleoside ethynyl uridine ( EU ) ( Jao and Salic , 2008 ) , centrifuging mitotic chromosomes onto coverslips , and fluorescently labeling the RNA by coupling an azido-modified fluorophore to the alkyne group using copper-catalyzed cycloaddition ( click chemistry ) . The EU-RNA signal we detected on human mitotic chromosomes , although distributed in distinct puncta along chromosome arms , was particularly concentrated around centromeres ( Figure 1A ) . The RNA signal was sensitive to treatment with Ribonuclease ( RNase ) A , confirming that our EU treatment specifically labeled RNA; but not sensitive to RNase III or RNase H , indicating that this RNA possesses single-stranded ( ssRNA ) regions but not nuclease-accessible double-stranded RNA ( dsRNA ) or RNA-DNA hybrids ( Figure 1—figure supplement 1A ) . Four out of eight different cell lines showed RNA enrichment at pericentric regions ( Figure 1—figure supplement 1B , C and D ) . Both primary fibroblasts and HeLa cells exhibited pericentric RNA localization , indicating that this phenomenon is not specific to immortalized cells ( Figure 1—figure supplement 1B ) . 10 . 7554/eLife . 25299 . 003Figure 1 . RNA associates with the pericentric regions of human mitotic chromosomes . ( A ) RNA localization on human mitotic chromosomes . A schematic of the chromosome-associated RNA labeling approach is shown at top . Mitotic HeLa cells were spun onto coverslips and stained for DNA ( blue ) , ethynyl uridine labeled RNA ( EU-RNA , green ) , and H3K9me3 ( red ) or CENP-T ( red ) to mark centromeres . Top row of images: cells were not treated with EU ( -EU ) . Bottom row of images: cells were labeled with EU for 12 hr ( +EU ) . ( B ) α-satellite RNA localization on mitotic chromosomes . Mitotic DLD-1 cells were spun onto coverslips , then α-satellite RNA was detected with a probe recognizing the pericentric D1Z5 array on human chromosome 1 ( green ) . Chromosomes were also stained for DNA ( blue ) , H3K9me3 ( red ) , and HEC1 ( red ) antibodies to mark pericentric heterochromatin and the core centromere/kinetochore region . ( C ) D1Z5 α-satellite RNA overlaps with pericentric heterochromatin , but not the core centromere/kinetochore . α-satellite RNA FISH with the D1Z5 probe ( green ) on a stretched DLD-1 mitotic chromosome , co-stained for DNA ( blue ) , H3K9me3 ( red ) , and HEC1 ( red ) . Line scans show α-satellite RNA overlaps with H3K9me3 , but not HEC1 . ( D ) RNase sensitivity of D1Z5 α-satellite RNA FISH signal . Spread mitotic DLD-1 cells were treated ± RNases , then stained for DNA ( blue ) , D1Z5 α-satellite RNA ( green ) , and H3K9me3 ( red ) . See also Figure 1—figure supplement 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 00310 . 7554/eLife . 25299 . 004Figure 1—figure supplement 1 . Characterization of chromosome-associated pericentric RNA . ( A ) RNase sensitivity of pericentric RNA . EU-labeled mitotic HeLa cells were spun onto coverslips and incubated with or without RNase A , RNase H , or RNase III as indicated , then stained for DNA ( blue ) , EU-RNA ( green ) and CENP-A ( red ) to mark centromeres . ( B , C ) Representative images of different types of human cell lines , showing RNA localization on mitotic chromosomes . Cells were labeled with EU for 12 hr , then mitotic cells were spun onto coverslips and stained for DNA ( blue ) , EU-RNA ( green ) , and H3K9me3 ( red ) . ( B ) shows cell lines tested that , like HeLa cells , show RNA concentrated around centromeres . ( C ) shows cell lines tested that show no apparent concentration of RNA around centromeres . ( D ) EU-RNA staining on DLD-1 chromosomes . Cells were labeled with EU and stained as described above . Although α-satellite RNA is detected on DLD-1 chromosomes by RNA FISH , there is little detectable EU-RNA signal . ( E ) β-satellite and Satellite III DNA and RNA on human mitotic chromosomes . Mitotic DLD-1 cells were spread onto coverslips , RNA or DNA FISH was performed to detect Satellite III ( green ) and β-satellite ( red ) sequences , and then chromosomes were stained for DNA ( blue ) . The β-satellite probe recognizes sequences on chromosomes 13 , 14 , 15 , 21 , and 22 , and the Satellite III probe recognizes sequences on chromosomes 14 and 22 . ( F ) RNA FISH for D1Z5 , a chromosome 1 specific α-satellite array , on HeLa mitotic chromosomes . Mitotic HeLa cells were spread onto coverslips , RNA FISH was performed to detect chromosome 1-specific D1Z5 α-satellite sequences ( green ) , and chromosomes were stained for HEC1 to mark centromeres ( red ) and with Hoechst to stain for DNA ( blue ) . ( G ) RNA FISH on DLD-1 chromosomes with a probe specific to an α-satellite array on chromosomes 13 and 21 . Mitotic DLD-1 cells were spread onto coverslips , RNA FISH was performed to detect α-satellite 13/21 ( green ) , and then chromosomes were stained for DNA ( blue ) and H3K9me3 ( red ) . Line scans of all four 13 and 21 chromosomes show localization of α-satellite RNA and H3K9me3 . White arrows delineate direction of line scan , and the labeling of homologous chromosomes as ‘homolog 1’ or ‘homolog 2’ is arbitrary . The Y-axis represents the pixel intensity along the drawn line . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 00410 . 7554/eLife . 25299 . 005Figure 1—figure supplement 2 . Identifying the transcriptional requirements for chromosome-associated RNA . ( A ) Representative images showing chromosome-associated EU-RNA signal after inhibiting RNA polymerases . HeLa cells were incubated with 0 . 5 mM EU and 50 μg/mL α-amanitin , 1 μM triptolide , 50 ng/mL actinomycin D , or 1 μM CX-5461 for 6 hr , then mitotic cells were spun onto coverslips and stained for DNA ( blue ) , EU-RNA ( green ) , and HEC1 to mark centromeres ( red ) . ( B ) Quantificaton of EU-RNA at pericentric regions and RT-qPCR controls to assess inhibition of transcription by different RNA polymerase inhibitors . Far left panel: images from the experiment shown in A were quantified using pericentromere finder software and HEC1 staining as a centromere marker . Shown are the means of 3 separate experiments , 15 images quantified per condition per experiment . Four rightmost panels: total RNA was purified from cells treated with the indicated polymerase inhibitors ( same cells shown in A ) , and RT-qPCR was performed with primers for control RNAs to check for inhibition of specific polymerases and to measure α-satellite RNA levels . All values were normalized to GAPDH RNA levels . Shown are the means of 3 separate experiments . All error bars represent standard error . ( C ) Diagram describing EU pulse/chase experiment to assess when chromosome-associated RNA is being transcribed . HeLa cells were incubated with 2 mM thymidine for 19 hr , then thymidine was washed out and 0 . 5 mM EU was added at the indicated time intervals before mitotic shake off . ( D ) Resulting mitotic spreads from experiment outlined in C , showing that chromosome-associated RNA is transcribed in the few hours before mitosis . Cells were harvested by mitotic shake off , spun onto coverslips , and stained for DNA ( blue ) , and EU-RNA ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 005 A previous study demonstrated that transcription by RNA polymerase II ( Pol II ) can occur from the core centromere regions of human mitotic chromosomes ( Chan et al . , 2012 ) . This RNA is presumably transcribed directly from the underlying centromeric DNA sequences , which in humans are composed of α-satellite repeats . α-satellite sequences , as well as other repeat classes – such as β-satellite and Satellite III – are also present at human pericentric regions ( Choo et al . , 1992; Greig and Willard , 1992 ) . To determine which RNA sequences are localized at pericentric heterochromatin , we performed RNA FISH with different repeat-specific probes . We used DLD-1 cells for this analysis , as they have a more stable karyotype than HeLa cells and allow for more reliable chromosome identification . We saw no RNA FISH signal with β-satellite ( chromosomes 13 , 14 , 15 , 21 , and 22 ) or Satellite III ( chromosomes 14 and 22 ) probes despite observing clear DNA FISH signal ( Figure 1—figure supplement 1E ) . However , we detected RNA FISH signal with two different α-satellite probes , one of which was specific for the pericentric α-satellite array D1Z5 ( Figure 1B and C , Figure 1—figure supplement 1F ) ( Pironon et al . , 2010 ) . This signal was sensitive to RNase treatment ( Figure 1D ) and colocalized with H3K9me3 staining ( Figure 1C , Figure 1—figure supplement 1G ) , but not with HEC1 staining that labels the core centromere/kinetochore ( Figure 1C ) . We also observed that α-satellite RNA was localized in cis ( i . e . at the same site of its transcription ) , as we saw that the RNA FISH signal was constrained to only the chromosomes known to contain the specific α-satellite DNA sequences recognized by our probes ( Figure 1D , Figure 1—figure supplement 1F and G ) . Interestingly , even though we detected EU-RNA signal enriched at pericentric regions on HeLa chromosomes and not on DLD-1 chromosomes ( Figure 1A , Figure 1—figure supplement 1B and D ) , we detected α-satellite RNA by RNA FISH in both cell types ( Figure 1C and D , Figure 1—figure supplement 1F ) , indicating that α-satellite RNA is localized at pericentric regions even in cells lines with low pericentric EU-RNA signal . To determine which RNA polymerase transcribes the pericentric RNA , we added different polymerase inhibitors to cells during EU treatment . We measured a reduction of approximately 50% in pericentric RNA signal after treatment with α-amanitin or triptolide – which preferentially inhibit Pol II – but observed a complete loss of signal after treatment with actinomycin D or CX-5461 – which preferentially inhibit Pol I , and show no detectable Pol II inhibition ( Figure 1—figure supplement 2A and B ) . Thus , most of the pericentric RNA signal we observe appears to be transcribed by Pol I . Total α-satellite RNA levels were not decreased by any RNA polymerase inhibition , suggesting that the amount of α-satellite transcribed in our 6 hr treatment window is likely a small percentage of the total α-satellite RNA ( Figure 1—figure supplement 2B ) . Our data does not exclude the possibility that some Pol II-dependent transcription is also occurring at the core centromere , as has been previously reported in human cells ( Chan et al . , 2012 ) . In that study , incorporation of fluorescent nucleotides was only assayed on mitotic chromosomes , and because we label cells with EU for a longer time period ( 4–12 hr ) , we are likely observing a broader set of RNAs that are being transcribed before and during mitosis . To investigate the cell cycle timing of pericentric RNA synthesis , we either labeled cells continuously prior to mitosis , or for a short pulse followed by washout prior to mitosis . We found that continuous EU labeling for 8 or 4 hr before mitosis led to pericentric EU-RNA signal , but a 4 hr EU pulse followed by a 4 hr chase resulted in no EU-RNA signal on mitotic chromosomes ( Figure 1—figure supplement 2C and D ) . This suggests that the pericentric RNA we observe on mitotic chromosomes is transcribed within the 4 hr before mitosis , or that it is transcribed earlier but dissociates from chromosomes in the 4 hr washout period after labeling . The finding that RNA is bound to the pericentric regions of human mitotic chromosomes is intriguing in light of previous observations that direct RNA binding by the heterochromatin factors Polycomb Repressive Complex 2 ( PRC2 ) and HP1α is important for their proper localization and function ( Muchardt et al . , 2002; Zhao et al . , 2008 ) . Because SUV39 enzymes are the primary H3K9 methyltransferases acting at pericentric regions , we tested if the association of SUV39H1 with pericentric heterochromatin also depends on RNA . We digested chromosomes with RNase A and assessed SUV39H1 localization . In a stable HeLa cell line expressing GFP-tagged SUV39H1 under doxycycline inducible control , we induced expression with a 6 hr pulse of doxycycline to avoid SUV39H1 mislocalization to chromosome arms caused by overexpression ( Melcher et al . , 2000 ) . We found that RNase A treatment reduced the pericentric localization of SUV39H1 to 43 ± 4% of the untreated control , and that SUV39H1 localization could be completely rescued by adding specific RNase inhibitors ( Figure 2A and B ) . DNA staining at pericentric regions was not reduced in this experiment , indicating that the general chromosome structure at pericentric heterochromatin was not disrupted by RNase A treatment ( Figure 2C ) . We also determined that total SUV39H1 protein levels are not affected by RNase A treatment ( Figure 2—figure supplement 1M ) . These data demonstrate that SUV39H1 depends on the presence of RNA for its localization on human mitotic chromosomes . 10 . 7554/eLife . 25299 . 006Figure 2 . SUV39H1 localization on mitotic chromosomes is disrupted by RNase treatment . ( A ) Mitotic HeLa cells expressing or not expressing SUV39H1-GFP ( +/-dox ) were spread onto coverslips and incubated without RNase , with RNase A , or with RNase A plus RNase inhibitors . Cells were then stained for DNA ( blue ) , anti-GFP to detect SUV39H1 ( green ) , and HEC1 to mark centromeres ( red ) . ( B ) Quantification of SUV39H1-GFP pericentric signal . The graph shows the average signal after subtracting background ( -dox ) and normalizing to control levels ( +dox , -RNase A ) . n = 5 separate experiments , 15 cells quantified per condition per experiment , error bars are standard error . ( C ) Quantification of DNA pericentric signal . The graph shows the average signal after normalizing to control levels ( +dox , -RNase A ) . n = 5 separate experiments , 15 cells quantified per condition per experiment , error bars are standard error . See also Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 00610 . 7554/eLife . 25299 . 007Figure 2—figure supplement 1 . Characterization of SUV39 DKO human cells . ( A ) Western blot analysis of HeLa SUV39 DKO cells , assessing total SUV39H1 , HP1α , tubulin , and H3K9me3 levels . ( B ) Analysis of SUV39H2 loci in HeLa SUV39 DKO cells . A section of the SUV39H2 gene surrounding the targeted Cas9 cut site was PCRed from genomic DNA and sequenced . Two mutant alleles were identified , and both lead to an early translation termination of SUV39H2 . ( C ) α-satellite RNA levels in HeLa SUV39 DKO cells were measured by reverse transcription followed by quantitative PCR ( RT-qPCR ) . Total RNA was isolated from HeLa control and HeLa SUV39 DKO cells , then reverse transcribed and amplified with α-satellite or GAPDH primers . Shown is average of 5 independent measurements , normalized to control cells , error bars are standard error . ( D ) H3K9me3 localization in control or SUV39 DKO HeLa cells . Mitotic cells were spread onto coverslips , then stained for DNA ( blue ) , H3K9me3 ( green ) , and HEC1 ( red ) to mark core centromere/kinetochore regions . ( E ) HP1α localization in control or SUV39 DKO HeLa cells . Mitotic cells were spread onto coverslips , then stained for DNA ( blue ) , HP1α ( green ) , and HEC1 ( red ) to mark core centromere/kinetochore regions . ( F ) Analysis of repetitive RNAs in HeLa SUV39 DKO cells . Total RNA was isolated from HeLa control and HeLa SUV39 DKO cells , and a cDNA library was generated and sequenced . Fold change in RNAs ( SUV39 DKO / control ) transcribed from different repeat types were plotted as a rank order from highest to lowest . Repeat types with over 300 reads were included . The horizontal gray dotted line represents a cutoff of 2 standard deviations from the mean of the dataset . Brown dots represent repeat types that fall under two standard deviations from the mean . Repeats that were enriched more than two standard deviations from the mean are labeled , and colors represent the RepeatMasker broad repeat class to which that repeat type belongs . ( G ) Comparative analysis of non-repetitive RNA levels in HeLa control and HeLa SUV39 DKO cells . RNA was isolated and sequenced as described in F , and non-repetitive ( uniquely mapping ) RNAs were analyzed . FPKM: Fragments Per Kilobase of transcript per Million mapped reads . ( H ) Western blot analysis of DLD-1 SUV39 DKO cells , assessing total SUV39H1 , HP1α , tubulin , and H3K9me3 levels . ( I ) Analysis of SUV39H2 loci in DLD-1 SUV39 DKO cells . A section of the SUV39H2 gene surrounding the targeted Cas9 cut site was PCRed from genomic DNA and subjected to MiSeq sequencing . Two mutant alleles were identified , and both lead to an early stop in the translation of SUV39H2 . ( J ) α-satellite RNA levels in DLD-1 SUV39 DKO cells were measured by reverse transcription followed by quantitative PCR ( RT-qPCR ) . Total RNA was isolated from DLD-1 control and DLD-1 SUV39 DKO cells , then reverse transcribed and amplified with α-satellite or GAPDH primers . Shown is average of 5 independent measurements , normalized to control cells , error bars are standard error . ( K ) H3K9me3 localization in control or SUV39 DKO DLD-1 cells . Mitotic cells were spread onto coverslips , then stained for DNA ( blue ) , H3K9me3 ( green ) , and HEC1 ( red ) to mark core centromere/kinetochore regions . ( L ) α-satellite RNA localization in control or SUV39 DKO DLD-1 cells . Mitotic cells were spread onto coverslips , then stained for DNA ( blue ) , α-satellite ( D1Z5 probe , green ) , and HEC1 ( red ) to mark core centromere/kinetochore regions . ( M ) Western blot showing SUV39H1 and histone H3 protein levels in cell lysates treated with RNase A or RNase A plus RNase inhibitors . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 007 SUV39H1 binds directly to HP1 proteins ( Yamamoto and Sonoda , 2003 ) , and mouse HP1α protein can directly bind RNA ( Muchardt et al . , 2002 ) . Given that SUV39H1 , HP1α , and RNA interact with one another , we tested the effect of SUV39 protein loss on HP1α and RNA localization . Using CRISPR/Cas9-mediated gene targeting , we generated SUV39H1 and SUV39H2 double knockout ( SUV39 DKO ) human HeLa and DLD-1 cell lines ( Figure 2—figure supplement 1 ) that lack expression of both SUV39H1 and the partially redundant SUV39H2 proteins ( Figure 2—figure supplement 1A , B , H and I ) . As observed in SUV39 double null mouse cells ( Lachner et al . , 2001; Peters et al . , 2001 ) , both H3K9me3 levels and HP1α localization were substantially reduced in human SUV39 DKO cells ( Figure 2—figure supplement 1A , D , E , H and K ) . However , the levels of centromeric satellite RNA , including α-satellite RNA , increased in SUV39 DKO cells ( Figure 2—figure supplement 1C , F and J ) ( Lee et al . , 1997 ) , consistent with SUV39H1/SUV39H2 loss causing defective heterochromatin and transcriptional depression of centromeric satellites . In contrast to the de-repression of satellite RNA , we saw no significant changes in global RNA expression levels ( Figure 2—figure supplement 1G ) . In the absence of SUV39H1 and SUV39H2 , α-satellite RNA continued to localize to pericentric heterochromatin as assayed by RNA FISH ( Figure 2—figure supplement 1L ) . Together , our data is consistent with a model in which RNA transcribed from pericentric α-satellite sequences maintains an association with the site of transcription during mitosis . In the absence of the SUV39 enzymes , α-satellite transcripts remain localized at pericentric regions , but transcriptional silencing , HP1α localization , and H3K9me3-dependent heterochromatin are disrupted . To determine whether SUV39H1 can bind RNA directly , we performed electrophoretic mobility shift assays ( EMSAs ) with a random 19-mer ssRNA oligonucleotide and MBP-tagged human SUV39H1 protein purified from E . coli . Because we were unable to purify full-length SUV39H1 protein free of degradation products , we purified truncations of SUV39H1 containing either the N-terminal extension alone ( amino acids 1–41 ) , the chromodomain alone ( amino acids 42–106 ) , the N-terminal extension plus the chromodomain ( amino acids 1–106 ) , or the C-terminus consisting of the pre-SET , SET , and post-SET domains ( amino acids 107–412 ) ( Figure 3A and B ) . We found that the SUV39H1 truncations containing the chromodomain bound to RNA with the highest affinity , whereas the C-terminal fragment showed no detectable binding ( Figure 3C and D ) . The addition of the N-terminal extension to the chromodomain of SUV39H1 ( 1–106 ) provided a 16-fold increase in affinity relative to the chromodomain alone ( 42-106 ) , with dissociation constants ( Kd ) of 0 . 15 ± 0 . 01 µM and 2 . 31 ± 0 . 17 µM , respectively . The N-terminal extension alone bound to RNA with a significantly reduced Kd ( 51 ± 30 µM ) compared to the truncations containing the chromodomain . Excess unlabeled RNA competed with radiolabeled RNA for binding to SUV39H1 ( Figure 3—figure supplement 1A ) , confirming that the RNA gel shift was not due to irreversible aggregation . From these data , we conclude that the chromodomain , SUV39H1 42–106 , is sufficient for RNA binding activity . This is consistent with previous findings that chromodomains can act as RNA binding domains ( Akhtar et al . , 2000 ) , and that the telomeric TERRA RNA associates with SUV39H1 ( Porro et al . , 2014 ) . 10 . 7554/eLife . 25299 . 008Figure 3 . SUV39H1 directly binds nucleic acids through its chromodomain . ( A ) Domain schematic of SUV39H1 truncations . NTE , N-terminal extension; chromo , chromodomain; pre , pre-SET; post , post-SET . Amino acid residues are listed above and to the left of each truncation . ( B ) Coomassie-stained gel of purified human SUV39H1 truncations fused to MBP . ( C ) Quantification of SUV39H1 domains binding to 19mer RNA . Binding curves are from quantifying EMSAs shown in 3D . Error bars are standard deviation from two independent experiments . Dissociation constants ( Kd , μM ) displayed on graph are determined by non-linear fitting of the binding curves . Standard error represents the error of the curve fitting to the average of the two experimental replicates . ( D ) Representative EMSAs showing binding of purified MBP-SUV39H1 truncations with a 19mer RNA oligo ( * ) , 1–41 , 42–106 and 1–106 diluted 2-fold from 100 μM , 107–412 diluted 2-fold from 62 μM . Quantified in 3C . ( E ) Binding of SUV39H1 to all nucleic acid types . Binding curves are from quantifying EMSAs ( Figure 3—figure supplement 1C ) of MBP-tagged SUV39H1 1–106 binding to 50mer nucleic acids ( ssRNA , ssDNA , dsRNA , dsDNA , or RNA/DNA ) ( Figure 3—figure supplement 1B ) . Various nucleic acids are composed of the first 50 bases of E . coli maltose binding protein ( MBP ) : ssRNA1 , sense MBP 1–50; ssRNA 2 , anti-sense MBP 1–50; ssDNA , sense MBP 1–50 . Error bars , dissociation constants ( Kd , μM ) , and standard error calculated as in 3C . ( F ) Quantification of MBP-SUV39H1 1–106 binding to 180mer α-satellite or β-actin ssRNA ( representative EMSA in Figure 3—figure supplement 1E ) . Error bars , dissociation constants ( Kd , μM ) , and standard error calculated as in 3C . See also Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 00810 . 7554/eLife . 25299 . 009Figure 3—figure supplement 1 . Characterization of SUV39H1-RNA binding in vitro . ( A ) Competitive binding assay measuring the ability of unlabeled 19mer ssRNA to compete with the interaction of labeled 19mer ssRNA binding to MBP-SUV39H1 42–106 . MBP-SUV39H1 42–106 present at 12 . 4 μM , labeled RNA at approximately 3 nM , and unlabeled RNA at concentrations ranging from 30 to 1500 nM . Percent labeled RNA bound is quantified and listed . ( B ) Various nucleic acids oligonucleotides consisting of the first 50 bases of E . coli MBP run out on a native polyacrylamide gel . Oligonucleotides were annealed to create various nucleic acids and end labeled with radioactive 32P . ssRNA 1 , sense MBP 1–50; ssRNA 2 , anti-sense MBP 1–50; ssDNA , sense MBP 1–50 . ( C ) Representative EMSAs showing the binding of purified MBP-SUV39H1 1–106 and to various nucleic acids , all composed of the E . coli MBP 1–50 sequence , shown in B . Protein is diluted 2-fold from 25 μM . Quantification in Figure 3E . ( D ) Left , binding curves showing binding of MBP-SUV39H1 1–106 and either sense or anti-sense 19mer ssRNA , protein diluted 2-fold from 100 µM . Error bars are standard deviation from two independent experiments . Right , dissociation constants ( Kd , μM ) determined by non-linear fitting of the binding curves . Standard error represents the error of the curve fitting to the average of two experimental replicates . ( E ) Representative EMSAs showing binding of purified MBP-SUV39H1 1–106 to 180 bases of either α-satellite or β-actin ssRNA . Protein diluted 2-fold from 2 . 5 μM . Quantification in Figure 3F . ( F ) SUV39H1 affinity increases as length of nucleic acid increases . Binding curves compiled from Figure 3E and F ( 50mers and 180mers , respectively ) and D ) ( 19mers ) showing the binding of MBP-SUV39H1 1–106 to various nucleic acid types . 19mer random sequence: sense and antisense ssRNA; 50mer MBP 1–50: sense and antisense ssRNA , ssDNA , dsRNA , dsDNA , and RNA/DNA hybrid; 180mers: α-satellite and β-actin ssRNA . All sequences are described in the materials and methods . Error bars are standard deviation from two independent experiments . Dissociation constants ( Kd , μM ) displayed on graph are determined by non-linear fitting of the binding curves . Standard error represents the error of the curve fitting to the average of two experimental replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 009 To determine if purified SUV39H1 exhibited a binding preference for different types of nucleic acids , we tested its binding to ssRNA , ssDNA , dsRNA , dsDNA and RNA/DNA hybrids using 50 nucleotides of the E . coli MBP sequence to generate each nucleic acid type ( Figure 3—figure supplement 1B ) . We found that SUV39H1 1–106 bound to all five types of nucleic acids with similar affinities , ranging from 61 ± 5 nM for ssRNA to 185 ± 17 nM for dsDNA ( Figure 3E , Figure 3—figure supplement 1C ) . SUV39H1 1–106 bound to the sense and antisense strand of MBP 1–50 ssRNA with similar affinities ( 61 ± 5 nM ssRNA1 vs . 64 ± 3 nM ssRNA2 ) ( Figure 3E ) , and showed a 2 . 5-fold difference in affinity for the sense and anti-sense 19mer ssRNA used in Figure 3C and D ( 145 ± 9 nM sense vs . 366 ± 22 nM anti-sense ) ( Figure 3—figure supplement 1D ) . The pericentric regions of human chromosomes are composed of repetitive α-satellite DNA sequences , which are weakly transcribed in humans ( Ideue et al . , 2014; Wong et al . , 2007 ) . Because SUV39H1 localizes to pericentric regions ( Aagaard et al . , 1999 ) and we detected chromosome-associated α-satellite RNA at these sites ( Figure 1C and D , Figure 1—figure supplement 1F ) , we tested whether SUV39H1 preferentially binds to α-satellite RNA . We measured SUV39H1 binding to a ssRNA containing a single monomeric unit of α-satellite ( 180 nucleotides ) , compared to a segment of β-actin mRNA of equal length . We found that SUV39H1 1–106 bound α-satellite or β-actin RNA with similar affinities: 23 ± 2 nM and 32 ± 2 nM , respectively ( Figure 3F , Figure 3—figure supplement 1E ) . This is consistent with the sequence-independent binding we observed for 19mer and 50mer ssRNA oligonucleotides . We conclude that SUV39H1 can bind both double- and single-stranded nucleic acids , has minimal sequence preference for the sequences tested thus far , and that binding affinity increases with increasing nucleic acid length ( Figure 3—figure supplement 1F ) . Our observation that RNA is bound to pericentric heterochromatin at the same sites where SUV39H1-dependent histone methylation occurs suggests that direct RNA binding may regulate the localization and function of SUV39H1 . Recent studies have proposed that direct RNA binding by SUV39H1 is necessary for its localization to specific genomic locations ( Porro et al . , 2014; Scarola et al . , 2015 ) , but this model has not been tested directly . To test whether RNA binding by SUV39H1 controls its interaction with chromatin , we identified RNA binding deficient mutants of SUV39H1 by alanine-scanning mutagenesis of the SUV39H1 chromodomain ( 42-106 ) ( Figure 4—figure supplement 1A and B ) . We identified mutants that either decreased ( 12 mutants ) or increased ( 11 mutants ) the affinity of SUV39H1 for ssRNA more than 10-fold when compared to the wild-type ( WT ) SUV39H1 chromodomain ( Figure 4—figure supplement 1B ) . Interestingly , 6 of the 12 mutants that decreased RNA binding affinity removed a positively charged residue ( R or K ) , and 7 of the 11 mutants that increased RNA binding removed a negatively charged residue ( D or E ) ( Figure 4—figure supplement 1B ) . By performing a more comprehensive analysis of the binding of these 23 mutants ( Figure 4A , Figure 4—figure supplement 1C ) , we discovered two mutations , R55A and R84A , that almost completely abolished RNA binding by SUV39H1 , reducing the affinity over 40-fold compared to WT ( Figure 4C , Figure 4—figure supplement 1C and D ) . 10 . 7554/eLife . 25299 . 010Figure 4 . Identification and biochemical characterization of SUV39H1 RNA binding-deficient mutants . ( A ) Identification of SUV39H1 mutants that affect RNA binding . Binding curves from second round of EMSA screening showing the binding of purified MBP-SUV39H1 42–106 point mutants to 19mer RNA ( table of measured dissociation constants in Figure 4—figure supplement 1B ) . ( B ) Crystal structures of the chromodomains of human SUV39H1 ( aa 44–91 ) ( Wang et al . , 2012 ) and human HP1α ( aa 18–68 ) bound to H3K9me3 peptide ( yellow ) ( Kaustov et al . , 2011 ) . Residues in SUV39H1 important for RNA binding ( red and blue ) or H3K9me3 binding ( green and orange ) are highlighted . The H3K9me3 peptide was co-crystalized with HP1α , but not SUV39H1 . ( C ) Quantification of WT SUV39H1 , SUV39H1R55A , and SUV39H1R84A binding to 19mer RNA . Binding curves generated by quantifying filter binding assays shown in Figure 4—figure supplement 1D . Error bars are standard deviation from three independent experiments . Dissociation constants ( Kd , μM ) displayed on graph are determined by non-linear fitting of the binding curves . Standard error represents the error of the curve fitting to the average of the three experimental replicates . ( D ) Coomassie stained gel of purified MBP-SUV39H1 42–106 proteins – WT , F43A , R55A , or R84A . ( E ) Binding of SUV39H1 mutants to H3K9me3 . Representative α-MBP western blot showing the amount of purified MBP-SUV39H1 42–106 protein , WT or indicated mutant , bound to streptavidin beads conjugated to either H3K9me0 ( me0 ) , H3K9me3 ( me3 ) , or no peptide ( - ) added . ( F ) Quantification of western blot shown in 4E , error bars are standard deviation , n = 3 , p<0 . 03 . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 01010 . 7554/eLife . 25299 . 011Figure 4—figure supplement 1 . In vitro characterization of SUV39H1 chromodomain mutants . ( A ) Coomassie stained gels showing purified MBP-SUV39H1 42–106 alanine point mutants . Percent full length protein calculated by dividing ( signal of the top band ) / ( signal of the top band + signal of the bottom band ) . Res . # , residue number; res . > A , residue mutated to alanine; % full , percent full-length protein . ( B ) Table displaying the results from the first round of EMSA screening to measure the binding of MBP SUV39H1 42–106 to a random 19mer ssRNA ( sense ) . Dissociation constants ( Kd , μM ) displayed on graph are determined by non-linear fitting of the binding curves . Standard error represents the error of the curve fitting to one experiment consisting of 6 protein concentrations analyzed for each mutant . Highlighted mutants ( yellow ) have either 10-fold greater or less RNA binding affinity compared to WT , and were re-measured in the second round of EMSAs ( Figure 4A , Figure 4—figure supplement 1C ) . ( C ) Table showing dissociation constants ( Kd , μM ) , standard error , Hill slope , and affinity fold-change compared to WT SUV39H1 , of point mutants during the second round of EMSAs . Standard error represents the error of the curve fitting to one experiment consisting of 9 protein concentrations analyzed for each mutant and three independent experiments for WT SUV39H1 . ( D ) Representative filter binding assay measuring RNA binding of WT SUV39H1 , SUV39H1R55A , and SUV39H1R84A to random sense 19mer RNA . Quantification shown in Figure 4C . Purified proteins shown in Figure 4D , diluted 2-fold from 200 μM . ( E ) Histone methyltransferase assay comparing the activity of WT SUV39H1 to SUV39H1 point mutants . Full length purified MBP-SUV39H1 was incubated with purified H3/H4 tetramer and radioactive methyl donor ( 14C S-adenosyl methionine ) , +/- RNA , then run on a denaturing gel . Histone H3 methylation was detected with a phosphorimager ( top panels ) , and proteins were visualized by Coomassie staining ( bottom panels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 011 Because SUV39H1 can bind all nucleic acids with comparable affinities , it is likely that the R55A and R84A mutations that disrupt RNA binding also prevent binding to DNA . Therefore , we refer to these mutations as nucleic acid binding mutations . The chromodomains of HP1 proteins specifically recognize methylated H3K9 through three conserved aromatic residues , called the ‘aromatic cage’ ( Bannister et al . , 2001; Jacobs and Khorasanizadeh , 2002; Lachner et al . , 2001; Nielsen et al . , 2002 ) . The SUV39H1 chromodomain also specifically binds to methylated H3K9 peptides , most likely via three similar aromatic cage residues ( Figure 4B ) ( Wang et al . , 2012 ) . The R55A and R84A mutations that disrupt nucleic acid binding are distant from the aromatic cage in the 3D structure of SUV39H1 , and thus we hypothesized that they do not affect H3K9me3 binding ( Figure 4B ) ( Wang et al . , 2012 ) . We tested this by measuring the binding of WT and mutant SUV39H1 chromodomains to unmethylated ( me0 ) or trimethylated ( me3 ) H3K9 peptides ( Figure 4D , E and F ) . As previously reported , we found that the WT SUV39H1 chromodomain preferentially binds to H3K9me3 compared to H3K9me0 , and this binding is completely disrupted by the aromatic cage mutant SUV39H1F43A ( Figure 4E and F ) ( Wang et al . , 2012 ) . Importantly , neither of the nucleic acid binding mutants ( SUV39H1R55A and SUV39H1R84A ) alter the binding of SUV39H1 to H3K9me3 ( Figure 4E and F ) . We conclude that SUV39H1R55A and SUV39H1R84A mutants specifically disrupt the interaction between SUV39H1 and nucleic acids without perturbing binding to H3K9me3 . We also confirmed that the R55A , R84A , and F43A mutations had no effect on the methyltransferase activity of full-length SUV39H1 ( Figure 4—figure supplement 1E ) . Interestingly , the addition of RNA inhibits SUV39H1 activity , as has been previously shown for PRC2 and several other SET domain histone methyltransferases ( Cifuentes-Rojas et al . , 2014; Kaneko et al . , 2014 ) . However , SUV39H1 nucleic acid binding mutants were equally inhibited by RNA , suggesting that this inhibition is largely independent of direct RNA binding . Our observations that RNA is associated with pericentric heterochromatin and that SUV39H1 directly binds RNA in vitro prompted us to test if SUV39H1 associates with RNA in human cells . In SUV39H1-GFP expressing HeLa cells , SUV39H1 colocalizes with EU-labeled RNA at pericentric regions of mitotic chromosomes ( Figure 5A ) . To test if SUV39H1 can bind to α-satellite RNA , we expressed GFP-tagged WT SUV39H1 , the nucleic acid binding deficient mutant SUV39H1R55A , or GFP alone under doxycycline inducible control . We then crosslinked the cells with formaldehyde , immunoprecipitated the GFP-tagged protein , and analyzed SUV39H1-associated RNAs by reverse transcription followed by quantitative PCR ( RT-qPCR ) . We found that α-satellite RNA was enriched in the WT SUV39H1 IP 13 ± 2 fold when compared to GFP alone ( Figure 5B and C , Figure 5—figure supplement 1A ) and that R55A mutation reduced the amount of α-satellite RNA to 45 ± 16% of the levels bound to WT ( Figure 5C ) . The expression of WT SUV39H1 or SUV39H1R55A did not lead to changes in total α-satellite RNA levels in the presence of endogenous SUV39H1 , indicating that the difference in immunoprecipitated α-satellite RNA was not due to changes in the overall levels of α-satellite RNA ( Figure 5D ) . In addition , the R55A mutation did not disrupt SUV39H1’s association with HP1α , indicating that α-satellite RNA binding to SUV39H1 is not occurring indirectly through HP1α ( Figure 5—figure supplement 1B ) . These results demonstrate that SUV39H1 associates with α-satellite RNA in cells , and that this association is facilitated by the nucleic acid binding activity of the SUV39H1 chromodomain . 10 . 7554/eLife . 25299 . 012Figure 5 . SUV39H1 association with α-satellite RNA in vivo depends on direct nucleic acid binding . ( A ) Colocalization of SUV39H1 and RNA at pericentric regions . SUV39H1-GFP expressing HeLa cells were labeled with EU and induced with doxycycline for 6 hr , then mitotic cells were spun onto coverslips . Cells were stained for DNA ( blue ) , with anti-GFP to detect SUV39H1-GFP ( green ) , for EU-RNA ( red ) , and CENP-A to mark centromeres . ( B ) Silver stained gel showing protein immunoprecipitated by an anti-GFP antibody from cell lines expressing GFP , SUV39H1-GFP , or SUV39H1R55A-GFP . ( C ) Quantification of α-satellite RNA immunoprecipitated with SUV39H1 . RNA was isolated from IPs , and RT-qPCR was performed to detect α-satellite RNA sequence . Enrichment values are the ratio of α-satellite/β-actin RNA in the IP over the ratio of α-satellite/β-actin RNA in total lysate , normalized to the GFP values . Error bars are standard error , n = 3 , p<0 . 04 . ( D ) Total α-satellite RNA levels in GFP , SUV39H1-GFP , and SUV39H1R55A-GFP cell lines , divided by the amount of β-actin RNA , normalized to the GFP values . See also Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 01210 . 7554/eLife . 25299 . 013Figure 5—figure supplement 1 . Immunoprecipitation of SUV39H1-GFP from human cells . ( A ) Anti-GFP western blot of total input lysate ( T ) , lysate after IP depletion ( D ) , and IP samples ( IP ) from three different HeLa Flp-In T-REx cell lines . RNA was purified from anti-GFP immunoprecipitated material and analyzed by RT-qPCR ( Figure 5 ) . ( B ) Anti-SUV39H1 and anti-HP1α western blot of total input lysate ( T ) , lysate after IP depletion ( D ) , and IP samples ( IP ) from three different HeLa Flp-In T-REx cell lines , after performed RIP protocol immunoprecipitation . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 013 To test how the SUV39H1 chromodomain interactions with RNA and H3K9me3 contribute to its localization in human cells , we measured the extent of localization of GFP-tagged WT or mutant SUV39H1 on spread mitotic chromosomes . The R55A and R84A nucleic acid binding mutants of SUV39H1 showed defects in localization , reducing SUV39H1 signal by about half ( R55A: 42 ± 8% of WT levels , R84A: 45 ± 9% of WT levels ) ( Figure 6A , B and C ) ; indicating that the direct interaction of SUV39H1 with nucleic acids is important for its localization to pericentric heterochromatin . We also found that the aromatic cage mutants W64A and F43A led to more pronounced localization defects than those caused by the nucleic acid binding mutants ( W64A: 10 ± 4% of WT levels , F43A: 15 ± 5% of WT levels ) ( Figure 6A , B and C ) . These results suggest that both H3K9me3 binding and nucleic acid binding contribute to SUV39H1 localization to pericentric heterochromatin during mitosis . 10 . 7554/eLife . 25299 . 014Figure 6 . Direct nucleic acid binding by SUV39H1 regulates its localization on mitotic chromosomes . ( A ) Localization of WT or mutant SUV39H1-GFP on human mitotic chromosomes . Expression of SUV39H1-GFP was induced in HeLa cell lines for 6 hr , then mitotic cells were spun onto coverslips and stained for DNA ( blue ) , with anti-GFP to detect SUV39H1-GFP ( green ) , and for HEC1 to mark centromeres ( red ) . Shown are representative images . ( B ) Western blot assessing expression of inducible SUV39H1-GFP in HeLa cell lines after 6 hr of +/- doxycycline induction . ( C ) Quantification of SUV39H1-GFP levels at pericentric regions , from experiment shown in 6A . Each experiment was normalized to the WT measurement . Bars show the average of n = 3 separate experiments , 15 cells quantified per condition per experiment , error bars represent standard error . ( D ) Localization of WT or mutant SUV39H1-GFP on human mitotic chromosomes after RNase treatment . Mitotic HeLa cells expressing SUV39H1-GFP ( WT or R55A ) were spread onto coverslips and incubated without RNase , with RNase A , or with RNase A plus RNase inhibitors . Cells were then stained for DNA ( blue ) , anti-GFP to detect SUV39H1 ( green ) , and HEC1 to mark centromeres ( red ) . Shown are representative images . ( E ) Quantification of SUV39H1-GFP levels at pericentric regions after RNase treatment , from experiment shown in 6D . Each experiment was normalized to the untreated WT measurement . Shown are the averages of n = 5 separate experiments , 15 cells quantified per condition per experiment , error bars represent standard error . See also Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 01410 . 7554/eLife . 25299 . 015Figure 6—figure supplement 1 . HP1α localization correlates with SUV39H1 localization , despite no change in H3K9me3 levels . ( A ) Representative images of HP1α and SUV39H1-GFP localization on HeLa mitotic chromosomes after RNase treatment . Expression of SUV39H1-GFP was induced or not for 6 hr ( +/- dox ) , then mitotic cells were spun onto coverslips and incubated without RNase , with RNase A , or with RNase A plus RNase inhibitors . Cells were then stained for DNA ( blue ) , with an anti-GFP antibody to detect SUV39H1-GFP , HP1α ( green ) , and HEC1 to mark centromeres ( red ) . ( B ) Quantification of HP1α at pericentric regions from experiment shown in A . Each experiment was normalized to the untreated -dox measurement . Bars are the average of n = 5 separate experiments , 15 cells quantified per condition per experiment , error bars represent standard error . ( C ) Quantification of H3K9me3 at pericentric regions in HeLa cells with or without exogenous expression of SUV39H1-GFP for 6 hr ( +/- dox ) . Bars are the average of n = 3 separate experiments , 15 cells quantified per condition per experiment , error bars represent standard error . ( D ) HP1α localization is reduced in cells expressing mutant SUV39H1-GFP compared to cells expressing WT SUV39H1-GFP . Expression of SUV39H1-GFP , WT or mutants , was induced or not for 6 hr ( -/+ dox ) in HeLa Flp in TREx cell lines . Mitotic cells were spun onto coverslips and stained for DNA ( blue ) , HP1α ( green ) , and HEC1 to mark centromeres ( red ) . ( E ) Quantification of HP1α at pericentric regions from experiment shown in D . Each experiment was normalized to the wildtype measurement . Bars are the average of n = 5 separate experiments , 15 cells quantified per condition per experiment , error bars represent standard error . P values were calculated with paired , two-tailed t tests . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 015 To specifically test the contribution of RNA to SUV39H1 localization , we digested mitotic chromosomes with RNase A , a treatment we previously observed removed both pericentric RNA ( Figure 1D , Figure 1—figure supplement 1A ) and WT SUV39H1 ( Figure 2A and B ) from chromosomes . RNase A treatment reduced the pericentric localization of SUV39H1 to 43 ± 4% of the undigested control , similar to the levels of the untreated SUV39H1R55A mutant in the same experiments ( 52 ± 9% of WT ) ( Figure 6D and E ) . This indicates that the effect of the R55A mutation on SUV39H1 localization can be fully attributed to a loss in RNA binding , independent of DNA binding . This is further supported by the fact that RNase A treatment did not significantly reduce the amount of SUV39H1R55A at pericentric chromatin ( untreated: 52 ± 9% , RNase A: 42 ± 5% ) ( Figure 6D and E ) . Taken together , these results demonstrate that direct binding to both chromosome-associated RNA and H3K9me2/3 are required for proper localization of SUV39H1 to pericentric heterochromatin . Previous studies found that HP1α localization at heterochromatin is sensitive to RNase treatment in both mouse interphase cells and on human mitotic chromosomes ( Maison et al . , 2002; Wong et al . , 2007 ) . We also observed a reduction in HP1α levels on chromatin after RNase A treatment , to 77 ± 13% of control levels ( Figure 6—figure supplement 1A and B ) . We wondered if the RNase sensitivity of HP1α could be due in part to its interaction with SUV39H1 . Consistent with this , expressing SUV39H1-GFP for 6 hr , which causes an increase in SUV39H1-GFP at pericentric regions , also caused a 1 . 5-fold increase in HP1α localization ( Figure 6—figure supplement 1A and B ) , in the absence of any detectable change in H3K9me3 levels ( Figure 6—figure supplement 1C ) . Additionally , we observed that RNase A treatment abolished the increase in HP1α localization following exogenous SUV39H1-GFP expression ( Figure 6—figure supplement 1A and B ) . Finally , after expressing SUV39H1 mutants , the extent of HP1α localization positively correlated with SUV39H1 localization ( Figure 6A and C; Figure 6—figure supplement 1D and E ) . These results are consistent with a model in which the pericentric localization of HP1α during mitosis depends on the SUV39H1 protein – independent of H3K9 methylation – and that HP1α localization is sensitive to RNase treatment because SUV39H1 localization depends on direct binding to RNA . The requirement for RNA for SUV39H1 localization in mitosis suggests that RNA binding might generally stabilize the interaction of SUV39H1 with chromatin . To assess this in interphase cells , we measured the turnover rate of GFP-tagged WT SUV39H1 , SUV39H1R55A , SUV39H1R84A , SUV39H1F43A , and SUV39H1W64A using fluorescence recovery after photobleaching ( FRAP ) ( Figure 7A , B and C ) . After photobleaching , WT SUV39H1 recovered with an average half-time of 27 ± 7 s ( Figure 7D ) , with a stable , immobile fraction of SUV39H1 that did not recover during the time frame of our experiment ( Figure 7B , C and E ) . The mobile fraction represented 58 ± 7% of the total nuclear SUV39H1 ( Figure 7E ) , consistent with previous measurements ( Hahn et al . , 2013; Krouwels et al . , 2005 ) . The aromatic cage mutants SUV39H1W64A ( Figure 7B ) and SUV39H1F43A ( Figure 7C ) , which disrupt binding to H3K9me3 , had a much faster recovery half-time of 6 ± 2 and 5 ± 1 s , respectively ( Figure 7D ) . Unlike WT SUV39H1 , approximately 90–95% of each aromatic cage mutant was determined to be mobile within the nucleus ( Figure 7E ) . Together , this high level of mobility and rapid recovery is consistent with weak and/or transient associations with chromatin that are normally stabilized by SUV39H1’s interaction with methylated histones . 10 . 7554/eLife . 25299 . 016Figure 7 . SUV39H1 depends on direct nucleic acid binding for its stable association with chromatin in interphase cells . ( A ) Fluorescence recovery after photobleaching ( FRAP ) assay to measure SUV39H1-GFP association with chromatin . Montage of images of representative cells expressing either WT or mutant SUV39H1-GFP , as indicated , in our FRAP assay . Leftmost panel shows the ROIs used for quantification and analysis , with the innermost circle representing the diameter of the photobleached area . ( B , C ) Quantification of SUV39H1-GFP recovery after photobleaching . The fluorescence intensity of the bleached area was normalized relative to the whole nucleus , and plotted as a function of time for WT ( gray , n = 19 ) , R55A ( red , n = 19 ) , W64A ( green , n = 23 ) , R84A ( blue , n = 23 ) and F43A ( orange , n = 35 ) mutants . Black lines are the average between individual traces and colored shading represents standard deviation . WT curves are shown in both panels for clarity . ( D ) Half-time to recovery measurements from SUV39H1-GFP FRAP traces . Individual traces ( averages shown in 7B and 7C ) were fit to a single exponential , and the half-time to recovery for each curve was plotted for WT and mutant SUV39H1-GFP cells . ( E ) Quantification of the mobile fraction of SUV39H1-GFP . Bar graph of the mean amplitude of the curves from 7B and 7C . Error bars represent standard deviation . The differences between mobile fractions are statistically significant ( p<0 . 01 ) , except between R55A and R84A . ( F ) Half-time to recovery measurements from SUV39H1-GFP FRAP traces , in the presence of the RNA Pol I inhibitor CX5461 . Individual traces were fit to a single exponential , and the half-time to recovery for each curve was plotted for WT and mutant SUV39H1-GFP cells . ( G ) Quantification of the mobile fraction of SUV39H1-GFP , in the presence of the RNA Pol I inhibitor CX5461 . Bar graph of the mean amplitude of recovery curves . Error bars represent standard deviation . See also Figure 7—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 01610 . 7554/eLife . 25299 . 017Figure 7—figure supplement 1 . Effect of transcription inhibition on SUV39H1 fluorescence recovery , and model for RNA-dependent retention of SUV39H1 at heterochromatin . ( A ) Half-time to recovery measurements from SUV39H1-GFP FRAP traces , in the presence of two different RNA Pol II inhibitors , α-amanitin and triptolide . Individual traces were fit to a single exponential , and the half-time to recovery for each curve was plotted for WT and mutant SUV39H1-GFP cells . ( B ) Quantification of the mobile fraction of SUV39H1-GFP , in the presence of two different RNA Pol II inhibitors , α-amanitin and triptolide . Bar graph of the mean amplitude of recovery curves . Error bars represent standard deviation . ( C ) Model for RNA-dependent localization of SUV39H1 . Left: WT SUV39H1 binds both H3K9me3 and RNA through via its chromodomain ( CD ) , and these interactions properly localize SUV39H1 to heterochromatin . The SET domain of SUV39H1 ( SET ) methylates H3K9 , and the N-terminus of SUV39H1 ( N ) helps to recruit HP1α via its chromoshadow domain ( CSD ) . Right: When the interaction between SUV39H1 and RNA is disrupted , either by RNase or by the RNA binding mutants SUV39H1R55A or SUV39H1R84A , the association of SUV39H1 with heterochromatin is weakened . Although SUV39H1 can still bind H3K9me2/3 , SUV39H1’s interaction with chromatin is more unstable and dynamic , and the levels of both SUV39H1 and HP1α bound to heterochromatin are reduced . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 017 Interestingly , the nucleic acid binding deficient mutants SUV39H1R55A ( Figure 7B ) and SUV39H1R84A ( Figure 7C ) showed an average recovery time in between that of WT SUV39H1 and the H3K9me3 binding mutants , with half-lives of 14 ± 3 and 13 ± 2 s , respectively ( Figure 7D ) . Consistent with this trend , both SUV39H1R55A and SUV39H1R84A demonstrated a level of mobility ( 77% ± 6 ) that was greater than WT SUV39H1 , but less than the aromatic cage mutants SUV39H1W64A and SUV39H1F43A ( Figure 7E ) . We noted that the sub-nuclear localization of the aromatic cage mutants was more diffuse , and generally lacking the typical heterochromatin centers observed in the WT SUV39H1 , SUV39H1R55A and SUV39H1R84A cell lines ( Figure 7A ) . Combined with our in vitro biochemical analysis , these observations support a model in which direct binding to nucleic acids stabilizes the association of SUV39H1 with chromatin in interphase cells as well as in mitosis . If RNA stabilizes SUV39H1 on chromatin , then inhibiting RNA transcription to reduce the levels of chromatin-associated RNA is expected to destabilize SUV39H1 localization . When we added the RNA Pol I inhibitor CX-5461 – a treatment that we saw reduces RNA localization on mitotic chromosomes ( Figure 1—figure supplement 2A and B ) – the fluorescence recovery half-time of WT SUV39H1 decreased from 27 ± 7 s to 16 ± 3 s , comparable to the half-time of the nucleic acid binding mutant SUV39H1R55A ( Figure 7F ) . However , we also observed a slight effect of Pol I inhibition on SUV39H1R55A , with a decrease in half-time from 14 ± 3 to 11 ± 2 s . This suggests that RNA polymerase inhibition may have non-specific effects on SUV39H1 localization in addition to disrupting SUV39H1 binding to chromatin through chromatin-associated RNAs . We also saw similar effects with α-amanitin or triptolide treatment ( Figure 7—figure supplement 1A ) , suggesting that Pol II may also be playing a role in SUV39H1 localization; however , this effect may also be due to the partial inhibition of Pol I with these inhibitors ( Figure 1—figure supplement 2B ) . Although polymerase inhibition resulted in measureable changes in the recovery kinetics of SUV39H1 , we observed no significant changes in overall SUV39H1 mobility ( Figure 7G ) . Because RNA binding and H3K9me3 binding both contribute to SUV39H1 localization , we wanted to determine the effect of disrupting these interactions on SUV39H1-dependent heterochromatic silencing . Using CRISPR/Cas9 genome editing , we generate DLD-1 cell lines in which the endogenous SUV39H1 was altered to contain either a nucleic acid binding mutation ( R55A ) , an H3K9me3 binding mutation ( F43A ) or both ( R55A/F43A ) ( Figure 8A ) . We made these SUV39H1 mutations in a SUV39H2 knockout background to prevent redundant activities of SUV39H2 from masking effects of these mutations . We found that total H3K9me3 levels , as well as H3K9me3 localization at centromeres , were reduced in all SUV39H1 mutants compared to WT SUV39H1 ( Figure 8B , C , D and E ) , demonstrating a role for both H3K9me3 binding and nucleic acid binding by SUV39H1 in maintaining normal H3K9 methylation levels . To assess the role of SUV39H1 nucleic acid binding and H3K9me3 binding in repressing pericentromeric heterochromatin , we measured α-satellite RNA levels in the WT and mutant SUV39H1 cell lines . We found that α-satellite expression increased 1 . 35-fold , 1 . 56-fold , and 2 . 10-fold , in the R55A , F43A , and the R55A/F43A double mutant cells lines , respectively , compared to WT ( Figure 8F ) . Interestingly , although no change in H3K9me3 was detected between the F43A single mutant and the R55A/F43A double mutant ( Figure 8C and D ) , a significant increase in α-satellite RNA levels was observed in R55A/F43A compared to F43A ( Figure 8F ) , indicating that SUV39H1 nucleic binding may make H3K9me3-independent contributions to heterochromatin function . Importantly , because the defects in SUV39H1 localization caused by the R55A nucleic acid binding mutation were similar to defects caused by RNase treatment or RNA polymerase inhibition ( Figure 6E , Figure 7 ) , we support a model in which the heterochromatin defects we observe with this mutant are due to breaking interactions with chromatin-associated RNA . Taken together , these data demonstrate that both RNA binding and H3K9me3 binding by SUV39H1 contribute to silencing of constitutive heterochromatin in human cells . 10 . 7554/eLife . 25299 . 018Figure 8 . Nucleic acid binding and H3K9me3 binding both contribute to SUV39H1-mediated constitutive heterochromatin silencing . ( A ) Sanger DNA sequencing traces showing point mutations in the endogenous SUV39H1 locus in DLD-1 cells , generated by CRISPR/Cas9 gene editing . ( B ) Representative western blot showing H3K9me3 , histone H4 , SUV39H1 , and HP1α levels in normal unedited DLD-1 cells ( far left ) , or SUV39H2 KO DLD-1 cells with the indicated mutations in SUV39H1 . ( C ) Total H3K9me3 levels in mutant SUV39 DLD-1 cells , measured by quantitative western blot . Nuclear lysate was prepared and blotted with indicated antibodies , and H3K9me3 signal was normalized to histone H4 levels . Graphed are the means of 3 repeats , errors bars represent standard error . ( D ) Quantification of H3K9me3 localization at centromeres in mutant SUV39 DLD-1 cells . Cells were grown on coverslips , fixed and permeabilized , then stained with the indicated antibodies , imaged , and quantified using centromere finder software , using CREST staining as a centromere marker . Graphed are the means of 4 repeats , error bars represent standard error . Significance was determined using paired , two-tailed t-tests . p<0 . 05 , p<0 . 005 , *p<0 . 0005 . ( E ) Representative images of immunofluorescence staining of mutant SUV39 DLD-1 cells , used for quantification shown in 8D . Cells are stained for DNA ( blue ) , H3K9me3 ( green ) , and CREST centromere stain ( red ) . ( F ) Quantification of α-satellite RNA in mutant SUV39 DLD-1 cells by RT-qPCR . Total α-satellite RNA levels are normalized to GAPDH RNA levels . Graphed are the means of 3 repeats , error bars represent standard error . Significance was determined using paired , two-tailed t-tests . p<0 . 05 , **p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 25299 . 018 We show for the first time that RNAs associated with mitotic chromosomes are enriched at pericentric regions in several human cell lines ( Figure 1 , Figure 1—figure supplement 1 ) . Previous studies have shown that transcription by RNA polymerase II ( Pol II ) can occur at the centromeres of human mitotic chromosomes ( Chan et al . , 2012; Liu et al . , 2015 ) . The RNA we observe is distinct from this previously described centromeric RNA in that its localization is insensitive to Pol II inhibition ( Figure 1—figure supplement 2A and B ) , and exhibits broader pericentric localization . RNA FISH shows that the RNA we observe is transcribed from pericentric α-satellite sequences outside of the core centromere region ( Figure 1D , Figure 1—figure supplement 1G ) . Additionally , the α-satellite RNAs we detect are associated in cis with the chromosomes from which they are transcribed ( Figure 1D , Figure 1—figure supplement 1F and G ) , supporting the idea that their localization may provide direct feedback on the local transcriptional state . This differs from the phenomenon observed in Drosophila cells , where SAT III RNA transcribed from the X chromosome localizes to all centromeres in trans ( Rošić et al . , 2014 ) . It is interesting that we observe RNA on mitotic chromosomes , as transcription is silenced during mitosis ( Gottesfeld and Forbes , 1997 ) , and known noncoding RNAs , such as Xist , dissociate from chromatin during mitosis in human cells ( Hall et al . , 2014 ) . This may reflect the particular importance of centromeric and pericentric RNA in maintaining chromatin structure around the centromere for chromosome cohesion and proper chromosome segregation during mitosis ( Ekwall et al . , 1996; Melcher et al . , 2000; Peters et al . , 2001 ) . Centromeric and pericentric RNAs have been proposed to play functional roles in several different organisms ( Blower , 2016; Carone et al . , 2009; Choi et al . , 2011; Grenfell et al . , 2016; Probst et al . , 2010; Rošić et al . , 2014 ) , and the transcription of repetitive α-satellite sequences at human centromeres has been implicated in maintaining centromere identity ( Chan et al . , 2012; Quénet and Dalal , 2014; Wong et al . , 2007 ) , regulating the mitotic kinase Aurora B ( Ferri et al . , 2009; Ideue et al . , 2014; Jambhekar et al . , 2014; Mallm and Rippe , 2015 ) , and influencing chromosome stability ( Zhu et al . , 2011 ) and cohesion ( Liu et al . , 2015 ) . RNAs encoded by pericentric DNA may maintain their association with mitotic chromosomes to regulate these processes , and may interact with chromatin in a manner that is distinct from other regulatory RNAs that dissociate from chromatin during mitosis . We find that the human SUV39H1 protein directly binds to RNA through its chromodomain . This is consistent with the identification of other chromodomain-containing proteins that bind RNA ( Akhtar et al . , 2000 ) and the interaction of the SUV39H1 chromodomain with TERRA RNA ( Porro et al . , 2014 ) . However , we also detect binding of SUV39H1 to other types of nucleic acid species ( dsDNA , ssDNA , dsRNA , RNA:DNA ) ( Figure 3E ) , observe no strong preference for any sequence tested ( Figure 3E and F , Figure 3—figure supplement 1D ) , and see a positive correlation between nucleic acid length and binding affinity ( Figure 3—figure supplement 1F ) . Additionally , we find that a loss of positively charged residues generally decreases SUV39H1-RNA affinity , whereas a loss of negatively charged residues generally increases SUV39H1-RNA affinity ( Figure 4—figure supplement 1B and C ) . One hypothesis that accounts for these observations is that binding occurs via electrostatic interactions between the negatively charged phosphate backbone of nucleic acids and the positively charged residues within the SUV39H1 chromodomain . It is notable that the Polycomb Repressive Complex 2 ( PRC2 ) also exhibits relatively promiscuous , length-dependent , sequence-independent RNA binding; though specificity for certain RNA sequences can be observed in vivo and in vitro ( Davidovich et al . , 2013 , 2015 ) . It is possible that for SUV39H1 and PRC2 , and potentially for other chromatin modifiers , RNA binding may generally stabilize interactions with chromatin rather than target the proteins to specific sites . Although we observe no sequence specificity for binding α-satellite RNA in vitro ( Figure 3F ) , SUV39H1 associates with α-satellite RNA in human cells , and this interaction is impaired by mutations that disrupt the ability of SUV39H1 to directly bind RNA in vitro ( Figure 5C ) . It will be important to further explore the specificity of the SUV39H1-RNA interaction , as well as more broadly identify RNAs that associate with SUV39H1 in human cells , as had been done with PRC2 ( Davidovich et al . , 2013; Kaneko et al . , 2013 ) . By generating point mutations that separate the nucleic acid binding and H3K9me3 binding functions of the SUV39H1 chromodomain , we assessed the individual contributions of these two interactions on SUV39H1 localization and function . We show for the first time that direct nucleic acid binding contributes to SUV39H1 localization by stabilizing its interaction with heterochromatin , in both interphase and mitotic human cells ( Figures 6 and 7 ) , and that the localization defects we observe in RNA binding mutants of SUV39H1 are largely recapitulated by RNase treatment or RNA polymerase inhibition ( Figures 6D , E , 7F and G ) ; strongly suggesting that SUV39H1 requires direct RNA binding for its stable association with chromatin . Importantly , we demonstrate that breaking the nucleic acid binding of SUV39H1 leads to defects in HP1α localization ( Figure 6—figure supplement 1 ) , H3K9me3 methylation ( Figure 8B , C , D and E ) , and repression of repetitive α-satellite sequences ( Figure 8F ) . It is important to note that because of the high conservation between SUV39H1 and HP1α aromatic cages ( Figure 4B ) , it has been widely hypothesized that an intact aromatic cage is necessary for the localization of SUV39H1 ( Wang et al . , 2012 ) , but this hypothesis has not been directly assessed . Both aromatic cage mutations of SUV39H1 we tested resulted in an almost complete loss of SUV39H1 localization and retention on chromatin , confirming that binding to H3K9me3 through the aromatic cage is indeed a key determinant of SUV39H1 localization ( Figures 6 and 7 ) . It remains a major outstanding question how SUV39H1 specifically recognizes its genomic sites of action . Although it had been originally proposed that HP1 proteins directly recruit SUV39H1 ( Bannister et al . , 2001; Hall et al . , 2002; Lachner et al . , 2001 ) , subsequent studies in fission yeast demonstrated that SUV39/Clr4 localization is largely independent of HP1 proteins ( Motamedi et al . , 2008; Sadaie et al . , 2004 ) . A recent study also showed that the N-terminus of mammalian SUV39H1 contributes to its association with chromatin in vitro and in vivo through a zinc-finger like motif thought to directly bind DNA ( Müller et al . , 2016 ) . However , SUV39H1 binds a variety of repetitive sequences in mammals ( Bulut-Karslioglu et al . , 2014 ) with no apparent DNA consensus sequence , making it unlikely that DNA recognition is the sole factor determining SUV39H1 targeting . In addition , the chromodomain of SUV39H1 has been shown previously to be important for its localization ( Melcher et al . , 2000 ) , but the relative contributions of binding to methylated H3K9 versus other interactions mediated by the chromodomain were not assessed . Here , we identify both RNA and H3K9me2/3 as main determinants of SUV39H1 localization . Importantly , the RNA-dependent mechanism for SUV39 function appears to be present in both human and mouse cells ( Shirai et al . , 2017; Velazquez Camacho et al . , 2017 ) suggesting a conserved role for RNA in regulating constitutive heterochromatin . A possible advantage to binding chromatin-associated RNA is that it provides direct feedback on the transcriptional state of the locus , ensuring that over-transcribed repeats retain more SUV39H1 to promote repression . Because SUV39H1 displays no apparent sequence preference for binding RNA ( Figure 3E and F , Figure 3—figure supplement 1D ) , it is unlikely that RNA binding alone could confer selective binding for one genomic region over another . It is also unlikely that H3K9me3 binding alone targets SUV39H1 , for many H3K9me3 sites exist in the mammalian genome , but not all recruit SUV39H1 ( Bulut-Karslioglu et al . , 2014 ) . We propose a model in which SUV39H1 engages chromatin through several different interactions that in combination define its genomic distribution ( Figure 7—figure supplement 1C ) . When direct binding to RNA is disrupted , SUV39H1 can still bind to H3K9me3; however , SUV39H1’s interaction with chromatin is destabilized , and its localization at heterochromatin is reduced . RNA binding may help SUV39H1 distinguish between target and non-target sites in the genome that both have H3K9me2/3 . In addition to directly binding RNA ( Figure 3 ) ( Porro et al . , 2014 ) , H3K9me3 ( Figure 4E and F ) ( Wang et al . , 2012 ) , and DNA ( Figure 3E ) ( Müller et al . , 2016 ) , SUV39H1 also directly binds the deacetylase Sirtuin 1 ( SIRT1 ) ( Bosch-Presegué et al . , 2011; Vaquero et al . , 2007 ) , and associates with the retinoblastoma tumor suppressor ( Rb ) protein ( Nielsen et al . , 2001; Vandel et al . , 2001 ) . These and other interactions may influence SUV39H1 recruitment , stable association , and spread on chromatin – as well as its methyltransferase activity – to define its sites of action . Thus , a major question going forward is how the collective interactions between SUV39H1 and its binding partners give rise to locus specific targeting and heterochromatin formation . Towards this goal , we have demonstrated that H3K9me2/3 and chromatin-associated RNA are key determinants for SUV39H1 localization and are necessary for maintaining silenced constitutive heterochromatin in human cells . HeLa , U2OS , HFF , HT1080 , Huh-7 , and primary fibroblasts were grown in Dulbecco’s Modified Eagle Medium ( DMEM ) , hTERT-RPE cells were grown in DMEM/F12 media , and DLD-1 cells were grown in RPMI 1640 media . HeLa cells and U2OS cells were acquired from the ATCC . We obtained DLD-1 human colorectal cancer cells from Dr . Daniele Fachinetti and Dr . Don Cleveland , primary fibroblasts from Dr . Paul Khavari , HFF ( human foreskin fibroblast ) cells from Dr . Matthew Bogyo , HT-1080 fibrosarcoma cells from Dr . Jianghong Rao , hTERT RPE-1 ( retina pigmented epithelial ) cells from Dr . Tim Stearns , Huh-7 human hepatoma cells from Dr . Peter Sarnow , and Flp-In T-REx HeLa cells from Dr . Pat Brown . These cells were not independently authenticated . Mycoplasma contamination was monitored frequently by cytoplasmic DAPI staining . All cells were grown under American Tissue Culture Collection ( ATCC ) standard conditions . All media was purchased from Thermo Fisher and supplemented with 10% vol/vol FBS and 100 U/ml penicillin and streptomycin . Flp-In T-REx HeLa cell lines expressing SUV39H1-GFP proteins were made by co-transfecting pOG44 Flp-recombinase expression vector and pcDNA5/FRT/TO tet inducible Flp-In cloning vector ( Invitrogen ) containing SUV39H1-GFP sequences ( ASP 1751 , 2137 , 2668 , 2663 , 2670 , 2789 ) using Fugene 6 ( Promega ) . After 48–72 hr , cells were selected with 350 μg/mL hygromycin B ( Invitrogen ) and 15 μg/mL blasticidin S ( Invivogen ) until visible colonies formed , then the entire selected population of cells was saved . SUV39 DKO cells were generated essentially as described ( Ran et al . , 2013 ) . Briefly , HeLa or DLD-1 cells were transfected with pX458 ( pSpCas9 ( BB ) −2A-GFP ) containing SUV39H1 and SUV39H2 gRNA sequences using Fugene HD ( Promega ) . SUV39H1 gRNA sequence ( ASON 2868 , ASP 2985 ) : 5’-caccgGTTCCTCTTAGAGATACCGA-3’ . SUV39H2 gRNA sequence ( ASON 2878 , ASP 2990 ) : 5’-caccgAAAGCTCTACAAGATGGCGG-3’ . After 4–5 days , GFP-positive cells were single-cell sorted into 96-well plates using a Sony SH800 cell sorter . Clonal populations were expanded and screened by western blotting for SUV39H1 . Mutations in SUV39H1 and SUV39H2 were confirmed by purifying genomic DNA , PCR amplifying the targeted locus , and analyzing the amplicons by high throughput sequencing on a MiSeq system ( Illumina ) . DLD-1 cell lines with point mutations in the endogenous SUV39H1 locus were generated by first making a SUV39H2 KO cell line ( ASTC 269 , described above ) , then performing further genome editing by co-transfecting pX458 ( pSpCas9 ( BB ) −2A-GFP ) with a SUV39H1-targeting gRNA ( ASP 2987; 5’-caccgGGATCTTCTTGTAATCGCAC-3’ ) and a puc18 homology arm construct containing desired point mutations ( R55A: ASP 3223 , F43A: ASP 3222 , or R55A/F43A: ASP 3543 ) with Fugene HD . After 4–5 days , GFP-positive cells were single-cell sorted into 96-well plates using a Sony SH800 cell sorter . Clonal populations were expanded and screened by western blotting for SUV39H1 . Mutations in SUV39H1 were confirmed by purifying genomic DNA , PCR amplifying the targeted locus , and analyzing the amplicons by Sanger sequencing . For mitotic spreads , cells were grown to about 70% confluency in T-25 or T-75 flasks , then mitotic cells were selected by shaking off less adherent cells . Released cells were pelleted , resuspended in 100 μL 75 mM KCl , then counted on a hemocytometer and diluted with more 75 mM KCl to 2 × 10∧5 cells/mL . 100 μL cells were spun onto no . 1 . 5 glass coverslips using a Shandon Cytospin 4 at 2000 rpm for 5 min . Coverslips were hydrated with KCM buffer ( 10 mM Tris-HCl pH 8 . 0 , 120 mM KCl , 20 mM NaCl , 0 . 5 mM EDTA , 0 . 1% Triton X-100 ) , then cells were permeabilized for 10 min with KCM buffer with 0 . 5% Triton X-100 . Coverslips were blocked with KCM + 2% BSA for 30 min , then stained with specific antibodies diluted in KCM + 2% BSA , 30 min each . After staining , cells were washed with KCM buffer and fixed with 3 . 7% formaldehyde in KCM for 10 min . After fixation , cells were washed with PBS , stained with 10 μg/mL Hoescht for 10 min , washed with PBS + 0 . 1% Triton X-100 , washed with PBS , then mounted onto microscope slides . For staining interphase DLD-1 cells , cells were plated on glass coverslips , fixed and permeabilized with 1% formaldehyde , 0 . 5% Triton X-100 , 1X PBS for 10 min , then re-permeabilized with 0 . 5% Triton X-100 , 1X PBS for 10 min . Cells were blocked in antibody dilution buffer ( AbDil: 20 mM Tris-HCl pH 7 . 4 , 150 mM NaCl , 0 . 1% Triton X-100 , 2% BSA , 0 . 1% sodium azide ) for 30 min , then stained with primary and secondary antibodies diluted in AbDil for 30 min each . Cells were then stained with 10 μg/mL Hoechst in Abdil for 10 min , washed with PBS + 0 . 2% Triton X-100 , washed with PBS , then mounted onto slides and imaged . We used the following antibodies for immunofluorescence: mouse anti-HEC1 ( Abcam ab3613 , 1:1000 ) , rabbit anti-CENP-A-Alexa647 ( Straight laboratory , 1:1000 ) , rabbit anti-GFP-Alexa488 ( Straight laboratory , 2 μg/mL ) , rabbit anti-HP1α ( Straight laboratory , 2 μg/mL ) , rabbit anti-H3K9me3 ( Abcam ab8898 , 1:1000 ) , human anti-centromere protein ( CREST , Antibodies , inc . , 1:100 ) . All secondary antibodies were obtained from Invitrogen and diluted 1:1000 . For experiments where SUV39H1-GFP localization was assessed , expression of WT or mutant versions of SUV39H1-GFP was induced in HeLa Flp-In T-REx cells with 1 μg/mL doxycycline for 6 hr before mitotic shake off . To visualize RNA by EU labeling , cells were incubated with 0 . 5 mM EU ( synthesized as previously described [Jao and Salic , 2008] ) for 4–12 hr before mitotic shake off . In RNA polymerase inhibitor experiments , cells were incubated with 0 . 5 mM EU and 50 μg/mL α-amanitin ( Santa Cruz Biotech . ) , 1 μM triptolide ( Selleckchem ) , 50 ng/mL actinomycin D ( Calbiochem ) , or 1 μM CX-5461 ( Millipore ) for 6 hr . Cells were processed and stained as above . After fixation , cells were washed with PBS , and EU-containing RNA was labeled using a Click-iT RNA Imaging kit ( Invitrogen C10329 ) , essentially according to the manufacturer’s instructions . In RNase sensitivity experiments , 0 . 5 mg/mL RNase A ( Sigma ) , III , or H was included in the 30 min block step before antibody staining ( in KCM –EDTA + 5 mM MgCl2 in experiments with RNase H and RNase III ) . Plus inhibitor ( +inhibitor ) controls also containing 4 U/μL each of RNaseOUT ( Invitrogen ) and RNase inhibitor ( Ambion AM2684 ) . Stacks of fixed spread mitotic chromosomes and interphase nuclei were acquired at 0 . 2 μm axial steps using a motorized stage mounted on an Olympus IX70 microscope , using a 100X , 1 . 4 NA PlanApo objective and a CoolSnap-HQ CCD camera ( Photometrics ) on a DeltaVision Spectris system ( Applied Precision ) . Displayed images are maximum intensity projections of raw or deconvolved z-stacks . Fluorescent signal at pericentric regions of mitotic chromosomes or at centromeres in interphase nuclei was quantified using custom software . Pericentromere finder ( Fuller , 2014 ) : centromeres were localized using HEC1 staining , and pericentric regions were defined by segmenting on DNA on Hoescht-stained chromosomes within a 10-pixel radius around centromeres . Parameters used were min-size = 4 , max-intercentromere-dist = 17 , marker-channel = 2 ( HEC1 staining ) , and pericentromere-channel = 0 ( DNA staining ) . Centromere finder ( Fuller , 2016 ) : centromeres were localized using CREST staining , and fluorescence of each channel was measured . Parameters used were min_size = 5 , max_size = 60 , marker_channel_index = 3 ( CREST staining ) , decrease_speckle_background = false . Intensity measurements were performed on non-deconvolved maximum intensity projections of each z-series . GFP background signal was assessed by imaging spreads from cells that were not induced with doxycycline , and therefore were not expressing SUV39H1-GFP . Background in the HP1α channel was assessed by imaging spread cells not stained with HP1α primary antibody , and EU-RNA background was assessed with a no EU treatment control . Mitotic DLD-1 cells were selected by mitotic shake off , washed with PBS , resuspended in 75 mM KCl to 5 . 8 × 104 cells/mL , then swelled at 37°C for 15 min . 500 μL cells were cytospun onto RNase AWAY treated slides ( 2000 rpm , 10 min ) . Cells were permeabilized in KCM buffer for 5 min , fixed in 4% PFA in PBS for 10 min , then blocked for 15 min ( 1X PBS , 0 . 5% Triton X-100 , 1% BSA ) and stained with specific antibodies ( described above ) in block . Cells were permeabilized again in CSK buffer/2 mM RVC ( Ribonucleoside Vanadyl Complex ) /0 . 5% Triton X-100 , 10 min on ice , then fixed again in 4% PFA in PBS for 10 min . Slides were dehydrated with an ethanol series ( 70% , 80% , 95% , 100% , 2 min each on ice ) , then air-dried at 37°C . After precipitating and denaturing , probes were hybridized in hybridization mix ( +RVC ) at 37°C overnight in a humid chamber ( pSD1-1 and ASP851 probes: 65% formamide; pβ4 and pTRS-47 probes: 50% formamide ) . Slides were then washed with 2X SSC/0 . 05% Tween/formamide ( same formamide percentages used in hybridization mix ) , pH 7 at 37°C , 3 times 5 min each -- then with 2X SSC/0 . 05% Tween , at 37°C , 3 times 5 min each – then 1X SSC/0 . 05% Tween , at room temp , 10 min – then 4X SSC/0 . 1% Tween , at room temp , 5 min . For biotin-16-dUTP labeled probes , after post-hybridization washes , slides were blocked in blocking buffer ( PBS + 0 . 5% Triton X-100 + 1% BSA + 0 . 02% sodium azide ) and biotin was detected with Alexa Fluor 488 streptavidin diluted in blocking buffer for 2 hr at room temperature . Slides were then washed with 4X SSC/0 . 1% Tween three times at room temp , 5 min each . All slides were stained with DAPI , coverslips were mounted , and images were acquired as described above . CSK buffer: 100 mM NaCl , 300 mM sucrose , 3 mM MgCl2 , 10 mM PIPES pH 6 . 8 . Probe hybridization mix: 50% formamide ( 65–68% for repetitive ) , 10% dextran sulfate , 2X SSC , 1% Tween-20 . RNase-treated controls slides were incubated with Ambion RNase cocktail ( 7 . 5 μL/mL KCM buffer ) at 37°C for 2 hr up to overnight before hybridization . DNA FISH was performed as above , with the following differences: slides were fixed in 10% formalin for 10 min at room temp , permeabilized in KCM for 10 min , denatured in 70% formamide , 2X SSC , pH 7 at 70°C for 3 min , then hybridized overnight . All probes were labeled by nick translation of plasmids either indirectly with Roche biotin-16-dUTP and subsequently detected with AlexaFluor 488 streptavidin or directly with AlexaFluor 488-dUTP . D1Z5-specific RNA FISH probes were generated from the pSD1-1 plasmid , which contains a single copy of the 1 . 9 kb D1Z5 higher order repeat ( HOR ) . Probes recognizing chromosome 13 and chromosome 21 α-satellite arrays were generated using a template PCR amplified from human genomic DNA with the following primers: α-satellite I ( ASON 614 ) : 5’-CTTGCTAGCAATCTGCAAGTGG-3’ , and α-satellite II ( ASON 615 ) : 5’-CTTGTCGACTACAAAAAGAGTG-3’ . β-satellite is detected with nick-translated plasmid pβ4 , which recognizes β-satellite sequences on the acrocentric chromosomes ( chromosomes 13 , 14 , 15 , 21 , and 22 ) . Satellite III is detected with nick-translated plasmid pTRS-47 , which recognizes Satellite III subfamily sequences on chromosomes 14 and 22 . pSD1-1 , pβ4 , and pTRS-47 plasmids were a gift from the lab of Dr . Huntington Willard . 500 ng of 488 dUTP-labeled probes were precipitated with 0 . 5 uL salmon sperm DNA ( 10 mg/mL stock ) and 2 . 5 volumes 100% EtOH for each 18 × 18 cm2 area for at least 4 hr at −80° C or overnight at −20° C . Before use , probes were pelleted at maximum speed in an Eppendorf microcentrifuge , washed with 70% ethanol , resuspended in 65% formamide hybridization mix , and denatured at 74°C for 10 min . The ‘RGB profile plot’ ImageJ plugin was used for line scan analysis of RNA FISH images . All purified SUV39H1 proteins used in this study were expressed as N-terminal Maltose Binding Protein ( MBP ) fusions . pMAL-c2X ( NEB ) expression vectors , carrying sequences encoding truncated versions of human SUV39H1 , were transformed into BL21 E . coli and grown in 2–6 L of 2X YT at 37°C to an OD of 0 . 6–0 . 8 . Cells were induced with 0 . 3 mM IPTG for 4–6 hr at room temperature . Cells were harvested , frozen in liquid nitrogen and stored at −80°C . Each liter of pelleted E . coli culture was resuspended in 20 mL of ice cold 4X lysis buffer ( 80 mM Tris-HCl pH 7 . 4 , 0 . 8 M NaCl , 0 . 4% Triton X-100 , 5 mM 2-mercaptoethanol , 1 mM PMSF , 1 mM benzamidine hydrochloride , 0 . 2 mg/ml lysozyme ) . Lysates were sonicated and centrifuged at 96 , 000Xg for 1 hr at 4°C . Supernatants were passed over 2–5 mL of either amylose resin ( NEB ) or glutathione agarose ( Sigma Aldrich ) , followed by washing with approximately 10 column volumes of 4X lysis buffer , 30 column volumes of 1X lysis buffer ( 20 mM Tris-HCl pH 7 . 4 , 0 . 2 M NaCl , 0 . 1% Triton X-100 ) , and five column volumes of 1X lysis buffer without Triton X-100 at 4°C . Protein was eluted with approximately five col . volumes of 1X lysis buffer without Triton X-100 plus either 10 mM maltose or 5 mM glutathione . MBP fusions were dialyzed into 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol , 5 mM 2-mercaptoethanol . For alanine scanning point mutants , frozen pellets from 100 mL cultures were thawed and resuspended in 5 mL of 4X lysis buffer . Lysates were sonicated , divided into 5 × 1 mL aliquots and spun at approximately 20 , 000Xg for 20 min at 4°C . Supernatants were combined and incubated with 500 μL of amylose resin ( NEB ) for 30 min at 4°C . The amylose resin was washed 3 times with 1 mL of 4X lysis buffer , 3 times with 1 mL of 1X lysis buffer without Triton X-100 , and 3 times with 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol , 5 mM 2-mercaptoethanol . Protein was eluted with 750 μL of 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol , 5 mM 2-mercaptoethanol , 10 mM maltose for 30 min at 4°C . Amylose resin was pelleted by spinning at 20 , 000Xg for 2 min , and supernatants were collected . To remove contaminating RNases , proteins were loaded onto a 1 or 5 mL HiTrap SP sepharose FF column ( GE ) using an AKTA FPLC . The flowthrough fraction was taken and concentrated using Amicon Ultra 30 kDa centrifugal filters ( EMD Millipore ) . Protein concentrations were quantified by Bradford assay and A280 measurement using a NanoDrop 1000 spectrophotometer . The absence of co-purifying nucleic acids was monitored by ensuring the absorbance ratio at 260 nm/280 nm was less than 0 . 7 . RNA and DNA probes ≤50 bp were ordered from Integrated DNA Technologies . Sense 19mer , 5’-AUAUGGGAACCACUGAUCC-3’; antisense 19mer , 5’-GGGAUCAGUGGUUCCCAUA-3’; sense MBP 1–50 , 5’-ACCAAAAUCGAAGAAGGUAAACUGGUAAUCUGGAUUAACGGCGAUAAAGG-3’; antisense MBP 1–50 , 5’-CCUUUAUCGCCGUUAAUCCAGAUUACCAGUUUACCUUCUUCGAUUUUGGU-3’ . To synthesize 180 bases of α-satellite or β-actin ssRNA , cDNAs were obtained from HeLa cell total RNA using the SuperScript III First-Strand Synthesis System ( Life Technologies ) according to the manufacturer’s instructions , and the following primers: α-satellite forward , 5’-CTTGCTAGCAATCTGCAAGTGG-3’; antisense β-actin , 5’-CGTAGATGGGCACAGTGTGG-3’ . cDNA was amplified into dsDNA using Phusion high-fidelity DNA polymerase ( NEB ) and the following primer pairs; α-satellite forward ( sequence above ) and α-satellite reverse +T7 , 5’- GTAATACGACTCACTATAGGGcttgtcgactacaaaaagagtg-3’; antisense β-actin ( sequence above ) and sense β-actin + T7 , 5’- GTAATACGACTCACTATAGGGAGGCCCCCCTGAACCCCAAG-3’ . dsDNA was used for in vitro transcription as described below . Nucleic acids were end-labeled using ATP γ-32P ( Perkin Elmer ) and polynucleotide kinase ( NEB ) . Approximately 15 pmol of ATP γ-32P was added to 5–10 pmol of nucleic acid , polynucleotide kinase , and 10X reaction buffer , and incubated for 75 min at 37°C . The polynucleotide kinase was subsequently deactivated by incubating at 68°C for 20 min . Free ATP γ-32P was separated from labeled nucleic acids using Micro Bio-spin Columns with Bio-Gel P-6 ( Bio Rad ) according to the manufacturers protocol , and exchanged into 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol in DEPC H2O . Labeled nucleic acids were diluted into approximately 1 . 5 mLs of Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol in DEPC H2O to a final concentration of approximately 1–5 nM . MBP 1–50 nucleic acids were further purified by native gel electrophoresis ( 13 . 33% acrylamide 29:1 , 0 . 5X TBE ) . Bands were identified using HyBlot CL Autoradiography Film ( Denville Scientific ) , excised and eluted for 3 hr at 25°C into 1 mL of 20 mM Tris-HCl , pH 7 . 4 , 50 mM NaCl , 10% glycerol in DEPC H2O at a final concentration of ≤1 nM . α-satellite and β-actin ssRNAs were further purified by denaturing gel electrophoresis ( 10% urea , 10% acrylamide 29:1 , 1X TBE ) . Bands were identified using HyBlot CL Autoradiography Film ( Denville Scientific ) , excised and eluted for 3 hr at 25°C into 1 ml DEPC H2O with the addition of 160 units of RNase Inhibitor ( Ambion , AM2684 ) . RNA was concentrated using Amicon Ultra 30 kDa centrifugal filters and washed 3 times with 500 μL DEPC H2O . RNA was resuspended in 1 mL of 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol , made with DEPC-treated water at a final concentration approximately ≤1 nM . RNA was transcribed from α-satellite and β-actin dsDNA ( from cDNA described in previous section , nucleic acid sequences used in EMSAs ) using the MEGAscript T7 Transcription Kit ( Life Technologies ) according to the manufacturers protocol , with the addition of 40 units of RNase Inhibitor ( Ambion , AM2684 ) . Transcribed RNA was loaded onto a 1% agarose gel to verify lengths , and RNA concentrations were measured using a NanoDrop 1000 spectrophotometer . Approximately 2 μg of RNA was desphosphorylated for 60 min at 37°C using Antarctic Phosphatase ( NEB ) with the addition of 40 units of RNase Inhibitor ( Ambion , AM2684 ) in preparation for end-labeled using ATP γ-32P . Purified proteins and nucleic acids were thawed at room temperature and kept on ice . Proteins were serially diluted into binding buffer ( 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol in DEPC H2O ) and kept on ice until mixed with RNA . Proteins were mixed with 32P-labeled nucleic acids and approximately 4 units RNase Inhibitor ( Ambion , AM2684 ) , and allowed to equilibrate to at room temperature for 10 min before loading onto a native gel ( 7 . 5–13 . 33% acrylamide 29:1 , 0 . 5X TBE ) buffered with 0 . 5X TBE . Gel electrophoresis was carried out for 1 hr at 25°C at 10 mAmps . Gels were exposed to phosphorimaging screens for approximately 1 hr before signal acquisition was performed using a Typhoon 9400 Variable Mode Imager ( GE ) . Gels were quantified using FIJI software , and binding curves , dissociation constants , and standard error were plotted and calculated using Graphpad Prism software . Individual alanine point mutants were synthesized using PCR assembly of six overlapping DNA oligonucleotides ( Integrated DNA Technologies ) of approximately 60 bp , spanning a 5’ AscI site , the SUV39H1 chromodomain ( residues 42–106 ) , and a 3’ PacI site . Following PCR assembly , DNA was purified using Agencourt AMPure XP beads . 40 μL of each PCR product was mixed with 72 μL Ampure beads and incubated for 5 min at 25°C . Beads were collected using a 24-well magnetic stand for 5 min , and washed twice with 200 μL 70% ethanol . Beads were air-dried for 20 min at 25°C , and DNA was eluted with 40 μL of H2O for 5 min at 25°C . Recovered DNA was quantified using a NanoDrop 1000 spectrophotometer . Each point mutant was cloned into the AscI/PacI sites of the pMAL-c2X vector and verified by sequencing . Purified proteins and nucleic acids were thawed at room temperature and kept on ice . Proteins were serially diluted into binding buffer ( 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol in DEPC H2O ) and kept on ice until mixed with RNA . Proteins were mixed with 32P-labeled sense 19mer RNA and approximately 4 units RNase Inhibitor ( Ambion , AM2684 ) , and allowed to equilibrate to at room temperature for 10 min . 4 μL of each binding reaction was dotted onto a nitrocellulose membrane . The membrane was washed several times with binding buffer without glycerol , and membranes were exposed to phosphorimaging screens and imaged using a Typhoon 9400 Variable Mode Imager ( GE ) . Gels were quantified using FIJI software , and binding curves , dissociation constants , and standard error were plotted and calculated using Graphpad Prism . SUV39H1 42–106 proteins were diluted to a final concentration of 12 . 5 μM in 20 mM Tris-HCl pH 7 . 4 , 50 mM NaCl , 10% glycerol . Histone H3 1–20 peptides ( EpiCypher ) were diluted to 40 μM in peptide binding buffer ( 50 mM Tris-HCl pH 7 . 4 , 150 mM NaCl , 0 . 05% NP-40 , 10 mg/mL BSA ) . 5 μL of SUV39H1 42–106 ( 1 . 25 μM final concentration ) was mixed with 1 μL peptide ( 800 nM final concentration ) and 44 μL peptide binding buffer and incubated for 45 min at 4°C . 20 μL of Dynabeads M-280 Strepavidin ( Life Technologies ) were washed 3 times with 1 mL of peptide binding buffer , added to the binding reaction and incubated for 45 min at 4°C . Beads were washed 4 times with 1 mL of peptide binding buffer and eluted with 100 μL of protein SDS sample buffer ( 10% SDS , 50% glycerol , 0 . 05% bromophenol blue , 0 . 2 M Tris-HCl pH 6 . 8 , 40 mM EDTA pH 8 , 20% 2-mercaptoethanol ) . 20 μL was loaded onto a 12 . 5% polyacrylamide SDS denaturing gel and transferred onto a PVDF membrane for western blotting with anti-MBP antibody . Bands were quantified using FIJI software and p-values for an unpaired t-test were calculated using Graphpad Prism software . The structures of the chromodomains of human HP1α bound to H3K9me3 ( Kaustov et al . , 2011 ) ( PDB ID:3FDT ) and human SUV39H1 ( Wang et al . , 2012 ) ( PDB ID:3MTS ) were aligned using the Matchmaker Tool in the Chimera Software Package ( Pettersen et al . , 2004 ) . Residue F43 of SUV39H1 is not shown in the comparison because it is not present in the crystal structure . 150 μg/mL full-length SUV39H1 proteins were incubated with 70 μg/mL Xenopus H3/H4 tetramer ( Guse et al . , 2012 ) , 0 . 8 μCi/mL 14C-SAM ( Perkin Elmer ) , ±124 μg/mL β-actin RNA ( 360 bp , cDNA amplified with ASON 2584/2582 and reverse transcribed ) , in activity buffer ( 10 mM HEPES pH 7 . 7 , 100 mM KCl , 1 mM MgCl2 , 1 mM CaCl2 , 0 . 5 mM DTT ) in a final volume of 20 μL . Reactions proceeded at 30°C for the indicated time , then samples were boiled in SDS-PAGE loading buffer for 5 min and saved at −20°C . Samples were run on a 20% SDS-PAGE gel; the gel was fixed , Coomassie stained , and dried; and methylation was detected with a phosphorimaging screen scanned on a Typhoon 9400 Variable Mode Imager ( GE ) . Rabbit anti-MBP , anti-GFP , and anti-HP1α polyclonal antibodies were generated by Cocalico Biologicals and purified from rabbit serum . All primary antibodies were incubated for 1 hr ( except for anti-SUV39H1 which was incubated overnight ) at the following dilutions: rabbit anti-MBP ( Straight lab , 0 . 2 μg/mL ) , rabbit anti-GFP ( Straight lab , 1 μg/mL ) , rabbit anti-HP1α ( Straight lab , 1 μg/mL ) , mouse anti-SUV39H1 ( Millipore clone MG44 , cat . #05–615 , 1:500 ) , rabbit anti-H3K9me3 ( AbCam ab8898 , 1:1000 ) , rabbit anti-histone H4 ( AbCam ab7311 , 1:2000 ) , and mouse monoclonal anti-tubulin ( clone DM1α , Sigma T6199 , 0 . 5 μg/mL ) . To western blot for endogenous SUV39H1 in DLD-1 cells , nuclear lysates were prepared to enrich for the SUV39H1 containing fraction . Cells were incubated in hypotonic buffer ( 10 mM Tris-HCl pH 8 , 1 . 5 mM MgCl2 , 10 mM KCl ) for 10 min on ice , dounced 10 times with a Kontes glass dounce ( B pestle ) , nuclei were pelleted at 5000xg for 10 min , then lysed in DLB ( Singh et al . , 2014 ) . If necessary , lysates were sonicated with a Branson Sonifier S-250A using a microtip until no longer viscous . To test the effect of RNase A on SUV39H1 protein levels , pelleted nuclei were incubated with 0 . 5 mg/mL RNase A , ±4 U/μL RNaseOUT , for 30 min before lysis . All buffers listed here are described in Singh et al . ( 2014 ) . HeLa Flp-In T-REx cell lines expressing GFP , WT SUV39H1-GFP , or R55A SUV39H1-GFP were expanded to six 15 cm dishes each , and protein expression was induced with 1 μg/mL doxycycline for 6 hr . Cells were trypsinized , pelleted , and then resuspended in 30 mL PBS . To crosslink cells , formaldehyde was added to a final concentration of 0 . 1% , and cells were rotated at room temperature for 10 min . Crosslinking was stopped by adding 3 mL quenching buffer , then rotating cells for 5 min . Cells were pelleted and lysed in 3 mL denaturing lysis buffer ( DLB ) on ice for 10 min . Lysates were sonicated with a Branson Sonifier S-250A using a microtip ( setting 4 . 5 , 2 rounds of 10 pulses ) , diluted up to 8 mL with DLB , then cleared by spinning at 12 , 800 rpm then 33 , 000 rpm in a Type 70 . 1 Ti rotor for 10 min each at 4°C . Lysates were all diluted to 0 . 3 mg/mL , 20 mL each , and samples were saved for western blotting and analysis of total input RNA . Remaining lysate was added to 40 μL NHS-activated sepharose coupled to a llama anti-GFP nanobody ( Rothbauer et al . , 2008 ) and rotated at 4°C for 2 hr . Beads were collected by brief centrifugation , and supernatant sample was saved for western analysis . Beads were washed 2 × 10 mL with denaturing wash buffer ( DWB ) , 2 × 10 mL with isotonic wash buffer ( IsoWB ) , then resuspended in 100 μL of clear sample buffer ( CSB ) . Crosslinks were reversed in both input lysate and IP samples by heating at 75°C for 40 min . 5 μL of each IP supernatant was saved for western blotting and silver staining , and the remaining supernatant and input lysates were mixed with 1 mL TriPure isolation reagent ( Roche ) . RNA was purified using TriPure , then treated with 10 units of TURBO DNase ( Life Technologies ) at 37°C overnight according to the manufacturer’s instructions . RNA was re-purified with TriPure , resuspended in 20 μL DEPC-treated water , then analyzed by RT-qPCR . Silver staining of immunoprecipitated material was performed essentially as described ( Chevallet et al . , 2006 ) . RNA was reversed transcribed into cDNA using the SuperScript III First-Strand Synthesis System ( Life Technologies 18080–051 ) per manufacturer’s instructions , using the following RT primers: α-satellite I ( ASON 614 ) : 5’-CTTGCTAGCAATCTGCAAGTGG-3’ , β-actin rev ( ASON 2430 ) : 5’-ATGTCCACGTCACACTTCAT-3’ , and GAPDH rev: 5’-CCTGCTTCACCACCTTCTT-3’ . Quantitative PCR ( qPCR ) was then performed with the same reverse primers plus the following forward primers: α-satellite II ( ASON 615 ) : 5’-CTTGTCGACTACAAAAAGAGTG-3’ , β-actin fwd ( ASON 2392 ) : 5’-AGAGCTACGAGCTGCCTGAC-3’ , and GAPDH fwd: 5’-AGATCATCAGCAATGCCTCC-3’ . qPCR mix recipe: 1X Phusion HF buffer ( NEB ) , 3% DMSO , 200 µM dNTPs , 1X ROX reference dye ( Life Technologies ) , 1X SYBR green I ( Invitrogen S-7563 ) , Phusion polymerase , and 100 nM each primer . The following programs were used: α-satellite ( 98°C 30 s , then 40X cycles of 98°C 15 s , 60°C 15 s , 72°C 15 s ) , β-actin ( 98°C 30 s , then 40X cycles of 98°C 10 s , 72°C 30 s ) . qPCR was performed on a 7900HT Fast Real-Time PCR System ( Applied Biosystems ) , and the number of cycles to reach the automatically set threshold ( Ct ) were determined using Sequence Detection Systems software . For each primer set , standard curves of diluted cDNA were used to assess the exponential amplification ( expAmp ) of the amplicon . The signal in each experimental sample was determined by signal = expAmp-Ct . For all α-satellite measurements , ‘-RT’ signal was subtracted from ‘+RT’ signal , as the repetitive nature of α-satellite DNA leads to high background . α-satellite signal was then divided by the β-actin or GAPDH signal to get final α-satellite normalized signal . For IP enrichment values , the α-satellite normalized signal for the IP was then divided by the α-satellite normalized signal of the input lysate . To test inhibition of transcription after adding RNA polymerase inhibitors , RT-qPCR was performed essentially as described above with the following primers: pre-tRNA ( tyr ) fwd ( ASON 1991 ) : CCTTCGATAGCTCAGCTGGTAGAG pre-tRNA ( tyr ) rev ( ASON 1990 ) : AAAAAACCGCACTTGTCTCCTTCG GAPDH pre-mRNA fwd ( ASON 4619 ) : CATGCCTTCTTGCCTCTTGT GAPDH pre-mRNA rev ( ASON 4620 ) : TGAGGTCAATGAAGGGGTCA 45S pre-rRNA fwd ( ASON 1891 ) : CCTGCTGTTCTCTCGCGCGTCCGAG 45S pre-rRNA rev ( ASON 1892 ) : AACGCCTGACACGCACGGCACGGAG For total RNA-seq analysis of control and SUV39 DKO HeLa and DLD-1 cells , total RNA was purified using TriPure , then treated with 10 units of TURBO DNase ( Life Technologies ) at 37°C for 30 min per the manufacturer’s instructions . RNA was re-purified with TriPure , resuspended in DEPC-treated water , and quality checked by Bioanalyzer ( Agilent ) . RNA was then analyzed either by RT-qPCR to measure α-satellite RNA levels as described above , or a cDNA library was generated for RNA-Seq analysis . For RNA-Seq , ribosomal RNAs were first depleted using a NEBnext rRNA depletion kit ( New England Biolabs ) , RNA was purified using Agencourt Ampure XP beads ( Beckman Coulter ) , then cDNAs were generated , amplified , and indexed with the ScriptSeq v2 RNA-Seq Library Preparation Kit ( Epicentre ) per the manufacturer’s instructions . Indexed libraries were quantified by Bioanalyzer and qPCR , pooled , and sequenced on a NextSeq 500 ( Illumina ) . For analysis of repetitive RNAs , raw fastq files were adapter trimmed and converted to fasta format . The fasta files were then split into smaller files of 1 , 000 , 000 sequences each . All smaller fasta files of each original fasta file were analyzed by RepeatMasker 4 . 0 . 3 using the human species database ( Bao et al . , 2015 ) . The RepeatMasker output file for each smaller fasta file was merged together to give the repeat content of the original fasta file . Repeat class and type was summarized using the buildSummary . pl script in the RepeatMasker utility script folder . The number of reads for each repeat in the RepeatMasker output file was normalized to the total number of repeats detected by RepeatMasker to obtain a frequency of detection for each repeat type . To calculate the fold enrichment for each repeat type , the frequency of each read in the SUV39 DKO cell line was divided by the frequency of the same read in the control cell line . A cutoff threshold for inclusion in the analysis was set at 300 reads per repeat type . Analysis of non-repetitive RNA was performed by aligning reads ( after PCR duplicate removal and adaptor trimming ) to the UCSC hg38 build ( downloaded Aug 14 , 2015 and archived by Illumina iGenomes ) using tophat2 ( v 2 . 2 . 9 ) . The aligned reads were annotated and quantified using Cufflinks ( v 2 . 2 . 1 ) . Flp-In HeLa cells containing WT or mutant SUV39H1-GFP were grown on glass-bottom chambered coverglass ( LabTek ) under standard conditions . SUV39H1-GFP expression was induced by adding 1 µg/mL doxycycline six hours before data collection . For drug treatment experiments , cells were treated with 1 μM CX-5461 ( Millipore ) , 50 μg/mL α-amanitin ( Santa Cruz Biotech ) , or 1 μM triptolide ( Selleckchem ) for 2 hr before image acquisition . Triptolide and α-amanitin experiments were done in cells induced with 50 ng/mL doxycycline for 24 hr . Immediately before the experiment , the media was exchanged to pH-indicator free media ( DMEM , 10% FBS , constant doxycycline and drug concentration ) . Microscopy was performed with a Nikon Eclipse Ti at 100x magnification . Photobleaching was performed using an Andor Mosaic II patterned illumination system coupled to a 450 mW 405 nm laser . Photobleaching was performed using 100% power for one second using a circular spot with a diameter of ~3 µm . Images of the recovery were collected , and intensity profiles of the bleaching area and of the whole nucleus were calculated using ImageJ 2 . 0 . Intensity profiles were normalized using the following formula: ( 1 ) I=ROI ( t ) ROI ( 0 ) ∙total ( t ) total ( 0 ) ( 2 ) Inorm=I ( t ) −I ( bleach ) I ( 0 ) −I ( bleach ) Where ROI and total are the intensities of the photobleaching area and whole nucleus , respectively , at either time zero or at time t . The intensity at time zero is I ( 0 ) ( i . e . , before bleaching ) , I ( bleach ) is the intensity immediately after bleaching , and I ( t ) is the intensity at subsequent time intervals . The normalized intensity traces were then analyzed for each cell using the following single-exponential function: ( 3 ) Y=Ymax ∙ ( 1−e ( −ln2t1/2 ) ( x−x0 ) ) where Ymax is the plateau of the recovery curve and t1/2 is the halftime for recovery , in seconds . The amplitude of the plateau is equivalent to the fraction of mobile protein .	Each cell in a human body contains the same DNA sequence , which serves as a set of instructions for how the body should develop and operate . However , only certain sections of DNA are “active” at any particular time and in any given type of cell . When a section of DNA is active , cells make many copies of it using a molecule called RNA . When a section of DNA in inactive , very little RNA is made . Some sections of DNA must always be kept inactive to avoid damaging the cell . DNA is packaged around proteins called histones , and enzymes that modify histones control which sections of DNA are switched on or off . One such modifying enzyme , called SUV39H1 , is important for inactivating sections of DNA that could cause harm to the cell if they are active . Previous studies showed that the loss of SUV39H1 and related proteins cause abnormalities and cancer in mice . However , it is not clear how this enzyme identifies and inactivates the DNA it needs to target . Johnson , Yewdell et al . studied SUV39H1 in human cells . The experiments show that RNA binds to the SUV39H1 enzyme and controls how it interacts with DNA . Specifically , Johnson , Yewdell et al . found that sections of DNA that are inactive can still make a small amount of RNA , and that this RNA tethers SUV39H1 to the DNA to keep the DNA switched off . Mutant forms of SUV39H1 that are unable to interact with RNA fall off the DNA , which allows DNA sequences that are normally switched off to become active . The findings of Johnson , Yewdell et al . reveal a new role for RNAs in regulating whether DNA is switched on or off . The next step is to determine whether other enzymes that can also modify histones use the same mechanism to activate or inactivate DNA . Differences in how the activity of DNA is regulated between individuals plays a crucial role in generating the diversity we see in nature . Therefore , this work helps us to understand our basic biology and may provide new opportunities for treating disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "cell", "biology" ]	2017	RNA-dependent stabilization of SUV39H1 at constitutive heterochromatin
Epstein-Barr virus ( EBV ) infection is associated with rheumatoid arthritis ( RA ) in adults , though the nature of the relationship remains unknown . Herein , we have examined the contribution of viral infection to the severity of arthritis in mice . We have provided the first evidence that latent gammaherpesvirus infection enhances clinical arthritis , modeling EBV’s role in RA . Mice latently infected with a murine analog of EBV , gammaherpesvirus 68 ( γHV68 ) , develop more severe collagen-induced arthritis and a Th1-skewed immune profile reminiscent of human disease . We demonstrate that disease enhancement requires viral latency and is not due to active virus stimulation of the immune response . Age-associated B cells ( ABCs ) are associated with several human autoimmune diseases , including arthritis , though their contribution to disease is not well understood . Using ABC knockout mice , we have provided the first evidence that ABCs are mechanistically required for viral enhancement of disease , thereby establishing that ABCs are impacted by latent gammaherpesvirus infection and provoke arthritis . Rheumatoid arthritis ( RA ) is one of the most common autoimmune diseases in adults , though the etiology and pathophysiology are not fully understood ( Humphreys et al . , 2013; Firestein and McInnes , 2017 ) . RA , as well as other autoimmune diseases including multiple sclerosis ( MS ) and systemic lupus erythematosus ( SLE ) , is associated with Epstein-Barr virus ( EBV ) infection ( Balandraud et al . , 2003; Ascherio and Munger , 2007; Gross et al . , 2005 ) . EBV infection typically takes place during childhood or adolescence , while RA generally becomes symptomatic during middle age , indicating that the latent EBV infection likely modulates the immune system over time in a manner that contributes to the development of RA ( Humphreys et al . , 2013; Dowd et al . , 2013; Fourcade et al . , 2017; Alamanos et al . , 2006 ) . The circulating EBV load is higher in individuals with RA than in otherwise healthy adults ( Balandraud et al . , 2003 ) , and RA patients have increased levels of antibodies specific to multiple EBV-encoded proteins ( Blaschke et al . , 2000; Ferrell et al . , 1981; Catalano et al . , 1979; Alspaugh et al . , 1981; Hazelton et al . , 1987 ) . Further , RA patients have increased EBV-specific CD8+ T cells ( Lünemann et al . , 2008 ) , yet these cells have a reduced ability to kill EBV-infected B cells when compared to the same subset of EBV-specific CD8+ T cells from healthy controls ( Takei et al . , 2001 ) . However , the precise role of EBV in RA pathogenesis remains unknown . Evidence from in vivo models is scarce and previous studies have focused primarily on the direct relationship between EBV infection and damage to the joint capsule , with little attention given to systemic effects of EBV infection on immune modulation preceding and continuing throughout disease ( Kuwana et al . , 2011; LeBel et al . , 2018 ) . Mice with humanized immune systems , namely NOD/Shi-scid/IL-2Rγnull mice reconstituted with CD34+ hematopoietic stem cells , that were infected with EBV went on to spontaneously develop erosive arthritis , suggesting a causative role of EBV in arthritis development ( Kuwana et al . , 2011 ) . Related , a serum transfer‐induced arthritis model was used to demonstrate that Ly6Chigh monocytes play a role in transporting murine gammaherpesvirus 68 ( γHV68 ) , an EBV homolog , to the synovium ( LeBel et al . , 2018 ) . Our group has previously shown that latent γHV68 infection exacerbates experimental autoimmune encephalomyelitis ( EAE ) and leads to a disease that more closely resembles MS , with increased demyelination and infiltration of CD8 to cells of the central nervous system in γHV68-infected mice ( Casiraghi et al . , 2012 ) . Critically , this enhancement was specific to γHV68; other viruses , including lymphocytic choriomeningitis virus ( LCMV ) and murine cytomegalovirus ( MCMV ) , did not lead to enhancement of EAE . Additionally , enhancement took place without changes to autoantibody levels . An in vivo model that recapitulates the temporal and systemic immunological aspects of the relationship between EBV and RA is critical . To examine the relationship between EBV and RA , we have adapted in vivo models of both . γHV68 is a natural pathogen that is a well-established and widely-used murine model of EBV infection that shares an array of characteristics with human EBV infection , including latent persistence in B cells , viral reactivation from latency , a potent CD8 T cell response , and immune evasion tactics ( Olivadoti et al . , 2007; Wirtz et al . , 2016 ) . Type II collagen-induced arthritis ( CIA ) is a commonly used model of RA wherein mice are injected with type II collagen emulsified in an adjuvant . Here , we chose to use C57BL/6J mice due to the extensive past characterization of γHV68 infection in C57BL/6 mice and the numerous knockout ( KO ) strains available on this background . Multiple strains of mice are susceptible to CIA , including C57BL/6 mice that , despite displaying a less severe disease course than other strains , generate a robust T cell response ( Inglis et al . , 2007; Campbell et al . , 2000 ) . In C57BL/6 mice , CIA follows a chronic disease course with a sustained T cell response , presence of anti-collagen IgG , and infiltration of inflammatory lymphocytes into the joint capsule ( Inglis et al . , 2007 ) . EBV primary infection generally takes place in childhood or adolescence ( Dowd et al . , 2013; Fourcade et al . , 2017 ) , and RA can occur at any age , though the mean incidence is in the sixth decade of life ( Myasoedova et al . , 2010 ) . Accordingly , we have infected immunologically and sexually mature 6- to 8-week-old C57BL/6J mice with γHV68 and have induced CIA when the mice were adults at 11–13 weeks old . Here we have shown that C57BL/6J mice infected with latent γHV68 and induced for CIA develop a more severe clinical course and an altered immunological profile compared to uninfected CIA controls , with expanded CD8+ T cells and Th1 skewing . We have utilized γHV68 infection and CIA induction to investigate mechanism ( s ) by which EBV contributes to RA , in particular through the modulation of age-associated B cells ( ABCs ) . The role of B cells in the relationship between EBV and RA is intriguing because B cells contribute pathogenically to RA , and EBV infects B cells and alters the B cell profile ( Marston et al . , 2010; Hatton et al . , 2014 ) . ABCs are a subset of B cells that are of particular interest as they have been implicated in both autoimmunity and viral infection . When compared to healthy adults , the relative proportion and/or absolute circulating counts of ABCs are elevated in RA patients , a subset of individuals with MS , individuals with SLE , and a subset of people with common variable immune deficiency that displays autoimmune complications ( Adlowitz et al . , 2015; Rubtsov et al . , 2011; Thorarinsdottir et al . , 2019; Wang et al . , 2019; Claes et al . , 2016; Wang et al . , 2018; Zhang et al . , 2019; Rakhmanov et al . , 2009 ) . ABCs are required for disease development in mouse models of SLE ( Rubtsova et al . , 2017 ) . Also , ABCs are increased during viral infections in mice and/or humans including LCMV , γHV68 , vaccinia , hepatitis C virus , HIV , and influenza ( Rubtsova et al . , 2013; Chang et al . , 2017; Knox et al . , 2017; Johnson et al . , 2020 ) . ABCs display an array of functional capacities , including the secretion of anti-viral or autoantibodies , initiation of germinal centers , antigen presentation to T cells , and secretion of cytokines ( Rubtsova et al . , 2013; Chang et al . , 2017; Knox et al . , 2017; Johnson et al . , 2020 ) . It is yet to be examined whether ABCs play a role in viral contribution to autoimmunity . We found that ABC KO mice are unable to develop the γHV68-exacerbation of CIA and therefore act as a mediator between viral infection and autoimmunity . The development of RA often occurs years after initial infection with EBV when the virus is latent . To mimic this temporal relationship , we infected mice with γHV68 , waited 5 weeks for the lytic infection to clear and the virus to establish latency , and induced CIA . Clearance of the acute virus and establishment of latency have previously been shown by plaque assay on spleens collected 35 days post-infection ( Casiraghi et al . , 2012; Barton et al . , 2014 ) . Following CIA induction , mice were assessed three times per week for redness and swelling in the two hind paws ( Figure 1—figure supplement 1A ) , which informed a clinical score for each mouse . We observed that CIA in latent γHV68-infected mice ( herein referred to as γHV68-CIA mice ) had a more severe clinical course than uninfected mice ( herein referred to as CIA mice ) , as evidenced by consistently higher clinical scores and changes in paw heights throughout the clinical course ( Figure 1A–B , Figure 1—figure supplement 1B ) . γHV68-CIA mice also developed onset of disease symptoms an average of 7 days earlier than CIA mice , reached a higher score at endpoint , and displayed a higher cumulative score ( Figure 1C , Figure 1—figure supplement 1C-D ) . In agreement with other research groups ( Campbell et al . , 2000 ) , male and female mice displayed similar clinical scores during CIA , and we also did not observe a sex difference in γHV68-CIA mice ( Figure 1—figure supplement 1E ) . As expected , latent γHV68-infected mice ( without CIA ) did not display any signs of disease ( Figure 1A–B ) . Titers of anti-type II collagen autoantibodies ( total IgG , IgG1 , and IgG2c ) were elevated in sera from mice with CIA compared to naive mice without CIA , yet were similar in mice with CIA regardless of infection ( Figure 1—figure supplement 1F–H ) . Additionally , we found that inducing CIA in γHV68-infected mice did not impact viral load ( Figure 1—figure supplement 1I ) , indicating that γHV68 is not reactivating . These findings are in line with our previous work showing that latent γHV68 infection enhances EAE without influencing autoantibody levels or reactivating γHV68 ( Casiraghi et al . , 2012 ) . These data demonstrate that latent γHV68 infection leads to earlier onset and more severe CIA , though the exacerbation is not due to higher titers of autoantibodies against type II collagen or changes in abundance of particular immunoglobulin isotypes . To assess the types and relative proportions of immune cells infiltrating the joint synovium , synovial fluid cells were collected on day 56 post-CIA induction . Synovial cells were collected from the knee and ankle joints by flushing each joint with phosphate-buffered saline ( PBS ) and subsequently analyzing isolated cells by flow cytometry . Synovial cells were not collected from naive or γHV68-infected mice without CIA because we would not expect there to be sufficient infiltration of immune cells for analysis . The number of CD8+ T cells infiltrating the synovium during γHV68-CIA was increased compared to CIA ( 3 . 6-fold change ) , while there was no significant difference in the number of CD4+ T cells ( Figure 2A ) . Additionally , the CD8+ and CD4+ T cells in γHV68-CIA synovium displayed a significant increase in Tbet expression compared to those in CIA ( Figure 2B ) , indicating Th1 skewing . As further evidence that infiltrated T cells are immunologically active , we used real-time quantitative PCR ( RT-qPCR ) to evaluate the expression of key T cell-derived cytokines Ifng and Il17 . The relative expression of Ifng in synovium cells of γHV68-CIA mice compared to CIA mice was increased ( 129-fold change ) , while the relative expression of Il17a was trending down in infected mice ( Figure 2C; 3 . 8-fold change ) , though the sample size was low due to the difficulty of obtaining these samples . Together , these results indicate that IFNγ-producing T cells were preferentially infiltrating the synovium in our model of γHV68-CIA , which is consistent with what was observed in the synovium of RA patients ( Yamada et al . , 2008 ) . Our data also demonstrate a skewing toward cytotoxic CD8+ T cells in mice latently infected with γHV68 prior to CIA . To examine how latent γHV68 might contribute to CIA , we specifically examined the systemic T cell profile . It is known that latent γHV68 infection expands cytotoxic T cells and reduces Tregs ( Casiraghi et al . , 2015 ) . Both cell types play a role in CIA with cytotoxic T cells being crucial mediators of CIA while Tregs play a protective role ( Tada et al . , 1996; Morgan et al . , 2003 ) . We examined T cells in the spleen and inguinal lymph nodes ( ILNs ) , a draining lymph node in which we observed a significant increase in overall abundance of immune cells during CIA ( Figure 3—figure supplement 1A ) . γHV68-CIA mice displayed a decrease in relative proportion of FoxP3+ Tregs and an increase in relative proportion of CD8+ T cells in the spleen compared to control CIA mice ( Figure 3A–B , Figure 3—figure supplement 1A , B-C ) . This is similar to what was observed in people with RA , as activated CD8+ T cells were increased and Tregs were decreased in the circulation of RA patients compared to otherwise healthy people ( Morita et al . , 2016; Ramwadhdoebe et al . , 2016 ) . In the ILNs of γHV68-CIA mice , we observed a nonsignificant trend of decreased CD8+ and CD4+ T cell relative proportions , indicating potential T cell egress from the ILNs during disease , and found that the proportion of regulatory T cells was unchanged between CIA and γHV68-CIA mice ( Figure 3A–B , Figure 3—figure supplement 1E–G ) . We also observed a significant increase in relative proportion of CD11c+CD8+dendritic cells ( DCs ) in γHV68-CIA mice compared to CIA ( Figure 3—figure supplement 1H–J ) . These data show that the T cell profile of γHV68-CIA mice is skewed pathogenically , with decreased Tregs and increased cytotoxic T cells . Although IL17 has been highly studied due to its predominance in animal models of arthritis , both IL17 and IFNγ are involved in RA ( Yamada et al . , 2008; Ramwadhdoebe et al . , 2016; Shen et al . , 2009; Nistala et al . , 2010 ) . As expected from our previous work with γHV68-EAE , we found that in γHV68-CIA , greater numbers of splenic CD8+ and CD4+ T cells express IFNγ compared to CIA alone ( Figure 3C–D ) . There is a maintenance of Th17 cells in the spleen , with a similar proportion of CD4+ T cells expressing IL17A in CIA and γHV68-CIA ( Figure 3D ) . In the ILNs , we observed a significant increase in Il17a by RT-qPCR ( Figure 3E ) and a corresponding trend toward more IL17A-expressing CD4+ T cells . We propose that the combined Th1 and Th17 profile observed in γHV68-CIA is more reminiscent of what is observed in people with RA than in CIA without γHV68 infection . To examine the requirement of γHV68 latency , as opposed to residual effects from acute infection , for exacerbating CIA , we used a recombinant γHV68 strain that does not develop latency , ACRTA-γHV68 . In ACRTA-γHV68 , the genes responsible for latency were deleted and a lytic gene , RTA , was constitutively expressed , resulting in clearance of the acute virus by day 14 post-infection ( Rickabaugh et al . , 2004 ) . We infected mice with ACRTA-γHV68 , waited 35 days for clearance of the acute infection , and induced CIA . We found that ACRTA-γHV68-infected mice did not develop the CIA clinical enhancement that we observed in latently γHV68-infected mice , with the clinical course and day of onset resembling that of uninfected CIA mice ( Figure 4A–B ) . Furthermore , the immunological changes observed in γHV68-CIA mice , when compared to CIA mice , were absent in ACRTA-γHV68 CIA mice . The increase in relative proportion of CD8+ T cells in the spleen was less pronounced in ACRTA-γHV68 CIA compared to γHV68-CIA , while there was no change in relative proportion of CD4+ T cells ( Figure 4C–D ) . In ACRTA-γHV68 CIA mice , there was abolishment of the γHV68-induced upregulation of IFNγ in CD8+ and CD4+ T cells , reflecting altered functional capacity and possibly specificity , and no change in IL17A expression by CD4+ T cells ( Figure 4E–G ) . The decrease in relative proportion of splenic Tregs and CD8+ infiltration into the synovial fluid observed in γHV68 CIA mice was not present in ACRTA-γHV68 CIA mice ( Figure 4H–I ) . Together , these data show that ACRTA-γHV68 CIA mice displayed a similar clinical and immunological profile to uninfected CIA mice . This demonstrates that the enhancement is not due to innate immune stimulation during the acute infection , but , rather , the latency phase of γHV68 infection is critical for the clinical and immunological exacerbation of CIA . The requirement of γHV68 latency mirrors the RA patient clinical course , wherein patients are infected with EBV years before the onset of disease . As the number of ABCs was expanded in the contexts of both viral infection and autoimmunity , including RA ( Rubtsov et al . , 2011; Claes et al . , 2016; Wang et al . , 2018; Rubtsova et al . , 2013; Knox et al . , 2017 ) , we investigated the role of ABCs in facilitating γHV68-exacerbation of CIA . We began by examining the proportion and phenotype of ABCs in uninfected CIA mice and CIA mice previously infected with latent γHV68 ( γHV68-CIA ) by flow cytometry ( Figure 5—figure supplement 1A ) . We found that CIA induction increased the proportion and total number of ABCs ( CD19+CD11c+Tbet+ ) in the spleen , and γHV68-CIA mice had further increased proportions of ABCs in the spleen compared to CIA ( Figure 5A–B ) . The proportion of ABCs in the ILNs was not significantly different between γHV68-CIA and CIA mice ( Figure 5—figure supplement 1C ) . The number of ABCs was substantially lower in the ILNs than in the spleen , concurring with other studies that found that ABCs primarily reside in the spleen ( Johnson et al . , 2020 ) . During CIA and γHV68-CIA , we did not observe differences in the proportions of ABCs between male and female mice ( Figure 5—figure supplement 1B ) . We next examined the phenotypic characteristics and found that ABCs in the spleen were phenotypically distinct in γHV68-CIA compared to CIA . We examined a series of markers previously shown to be expressed by ABCs , including cytokines IL10 , IFNγ , and TNFα ( Rubtsov et al . , 2011; Hao et al . , 2011; Russell Knode et al . , 2017; Ratliff et al . , 2013 ) , an array of inhibitory receptors ( Rubtsov et al . , 2011; Wang et al . , 2018; Knox et al . , 2017 ) , maturity and memory markers IgD , IgM , and CD27 ( Rubtsov et al . , 2011; Hao et al . , 2011 ) , and MHCII ( Rubtsov et al . , 2011; Knox et al . , 2017; Aranburu et al . , 2018 ) . We found that fewer ABCs in the spleens of γHV68-CIA mice expressed IL10 , while an increased proportion expressed IFNγ ( Figure 5C ) , indicating that they are skewed toward a pathogenic Th1 phenotype . Further , fewer splenic ABCs in γHV68-CIA mice expressed inhibitory receptors CTLA4 , PDL1 , and PD1 ( Figure 5D–F ) , and thus ABCs in CIA displayed a more regulatory phenotype than those in γHV68-CIA mice . Additionally , the ABCs in γHV68-CIA mice displayed a more mature phenotype , with fewer IgD+IgM+ naive B cells and increased MHCII expression , though the expression of memory marker CD27 was unchanged ( Figure 5G–I ) . These results indicate that ABCs in γHV68-CIA mice are more mature and may have increased antigen presentation capacities but are not primarily a memory subset . There were no differences in the expression of CD20 , TNFα , CD95 ( Fas ) , nor IDO expression ( Figure 5—figure supplement 1D–G ) . Collectively , these results indicate that ABCs in γHV68-CIA mice display a more pathogenic phenotype than those in CIA , with decreased expression of regulatory cytokine IL10 and inhibitory markers , and increased expression of IFNγ . To determine whether ABCs are a subset mediating the viral enhancement of CIA , we utilized ABC KO mice that harbor a B cell-specific Tbet deletion . The clinical course and immune profile of CIA and γHV68-CIA mice were compared in littermate controls of Tbx21fl/flCd19cre/+ ( KO ) and Tbx21fl/flCd19+/+ ( Ctrl ) mice ( Figure 6A ) . We observed that the clinical course was unchanged in CIA between Ctrl and KO mice , indicating that ABCs are not contributing to the disease course in CIA ( Figure 6B ) . Alternatively , when induced with CIA , γHV68-infected KO mice did not display the γHV68-exacerbated clinical course compared to γHV68-CIA Ctrl mice ( Figure 6C ) , indicating that ABCs are a pathogenic subset in γHV68-CIA . Without ABCs , γHV68-CIA mice did not display clinical exacerbation , but rather appeared similar to uninfected CIA mice in terms of disease severity and day of onset ( Figure 6B–D ) . We observed that the ablation of ABCs does not significantly alter the proportion of CD8 , CD4 , or Treg populations in the spleen during CIA or γHV68-CIA , nor the expression of IFNγ or IL17A ( Figure 6—figure supplement 1 ) . These results indicate that ABCs are a critical pathogenic population in γHV68-CIA though more work is needed to fully elucidate the mechanism by which ABCs are contributing to disease . In this report , we demonstrate that latent γHV68 exacerbates CIA clinically and immunologically , and Tbet+ B cells , known as ABCs , are critical for this exacerbation . Investigation of the mechanism by which EBV contributes to RA has previously been challenging due to the lack of a murine model to examine the systemic immune modulation caused by latent gammaherpesvirus infection and resulting influence on arthritis . Here , we show that infecting mice with latent γHV68 prior to CIA induction results in an immune course more similar to that of RA patients than CIA alone and is a suitable model for examining the contribution of EBV to RA . Elucidating how EBV infection contributes to the development of RA is critical to understanding the underlying pathophysiology of the disease . As EBV is associated with several autoimmune diseases , it is important to examine whether there are conserved mechanisms of contribution . The overlap in etiology and pathophysiology between these autoimmune diseases may help to explain the cross-efficacy of immunotherapies between MS and RA , including B cell-depletion therapies . Our lab has previously demonstrated that latent γHV68 infection enhances EAE , a common model of MS , clinically and immunologically ( Casiraghi et al . , 2012 ) . In both the γHV68-CIA and γHV68-EAE models , we observed an increase in CD8+ T cells at the site of disease and increased expression of IFNγ by cytotoxic and helper T cells . Latent gammaherpesvirus infection of mice clearly alters autoimmune disease onset and severity reminiscent of the strong association of latent EBV infection in RA patients . As such , these investigative models will serve to identify common mechanisms in which EBV contributes to multiple autoimmune diseases . Due to EBV infection often taking place years before the onset of arthritis , we posit that latent EBV infection modulates the peripheral immune response in a manner that contributes to the development of RA . We suggest that latently EBV-infected B cells alter , either directly or indirectly , lymphocytes that go on to contribute to disease onset , likely through expanding and activating CD8+ T cells and skewing toward a Th1 response . CD11c+CD8+ DCs may play a role in priming the pathogenic CD8+ T cell response , as they have been shown to cross-present antigen ( den Haan et al . , 2000; Schulz and Reis e Sousa , 2002 ) . By acting as a mediator between infected cells and pathogenic T cells , ABCs are likely critical moderators in driving the heightened Th1 immune response to latent viral infection . Accumulating evidence shows that ABCs are expanded in multiple autoimmune diseases and function pathogenically in mouse models of lupus ( Rubtsov et al . , 2011; Thorarinsdottir et al . , 2019; Wang et al . , 2019; Claes et al . , 2016; Wang et al . , 2018; Zhang et al . , 2019; Rakhmanov et al . , 2009; Rubtsova et al . , 2017 ) . Precisely how ABCs are contributing to pathogenicity is unclear , and ABCs are known to display multiple functional capacities that could contribute to disease . In models of SLE , ABCs have been shown to secrete autoantibodies and compromise germinal center responses ( Zhang et al . , 2019 ) . Additionally , ABCs function as excellent antigen-presenting cells ( Rubtsov et al . , 2015 ) . In a model of SLE , the ablation of ABCs decreases activated CD4+ T cells and IFNγ-CD8+ T cells ( Rubtsova et al . , 2017 ) . How precisely ABCs alter the CD8+ T cell population , whether they are cross-presenting antigen or impacting the CD8+ T cells indirectly , warrants further investigation . Alternatively , ABCs have been shown to secrete regulatory IL10 ( Rubtsov et al . , 2011; Hao et al . , 2011 ) , suggesting that a portion of the ABC population , in some individuals or contexts , could function in a protective manner . Further characterization of the phenotype and functional capacities of ABCs in autoimmune patients may help to elucidate their functional role . RA patients who experience a relapse following B cell-depletion therapy are more likely to display a reconstitution profile with increased numbers of memory B cells ( Leandro et al . , 2006 ) . Whether existing therapeutics , such as B cell-depletion therapies or other approved drugs for RA , such as Abatacept ( CTLA4 Ig ) , impact the ABC repertoire remains unknown . Further evaluation of the influence of viral infection on ABC pathogenicity is needed . It is intriguing that ABCs are pathogenic in a genetic model of SLE without the presence of a virus ( Rubtsova et al . , 2017 ) , though we observe that latent γHV68 is necessary for the pathogenicity of ABCs in CIA . This discrepancy indicates that ABCs may be contributing to disease through various mechanisms or that different contexts can prime ABCs for pathogenicity . The role of ABCs in controlling viral infections is an ongoing topic of study , with multiple papers recently providing compelling evidence that ABCs are critical for an effective anti-influenza response ( Johnson et al . , 2020 ) and are required to control LCMV infection ( Barnett et al . , 2016 ) , in part through their secretion of antiviral IgG2a . Additionally , the influence of aging on ABC population and on autoimmunity development and progression warrants further study . In summary , we have developed an in vivo model of EBV’s contribution to RA that recapitulates aspects of human disease . Further , we have examined the role of ABCs and found that they are critical mediators of the viral enhancement of arthritis . Tbx21fl/flCd19cre/+ mice were generated by crossing Tbx21fl/flCd19cre/+ and Tbx21fl/flCd19+/+ mice . Tbx21fl/fl and Cd19cre/+ mice were provided by Dr . Pippa Marrack ( Rubtsova et al . , 2017 ) . C57BL/6J mice were originally purchased from The Jackson Laboratory . All animals were bred and maintained in the animal facility at the University of British Columbia . All animal work was performed per regulations of the Canadian Council for Animal Care ( Protocols A17- 0105 , A17-0184 ) . γHV68 ( WUMS strain , purchased from ATCC ) and ACRTA-γHV68 ( originally developed by Dr . Ting-Ting Wu , the generous gift of Dr . Marcia A Blackman ) ( Jia et al . , 2010 ) were propagated in Baby Hamster Kidney ( BHK , ATCC ) cells . Prior to infection , viruses were diluted in Minimum Essential Media ( MEM , Gibco ) and maintained on ice . Mice ( 6- to 8-week-old ) were infected intraperitoneally ( i . p . ) with 104 plaque-forming unit ( PFU ) of γHV68 or ACRTA-γHV68 or mock-infected with MEM . No clinical symptoms were observed from viral infections . On day 35 post-γHV68 or -ACRTA-γHV68 infection , CIA was induced by injection of immunization-grade , chick type II collagen emulsified in complete Freund’s adjuvant ( CFA; Chondrex , Inc ) intradermally at the base of the tail , followed by a booster injection of the same emulsion on day 14 , as adapted from Inglis et al . , 2008 . Each mouse received 0 . 1 mg chick type II collagen and 0 . 25 mg CFA at days 0 and 14 . Clinical signs of CIA were assessed and scored three times per week beginning at the day of CIA induction: 0 = no symptoms; 1 = slight swelling and/or erythema; 2 = pronounced swelling and erythema; and 3 = severe swelling , erythema , and ankylosis , as adapted from Brand et al . , 2007 . Hind paws were scored individually by a blinded scorer and added for a single score . Day of onset considered two consecutive scoring days of a score of at least 1 . The thickness of each hind paw was measured using a digital caliper and the size was expressed as the average thickness of the two paws . Mice were anesthetized with isoflurane and euthanized by cardiac puncture . Blood was collected by cardiac puncture into empty sterile tubes and placed on ice until processing , and mice were perfused with 20 ml sterile PBS to allow for synovial fluid harvesting without blood contamination . ILNs and spleen were extracted and placed into 2 ml sterile PBS and stored temporarily on ice until processing . Synovial fluid was collected by exposing the knee and ankle joints , removing the patellar ligament , and flushing each flexed ankle and knee joint with sterile RNase/DNase-free PBS ( Invitrogen ) using an 18-gauge needle , adapted from Barton et al . , 2007; Futami et al . , 2012 . Using a 70-μm cell strainer and a 3-ml syringe insert , spleens and ILNs were each mashed through the cell strainer mesh and a single-cell suspension was prepared for each sample . Splenocytes were incubated in ACK lysing buffer for 10 min on ice to lyse red blood cells , and remaining cells were kept on ice until further use . To evaluate cytokine production by various cell types , 4 million isolated splenocytes or ILNs were stimulated ex vivo for 3 hr in 5% CO2 at 37°C in Minimum Essential Media ( Gibco ) containing 10% fetal bovine serum ( FBS; Sigma-Aldrich ) , 1 μl/ml GolgiPlug ( BD Biosciences ) , 10 ng/ml phorbol 12-myristate 13-acetate ( PMA , Sigma-Aldrich ) , and 500 ng/ml ionomycin ( Thermo Fisher ) . Stimulated cells were then washed prior to staining . For each spleen and ILN sample , 4 million cells were stained in two wells , with 2 million cells per well . All collected synovial fluid cells were resuspended in flow cytometry staining buffer ( FACS , PBS with 2% newborn calf serum , Sigma-Aldrich ) and stained in a single well . Before staining , samples were incubated at 4°C covered from light for 30 min with 2 ul/ml Fixable Viability Dye eFluor506 ( Thermo Fisher ) while in FACS buffer ( PBS with 2% newborn calf serum; Sigma-Aldrich ) . Cells were then incubated with a rat anti-mouse CD16/32 ( Fc block ) ( BD Biosciences ) antibody for 10 min . Fluorochrome-labeled antibodies against cell surface antigens were then added to the cells for 30 min covered from light at 4°C . After washing , cells were suspended in Fix/Perm buffer ( Thermo Fisher ) for 30 min to 12 hours covered from light at 4°C , washed twice with Perm buffer , and incubated 40 min with antibodies for intracellular antigens in Perm buffer . Cells were then washed and resuspended in FACS buffer with 2 mM ethylenediaminetetraacetic acid . Cells were stained with anti-mouse CD45 ( Clone 30-F11; Thermo Fisher Scientific ) , CD3 ( Clone eBio500A2; Thermo Fisher Scientific ) , CD19 ( Clone eBio1D3; Thermo Fisher Scientific ) , CD4 ( Clone RM4-5; Thermo Fisher Scientific ) , CD8 ( Clone 53–6 . 7; Thermo Fisher Scientific ) , FoxP3 ( Clone FJK-16S; Thermo Fisher Scientific ) , IFNγ ( Clone XMG1 . 2; Thermo Fisher Scientific ) , IL17A ( Clone TC11-18H10 . 1; Thermo Fisher Scientific ) , IL10 ( Clone JES5-16E3; Thermo Fisher Scientific ) , CD11c ( Clone 418; Thermo Fisher Scientific ) , Tbet ( Clone eBio4B10; Thermo Fisher Scientific ) , CD11b ( Clone M1/70; Thermo Fisher Scientific ) , IgD ( Clone 11–26 c; Thermo Fisher Scientific ) , CTLA4 ( Clone UC10-4B9; Thermo Fisher Scientific ) , PDL1 ( Clone MIH5; Thermo Fisher Scientific ) , PD1 ( Clone J43; Thermo Fisher Scientific ) , IgM ( Clone RMM-1; BioLegend ) , MHCII ( Clone M5/114 . 15 . 2; BioLegend ) , CD27 ( Clone LG . 3A10; BioLegend ) , CD20 ( Clone SA275A11; BioLegend ) , TNFα ( Clone MP6-XT22; BioLegend ) , CD95 ( Clone SA367H8; BioLegend ) , and IDO ( Clone mIDO-48; Thermo Fisher Scientific ) . The entirety of each sample was collected on an Attune NxT Flow Cytometer ( Thermo Fisher ) and analyzed with FlowJo software v10 ( FlowJo LLC ) . Full-minus-one ( FMO ) controls were used for gating . RNA was extracted from synovial fluid and ILNs with a Qiagen AllPrep DNA/RNA Micro kit . cDNA was synthesized using Applied Biosystems High-Capacity cDNA Reverse Transcription Kit ( Thermo Fisher ) . qPCR was performed using iQTM SYBR Green supermix ( Bio-Rad ) on the Bio-Rad CFX96 Touch Real Time PCR Detection system . Primer sets from Integrated DNA Technologies were Il17a 5’-GCT CCA GAA GGC CCT CAG-3’ ( forward ) and 5’-AGC TTT CCC TCC GCA TTG-3’ ( reverse ) and Ifng 5’-ACT GGC AAA AGG ATG GTG AC-3’ ( forward ) and 5’-TGA GCT CAT TGA ATG CTT GG-3’ ( reverse ) . Normalized to the ribosomal housekeeping gene 18 s 5’-GTAACCCGTTGAACCCCATT-3’ ( forward ) and 5’- CCATCCAATCGGTAGTAGCG-3’ ( reverse ) and expression determined relative to control group . The sera were isolated by centrifugation 2000 × g for 10 min , aliquoted , and stored for up to 14 months at −80°C prior to running the enzyme-linked immunosorbent assay ( ELISA ) . Anti-type II collagen antibodies were quantified by standard indirect ELISA . Briefly , ELISA plates ( NUNC; Thermo Fisher ) were coated with 5 μg/ml ELISA-grade type II collagen ( Chondrex , Inc ) overnight at 4°C , washed 4x with wash buffer ( PBS , 0 . 05% Tween-20 ) , blocked with 5% newborn calf serum ( NBCS; Sigma-Aldrich ) for 1 hr at 37°C , incubated with serial dilutions ( 1:100 to 1:12800 ) of test sera diluted in blocking buffer for 2 hr at 37°C , and washed 4x with wash buffer . Bound ( anti-collagen II ) antibody was incubated with HRP-conjugated goat anti-mouse IgG ( Thermo Fisher ) , rat anti-mouse IgG1 ( BD Biosciences ) , or goat anti-mouse IgG2c ( Thermo Fisher ) , all diluted 1:500 in blocking buffer , for 1 hr at 37°C , washed 4x with wash buffer , and detected by TMB substrate ( BD Biosciences ) . Absorbance was read at 450 nm on a VarioSkan Plate Reader ( Thermo Fisher ) . Quantification of γHV68 load was done as previously described ( Márquez et al . , 2020 ) . Genomic DNA ( gDNA ) was extracted from 4 × 106 splenocytes with PureLink Genomic DNA mini kit ( Thermo Fisher ) , according to the manufacturer’s instructions , and stored at −20°C . For qPCR , 150 ng DNA per reaction was amplified in duplicate using primers and probes specific to γHV68 ORF50 and mouse PTGER2 with QuantiNova Probe Mastermix ( Qiagen ) . Primers and probes used from Integrated DNA Technologies were PTGER2: forward primer: 5′-TACCTTCAGCTGTACGCCAC-3′; reverse primer: 5′-GCCAGGAGAATGAGGTGGTC-3′; probe: 5′-/56-FAM/CCTGCTGCT/ZEN/TATCGTGGCTG/3IABkFQ/-3′; ORF50: forward primer: 5′-TGGACTTTGACAGCCCAGTA-3′; reverse primer: 5′-TCCCTTGAGGCAAATGATTC-3′; probe: 5′-/56-FAM/TGACAGTGC/ZEN/CTATGGCCAAGTCTTG/3IABkFQ/-3′ . Standard curves were obtained by serial dilutions of ORF50 and PTGER2 gBlocks ( ORF50: 2 × 106 – 2 × 101; PTGER2: 5 × 107–5×102 ) . Reactions were run on the Bio-Rad CFX96 Touch Real Time PCR Detection system . Data and statistical analyses were performed using GraphPad Prism software 8 . 4 . 2 ( GraphPad Software Inc ) . Results are presented as mean ± SEM . Statistical tests , significance ( p-value ) , sample size ( n , number of mice per group ) , and number of experimental replicates are stated in figure legends . Statistical analyses included two-way ANOVA with Geisser-Greenhouse's correction , Mann-Whitney test , or one-way ANOVA . P-values are indicated by asterisks as follows: **p<0 . 0001 , p<0 . 001 , p<0 . 01 , p<0 . 05 . All work was approved by the Animal Care Committee ( ACC ) of the University of British Columbia ( Protocols A17- 0105 , A17-0184 ) .	Rheumatoid arthritis is one of the most common autoimmune diseases , leaving patients in pain as their immune system mistakenly attacks the lining of their joints . The precise cause is unknown , but research suggests a link to the Epstein-Barr virus , the agent responsible for mononucleosis ( also known as glandular fever ) . After infection and recovery , the virus remains in the body , lying dormant inside immune ‘B cells’ which are often responsible for autoimmune diseases . Of particular interest are a sub-group known as ‘age-associated B-cells’ , which are mostly cells left over from fighting past infections such as mononucleosis . Yet , the link between Epstein-Barr virus and rheumatoid arthritis remains hard to investigate because of the long gap between the two diseases: the virus mostly affects children and young people , while rheumatoid arthritis tends to develop in middle age . To investigate how exactly the two conditions are connected , Mouat et al . created a new animal model: they infected young mice with the murine equivalent of the Epstein-Barr virus , and then used a collagen injection to trigger rheumatoid arthritis-like disease once the animals were older . Next , Mouat et al . monitored the paws of the mice , revealing that viral infection early in life worsened arthritis later on . These animals also had more age-associated B cells than normal , and the cells showed signs of participating in inflammation . On the other hand , early viral infection did not make arthritis worse in mice unable to produce age-associated B cells . Taken together , these results suggest that the immune cells are required to enhance the effect of the viral infection on rheumatoid arthritis . This new insight may help to refine current treatments that work by reducing the overall number of B cells . Ultimately , the animal model developed by Mouat et al . could be useful to identify better ways to diagnose , monitor and treat this debilitating disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2021	Latent gammaherpesvirus exacerbates arthritis through modification of age-associated B cells
The combined-immunotherapy of adoptive cell therapy ( ACT ) and cyclophosphamide ( CTX ) is one of the most efficient treatments for melanoma patients . However , no synergistic effects of CTX and ACT on the spatio-temporal dynamics of immunocytes in vivo have been described . Here , we visualized key cell events in immunotherapy-elicited immunoreactions in a multicolor-coded tumor microenvironment , and then established an optimal strategy of metronomic combined-immunotherapy to enhance anti-tumor efficacy . Intravital imaging data indicated that regulatory T cells formed an 'immunosuppressive ring' around a solid tumor . The CTX-ACT combined-treatment elicited synergistic immunoreactions in tumor areas , which included relieving the immune suppression , triggering the transient activation of endogenous tumor-infiltrating immunocytes , increasing the accumulation of adoptive cytotoxic T lymphocytes , and accelerating the infiltration of dendritic cells . These insights into the spatio-temporal dynamics of immunocytes are beneficial for optimizing immunotherapy and provide new approaches for elucidating the mechanisms underlying the involvement of immunocytes in cancer immunotherapy . Cancer immunotherapy has been an area of significant process in recent decades , and is considered a breakthrough in cancer therapy ( Couzin-Frankel , 2013 ) . Current promising techniques in cancer immunotherapy are the use of monoclonal antibodies ( mAbs ) , tumor vaccines , and adoptive cell therapy ( ACT ) ( Elert , 2013; Couzin-Frankel and McNutt , 2013 ) . ACT has been reported to be one of the most efficient treatments for patients with melanoma because it effectively elicits anti-tumor immune responses and leads to objective tumor regression in more than 50% of patients ( Pittet et al . , 2007; Rosenberg et al . , 2008 ) . During the immunotherapy-elicited immune response , various immune cells , including endogenous and adoptive immunocytes , are activated in the tumor microenvironment ( Schreiber et al . , 2011; Gajewski et al . , 2013 ) . However , the underlying process remains a ‘black box’ because , whereas information is available to describe the input ( immunotherapy ) and the output ( tumor elimination ) , relatively limited information is available to describe the spatio-temporal changes in the participating immune cells in the tumor microenvironment ( Krummel , 2010 ) . Traditional biological methods , such as immunohistochemistry methods or flow cytometry , have revealed the type and function of immune cells that play a role at certain time-points during the anti-tumor reaction ( Perentes et al . , 2009; Boissonnas et al . , 2013 ) . Visualization of the dynamic characteristics of immunocytes , including their movement , migration , and recruitment , as well as of the interaction between immunocytes and tumor cells in the systemic environment , is critical for understanding the success or failure of cancer immunotherapy on a mechanistic level ( Krummel , 2010; Bourzac , 2013 ) . The common strategy used in clinical trials of ACT is to isolate tumor antigen-specific lymphocytes from patients , robustly expand these cells in vitro , and infuse them into cancer patients ( Yee et al . , 2002; Mackensen et al . , 2006; Benlalam et al . , 2007 ) . The activity and cytotoxicity of tumor-specific cytotoxic T lymphocytes ( CTLs ) have long been considered the crucial factors for the efficacy of ACT against solid tumors ( Boon , 1994 ) . However , recent reports have indicated that despite the detection of tumor-reactive CTLs in the bloodstream after the adoptive transfer , most CTLs lost their anti-tumor functions as a result of endogenous immunosuppressive networks in the tumor microenvironment ( Zippelius et al . , 2004; Baitsch et al . , 2011 ) . Foxp3+ regulatory T cells ( Tregs ) play a key immunosuppressive role in the endogenous immunosuppressive networks ( Curiel et al . , 2004; Vignali et al . , 2008 ) . Various immunosuppressive mechanisms associated with Tregs have been revealed , including their specific recruitment and accumulation in the tumor stroma ( Wang et al . , 2004; Nishikawa and Sakaguchi , 2014 ) , suppression of the CTL function by TGF-β ( Chen et al . , 2005 ) , and induction of the CTL dysfunction by interactions of Tregs with intratumoral dendritic cells ( DCs ) ( Bauer et al . , 2014 ) . However , the details of the spatio-temporal orchestration of Tregs associated with tumor progression , the contributions of the Treg distribution in tumor stroma to immunosuppressive networks , and the effect of Tregs on the infiltration of adoptive CTLs into solid tumors remain unclear . Treg elimination is considered to be an essential component for disrupting the immunosuppressive networks in the tumor microenvironment before ACT ( Rosenberg et al . , 2008; Antony et al . , 2005 ) . Cyclophosphamide ( CTX ) is a commonly used alkylating agent in chemotherapy , which has been applied in recent years to selectively suppress , abrogate and deplete Tregs before ACT treatment ( Rosenberg et al . , 2008; Nishikawa and Sakaguchi , 2014; Oleinika et al . , 2013; Zhao et al . , 2010 ) . Lymphodepletion before ACT by CTX treatment further promoted tumor-specific CTL infiltration into the target tumors ( Bracci et al . , 2007; Rosenberg et al . , 2008; Boissonnas et al . , 2013 ) . The migratory behavior of effector T cells in solid tumors has been analyzed by two-photon microscopy , proving that migration presents the critical step for the sufficient infiltration of T cells into the solid tumor and promotes tumor elimination ( Mrass et al . , 2006; Boissonnas et al . , 2007 ) . However , the influence of CTX treatment on the long-term precise migratory behavior of adoptive CTLs in solid tumors has not been directly observed in vivo . In addition to the selective depletion and suppression of Tregs , other immunomodulatory mechanisms of CTX treatment include the systemic activation of DCs ( Radojcic et al . , 2010; Veltman et al . , 2010; Nakahara et al . , 2010 ) , increase of the infiltration of DCs into the tumor area ( Schiavoni et al . , 2011 ) , and promotion of the penetration of CTLs into the tumor parenchyma ( Boissonnas et al . , 2013; Bracci et al . , 2007 ) . Although certain immunomodulatory functions of CTX alone or in combination with immunotherapy have been proposed previously , the synergistically enhanced effects of the combined CTX and ACT ( CTX-ACT ) treatment on the adoptive and endogenous anti-tumor immunoreactions have not been revealed . In particular , the dynamic effects of the CTX-ACT treatment on endogenous immune cells ( such as CTLs , neutrophils and DCs ) in the tumor microenvironment have not been well described . Using intravital microscopy and a skin-fold window chamber model , we captured a long-term comprehensive sequence of cellular events associated with multicolor-coded tumor cells , adoptive CTLs , endogenous CTLs , Tregs , DCs and tumor-infiltrating immunocytes in the tumor microenvironment during CTX-ACT combined-immunotherapy for B16 melanoma . The long-term intravital imaging data provided direct evidence for the synergetic anti-tumor immune-enhanced effects of the CTX-ACT combination treatment , including the elimination of Tregs , activation of the endogenous tumor-infiltrating immunocytes , and infiltration of adoptive and endogenous CTLs and DCs . To assess the curative efficacy of ACT on murine melanoma in vivo , CTLs were obtained from the splenocytes of C57BL/6 mice immunized with mitomycin C pre-treated B16 melanoma cells according to previous protocols ( Restifo and Nicholas , 2011 ) and reports ( Liu et al . , 2006; Bauer et al . , 2014 ) . Here , the activity and cytotoxicity of CTLs against target tumor cells were demonstrated in vitro by flow cytometry of B16 tumor cells that stably express mutants of the cyan fluorescent protein mCerulean ( CFP-B16 tumor cells , Figure 1—figure supplement 1 ) , which is minimally immunogenic in C57BL/6 mice ( Skelton et al . , 2001; Yang et al . , 2016 ) . However , the ACT treatment failed to control the tumor growth ( Figure 1A ) , which is consistent with the results reported for certain clinical cases ( Zippelius et al . , 2004; Mukai et al . , 1999 ) . 10 . 7554/eLife . 14756 . 003Figure 1 . ‘Immunosuppressive ring’ formed by Tregs in CFP-B16 tumor-bearing mice , which inhibited the anti-tumor efficacy of the adoptive CTLs . ( A ) Tumor growth curves for CFP-B16 tumors of mice treated with ACT or PBS control . The data are represented as the mean ± SEM tumor volume ( n = 12–14 , three independent experiments ) . ns: not significant , ( Figure 1—source data 1 ) . ( B ) Density of carboxyfluorescein succinimidyl ester ( CFSE ) -labeled CTLs within different organs ( tumor-draining lymph nodes ( TDLNs ) , non-tumor-draining lymph nodes ( NDLNs ) , spleens , and tumors ) . The density was determined by counting the number of CFSE-labeled CTLs per mm2 on frozen tissue sections . The data are represented as the mean ± SEM ( n = 18–22 fields , 0 . 18 mm2 per field ) from three independent experiments . p<0 . 05 , p<0 . 01 , p<0 . 001 , ( Figure 1—source data 2 ) . ( C ) Large-field intravital images of an ‘immunosuppressive ring’ around the CFP-B16 tumor . Blue – CFP-B16 tumor; red – Tregs ( Foxp3-mRFP cells ) ; green – CFSE-labeled CTLs . The left panel shows different single color channels of the tumor microenvironment , and the right panel shows the three color channels merged . Scale bar: 500 µm . ( D ) Schematic diagram of the ‘immunosuppressive ring’ in the tumor microenvironment . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 00310 . 7554/eLife . 14756 . 004Figure 1—source data 1 . Tumor growth curves for CFP-B16 tumors of mice treated with ACT or PBS . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 00410 . 7554/eLife . 14756 . 005Figure 1—source data 2 . Density of CFSE-labeled CTLs within different organs . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 00510 . 7554/eLife . 14756 . 006Figure 1—figure supplement 1 . Establishment and characterization of CFP-B16 tumor-specific CTLs . ( A ) CFSE- and propidium iodine ( PI ) -based assays to assess the cytotoxicity of CTLs added to the CFP-B16 tumor cells or added to the splenocytes by flow cytometry , and the control group which was composed of splenocytes added to the CFP-B16 tumor cells . The data are represented as the mean ± SD from three independent experiments . ( B ) Analysis and characterization of the in vitro-activated CTLs . The data are representative of similar results from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 00610 . 7554/eLife . 14756 . 007Figure 1—figure supplement 2 . Fluorescence microscopy images showing the distribution of CFSE-labeled CTLs in different organs . Representative sections of tumor-draining lymph node ( TDLN ) , non-tumor-draining lymph node ( NDLN ) , spleen and tumor tissues were collected from tumor-bearing mice on Days 1–3 ( early stage ) and Days 4–6 ( late stage ) after the adoptive transfer of CTLs . The top row represents the early stage , and the bottom row represents the late stage . CFSE-labeled adoptively transferred CTLs are shown in green . Scale bars: 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 00710 . 7554/eLife . 14756 . 008Figure 1—figure supplement 3 . Long-term and large-field imaging of the process by which Tregs formed an immunosuppressive ring . CFP-B16 tumor is shown in blue ( CFP ) , Tregs are shown in red ( mRFP ) , and adoptively transferred CTLs are shown in green ( CFSE , rarely observed ) . Top row: large-field images; scale bar: 500 µm . Bottom row: images of the region of interest; scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 008 To determine the reasons for the failure of adoptive CTLs , we detected the distribution of these cells in different organs of CFP-B16 tumor-bearing mice at an early stage ( 1–3 days after the adoptive transfer of CTLs ) and a later stage ( 4–6 days after the adoptive transfer of CTLs ) . Adoptive CTLs stained with the fluorescent dye CFSE ( 5- ( and-6 ) -carboxyfluorescein diacetate succinimidyl ester , CFDA SE ) were intravenously transferred into tumor-bearing mice on the sixth day after the implantation of tumor cells , which we defined as Day 0 . The distribution of the CTLs was assessed in vivo by fluorescence imaging of frozen tissue sections from the naïve lymph nodes ( NLNs ) , tumor-draining lymph nodes ( TDLNs ) , spleens and tumors . The data indicated that the CTLs primarily accumulated in the TDLNs but rarely accumulated in the tumors during either the early or the late stage ( Figure 1B , Figure 1—figure supplement 2 ) . To visualize the immune reaction in the tumor microenvironment under immunotherapy dynamically , we established a murine model of the multicolor-coded tumor microenvironment using fluorescent protein ( FP ) transgenic C57BL/6 mice with a skin-fold window chamber implanted with CFP-B16 tumor cells . Here , several types of FP transgenic C57BL/6 mice were used to represent the host immunocytes: Foxp3+ regulatory T cells were labeled with red FP ( mRFP ) , endogenous tumor-infiltrating lymphocytes were labeled with green FP ( GFP ) , CD11c+ dendritic cells were labeled with yellow FP ( YFP ) , and tumor-infiltrating immunocytes ( TIIs ) were labeled with enhanced green FP ( EGFP ) . The intravital imaging data revealed that only a few CFSE-labeled adoptive CTLs infiltrated into the tumor area , regardless of an early or late stage post-adoptive CTL transfer ( Figure 1C , Figure 1—figure supplement 3 ) . Compared with the rare infiltration of adoptive CTLs , Tregs ( Foxp3-mRFP cells ) were clearly observed infiltrating the tumor microenvironment on Day 1 , before they gradually accumulated at the edge of the tumor to form a ring on Day 5 ( Figure 1C , Figure 1—figure supplement 3 ) . This observation suggested that an ‘immunosuppressive ring’ formed by Tregs surrounded the solid tumor to prevent the recruitment of adoptive CTLs , as illustrated in the schematic diagram in Figure 1D . To verify the immunosuppressive effect of Tregs on the infiltration of adoptive CTLs into the tumor areas , CTX ( 150 mg kg−1 ) was used to specifically abrogate Tregs on the fourth day after the implantation of CFP-B16 tumor cells . This CTX dose was selected on the basis of CFP-B16 tumor inhibition experiments with different CTX doses in immune-competent C57BL/6 mice and immune-deficient BALB/c nude mice . The results of these experiments showed that , compared with other CTX doses ( 50 mg kg−1 and 100 mg kg−1 ) usually used in tumor immunotherapy , only the dose of 150 mg kg−1 CTX controlled the CFP-B16 tumor growth successfully when CTX was used alone as well as when it was used in combination with ACT ( Figure 2—figure supplement 1A ) . Furthermore , a tumor growth inhibition experiment performed on the BALB/c nude mice , confirmed that the effective anti-tumor effect of CTX at the dose of 150 mg kg−1 is not due to its direct cytotoxicity , because no difference in the inhibition of tumor growth was observed between the PBS control group and the mice treated with different doses of CTX ( 50 , 100 or 150 mg kg−1 , Figure 2—figure supplement 1B ) . These results suggest that the efficacy of the 150 mg kg−1 CTX treatment for CFP-B16 tumor in C57BL/6 mice depends on the enhanced anti-tumor immune response , and not on the direct cytotoxicity of the treatment . Two days later , in-vitro-expanded and activated CTLs were intravenously transferred into the tumor-bearing mice; this day was defined as Day 0 ( six days after tumor cells implantation ) . The CTX-ACT combined treatment significantly inhibited tumor growth compared with the ACT treatment alone , but produced results that were not significantly different from those produced by the CTX treatment alone ( Figure 2A ) . The spatio-temporal changes in immunocytes in the tumor microenvironment were monitored on Days 1–7 using large-field confocal microscopy . 10 . 7554/eLife . 14756 . 009Figure 2 . Synergistic effect of CTX and the adoptive CTLs ( ACT ) on CFP-B16 tumor immunotherapy . ( A ) Growth curves for the CFP-B16 tumors treated with ACT , CTX or CTX-ACT and the PBS control . The data are represented as the mean ± SEM tumor volume ( n = 12–14 , three independent experiments ) . ns: not significant , p<0 . 001 , ( Figure 2—source data 1 ) . ( B ) Intravital confocal fluorescence imaging of Tregs ( red ) at the tumor periphery ( left panel ) , and immunohistochemistry ( right panel ) of the CFP-B16 tumor tissues after ACT ( top row ) or CTX-ACT ( bottom row ) treatment . Scale bars: 100 µm . The arrows indicate Tregs . ( C ) Long-term intravital imaging of the multicolor-coded tumor environment in CTX-ACT-treated mice . Red – Tregs ( Foxp3-mRFP ) ; green –CSFE-labeled CTLs; blue –CFP-B16 tumor . Top row: large-field images; scale bar: 500 µm . Bottom row: images from the region of interest in the top row; scale bar: 100 µm . The imaging data are representative of similar results from 3–5 mice in two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 00910 . 7554/eLife . 14756 . 010Figure 2—source data 1 . Growth curves for the CFP-B16 tumors treated with ACT , CTX , CTX-ACT or PBS control . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01010 . 7554/eLife . 14756 . 011Figure 2—figure supplement 1 . Evaluation of the effect of different doses of CTX and CTX combined with ACT on CFP-B16 tumor growth in vivo . ( A , B ) Growth curves for the CFP-B16 tumors in ( A ) C57BL/6 and ( B ) BALB/c nude mice treated with different doses of CTX ( 50 , 100 or 150 mg kg−1 ) , different doses of CTX ( 50 , 100 or 150 mg kg−1 ) combined with ACT , and PBS control . The data are represented as the mean ± SEM tumor volume ( n = 9–10 , two independent experiments ) . p<0 . 001 , ns: not significant , ( Figure 2—figure supplement 1—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01110 . 7554/eLife . 14756 . 012Figure 2—figure supplement 1—source data 1 . Growth curves for the CFP-B16 tumors in C57BL/6 and BALB/c nude mice treated with different doses of CTX , different doses of CTX combined with ACT treatment and PBS control . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01210 . 7554/eLife . 14756 . 013Figure 2—figure supplement 2 . Quantification of the intravital imaging of Tregs and adoptive CTLs in ACT- and CTX-ACT-treated mice . ( A ) Density of Tregs and ( B ) adoptive CTLs were determined by counting the number of mRFP-Tregs and CFSE-labeled CTLs per mm2 in the tumor area . The data are represented as the mean ± SEM ( n = 10–12 fields , 0 . 40 mm2 per field ) from 3–5 mice in two independent experiments . p<0 . 001 , ( Figure 2—figure supplement 2—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01310 . 7554/eLife . 14756 . 014Figure 2—figure supplement 2—source data 1 . Density of Tregs and adoptive CTLs in the tumor area . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 014 First , we assessed whether the CTX treatment effectively depleted Tregs . Intravital imaging ( Figure 2B ) demonstrated that without CTX treatment , the margin of the tumor was surrounded by a layer of Tregs on Day 5 ( Figure 2B , top left corner ) . The CTX treatment successfully eliminated most of the Tregs in the tumor area and removed the ‘immunosuppressive ring’ formed by Tregs ( Figure 2B , bottom left corner ) . These imaging results were further confirmed by immunohistochemical staining using the anti-Foxp3 antibody ( Figure 2B , right panel ) . In addition , the number of CFSE-labeled CTLs in the tumor area was clearly increased in CTX-ACT-treated mice , compared with that in mice received the ACT treatment alone ( Figure 2B , left panel , Figure 2—figure supplement 2 ) . Next , long-term large-field intravital imaging revealed the entire immune reaction in the tumor microenvironment after CTX-ACT treatment , from the accumulation of adoptive CTLs to the destruction of the tumor ( Figure 2C ) . These images showed that in the combined treatment , few adoptive CTLs infiltrated into the tumor microenvironment during the early stage ( Days 1–3 ) . Subsequently , the accumulation of CTLs increased significantly during the late stage ( Days 4–6 , Figure 2—figure supplement 2 ) . By contrast , mRFP-Tregs were rarely observed regardless of the respective stage ( Figure 2—figure supplement 2 ) . The elimination of Tregs and the infiltration of CTLs resulted in large-scale tumor death consisting of tumors 'shrinking' from the outside and 'melting' from the inside ( Figure 2C , Days 5–6 ) . This finding suggested that the infiltrating CTLs destroyed the solid tumor by applying two approaches: attacking the tumor cells at the periphery according to an 'outside-in' pattern and eliminating tumor cells at the parenchyma according to an 'inside-out' pattern . The anti-tumor effect of the adoptive CTLs peaked at Days 5–6 . One day later ( Figure 2C , Day 7 ) , the number of adoptive CTLs decreased rapidly in the tumor area because of the death or dysfunction of CTLs , leading to tumor regrowth . The migration of adoptive CTLs is the crucial step in accelerating the penetration of CTLs into the solid tumors . To understand the migratory behavior of the adoptive CTLs in the tumor microenvironment , we obtained dynamic information about the adoptive CTLs using confocal laser scanning microscopy ( CLSM ) from Day 1 to Day 6 after the CTX-ACT treatment ( Figure 3A–D ) . Subsequently , we quantified the motility of the adoptive CTLs in detail over four stages . Three parameters were used to describe the motility properties of the CTLs in vivo ( Figure 3F–H ) : the mean velocity , which represents the migratory speed; the confinement ratio , which indicates the ratio of the maximum displacement of each cell from its path length within a given time ( Boissonnas et al . , 2007; Cahalan and Parker , 2008 ) ; and the arrest coefficient , which denotes the percentage of time that each cell remained arrested ( Boissonnas et al . , 2007 ) . 10 . 7554/eLife . 14756 . 015Figure 3 . Migratory behavior of the adoptive CTLs in the tumor microenvironment of mice treated with CTX-ACT . ( A–D ) Time-lapse images of CTLs with time-coded motion trajectories ( color scale represents the duration ) . ( A , C ) Images of CTLs ( green ) at the periphery ( near the blue area , A ) or in the parenchyma ( in the blue area , C ) of the CFP-B16 tumors on different days after CTX-ACT treatment . Scale bar: 100 µm . ( B , D ) Trajectories of the individual CTLs at the periphery or in the parenchyma were plotted following the alignment of their starting positions . ( E ) Random walking analysis of the adoptive CTLs . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs at the periphery of ( top ) or in the parenchyma ( bottom ) on different days , ( Figure 3—source data 1–9 ) . ( F–H ) Scatter plots of ( F ) the mean velocity , ( G ) the confinement ratio , and ( H ) the arrest coefficient of the CTLs at the tumor periphery or in the tumor parenchyma on different days after CTX-ACT treatment . Each data point represents a single cell , and the red bars indicate mean values . p<0 . 05 , p<0 . 01 , p<0 . 001; ns – not significant , ( Figure 3—source data 10 ) . The data from 4–6 mice in three independent experiments were pooled . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01510 . 7554/eLife . 14756 . 016Figure 3—source data 1 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs at the tumor periphery on Day 1 . Data listed in the excel files are displacements . Displacement is a function of time . Each column in the excel files lists displacements corresponding to one time . The time values are 0 . 5 min , 1 min , 1 . 5 min , and so on , in turn . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01610 . 7554/eLife . 14756 . 017Figure 3—source data 2 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs at the tumor periphery on Day 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01710 . 7554/eLife . 14756 . 018Figure 3—source data 3 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs at the tumor periphery on Day 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01810 . 7554/eLife . 14756 . 019Figure 3—source data 4 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs at the tumor periphery on Day 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 01910 . 7554/eLife . 14756 . 020Figure 3—source data 5 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs in the tumor parenchyma on Day 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 02010 . 7554/eLife . 14756 . 021Figure 3—source data 6 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs in the tumor parenchyma on Day 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 02110 . 7554/eLife . 14756 . 022Figure 3—source data 7 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the CTLs in the tumor parenchyma on Day 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 02210 . 7554/eLife . 14756 . 023Figure 3—source data 8 . Linear fitting results of MD ( mean displacement ) of adoptive CTLs at the tumor periphery on Day 1 , Day 3 , Day 5 and Day 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 02310 . 7554/eLife . 14756 . 024Figure 3—source data 9 . Linear fitting results of MD ( mean displacement ) of adoptive CTLs in the tumor parenchyma on Day 3 , Day 5 and Day 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 02410 . 7554/eLife . 14756 . 025Figure 3—source data 10 . Scatter plots of the mean velocity , confinement ratio , and arrest coefficient of the adoptive CTLs at the tumor periphery or in the tumor parenchyma on different days . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 025 The first stage corresponds to Days 1–2 . During this stage , a number of adoptive CTLs migrated to the tumor periphery ( Figure 3A , Day 1 ) with a mean velocity of 3 . 23 ± 2 . 18 µm min−1 ( n = 80 cells from 4–5 mice , Figure 3F ) and only a few CTLs infiltrated into the tumor parenchyma ( Video 1 ) . 10 . 7554/eLife . 14756 . 026Video 1 . In vivo sequential imaging of adoptive CTLs at the periphery and in the parenchyma of CFP-B16 tumors on Day 1 ( after CTX-ACT treatment ) . The 3D time-lapse images were acquired as a 30 μm z-stack . Adoptively transferred CTLs are shown in green ( CFSE-labeled ) , and the B16 tumor cells are shown in blue ( CFP ) . Scale bar: 70 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 026 The second stage corresponds to Days 3–4 . During this stage , adoptive CTLs accumulated at the periphery of the tumor ( mean velocity – 3 . 13 ± 2 . 43 µm min−1 , confinement ratio – 0 . 57 ± 0 . 29 , and arrest coefficient – 48 ± 36% , n = 305 cells; Figure 3A , F–H , Day 3 ) and their motility was similar to that at Day 1 . Nevertheless , a few of adoptive CTLs were able to infiltrate into the tumor parenchyma , and remained confined with constrained trajectories ( Figure 3C , D ) and slow speed ( mean velocity – 0 . 87 ± 0 . 28 µm min−1 , confinement ratio – 0 . 32 ± 0 . 21 , and arrest coefficient – 91 ± 8%; n = 118 cells; Figure 3F–H ) , suggesting that these CTLs formed stable interactions with neighboring tumor cells ( Video 2 ) . 10 . 7554/eLife . 14756 . 027Video 2 . In vivo sequential imaging of adoptive CTLs at the periphery and in the parenchyma of CFP-B16 tumors on Day 3 . The 3D time-lapse images were acquired as a 30 μm z-stack . Adoptively transferred CTLs are shown in green ( CFSE-labeled ) , and the B16 tumor cells are shown in blue ( CFP ) . Scale bar: 70 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 027 During the third stage ( Day 5 ) , a larger number of adoptive CTLs migrated to the tumor periphery and their mean velocity dramatically increased to 5 . 79 ± 3 . 12 µm min−1 ( n = 428 cells ) , which was much faster than the mean velocities calculated during the first two stages ( p<0 . 001; Figure 3F ) . Consistently , the migration trajectories at this stage were less confined ( confinement ratio: 0 . 67 ± 0 . 26; Figure 3B , G ) , and the arrest coefficient decreased to 22 ± 30% ( Figure 3H ) . The CTLs in the deep tumor parenchyma also displayed a trend toward greater speed with a marked increase in their mean velocity to 3 . 22 ± 2 . 58 µm min−1 ( n = 193 cells; Figure 3F ) , less confined trajectories ( confinement ratio – 0 . 60 ± 0 . 25; Figure 3B , G ) and a decrease in the arrest coefficient ( 50 ± 39%; Figure 3H ) . The changes in the migratory behavior of the CTLs in both tumor periphery and parenchyma were consistent with the observed tumor destruction ( Figures 2C and 3C , Video 3 ) , indicating that when most tumor cells were killed , the adoptive CTLs resumed their high-speed movement to search for new targets . 10 . 7554/eLife . 14756 . 028Video 3 . In vivo sequential imaging of adoptive CTLs at the periphery and in the parenchyma of CFP-B16 tumors on Day 5 . The 3D time-lapse images were acquired as a 30 μm z-stack . Adoptively transferred CTLs are shown in green ( CFSE-labeled ) , and the B16 tumor cells are shown in blue ( CFP ) . Scale bar: 70 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 028 During the fourth stage ( Day 6 ) , the motility of CTLs near the tumor periphery remained high ( mean velocity – 5 . 89 ± 3 . 9 µm min−1 , confinement ratio – 0 . 69 ± 0 . 24 , and arrest coefficient – 26 ± 34%; n = 147 cells; Figure 3F–H ) . By contrast , the motility of CTLs in the parenchyma significantly decreased as evidenced by their more confined trajectories and an obvious increase in the arrest coefficient ( mean velocity – 0 . 86 ± 1 . 49 µm min−1 , confinement ratio – 0 . 46 ± 0 . 22 , and arrest coefficient – 88 ± 29%; Figure 3F–H ) . The decreased number and motility of the CTLs and tumor regrowth suggested that the adoptive CTLs were dysfunctional or died during this stage ( Video 4 ) . The mean displacement analysis revealed that all of the adoptive CTLs in both the tumor periphery and the tumor parenchyma displayed random walking during different stages ( Figure 3E ) . 10 . 7554/eLife . 14756 . 029Video 4 . In vivo sequential imaging of adoptive CTLs at the periphery and in the parenchyma of CFP-B16 tumors on Day 6 . The 3D time-lapse images were acquired as a 30 μm z-stack . Adoptively transferred CTLs are shown in green ( CFSE-labeled ) , and the B16 tumor cells are shown in blue ( CFP ) . Scale bar: 70 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 029 These data suggested that the migratory behavior of adoptive CTLs in the tumor areas at different stages is correlated to their anti-tumor efficacy . The dysfunction and death of adoptive CTLs and tumor regrowth during the fourth stage suggested that a subsequent round of CTX-ACT combined treatment is required to maintain the anti-tumor immune activation . Next , we studied the dynamic behavior of endogenous CTLs in the tumor area . Here , we used Cxcr6+/gfp transgenic mice that carried the CXCR6 sequence on one allele but had the GFP gene replacing the Cxcr6 coding region on the other allele , in which GFP cells with the CD8+ marker represent endogenous CTLs ( Unutmaz et al . , 2000; Ruocco et al . , 2012 ) . First , the ex vivo analysis of endogenous GFP cells in the CFP-B16 tumors of mice treated with PBS , ACT , CTX , and CTX-ACT on Day 5 ( 11 days after implantation of CFP-B16 tumor cells ) was performed by flow cytometry . The data showed that the endogenous GFP cells in the tumors decreased after CTX-ACT treatment ( Figure 4—figure supplement 1A , B ) . Importantly , most of the GFP cells in the tumors of the CTX-ACT-treated mice were CD8+ CTLs with expression of the activation marker CD69 ( more than 60% , Figure 4—figure supplement 1C ) . These results suggest that the CTX-ACT combined treatment decreased the endogenous T cells but selectively retained the activated endogenous CD8+ CTLs in the tumors . In order to distinguish adoptive and endogenous CTLs by intravital imaging , the adoptive CTLs were labeled with the red fluorescent dye CMTPX and then transferred into Cxcr6+/gfp transgenic mice with CFP-B16 tumor cells implanted into the window chamber on Day 0 ( six days after tumor implantation ) . The intravital imaging was performed on Day 5 ( five days after adoptive transfer of CTLs ) because tumor elimination reached its peak in the CTX-ACT-treated mice on this day . Intravital imaging showed that the number of endogenous GFP T cells in the tumor area of CTX-ACT-treated mice was much smaller than that in ACT-treated mice ( Figure 4A , B ) . The analysis of the migratory behavior of the endogenous T cells showed that , compared with the ACT group , the motility of the endogenous GFP T cells in the tumor areas of the CTX-ACT group was significantly decreased ( Video 5 ) , with more confined trajectories and a significantly increased arrest coefficient ( mean velocity – 3 . 80 ± 2 . 56 µm min−1 in the CTX-ACT group versus 4 . 6 ± 2 . 0 µm min−1 in the ACT group; confinement ratio – 0 . 47 ± 0 . 26 in the CTX-ACT group versus 0 . 60 ± 0 . 25 in the ACT group; and arrest coefficient – 43 ± 35% in the CTX-ACT group versus 25 ± 26% in the ACT group; Figure 4C–F ) . 10 . 7554/eLife . 14756 . 030Figure 4 . Migratory behavior of endogenous CTLs in the tumor microenvironment of mice treated with ACT and CTX-ACT on Day 5 . ( A ) In vivo time-lapse images of the endogenous GFP T cells ( green ) and adoptive CTLs ( red ) in the CFP-B16 tumor area ( blue ) . Mice were treated with ACT or CTX-ACT . Scale bar: 100 μm . ( B ) Quantification of endogenous GFP T cells in the differently treated groups on Day 5 . Results are represented as the mean ± SEM ( n = 11–19 fields , 0 . 40 mm2 per field ) from three mice per group . *p<0 . 001 ( Figure 4—source data 1 ) . ( C ) Trajectories of GFP T cells in the differently treated groups were plotted following the alignment of their starting positions . ( D–F ) Scatter plots of ( D ) the mean velocity , ( E ) the confinement ratio , and ( F ) the arrest coefficient of the GFP T cells in tumor areas in the differently treated groups on Day 5 . Each data point represents a single cell , and the red bars indicate mean values . p<0 . 05 , p<0 . 01 , p<0 . 001; ns: not significant ( Figure 4—source data 2 ) . The data from 3–5 mice in two independent experiments were pooled . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03010 . 7554/eLife . 14756 . 031Figure 4—source data 1 . Quantification of endogenous GFP T cells in the differently treated groups on Day 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03110 . 7554/eLife . 14756 . 032Figure 4—source data 2 . Scatter plots of the mean velocity , confinement ratio , and arrest coefficient of the GFP T cells in tumor areas in the differently treated groups on Day 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03210 . 7554/eLife . 14756 . 033Figure 4—figure supplement 1 . Characterization of GFP cells in differently treated Cxcr6+/gfp mice . ( A ) Ex vivo analysis and characterization of the GFP cells in the tumors of Cxcr6+/gfp mice that were treated differently . ( B ) Percentage of GFP cells in the tumors of mice following the different treatments . ( C ) Percentage of CD8+CD69+ CTLs in the GFP cells from ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03310 . 7554/eLife . 14756 . 034Video 5 . In vivo sequential imaging of endogenous T cells and adoptive CTLs in CFP-B16 tumor areas in mice exposed to ACT and CTX-ACT combined treatments on Day 5 . Time-lapse images were collected for 15 min . Endogenous T cells are shown in green ( GFP labeled ) , adoptive CTLs are shown in red ( CMTPX labeled ) and the B16 tumor cells are shown in blue ( CFP labeled ) . Scale bar: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 034 Both ex vivo characterization experiments and intravital imaging suggested that the CTX-ACT combined treatment deleted most of the endogenous T cells but retained the activated endogenous CTLs in the tumor area . The activated endogenous CTLs arrested in the tumor area for a long time and built stable , long-lasting interactions with the tumor cells to kill them efficiently . Next , we used EGFP-transgenic C57BL/6 mice to observe the activated endogenous immunocytes in the tumor areas . CFP-B16 tumor cells were implanted into EGFP mice , in which all of the nucleated cells expressed EGFP and most of the mobile cells in vivo were immunocytes . Using a multi-photon excitation microscope , the tumor microenvironments of the CFP-B16 tumors and TIIs were continuously observed through a skin-fold window chamber from the day prior to the adoptive transfer of CTLs ( Day 0–4 , Figure 5A , B ) . 10 . 7554/eLife . 14756 . 035Figure 5 . Migratory behavior of endogenous tumor-infiltrating immunocytes ( TIIs ) induced by the CTX-ACT treatment . ( A ) Experimental procedure for long-term intravital imaging of TIIs in the tumor microenvironment . ( B ) In vivo time-lapse images of EGFP TIIs ( green ) in the CFP-B16 tumor area ( blue ) from Day 0 to Day 4 after CTX-ACT treatment . Green arrows represent TIIs displacement , and blue areas represent CFP-B16 tumors . Scale bar: 100 µm . ( C ) Trajectories of individual TIIs on different days were plotted following the alignment of their starting positions . ( D–F ) Scatter plots of the ( D ) mean velocity , ( E ) confinement ratio , and ( F ) arrest coefficient of EGFP TIIs in tumor areas on different days . Each data point represents a single cell , and the red bars indicate mean values . p<0 . 05 , p<0 . 01 , p<0 . 001; ns , not significant , ( Figure 5—source data 1 ) . ( G ) Random walking analysis of the TIIs on different days . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs , ( Figure 5—source data 2–7 ) . The data from 4–7 mice in three independent experiments were pooled . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03510 . 7554/eLife . 14756 . 036Figure 5—source data 1 . Scatter plots of the mean velocity , confinement ratio , and arrest coefficient of EGFP TIIs in tumor areas on different days . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03610 . 7554/eLife . 14756 . 037Figure 5—source data 2 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs on Day 0 . Data listed in the excel files are displacements . Displacement is a function of time . Each column in the excel files lists displacements corresponding to one time . The time values are 0 . 5 min , 1 min , 1 . 5 min , and so on , in turn . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03710 . 7554/eLife . 14756 . 038Figure 5—source data 3 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs on Day 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03810 . 7554/eLife . 14756 . 039Figure 5—source data 4 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs on Day 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 03910 . 7554/eLife . 14756 . 040Figure 5—source data 5 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs on Day 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04010 . 7554/eLife . 14756 . 041Figure 5—source data 6 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs on Day 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04110 . 7554/eLife . 14756 . 042Figure 5—source data 7 . Linear fitting results of MD ( Mean displacement ) of TIIs at tumor areas on Day 0–Day 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 042 On Day 1 , the endogenous TIIs in the CTX-ACT-treated mice migrated rapidly toward the tumor parenchyma ( mean velocity – 4 . 92 ± 2 . 80 µm min−1; n = 435 cells ) . The velocity on Day 1 ( 24 hr after CTX-ACT treatment , Figure 5B–D , Video 6 ) was approximately 2 . 4-fold faster than that on Day 0 ( 2 . 07 ± 2 . 18 µm min−1 , n = 477 cells; Figure 5D ) . Consistently , the TIIs on Day 1 displayed more expanded migration trajectories ( Figure 5C ) and less arrest ( confinement ratio – 0 . 66 ± 0 . 25 on Day 1 , 0 . 43 ± 0 . 26 on Day 0; arrest coefficient – 21 ± 21% on Day 1 , 68 ± 37% on Day 0 , Figure 5E , F ) . 10 . 7554/eLife . 14756 . 043Video 6 . In vivo sequential imaging of endogenous TIIs at the periphery of the CFP-B16 tumors on different days following treatment with CTX and ACT . Time-lapse images were collected for 10 or 15 min . Endogenous TIIs are shown in green ( EGFP ) , and the B16 tumor cells are shown in blue ( CFP ) . Scale bar: 70 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 043 On Day 2 , the motility of the endogenous TIIs decelerated markedly ( Figure 5B–D ) , demonstrating a decreased mean velocity ( 1 . 51 ± 1 . 78 µm min−1; n = 136 cells ) and confinement ratio ( 0 . 48 ± 0 . 29 ) and an increased arrest coefficient ( 75 ± 36% , Figure 5D–F ) . On Day 3 and Day 4 , the speed of the TIIs further decreased ( 0 . 72 ± 0 . 31 µm min−1 , n = 164 cells; 0 . 62 ± 0 . 72 µm min−1 , n = 124 cells , respectively ) , and all other motility parameters of the TIIs differed from those on Day 2 ( Figure 5D–F ) . The mean displacement of the TIIs on different days was consistent with random walking ( Figure 5G ) . Thus , a transient increase in the motility of the endogenous TIIs was induced within 24 hr after the CTX-ACT treatment . These findings indicate that Day 1 is a key point in time for the activation of endogenous anti-tumor immune reactions in response to the CTX-ACT combined treatment . We speculated that the tumor destruction was caused by both adoptive CTLs and endogenous TIIs . To further investigate the role of CTX-ACT treatment in the activation of endogenous TIIs , we analyzed the movement of TIIs in the tumor microenvironment of mice exposed to different treatments on the same day ( Day 1 after the CTX-ACT treatment and seven days after tumor implantation ) . Analysis of the imaging data and migratory path indicated that the chemotaxis of the endogenous TIIs toward the tumor parenchyma was elicited only by the CTX-ACT treatment ( Figure 6A , B , Video 7 ) , which caused the TII movement to display long and linear trajectories ( confinement ratio – 0 . 66 ± 0 . 25 , Figure 6B , D ) . Compared with those in the CTX-ACT treatment group , the directions of TII movement in the ACT and PBS groups were disordered with confined trajectories ( confinement ratio – 0 . 57 ± 0 . 27 in the ACT group and 0 . 47 ± 0 . 24 in the PBS group , Figure 6A , B and D ) . Although the trajectories of the TIIs in the CTX group were less confined with a high confinement ratio ( 0 . 63 ± 0 . 25 ) , the directional movement of the TIIs was disordered ( Figure 6A , B ) . 10 . 7554/eLife . 14756 . 044Figure 6 . Migratory behavior of the TIIs following different treatments . ( A ) In vivo time-lapse images of the EGFP TIIs in the CFP-B16 tumor area on Day 1 . Mice were treated with ACT , CTX , CTX-ACT or PBS control . Green arrows represent TIIs displacement , and blue areas represent CFP-B16 tumors . Scale bar: 100 µm . ( B ) The trajectories of individual EGFP TIIs in different treated groups were plotted following the alignment of their starting positions . ( C–E ) Scatter plots of the ( C ) mean velocity , ( D ) confinement ratio , and ( E ) arrest coefficient of the EGFP TIIs in the differently treated groups . Each data point represents a single cell , and the red bars indicate mean values . p<0 . 05 , p<0 . 01 , p<0 . 001; ns , not significant ( Figure 6—source data 1 ) . ( F ) Random walking analysis of the TIIs in the different groups . Mean displacement ( μm ) versus the square root of time ( min1/2 ) of TIIs in different treatment groups ( Figure 6—source data 2–6 ) . The data from 12–15 mice in three independent experiments were pooled . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04410 . 7554/eLife . 14756 . 045Figure 6—source data 1 . Scatter plots of the mean velocity , confinement ratio , and arrest coefficient of the EGFP TIIs in the different treatment groups . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04510 . 7554/eLife . 14756 . 046Figure 6—source data 2 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs in the PBS group . Data listed in the excel files are displacements . Displacement is a function of time . Each column in the excel files lists displacements corresponding to one time . The time values are 0 . 5 min , 1 min , 1 . 5 min , and so on , in turn . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04610 . 7554/eLife . 14756 . 047Figure 6—source data 3 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs in the ACT group . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04710 . 7554/eLife . 14756 . 048Figure 6—source data 4 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs in the CTX group . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04810 . 7554/eLife . 14756 . 049Figure 6—source data 5 . Mean displacement ( μm ) versus the square root of the time ( min1/2 ) of the TIIs in the CTX-ACT group . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 04910 . 7554/eLife . 14756 . 050Figure 6—source data 6 . Linear fitting results of MD ( Mean displacement ) of TIIs at tumor areas in the different treatment groups on Day 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 05010 . 7554/eLife . 14756 . 051Figure 6—figure supplement 1 . Phenotype of EGFP TIIs in CFP-B16 tumors of mice following different treatments . ( A ) Representative tumor sections stained with CD3 , Ly6G and F4/80 . Most of the EGFP TIIs at the tumor periphery were Ly6G+ and F4/80+ . Scale bar: 50 μm . ( B ) Histopathology of HE-stained tumor sections from tumor-bearing mice exposed to different treatments . Black arrows indicate neutrophils . Scale bar: 50 μm . ( C ) Percentage of neutrophils among TIIs at the periphery of the tumors in mice that received different treatments . The data are represented as the mean ± SEM ( n = 14–20 fields ) results from three mice per group . p<0 . 05 , *p<0 . 01 , ( Figure 6—figure supplement 1—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 05110 . 7554/eLife . 14756 . 052Figure 6—figure supplement 1—source data 1 . Percentage of neutrophils among TIIs at the periphery of the tumors in mice that received different treatments . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 05210 . 7554/eLife . 14756 . 053Video 7 . In vivo sequential imaging of endogenous TIIs at the periphery of the CFP-B16 tumors on Day 1 in mice exposed to different treatments . Time-lapse images were collected for a duration of 10 or 15 min . Mice were mock-treated ( with PBS ) or treated with CTX , ACT , or CTX-ACT as indicated . Endogenous TIIs are shown in green ( EGFP-labeled ) and the B16 tumor cells are shown in blue ( CFP-labeled ) . Scale bar: 70 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 053 The endogenous TIIs in the CTX-ACT treatment group displayed the greatest motility with the fastest mean velocity ( 4 . 92 ± 2 . 80 µm min−1; Figure 6C ) and the lowest arrest coefficient ( 21 ± 21% , Figure 6E ) . The TIIs in both the ACT and PBS groups displayed similarly low mean velocities and high arrest coefficients ( 2 . 51 ± 2 . 53 µm min−1 and 61 ± 40% in the ACT group , respectively; 2 . 18 ± 2 . 31 µm min−1 and 66 ± 39% in the PBS group , respectively; Figure 6C , E ) . The motility of the TIIs in the CTX group was higher than that in the ACT or PBS groups but was still lower than that in the CTX-ACT group ( mean velocity – 4 . 1 ± 3 . 1 µm min−1; arrest coefficient – 39 ± 38% , Figure 6C , E ) . The mean displacement of TIIs in the different groups was consistent with random walking except for that in the CTX group , where the TIIs displayed slightly constrained motility ( Figure 6F ) . Among all of these groups , only the CTX-ACT treatment group exhibited the movement of TIIs toward the tumor parenchyma with high motility and expanded trajectories . These findings indicate that the CTX-ACT treatment had a synergistic effect on the activation of endogenous TIIs , which in return accelerated the anti-tumor effects of the adoptive CTLs . Immunofluorescence analysis of the frozen tumor sections showed that most of the tumor-surrounding EGFP cells were Ly6G+ cells ( Figure 6—figure supplement 1A ) . Hematoxylin and eosin ( HE ) staining further confirmed that more than 50% of the immunocytes surrounding the tumors were neutrophils ( Figure 6—figure supplement 1B and C ) . These results indicated that the CTX-ACT treatment triggered the activity of endogenous neutrophils in the anti-tumor immune response rapidly , just 24 hr after the CTX-ACT treatment ( Day 1 ) . DCs also played a key role in activating anti-tumor immune responses . To investigate whether the CTX-ACT treatment accelerated the infiltration of DCs into the tumors , we observed the tumor microenvironment using large-field microscopy . The imaging data showed that many DCs infiltrated into the tumor areas of the CTX-ACT-treated mice on Day 3 ( three days after CTX-ACT treatment , Figure 7A ) . The density of the DCs in the CTX-ACT group increased by 3 . 6-fold compared with that in the PBS group , by 2 . 0-fold compared with that in the CTL group and by 1 . 6-fold compared with that in the CTX group ( Figure 7B ) . Immunofluorescence analysis of the frozen tumor sections from CTX-ACT-treated mice revealed that certain YFP-labeled DCs displayed an mature phenotype characterized by the expression of MHC-II and CD86 , indicating that some mature DCs were present in the tumor areas ( Figure 7C ) . This finding suggests that the CTX-ACT treatment promotes the infiltration of DCs ( some of them were mature ) to the tumor areas to further enhance the anti-tumor effects . 10 . 7554/eLife . 14756 . 054Figure 7 . Intravital imaging of DCs infiltrating into the tumor areas of the mice following different treatments . ( A ) Large-field intravital imaging of DCs ( green ) and Tregs ( red ) in the CFP-B16 tumor area ( blue ) on Day 3 . Top row: large-field images; scale bar 500 µm . Bottom row: images of the region of interest from the top row; scale bar 100 µm . ( B ) Density of DCs in the tumor areas in the different treatment groups . The data are represented as the mean ± SEM ( n = 10–14 fields , 12 mm2 per field ( large-field images ) or 0 . 40 mm2 per field ) from three independent experiments . p<0 . 05 , p<0 . 01 , p<0 . 001; ns , not significant ( Figure 7—source data 1 ) . ( C ) Representative images of mature DCs in tumor sections were immunofluorescently labeled to detect MHC II ( top row ) and CD86 ( bottom row ) . Scale bar: 40 µm . Inserts are magnifications of the regions indicated with arrows . Scale bar: 15 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 05410 . 7554/eLife . 14756 . 055Figure 7—source data 1 . Density of DCs in the tumor areas in the different treatment groups . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 055 Although the CTX-ACT treatment only controlled the tumor growth for several days and displayed no significant difference compared with the CTX treatment group ( Figure 2A ) , intravital imaging revealed that the CTX-ACT combined treatment elicited a stronger anti-tumor immune response than the CTX treatment alone . To improve the anti-tumor efficacy of the combined immunotherapy , we adopted three rounds of CTX-ACT treatment using a metronomic schedule ( Figure 8A ) . According to the intravital imaging data , the motility of the adoptive CTLs remained at a high level on Day 5 both at the tumor periphery and in the parenchyma ( Figure 3A–D , Video 3 ) . Concomitant with this phenomenon , the anti-tumor response also peaked on Day 5 ( Figure 2C , Figure 3A , C ) , and then the tumor continued to grow ( Figure 2C ) . Before the adoptive CTLs were recruited and activated , the motility of the endogenous TIIs transiently increased on Day 1 and then the TIIs efficiently infiltrated into the tumor parenchyma ( Figure 5B , Video 6 ) . According to these results , an effective interval time of five days between treatment rounds was determined for the CTX-ACT combined therapy . When the anti-tumor effect of the adoptive CTLs decreased , a second round of treatment was implemented to elicit the transient activity of the endogenous TIIs . This approach maintained a specific immune reaction against CFP-B16 tumors in vivo at high levels . When administered three times successfully , the metronomic treatments controlled the growth of the tumor during its entire course , with a significantly more effective tumor growth control compared with the other control groups that received only a single round of the CTX-ACT treatment ( p<0 . 001 , Figure 8B ) and the metronomic CTX treatment ( p<0 . 05 , Figure 8B ) . 10 . 7554/eLife . 14756 . 056Figure 8 . Metronomic CTX-ACT therapy efficiently controlled the growth of CFP-B16 tumors in vivo . ( A ) Metronomic therapy schedule of the CTX-ACT treatment . ( B ) Growth curves for CFP-B16 tumors in the different treatment groups . The data are represented as the mean ± SEM tumor volume ( n = 12–15 , three independent experiments ) . p<0 . 05 , p<0 . 01 , *p<0 . 001 ( Figure 8—source data 1 ) . ( C ) Long-term and large-field intravital images of the tumor microenvironment during CTX-ACT metronomic therapy . Blue , a CFP-B16 tumor; green , EGFP host cells . Scale bar: 500 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 05610 . 7554/eLife . 14756 . 057Figure 8—source data 1 . Growth curves for CFP-B16 tumors in the different treatment groupsDOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 057 The entire anti-tumor immune response elicited by the three rounds of metronomic CTX-ACT treatments was captured by long-term intravital imaging . As shown in Figure 8C , a conflict occurred between the tumor cells and immunocytes . Tumor growth was controlled for four days after the first round of CTX-ACT treatment ( Figure 8C , top row ) . When the tumor began to regrow one day later , the second treatment was applied , and continued to control the growth of the tumor for an additional four days ( Figure 8C , middle row ) . The third treatment was applied when the tumor began to regrow , and controlled the growth of the tumor successfully again for several more days ( Figure 8C , bottom row ) . During this process , the continuous elimination of Tregs , transfusion of adoptive CTLs and activation of endogenous immunocytes ( e . g . , CTLs , neutrophils and DCs ) were required to control the tumor growth . To understand the cellular mechanisms underlying tumor immunotherapy , we globally monitored the sequence of events involving multicolor-coded immune cells during CTX-ACT immunotherapy using a tumor-implanted window chamber device ( Schietinger et al . , 2013 ) and large-field intravital imaging technology . To the best of our knowledge , this study is the first to elucidate the entire process and to provide sequential descriptions of the dynamic spatio-temporal changes in Tregs , adoptive CTLs , endogenous CTLs , endogenous TIIs and DCs in the tumor microenvironment ( Figure 9 ) . In particular , it is worth mentioning that we discovered two symbolic cell events that characterize immunosuppression and immunoactivation: the formation of an immunosuppressive ring of Tregs around the solid tumor , which indicates immune suppression , and the chemotactic movement of endogenous TIIs and infiltration of adoptive CTLs and DCs into the tumor parenchyma , which demonstrate immune activation . 10 . 7554/eLife . 14756 . 058Figure 9 . Timeline schematic showing the elicitation of anti-tumor immune responses in the tumor microenvironment by CTX-ACT treatment . Step 1: CFP-B16 tumor cells grew and Tregs accumulated in the tumor area before CTX treatment . Step 2: CTX treatment depleted most Tregs , blocking the formation of an ‘immunosuppressive ring’ . Step 3: A transient increase in the motility of the endogenous neutrophils is elicited by CTX-ACT treatment on Day 1 . Step 4: DCs present increased infiltration on Day 3 . Step 5: Adoptive CTLs present increased infiltration and motility and the CTX-ACT treatment retained activated endogenous CTLs in the tumor area on Day 5 . Step 6: Solid tumor ‘melting' occurs from the inside and 'shrinking’ occurs from the outside on Days 5–6 . DOI: http://dx . doi . org/10 . 7554/eLife . 14756 . 058 An immunosuppressive environment was caused by the accumulation of a large number of Tregs in the tumor areas . Tregs are considered to play a crucial role in mediating immunosuppression in anti-tumor immune responses ( Zou , 2006; Beyer and Schultze , 2006 ) . By using large-field intravital imaging , we observed that Tregs not only concentrated at the tumor periphery ( Figure 1—figure supplement 3 ) but also formed an immunosuppressive ring around a solid tumor ( Figure 1C , D ) , and adoptive CTLs barely infiltrated solid tumors in the presence of this immunosuppressive ring ( Figure 1C , D , Figure 2B , and Figure 1—figure supplement 3 ) . CTLs must recognize the cognate antigen of tumor cells before they are able to infiltrate into the tumor parenchyma sufficiently ( Boissonnas et al . , 2007; Mrass et al . , 2006 ) . Thus , we propose that the immunosuppressive ring of Tregs blocked the ability of adoptive CTLs to recognize the cognate antigen of tumor cells and decreased the migration of these CTLs into the tumor area . This conclusion is supported by the increased number of adoptive CTLs infiltrating into the tumor parenchyma after the CTX treatment ( Figure 2B , C ) , which depleted most of the Tregs in the tumor area and blocked the formation of an immunosuppressive ring around the solid tumor ( Figure 2B , C ) . Additionally , the depletion of the Tregs induced by CTX treatment contributed to the accumulation of adoptive CTLs in the tumor area in three ways: ( 1 ) it cleared the immunosuppressive barrier to facilitate the migration of adoptive CTLs into the tumor area; ( 2 ) it induced the elevated expression of some chemokines and cytokines in the tumor microenvironment to promote the accumulation of adoptive CTLs ( Bracci et al . , 2007; Schiavoni et al . , 2011 ) ; and ( 3 ) it provided space to allow the infiltration and homeostatic proliferation of adoptive CTLs ( Bracci et al . , 2007; Sistigu et al . , 2011 ) . As shown in the schematic in Figure 9 , specific endogenous and adoptive anti-tumor immune reactions were triggered by the synergistic effect of the CTX-ACT combined treatment . Endogenous TIIs were activated and underwent chemotactic movements toward the tumor parenchyma within 24 hr ( Figure 5B , Video 6 ) , before endogenous DCs infiltrated into the tumor parenchyma on Day 3 ( Figure 7A , C ) . In addition , adoptive CTLs accumulated at the tumor periphery , arrested in the tumor parenchyma to contact neighboring tumor cells on Day 3 and then resumed their high-speed migration on Day 5 ( Figure 3A , C and F ) . On Day 5 , the anti-tumor effect elicited synergistically by CTX and ACT in a combined treatment achieved its maximum and promoted external and internal tumor shrinking and melting ( Figure 2C , Figure 3A , C ) . Our results highlight the importance of metronomic treatments in CTX-ACT therapy , as a designed metronomic CTX-ACT treatment schedule ( Figure 8A ) alternately accelerates the endogenous and adoptive anti-tumor immune responses and successfully controls the tumor growth for a long duration ( Figure 8B , C ) . The migratory behavior of CTLs in the tumor microenvironment is complicated and changes dynamically . Certain researchers have indicated that T cells need to arrest for a long period at the tumor periphery and that their interaction with tumor cells is crucial for tumor elimination ( Boissonnas et al . , 2007 ) . However , other studies have shown that the random , high speed migration of lymphocytes from the tumor periphery to the tumor center is beneficial for tumor elimination ( Mrass et al . , 2006 ) . In our study , we observed four different stages of adoptive CTL migration in the same tumor region , which occurred during the tumor destruction process over a period of up to seven days after the CTX-ACT combined treatment ( Figure 3A , C , and Videos 1–4 ) . The adoptive CTLs accumulated at the tumor periphery but barely infiltrated into the tumor parenchyma during the first stage ( Figure 3A-D , Video 1 ) . Then , the CTLs formed a long-term interaction with the tumor cells in both the tumor periphery and the tumor parenchyma during the second stage ( Figure 3A-D , Video 2 ) . Simultaneously with tumor cell death and elimination , the adoptive CTLs resumed their motility during the third stage ( Figure 3A-D , Video 3 ) . Finally , the number and motility of the CTLs rapidly decreased , suggesting CTL dysfunction during the fourth stage ( Figure 3A-D , Video 4 ) . Thus , the differences in the behavior of the adoptive CTLs throughout the four stages provide an explanation for the variations observed in the previous reports . Besides the activities of the adoptive CTLs , the dynamic behavior of endogenous CTLs has also been observed in our study . The CTX-ACT combined treatment depleted most of the endogenous T cells in the tumor area , and most of the remaining endogenous T cells were activated CD8+ CTLs ( Figure 4—figure supplement 1A–C ) . While the adoptive CTLs had complex migratory behaviors during the four different stages , the endogenous CTLs arrested at the tumor area for a long time and showed confined movement so that they formed stable interactions with tumor cells and then effectively destroyed them ( Figure 4A , C–F ) . In recent years , the interest in tumor-associated neutrophils ( TANs ) has increased , and accumulating data have demonstrated that TANs play dichotomous roles because they exhibit both pro-tumor and anti-tumor effects during tumor progression ( Mantovani et al . , 2011; Piccard et al . , 2012; Brandau et al . , 2013 ) . Previous studies have demonstrated that during the early stage of tumor progression , accumulating TANs with high motility ( ( Yao et al . , 2015 ) , also described as TENs ( Granot et al . , 2011 ) ) , play an important role in the anti-tumor response by exerting a direct cytotoxic effect on tumor cells ( Hicks et al . , 2006 ) , enhancing antibody-induced anti-tumor effects ( Albanesi et al . , 2013 ) , and having a stimulatory effect on T cell proliferation ( Eruslanov et al . , 2014 ) . In the present study , more than 50% of the TIIs were neutrophils at a very early stage ( Day 1 ) after the CTX-ACT treatment ( Figure 6—figure supplement 1 ) and exhibited a high motility . Thus , we assumed that these neutrophils were TENs with anti-tumor effects . After the CTX-ACT combined treatment , the endogenous TENs transiently moved ( within 24 hr ) with high speed via chemotaxis toward the tumor parenchyma ( Figure 5 , 6 ) , and their motility increased before the accumulation of adoptive CTLs and DCs ( Figure 2C , 7A ) . Intravital imaging showed that the number of DCs increased significantly on Day 3 after the CTX-ACT treatment ( Figure 7A , B ) . Some DCs displayed mature phenotypes in the tumor tissues of CTX-ACT-treated mice and presented high MHC-II and CD86 expression ( Figure 7C ) . A large number of DCs infiltrating into the tumor parenchyma has been shown to have a positive association with longer survival of cancer patients ( Chaput et al . , 2008; Liu et al . , 2005; Pagès et al . , 2010 ) . Notably , the recruitment of DCs into the tumor area occurred earlier than the solid tumor ‘melting’ from the inside . This phenomenon suggests that the infiltrated DCs may take up tumor antigens and present them to T cells ( Restifo et al . , 2012 ) to further elicit the anti-tumor immune response . Most previous studies applied the OT-I system with Kb-OVA as a model antigen in T lymphoma cells to observe the migratory behavior of antigen-specific CTLs in solid tumors ( Boissonnas et al . , 2007; Breart et al . , 2008; Boissonnas et al . , 2013 ) . However , cancer cell antigens are expressed at a low level in most patients ( Restifo et al . , 2012 ) . Thus , the Kb-OVA antigen tumor model could not completely mimic the clinical presentation . In the present study , by contrast , we designed an experimental system to mimic ACT immunotherapy using whole tumor cell antigens to prime and activate the CTLs . In summary , the CTX-ACT combined treatment displayed synergistic anti-tumor effects by blocking the formation of an immunosuppressive ring around solid tumors by Tregs , eliciting the transient migratory activity of TIIs toward the tumor parenchyma , and accelerating the infiltration of adoptive CTLs and endogenous DCs into the tumor . The CTX-ACT treatment not only promoted the anti-tumor effects of adoptive CTLs but also elicited endogenous anti-tumor immune responses . Guided by the results obtained by intravital imaging , three rounds of metronomic CTX-ACT treatments successfully controlled the tumor growth for several weeks . Thus , our long-term intravital imaging of the multicolor-coded tumor microenvironment provides new perspectives regarding the cellular mechanisms underlying the success or failure of cancer immunotherapies , and will facilitate improvements in the efficacy of combination immunotherapy . Furthermore , the method of long-term intravital imaging of various immune cells in the tumor microenvironment in vivo is also suitable for monitoring multiple immune reactions , studying the spatio-temporal cellular events of other immunotherapies , such as checkpoint inhibitors ( CTLA-4 antibody and PD-1/PD-L1 antibodies ) , and comparing the efficacy of classical immunotherapies ( such as the CTX-ACT combined treatment ) and checkpoint inhibitors in the treatment of different tumor models . C57BL/6 female mice and BALB/c nude mice ( of 6–12 weeks old ) were obtained from Hunan Slack King of Laboratory Animal Co . , Ltd ( Hunan , China ) . B6 . Cg-Tg ( Itgax-Venus ) 1Mnz/J ( CD11c-YFP , RRID: IMSR_JAX:008829 ) mice , C57BL/6-Foxp3tm1Flv/J ( Foxp3-mRFP , RRID: IMSR_JAX:008374 ) mice and B6 . 129P2-Cxcr6tm1Litt/J ( Cxcr6+/gfp , RRID: IMSR_JAX:005693 ) mice were derived from breeding pairs that were originally obtained from Jackson Laboratory ( Bar Harbor , ME ) . To generate multicolor-coded transgenic mice , CD11c-YFP mice were hybridized with Foxp3-mRFP mice . C57BL/6-Tg ( CAG-EGFP ) /J mice , that express EGFP throughout the entire body , excluding erythrocytes and hair , were generously provided by Dr . Zhiying He ( Second Military Medical University , Shanghai , China ) . All of the mice were bred and maintained in a specific pathogen-free ( SPF ) barrier facility at Animal Center of Wuhan National Laboratory for Optoelectronics . All animal studies were approved by the Hubei Provincial Animal Care and Use Committee and followed the experimental guidelines of the Animal Experimentation Ethics Committee of Huazhong University of Science and Technology . B16 melanoma cells were purchased from Boshide Biology Ltd . China ( RRID: CVCL_F936 ) . The B16 cell line was stably transfected with the PB transposon system ( Ding et al . , 2005 ) ( a gift from Dr . Xiaohui Wu , Fudan University , Shanghai , China ) , which contained the sequence encoding mCerulean to generate the CFP-B16 tumor cell line . All cell lines were mycoplasma-negative as determined by screening using the MycoProbe Mycoplasma Detection Kit ( R and D Systems , Minneapolis , MN ) . The CFP-B16 cell line was authenticated using the Cell Line Authentication Service by short-tandem repeat ( STR ) profiling carried out by Beijing Microread Genetics Co . , Ltd . ( Beijing , China ) . These cells were cultured in RPMI-1640 medium ( HyClone , Beijing , China ) containing 1% penicillin-streptomycin ( HyClone ) and 10% fetal bovine serum ( FBS , HyClone ) . T cells that were reactive with CFP-B16 were established from C57BL/6 mice using an immunization strategy consisting of a series of tumor cellchallenges . First , the C57BL/6 mice were immunized subcutaneously in both flanks with 2 . 5 × 106 CFP-B16 cells ( pretreated with 50 μg ml−1 mitomycin C [Sigma-Aldrich , Saint Louis , MO] for 2 hr at 37°C ) . Seven days after the primary immunization , the mice were immunized using the same methods . Seven days after the rechallenge , the mice were euthanized , and their spleens were dissected to prepare immunocytes . Spleen-derived cells ( 1–2 × 106 per ml ) were cultured in 24-well plates ( Costar , Suzhou , China ) with RPMI-1640 medium ( HyClone ) , 10% FBS ( HyClone ) , IL-2 ( 50 U ml−1 , Peprotech , Rocky Hill , NJ , USA ) and CFP-B16 whole-cell antigen ( 50 μg ml-1 , supernatant from freeze-thawed tumor cell lysate ) . Three days later , when the CTLs became confluent , the cells were split 1:2 to 1:4 into new 24-well plates using fresh complete medium with whole-cell antigen and IL-2 . The method of generating CFP-B16 CTLs in vitro is discussed in the book ‘Current protocols in Immunology’ ( Restifo and Nicholas , 2011 ) and in previous reports ( Liu et al . , 2006; Bauer et al . , 2014 ) . The cytotoxicity of the CTLs was evaluated using CFSE and PI dual-staining assays . Target cells ( CFP-B16 and normal C57BL/6 splenocytes ) were labeled with 5 μM CFSE ( Invitrogen , Eugene , OR ) for 15 min at 37°C . The CTLs were collected from cultured lymphocytes by centrifugation in Histopaque−1 . 083 ( Sigma-Aldrich , Saint Louis , MO ) and co-cultivated with CFSE-labeled target cells in 96-well U-bottom plates ( Costar , Suzhou , China ) at different E:T ratios ( 25:1 , 12 . 5:1 , 5:1 , 2 . 5:1 , 1 . 25:1 , 0 . 5:1 , and 0 . 25:1 ) . After the cells were incubated for 4 hr at 37°C , the mixed cells were labeled with 5 μM PI ( Sigma-Aldrich , Saint Louis , MO ) in the dark for 10 min . The dual fluorescent signals of the target cells were analyzed using a FACS-Calibur flow cytometer ( Guava EasyCyte 8HT , EMD Millipore Corporation , Darmstadt , Germany ) . The percentage of CTL-specific lysis was calculated according to the following formula: Cytotoxicity ( % ) = [ ( experimental dead target cells ( % ) − spontaneous dead target cells ( % ) ) / ( 1 − spontaneous dead target cells ( % ) ) ]×100 . Before the adoptive transfer , the activity and quality of a number of CTLs cultured in vitro were tested . The CTLs were incubated with anti-mouse CD 16/32 ( Fc block , Cat# 101302 , Biolegend , San Diego , CA; RRID: AB_312801 ) for 10 min , and then they were stained with various antibodies at 4°C in the dark for 30 min and washed once with PBS ( HyClone ) . The supernatants were discarded , and the cell pellets were resuspended in 0 . 2 ml PBS before analysis using a FACS-Calibur flow cytometer ( Guava EasyCyte 8HT ) . Typically , 10 , 000 events were assessed . For the intracellular cytokine staining , the CTLs were stimulated with whole CFP-B16 cell antigen in the presence of Brefeldin A ( 10 μg ml−1 , eBioscience , San Diego , CA ) , and the cells were maintained in Brefeldin A until fixation and then prepared as previously described . CD3-PE ( Cat# 100308 , RRID: AB_312673 ) , CD3-APC/Cy7 ( Cat# 100329 , RRID: AB_1877171 ) , CD8-FITC ( Cat# 100705 , RRID: AB_312744 ) , CD8-PE/Cy7 ( Cat# 100722 , RRID: AB_312761 ) , CD4-APC ( Cat# 100412 , RRID: AB_312697 ) , CD25-FITC ( Cat# 102005 , RRID: AB_312854 ) , CD69-PerCP/Cy5 . 5 ( Cat# 104521 , RRID: AB_940497 ) , CD11b-APC ( Cat# 101211 , RRID: AB_312794 ) , Ly6c-PerCP/Cy5 . 5 ( Cat# 128012 , RRID: AB_1659241 ) , Ly6G-APC/Cy7 ( Cat# 127624 , RRID: AB_10640819 ) , Gr1-PerCP/Cy5 . 5 ( Cat# 108427 , RRID: AB_893561 ) , CD45-PE ( Cat# 103106 , RRID: AB_312971 ) and IFNγ-APC ( Cat# 505809 , RRID: AB_315403 ) ( all obtained from Biolegend , San Diego , CA ) , and Granzyme B-PerCP-eFluor 710 ( Cat# 46-8898–80 , eBioscience , San Diego , CA; RRID: AB_11217678 ) were used . CFP-B16 melanoma cells ( 5 × 105 ) were subcutaneously implanted in the right flank of C57BL/6 mice ( females , 6–8 weeks old ) . The tumor size was measured at five days post-implantation using a digital caliper . The volume of the tumor was calculated as V = L ( length ) × W ( width ) × H ( height ) /2 ( Huang et al . , 2013 ) . Four days after the tumor inoculation , the mice in the CTX ( Sigma-Aldrich , Saint Louis , MO ) treatment group received an intraperitoneal ( i . p . ) injection of CTX diluted in sterile distilled H2O at a concentration of 50 mg kg−1 , 100 mg kg−1 and 150 mg kg−1 . Six days after tumor inoculation , the mice of the ACT and CTX-ACT groups were intravenously ( i . v . ) injected with 5 × 106 adoptive CTLs ( 250 μL total volume ) . The CTLs ( five days after in vitro stimulation ) were collected and prepared by centrifugation in Histopaque−1 . 083 ( Sigma-Aldrich , Saint Louis , MO ) , washed three times and resuspended in ice-cold HBSS ( HyClone ) . For the metronomic CTX-ACT treatment , the CTX dose schedule was 150 mg kg−1 CTX injected i . p . on days 4 , 9 and 14 after CFP-B16 tumor cell implantation followed by 5 × 106 CTLs ( 250 μL total volume ) injected i . v . two days after CTX injection . Tumor tissues were fixed in 4% paraformaldehyde for 24–48 hr at 4°C , they were embedded in paraffin , sectioned , and stained with HE . For the immunohistochemical analysis , the sections were stained with markers of regulatory T cells ( anti-Foxp3 antibody , 1:800 , ab54501 , Abcam , Cambridge , United Kingdom; RRID: AB_880110 ) . For the immunofluorescence analysis , tumor tissues were fixed in 4% paraformaldehyde for 24–48 hr at 4°C and then subjected to sequential dehydration in 10% , 20% , and 30% sucrose solution . The tissues were then frozen in OCT ( Sakura , Torrance , CA ) compound and cut into 20 μm slices . OCT was removed by three washes in PBS , and the tumor tissues containing EGFP cells were immunostained with Alexa Fluor 594 anti-mouse CD3 ( 1:150 , Cat# 100240 , RRID: AB_2563427 ) , Alexa Fluor 647 anti-mouse CD4 ( 1:100 , Cat# 100424 , RRID:AB_389324 ) , Alexa Fluor 647 anti-mouse CD8 ( 1:100 , Cat# 100724 , RRID: AB_389326 ) , Alexa Fluor 700 anti-mouse Ly6G ( 1:200 , Cat# 127622 , RRID: AB_10643269 ) , and Alexa Fluor 647 anti-mouse F4/80 ( 1:200 , Cat# 123122 , RRID: AB_893480 ) antibodies ( all of which were obtained from Biolegend , San Diego , CA ) to identify the T cells , neutrophils and macrophages , respectively . The tumor tissues containing YFP-DCs were immunostained with anti-MHC II ( 1:200 , ab15630 , Abcam , RRID: AB_302007 ) and anti-CD86 ( 1:100 , ab25376 , Abcam , RRID: AB_470491 ) and then incubated with mouse monoclonal anti-rat IgG2b/IgG 2a conjugated to Alexa Fluor 647 ( 1:2000 , ab172333 and ab172335 , Abcam , Cambridge , United Kingdom ) . CLSM was performed using a LSM710 ( Carl Zeiss MicroImaging , Inc . , Jena , Germany ) with a 20× objective ( N . A . 0 . 8 ) . The window chambers were prepared as previously described ( Palmer et al . , 2011; Schietinger et al . , 2013 ) . Hair was removed from the back of the mouse one day before surgery . For the window chamber surgery , the mice were anesthetized by i . p . injecting a mix of ketamine ( 100 mg kg−1 , Sigma-Aldrich , Saint Louis , MO ) and xylazine ( 10 mg kg−1 , Sigma-Aldrich , Saint Louis , MO ) and positioned on a warmer plate at 37°C ( Thermo Plate , TOKAI HIT , Shizuoka-ken , Japan ) . The skin on the back of the mice was sterilized with 70% alcohol and iodine solution . Titanium window frames ( APJ Trading Co . , Inc . , Ventura , CA ) were implanted onto the back of the mouse , and then , a hole with a diameter of 1 cm was dissected by removing the skin and fascia on one side of the dorsal skin-fold flap while maintaining the integrity of the opposing dermis , fascial plane and vasculature . One day later , CFP-B16 tumor cells ( 5 × 105 resuspended in 25 μL PBS ) were injected at one site near the major vessel and between the fascia and dermis of the rear skin . The entire surgical process was conducted under sterile conditions to avoid infection . To relieve pain associated with surgery and inflammation , the mice received Tolfedine via i . p . injection ( 16 . 25 mg kg−1 , Vétoquinol , Québec , Canada ) immediately and within 24 hr after implantation . For intravital imaging , the adoptive CTLs were labeled with 25 μM CFSE ( Invitrogen , Eugene , OR or 25 μM CellTracker Red CMTPX Dye ( Invitrogen , Eugene , OR ) for 15–45 min at 37°C as the standardized protocols described . The window-chamber mice were anesthetized by inhalation of 1 . 0–3 . 0% isoflurane in oxygen flow using a Matrx VMS small animal anesthesia machine ( Midmark , Dayton , OH ) . The window was fixed on a warm plate ( Thermo Plate ) using a custom-made holder and then fastened to the microscope stage . Intravital images ( Qu et al . , 2012 ) were obtained using an A1R MP+ System ( Nikon , Tokyo , Japan ) with the large-field imaging function on a motorized stage . The images were captured using a 16× water immersion objective ( N . A . 0 . 8 ) or 20× objective ( N . A . 0 . 75 , Nikon , Japan ) . Throughout the intravital imaging process , the temperature of the mice was maintained at 37°C with a warm plate . Using large-field imaging technology combined with blood vessel imaging as a ‘position mark’ , the same imaging region could be focused on and images of the tumor could be obtained on different days . Confocal laser scanning microscopy ( CLSM ) was used to simultaneously image the CFP-B16 cells ( 405 nm laser , 400–500 nm emission ) , mRFP-Tregs and CMTPX-labeled adoptive CTLs ( 561 nm laser , 570–620 nm emission ) , CFSE-labeled adoptive CTLs and Cxcr6-GFP cells ( 488 nm laser , 500–550 nm emission ) . For the simultaneous imaging of the CFP-B16 cells and EGFP cells in vivo , multi-photon excitation microscopy was applied with an excitation wavelength of 860 nm . Intravital cell movement was tracked and analyzed with Image-Pro Plus ( Media Cybernetics , Inc . , Rockville , MD; RRID: SCR_007369 ) or Imaris 7 . 6 ( Bitplane AG , Zurich , Switzerland; RRID: SCR_007370 ) software . The mean velocity , arrest coefficient , confinement ratio and mean displacement were calculated using Post-TrackObject software ( custom-designed software ) as previously described ( Sumen et al . , 2004; Hugues et al . , 2007; Cahalan and Parker , 2008; Matheu et al . , 2011; Miller et al . , 2002 ) . The arrest coefficient was calculated as the percentage of time that the instantaneous velocity of each cell was less than 2 μm min−1 ( Boissonnas et al . , 2007 ) , and the confinement ratio was calculated as the ratio of the maximum displacement of a given cell to its path length within a given time ( Cahalan and Parker , 2008 ) . Cells with a mean velocity of less than 2 μm min−1 were defined as immotile ( Boissonnas et al . , 2007 ) . The mean displacement plotted against the square root of time was calculated as previously described ( Ruocco et al . , 2012 ) . Linear fitting was then performed on the plotted curves to determine whether the cells underwent random walking ( Cahalan and Parker , 2008 ) . An R2>0 . 95 was evaluated as a good fit and the corresponding cell population was considered to be executing random walking . Statistical analysis was performed using GraphPad Prism 5 ( GraphPad Software , Inc . , La Jolla , CA; RRID: SCR_002798 ) . For comparisons of three or more groups , the Kruskal-Wallis test was performed and followed by Dunn’s multiple comparison tests ( Mrass et al . , 2006 ) . For comparisons of two groups , the two-tailed unpaired t-test was performed . The statistical analysis is described in each figure legend .	Melanoma is a form of skin cancer that is particularly difficult to treat . A new approach that has shown a lot of promise in treating many different cancers , including melanoma , is called “immunotherapy” . This technique harnesses the immune system – the body’s natural defences that help to protect against infections and disease – to combat cancer . One powerful type of immunotherapy involves injecting patients with cells called lymphocytes , which form part of the immune system . This is known as adoptive cell therapy and can activate the immune system to fight cancer , helping to shrink tumors . This treatment can be made even more powerful by combining it with a drug called cyclophosphamide and this combination , known as CTX-ACT , is currently one of the most efficient treatments for melanoma . Yet , little information is available to indicate why this treatment is so effective . Using mice implanted with melanoma cells , Qi , Li et al . sought to understand how CTX-ACT treatment works , with the goal of optimising it to increase its success . The results showed that a protective barrier of immune cells that suppresses the anti-tumor immune response – called an “immunosuppressive ring” – surrounds untreated tumors . CTX-ACT treatment can breakdown these rings , helping to reactivate the anti-tumor immune reaction in the tumors . This allows both the injected and mouse’s own immune cells to move into the tumor and destroy cancer cells . Qi , Li et al . used their findings to optimise treatment and succeeded in controlling tumor growth in the mice for several weeks . These new insights could be used to improve current immunotherapies , and offer new approaches for investigating the involvement of immune cells in the treatment of a wide range of different cancers .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation", "cancer", "biology" ]	2016	Long-term intravital imaging of the multicolor-coded tumor microenvironment during combination immunotherapy
Protein polarization underlies differentiation in metazoans and in bacteria . How symmetric polarization can instate functional asymmetry remains elusive . Here , we show by super-resolution photo-activated localization microscopy and edgetic mutations that the bitopic zinc-finger protein ZitP implements specialized developmental functions – pilus biogenesis and multifactorial swarming motility – while shaping distinct nanoscale ( bi ) polar architectures in the asymmetric model bacterium Caulobacter crescentus . Polar assemblage and accumulation of ZitP and its effector protein CpaM are orchestrated in time and space by conserved components of the cell cycle circuitry that coordinate polar morphogenesis with cell cycle progression , and also act on the master cell cycle regulator CtrA . Thus , this novel class of potentially widespread multifunctional polarity regulators is deeply embedded in the cell cycle circuitry . Some regulatory proteins that execute important developmental , cytokinetic or morphogenetic functions are localized in monopolar fashion , whereas others are sequestered to both cell poles ( Dworkin , 2009; Martin and Goldstein , 2014; Shapiro et al . , 2002; St Johnston and Ahringer , 2010 ) . It is unclear if bipolar proteins can confer specialized functions from each polar site , but examples of proteins with a bipolar disposition have been reported for eukaryotes and prokaryotes ( Davis et al . , 2013; Martin and Berthelot-Grosjean , 2009; Tatebe et al . , 2008; Treuner-Lange and Sogaard-Andersen , 2014 ) . The synchronizable Gram-negative α-proteobacterium Caulobacter crescentus ( henceforth Caulobacter ) is a simple model system to study pole-specific organization and cell cycle control ( Tsokos and Laub , 2012 ) . The Caulobacter predivisional cell is overtly polarized and spawns two morphologically dissimilar and functionally specialized daughter cells , each manifesting characteristic polar appendages ( Figure 1A ) . The swarmer progeny is a motile and non-replicative dispersal cell that samples the environment in search of food . It harbours adhesive pili and a single flagellum at one pole and is microscopically discernible from the stalked cell progeny , a sessile and replicative cell that features a stalk , a cylindrical extension of the cell envelope , on one cell pole . While the stalked cell resides in S-phase , the swarmer cell is in a quiescent G1-like state from which it only exits concomitant with the differentiation into a stalked cell . During this G1→S transition , the polar flagellum and pili of the swarmer cell are eliminated and replaced by the stalk that elaborates from the vacated cell pole . Upon sequential transcriptional activation of developmental factors during the cell cycle ( Panis et al . , 2015 ) , the nascent stalked cell re-establishes polarization and ultimately gives rise to an asymmetric pre-divisional cell that yield a swarmer and a stalked progeny . 10 . 7554/eLife . 18647 . 003Figure 1 . Cell cycle profile and phylogeny of ZitP and CpaM . ( A ) Scheme depicting the polarized factors PopZ , ZitP and CpaM during the cell cycle of the dimorphic bacterium C . crescentus . ( B ) Pilus assembly pathways and global dependencies of the two master cell cycle regulators GcrA and CtrA on the expression of the polar factors PodJ , CpaE , ZitP , CpaM and CpaC that control pilus biogenesis . Red and black dashed lines highlight transcriptional activation and polar recruitment , respectively . ( C ) Schematic representation ( drawn to scale ) of ZitP ( blue ) and CpaM ( yellow ) . ZnR: zinc finger domain; TM: transmembrane domain , C: cysteine . Arrowheads below each protein pinpoint the site of truncation due to transposon insertion in the coding sequence . The large triangle on top of ZitP shows the 2 amino acid residues deleted in the ZitPGAP variant and the small triangle depicts the position of residue 133 where the ZitP coding sequence is truncated in the ZitP1-133 variant . ( D ) Conservation of ZitP ( blue ) , CpaM ( yellow ) and CpaC ( purple ) across the α-proteobacterial clades . The phylogenetic tree was built in CLC Main Workbench ( http://www . clcbio . com/products/clc-main-workbench/ ) from 16S RNA alignments based on the Neighbor Joining method ( Juke Cantor substitution model ) with 100 bootstrap replicates . Empty boxes mean that no ortholog was found in the genome . Scale bar , 0 . 15 substitution per site . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 003 The GcrA transcriptional regulator predominates in early S-phase ( Holtzendorff et al . , 2004 ) ( Figure 1A–B ) . It accumulates during the G1→S transition and activates expression of polarity factors that are required for pilus or flagellum biogenesis and cytokinetic components ( Davis et al . , 2013; Fioravanti et al . , 2013; Murray et al . , 2013; Quon et al . , 1996; Viollier et al . , 2002b ) ( Figure 1A–B ) . Among GcrA target promoters , is the promoter controlling expression of the PodJ polar organizer that localizes to the pole opposite the stalk and directs assembly of the Caulobacter pilus assembly ( cpa ) machine at that site . In this cascade , PodJ recruits the cytoplasmic CpaE protein that then promotes the localization and assembly of CpaC secretin localization ( Figure 1B ) ( Viollier , 2002a ) . Another key promoter controlled by GcrA is the one driving expression of the master cell cycle regulator CtrA that induces the synthesis of a second wave of polar and morphogenesis factors in late S-phase including the cpa operon ( Figure 1B ) . The abundance of CtrA and GcrA is regulated at the level of synthesis and degradation ( Collier et al . , 2006; Domian et al . , 1997 ) and as a result , cell division spawns a swarmer and stalked cell progeny containing CtrA and GcrA , respectively . An important polarity determinant in the α-proteobacteria is the conserved matrix protein PopZ ( Figure 1A ) that organizes poles by forming a molecular lattice that traps polar determinants and effectors ( Bowman et al . , 2008; Deghelt et al . , 2014; Ebersbach et al . , 2008; Grangeon et al . , 2015; Laloux and Jacobs-Wagner , 2013 ) . PopZ is bipolar in the Caulobacter predivisional cell and it interacts directly with numerous cell cycle kinases , the ParAB chromosome segregation proteins and cell fate determinants ( Holmes et al . , 2016 ) . Here , we dissect at the genetic and cytological level the polar localization and function of two poorly characterized trans-membrane proteins , the zinc-finger protein ZitP and the CpaM effector protein , that are polarly localized and that execute multiple regulatory functions . We unearthed two separate localization pathways for each cell pole , one PopZ-dependent and another that is PopZ-independent , and we provide evidence by photo-activated localization microscopy ( PALM ) and by genetic dissection that each polar cluster has a distinctive architecture and a specialized function . As pili are necessary for infection by the lytic caulophage CbK ( φCbK ) ( Skerker and Shapiro , 2000 ) , we specifically sought mutants in pilus assembly factors encoded outside of the major pilus assembly cpa gene locus ( pilA-cpaA-K ) ( Christen et al . , 2016; Skerker and Shapiro , 2000 ) . To this end , we conducted himar1-transposon ( Tn ) mutagenesis of wild-type ( WT ) Caulobacter in the presence of φCbK ( see Methods ) and recovered mutants with Tninsertions in CCNA_02298 , renamed here zitP ( zinc-finger targeting the poles ) because of the pleiotropic roles detailed below , or in cpaM ( CCNA_03552 ) ( Figure 1C ) ( Marks et al . , 2010 ) . While both genes have previously been implicated in polar functions and their transcription is cell cycle-regulated ( Christen et al . , 2016; Fioravanti et al . , 2013; Fumeaux et al . , 2014; Hughes et al . , 2010; McGrath et al . , 2007 ) , they are poorly characterized . The zitP gene is predicted to encode a 311-residue bitopic trans-membrane ( TM ) protein harbouring a CXXC- ( X ) 19-CXXC motif that binds a zinc ion ( zinc_ribbon_5 or PF13719 superfamily , residues 1-37 ) at the cytoplasmic N-terminus ( Bergé et al . , 2016 ) and a conserved domain-of-unknown function ( DUF3426 , residues 128-245 ) in the C-terminal region that is predicted to reside in the periplasm ( Figure 1C ) . The cpaM gene codes for a 394-residue protein harbouring a single N-terminal TM domain and a C-terminal CE4_DAC2-like polysaccharide deacetylase domain predicted to be periplasmic ( Figure 1C ) . ZitP and CpaM are not restricted to the Caulobacter lineage as BLASTP searches revealed orthologs in many α-proteobacterial clades ( Figure 1D ) . To confirm the phenoytpes of the Tn insertion mutants , we engineered strains with an in-frame deletion in zitP ( ΔzitP ) or cpaM ( ΔcpaM ) and found that the mutants no longer supported plaque formation ( lysis ) by the pilus-specific bacteriophage φCbK . By contrast , plaques were still formed by the S-layer specific caulophage φCr30 ( Edwards and Smit , 1991 ) ( Figure 2A ) , showing that mutations in cpaM or zitP prevent infection of φCbK , but not all phages . This defect was corrected upon expression of either ZitP or CpaM from an ectopic locus in ΔzitP or ΔcpaM cells , respectively ( Figure 2A ) . 10 . 7554/eLife . 18647 . 004Figure 2 . Functional dichotomy in ZitP and effects on polar morphogenesis . ( A ) Bacteriophage infection assays of WT , ΔzitP , ΔzitP;fliGD306G and ΔcpaM mutant cells . Cells harbour empty pMT335 or a complementing plasmid ( pMT335 backbone ) and were grown in the absence of vanillate . No xylose was added to the agar for the phage assay on ΔcpaM; Pxyl-dendra2-cpaM cells . The phages φCbK and φCr30 were spotted with serial dilution on C . crescentus embedded in top agar . Sensitivity to phages is indicated by plaques ( lysis ) . ( B ) Adsorption kinetics of φCbK to WT and mutant cells . ( C ) Steady-state levels of ZitP , CpaM , CpaC , modified CpaC ( CpaC* ) and PilA in WT and mutant cells as determined by immunoblotting . In the PilA immunoblots , the asterisk ( * ) points to a non-specific band . ( D ) Immunoblots showing the steady-state levels of monomeric CpaC and CpaC* in ΔzitP cells harbouring pMT335 or derivatives encoding ZitPWT , ZitPCS or ZitPGAP grown in the presence of vanillate ( 50 µM ) . ( E ) Immunoblots showing PilA and FljK abundance in supernatants of WT and various mutant cells . Supernatants were harvested from mid-log cultures after shearing . ( F ) Swarming motility test performed on soft ( 0 . 3% ) agar with WT , ΔzitP , ΔcpaM , ΔpilA and Δfljx6 mutant cells . ( G ) Complementation of the motility defect on swarm ( 0 . 3% ) agar displayed by the ΔzitP cells expressing Dendra2-ZitP variants from Pxyl at the xylX locus . Xylose was added to the swarm ( 0 . 3% ) agar as indicated . ( H ) Flow cytometry of exponential phase WT and ΔzitP cells . N refers to chromosome equivalents . ( I ) Suppression of the ΔzitP motility phenotype by fliGD306G point mutation as shown on a swarm ( 0 . 3% ) agar plate . ( J ) Phage spot tests with φCr30 and φCbK on WT or ΔzitP cells expressing Dendra2-ZitP variants from Pxyl at the xylX locus . Cells were embedded in top agar containing xylose ( 0 . 3% ) . ( K ) Motility assays of ΔzitP cells expressing WT ZitP ( ZitPWT ) , ZitPCS or ZitPGAP from pMT335 . Swarming motility was assessed in absence of vanillate on 0 . 3% agar . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 00410 . 7554/eLife . 18647 . 005Figure 2—figure supplement 1 . Master regulator-dependent promoters in ΔzitP . Relative β-galactosidase activity ( in percentage ) of various lacZ-fused promoters in WT and ΔzitP cells . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 00510 . 7554/eLife . 18647 . 006Figure 2—figure supplement 2 . CtrA- and ( p ) ppGpp-independent influence of the ΔzitP motility defect . ( A ) Relative β-galactosidase activity of lacZ-based promoter-probe reporters to the promoters of pilA and CC_1982 in WT , ΔzitP and cpaM cells . ( B ) Relative β-galactosidase activity of lacZ-based promoter probe reporters to the promoters of pilA and CC_1982 with Pxyl-relA’ ( pXTCYC-4-relA′-FLAG ) or the pXTCYC-4 control plasmid ( vector ) . ( C ) Motility test on swarm agar of WT , ΔzitP and ΔcpaM cells transformed with Pxyl-relA’ ( pXTCYC-4-relA′-FLAG ) or the pXTCYC-4 control plasmid ( vector ) . Xylose was added or not to the agar . ( D ) Motility test on swarm agar of WT and ΔzitP cells transformed with Pxyl-relA’ ( pXTCYC-4-relA′-FLAG ) or the pXTCYC-4 control plasmid ( vector ) . Xylose was added to the agar . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 006 Next , we conducted time-course adsorption assays and found the adsorption kinetics of ΔzitP and ΔcpaM cells to be substantially compromised compared to WT cells ( Figure 2B ) . The φCbK adsorption kinetics of the mutants closely resemble those for ΔcpaC cells that cannot assemble pili because they lack the CpaC secretin ( Skerker and Shapiro , 2000 ) . Moreover , immunoblotting revealed that ΔzitP and ΔcpaM cells do not accumulate the modified form of CpaC , CpaC* ( Figure 2C–D ) . A comparable reduction in CpaC* abundance has been previously reported for ΔcpaE , ΔpodJ and ΔpleA cells that no longer assemble a polar CpaC pilus channel in the outer membrane and cannot be infected by φCbK ( Viollier and Shapiro , 2003; Viollier et al . , 2002b ) . However , CpaC* accumulates in ΔpilA cells ( Figure 2C ) , suggesting that the CpaC channel forms independently of PilA . To test whether ΔzitP and ΔcpaM cells assemble a pilus filament on the cell surface , we conducted shearing assays followed by immunoblotting using antibodies to the PilA pilin , the subunit of the pilus filament ( Figure 2E ) ( Skerker and Shapiro , 2000 ) . Whereas PilA was efficiently released from WT cells into the supernatant by shearing , no PilA was detectable in the supernatants of ΔcpaE , ΔzitP and ΔcpaM cells after shearing ( Figure 2E ) , even though PilA is clearly expressed in these cells ( Figure 2C ) . As the major subunit of the flagellar filament , the FljK flagellin , accumulates in the supernatants in all samples ( Figure 2E ) , we conclude that ZitP and CpaM are required for the presentation of PilA on the cell surface and , as shown below , that they act in the same pathway ( Figure 1B ) . The φCbK adsorption kinetics hinted that motility might be altered in ΔzitP and ΔcpaM cells . This hypothesis is based on the comparison of the φCbK adsorption kinetics to WT , ΔpilA and Δfljx6 ( lacking all six flagellin genes: fljJ/K/L/M/N/O ) cells to ΔzitP and ΔcpaM cells . While pililess ΔpilA cells assemble a flagellum and are motile ( Figure 2F ) , Δfljx6 cells are flagellumless , but piliated ( φCbK sensitive ) ( Guerrero-Ferreira et al . , 2011 ) . The kinetics of adsorption of φCbK to ΔzitP and ΔcpaM cells was strongly reduced compared to WT , fitting halfway between the adsorption curves of φCbK to ΔpilA and Δfljx6 cells ( Figure 2B ) . Since it is known that φCbK first reversibly adsorbs to the flagellar filament rotating counter-clockwise , before the irreversibly attachment to the pilus portal is established for productive infection ( Guerrero-Ferreira et al . , 2011 ) , we wondered whether there are fewer motile cells in the ΔzitP and ΔcpaM populations than in WT or if motility in these mutants is altered in other ways . In fact , motility tests on swarm ( 0 . 3% ) agar revealed a mild reduction in motility of ΔcpaM cells and a severe reduction of ΔzitP cells compared to WT ( Figure 2F ) . However , ΔzitP cells still have residual motility that allows them to spread in swarm agar compared to Δfljx6 cells ( Figure 2F ) . Expression of Dendra2-ZitP from an ectopic locus confers near WT motility to ΔzitP cells ( Figure 2G ) , showing that this deficiency in motility is indeed due to the absence of ZitP . As Caulobacter divides into a motile G1-phase cell and a sessile S-phase cell , mutants accumulating fewer G1-phase cells in the population can exhibit reduced motility on soft agar ( Sanselicio et al . , 2015; Sanselicio and Viollier , 2015 ) . To test if ZitP controls the G1 cell number , we used flow cytometry to quantify the number of G1 cells and indeed observed fewer G1 cells in the ΔzitP population compared to WT ( Figure 2H ) . Knowing that the master cell cycle transcriptional regulator CtrA retains cells in G1-phase and activates many cell cycle-regulated promoters that fire in G1-phase ( Domian et al . , 1997; Fumeaux et al . , 2014; Quon et al . , 1996 ) , we then conducted promoter-probe assays using several CtrA-activated promoters fused to the promoterless lacZ gene and quantified CtrA-dependent promoter activity in WT and ΔzitP cells ( Figure 2—figure supplement 1 ) . While all such promoter-probe reporters for the CtrA regulon exhibited a decrease in activity by 30-40% in ΔzitP versus WT cells , promoter-probe reporters for the GcrA regulon or other reporters were unaffected . Thus , ZitP is required for optimal CtrA activity and G1 cell accumulation . The reduction in CtrA-dependent transcription does not appear to be solely responsible for the motility defect of ΔzitP cells . First , promoter-probe assays revealed that ΔcpaM cells also suffer from reduced CtrA-dependent activation ( Figure 2—figure supplement 2A ) , even though their motility exceeds that of ΔzitP cells ( Figure 2F ) . Second , we were able to mitigate the defect in CtrA-dependent transcription by ectopic expression of the ( p ) ppGpp alarmone , a signalling molecule that enhances CtrA function and stability via a poorly understood mechanism ( Gonzalez and Collier , 2014 ) . We accomplished this by heterologously expressing the truncated version of the E . coli ( p ) ppGpp-synthase RelA ( RelA’ ) from the xylose-inducible promoter at the xylX locus in WT and ΔzitP cells . LacZ-based promoter-probe assays revealed that ectopic induction of ( p ) ppGpp restores CtrA-dependent promoter activity to near WT levels ( Figure 2—figure supplement 2B ) . However , the motility of ΔzitP cells ectopically producing ( p ) ppGpp is still substantially lower than that of WT cells ( Figure 2—figure supplement 2C–D ) , indicating that ZitP also promotes motility through a CtrA- and ( p ) ppGpp-independent pathway . To reinforce this conclusion , we isolated a spontaneous motile suppressor of ΔzitP cells ( see Materials and Methods , Figure 2I ) with a single point mutation in the fliG flagellar gene ( fliGD306G ) that neither corrects the pilus assembly defect ( φCbK-resistance , Figure 2A ) , nor the reduction in G1 cell number of the ΔzitP mutant ( Figure 2H ) . As FliG encodes a component of the flagellar motor that is associated with the cytoplasmic membrane ( Macnab , 2003 ) , we conclude that ZitP controls pilus biogenesis and a multifactorial motility phenotype , with a minor contribution from a CtrA-dependent pathway and a major one from a CtrA-independent pathway ( s ) that can be bypassed by a mutant variant of FliG . To investigate if ZitP also controls its polar functions from the cell pole , we resorted to live-cell fluorescence imaging by epifluorescence microscopy ( Figure 3—figure supplement 1A–D ) and photo-activated localization microscopy ( PALM , Figure 3A–B and D–E ) ( Betzig et al . , 2006 ) using WT , ΔzitP or ΔcpaM cells expressing functional Dendra2-CpaM or Dendra2-ZitP . We observed Dendra2-ZitP to adopt a bipolar disposition in dividing cells , whereas Dendra2-CpaM is restricted to the pole opposite the stalk where the pilus biogenesis machinery assembles ( Figure 3A–B; Figure 3—figure supplement 2A–C ) . While Dendra2-ZitP localization is not noticeably perturbed in ΔcpaM cells ( Figure 3—figure supplement 1B–C ) , Dendra2-CpaM is dispersed in ΔzitP cells ( Figure 3A; Figure 3—figure supplement 1D and 2B ) . Moreover , biochemical pull-down experiments with ZitP-TAP ( Figure 3—figure supplement 3 ) and reciprocal co-immunoprecipitation experiments using antibodies to ZitP and CpaM ( Figure 3C ) showed that ZitP and CpaM reside in a complex . Since Dendra2-ZitP and Dendra2-CpaM localization is not affected in ΔpodJ , ΔcpaE or ΔcpaC cells ( Figure 3F , Figure 3—figure supplement 1C and D ) and since CpaE localization is not noticeably altered in ΔzitP and ΔcpaM cells ( Figure 3—figure supplement 4A–B ) , we conclude that ZitP and CpaM are part of a previously unknown ( PodJ/CpaE-independent ) polarization pathway for pilus assembly in Caulobacter in which ZitP recruits CpaM ( Figure 1B ) . 10 . 7554/eLife . 18647 . 007Figure 3 . Distinct ZitP nanoscale assemblies and localization determinants . ( A ) Photo-activated light microscopy ( PALM ) imaging of Dendra2-ZitP or Dendra2-CpaM expressed from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus in ΔzitP or ΔcpaM cells exposed to xylose 3 hours before imaging . Scale bar: 1 µm . ( B ) PALM imaging of Dendra2-ZitP in WT or ΔpopZ::Ω cells . We induced expression of Dendra2-ZitP from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus by the addition of xylose 3 hours before imaging . Scale bar: 1 µm . Scale bar of zoomed images: 0 . 5 µm . ( C ) Co-immunoprecipitation ( co-IP ) of ZitP or CpaM with polyclonal antibodies to CpaM or ZitP , respectively . Immunoprecipitates and cell lysates from WT , ΔzitP or ΔcpaM cells were probed for the presence of ZitP or CpaM . ( D ) Projected area of the Dendra2-ZitP polar complex as determined by PALM from Dendra2-ZitP expressed in WT and ΔpopZ::Ω cells . Black lines indicate medians . Statistical significance from Mood’s median test: n . s , p>0 . 05; *p<0 . 001 . ( E ) ZitP polar binding times in WT and ΔpopZ::Ω cells , measured via single particle tracking PALM . Error bars indicate 95% confidence interval of the fit to the data ( Figure 3—figure supplement 6D ) . Statistical significance from a 2 sample t-test: p=p<0 . 001 . ( F ) Epifluorescence ( Dendra2 ) and Nomarski ( DIC ) images depicting the localization of Dendra2-ZitP or Dendra2-CpaM in ΔpopZ::Ω , ΔdivJ , divKcs , ΔpleC , ΔcpaE or ΔpodJ cells . Expression of Dendra2-ZitP or Dendra2-CpaM was induced from the chromosomal xylX locus with xylose 4 hours before imaging . Scale bars: 1 µm . ( G ) ( H ) Epifluorescence ( Dendra2 ) and Nomarski ( DIC ) images depicting the localization of the motility-deficient and pilus-proficient Dendra2-ZitPCS variant ( G ) or the motility-proficient and pilus-deficient Dendra2-ZitP1-133 variant ( H ) in ΔzitP cells . Arrow heads pinpoint stalked poles . We induced expression of Dendra2-fusions from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus by the addition of xylose 4 hours before imaging . Scale bars: 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 00710 . 7554/eLife . 18647 . 008Figure 3—figure supplement 1 . Extrinsic determinant for the localization of ZitP and CpaM . ( A ) Epifluorescence ( Dendra2 ) and Nomarski ( DIC ) images depicting the localization of Dendra2-ZitP and Dendra2-CpaM variants in asynchronious ΔzitP or ΔcpaM cells , respectively . We induced expression of Dendra2 fusions expressed from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus . Scale bars: 1 µm . ( B ) Subcellular localisation of Dendra2-ZitP in the ΔcpaM mutant . Cells were imaged in epifluorescence ( GFP channel ) and bright field mode ( DIC ) . We induced expression of Dendra2-ZitP from Pxyl on plasmids integrated at the chromosomal xylX locus by the addition of xylose 4 hours before imaging . Scale bar: 1 µm . ( C ) ( D ) Quantification of Dendra2-ZitP ( C ) or Dendra2-CpaM ( D ) localization states ( diffuse , monopolar or bipolar ) in WT or polarity mutants cells . Dendra2 fusions were expressed from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus . The total cell count ( n ) for each strain is shown above related stacked bars . Values are expressed in percentage of whole cell population . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 00810 . 7554/eLife . 18647 . 009Figure 3—figure supplement 2 . ZitP and CpaM polar localization by PALM . ( A ) PALM images of ZitP or CpaM localization in ΔzitP and ΔcpaM cells , respectively . We induced expression of Dendra2-fusions from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus by the addition of xylose 3 hours before imaging . Scale bars: 0 . 5 µm . ( B ) PALM images of CpaM localization in WT and ΔzitP cells . We induced expression of Dendra2-fusions from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus by the addition of xylose 3 hours before imaging . Scale bars: 1 µm . ( C ) ZitP localization , with zoomed images of poles in WT and ΔpopZ::Ω cells . Scale bar 0 . 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 00910 . 7554/eLife . 18647 . 010Figure 3—figure supplement 3 . Tandem affinity purification of ZitP . Tandem affinity purification ( TAP ) performed on WT cell extracts bearing an empty pCWR512 plasmid ( control ) or the Pvan-zitP-TAP plasmid . After electrophoresis of TAP extracts , the gel was silver-stained as guided by the manufacturer ( SilverQuest , Invitrogen ) . Arrows indicate bands that were extracted and sent for mass spectrometry analyses . They indicate as well the main identified proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 01010 . 7554/eLife . 18647 . 011Figure 3—figure supplement 4 . CpaE localization in ΔzitP and ΔcpaM mutant cells . ( A ) Epifluorescence ( YFP-CpaE ) and Nomarski ( DIC ) images depicting the localization of the pilus component CpaE N-terminally fused to YFP expressed form the native cpaE locus ( yfp-cpaE ) in WT , ΔzitP and ΔcpaM cells . The scale bars represent 1 µm . ( B ) Quantification of YFP-CpaE localization profile ( diffuse , monopolar or bipolar ) in the aforementioned strains . The total cell count ( n ) for each strain is shown above related stacked bars . Values are expressed in percentage of whole cell population . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 01110 . 7554/eLife . 18647 . 012Figure 3—figure supplement 5 . Quantitative analysis of ZitP cluster shape and area . ( A ) Measurement of polar cluster area . Automated clustering of Dendra2-ZitP localization using DBSCAN . Red dots , identified clusters; black crosses , low-density localizations outside of polar clusters . ( B ) Image processing operations for area measurement . Identified clusters ( i ) were converted to a binary image ( ii ) which was then processed via morphological closing ( iii ) to make the cluster area measurement less sensitive to noise and molecule sampling rate . ( C ) Measured circularity , solidity and eccentricity of WT stalked pole , WT other ( swarmer ) pole and polar foci in ΔpopZ::Ω cells . ( D ) Measured area of WT stalked pole , WT other ( swarmer ) pole and polar foci in ΔpopZ::Ω cells compared to the observed area of simulated zero-area clusters . Observed zero-area cluster size is non-zero due to experimental noise . Stars indicate statistical significance: n . s , p>0 . 05; p<0 . 05; p<0 . 01; p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 01210 . 7554/eLife . 18647 . 013Figure 3—figure supplement 6 . Binding time estimation by stroboscopic single particle tracking of ZitP . ( A ) Exemplar histograms of effective on-time ( ie . combination of actual binding lifetime with photobleaching lifetime ) in different time-lapse conditions for Dendra2-ZitP in WT cells . Lines are fits by a single exponential model . ( B ) Observed effective on-time of polar Dendra2-ZitP as a function of time-lapse duration in WT and ΔpopZ cells . Fitting of the data with Eq 4 ( Gebhardt model ) shows large systematic errors . By accounting for finite camera integration time ( Eq 5 ) we obtained good fits to the data . Error bars indicate 1 s . d . ( C ) Simulation showing the effect of finite camera integration time on observed on-time . Observed on-time ( obs . on-time ) shows a strong plateau at a minimum observable on-time , compared to the true on-time ( sim . on-time ) , confirming that the plateau observed experimentally most likely arises from finite camera integration time . By accounting for this effect ( Eq 5 ) we obtained good fits to the data . Error bars indicate 1 s . d . ( D Comparison of observed effective on-time of polar Dendra2-ZitP as a function of time-lapse duration in WT and ΔpopZ cells , showing fitting with the revised binding time model ( Eq 5 ) . Error bars indicate 1 s . d . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 01310 . 7554/eLife . 18647 . 014Figure 3—figure supplement 7 . Intrinsic determinants for ZitP localization and function . ( A ) Epifluorescence ( Dendra2 ) and Nomarski ( DIC ) images depicting the localization profile of Dendra2-CpaM in ΔzitP cells complemented with ZitP-expression plasmids . ZitPWT , ZitPCS or ZitPGAP were expressed from pMT335 ( without vanillate ) . We used empty pMT335 ( - ) as a control . Scale bars: 1 µm . ( B ) Quantification of Dendra2-CpaM localization states ( diffuse , monopolar or bipolar ) in ΔzitP cells expressing ZitPWT , ZitPCS or ZitPGAP from pMT335 ( without vanillate ) . We used empty pMT335 ( - ) as a control . Dendra2 fusions were expressed from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus . The total cell count ( n ) for each strains are shown above related stacked bars . Values are expressed in percentage of whole cell population . ( C ) Reciprocal co-immunoprecipitation ( IP ) of ZitP variants or CpaM protein with polyclonal antibodies to CpaM or ZitP , respectively . Immunoprecipitates and cell lysates were probed for the presence of ZitP or CpaM by immunoblotting ( IB ) . Extracts were made from ΔzitP and ΔcpaM cells expressing ZitPWT , ZitPCS , ZitPGAP or CpaM from pMT335 grown without vanillate . ( D ) Quantification of Dendra2-ZitPCS and Dendra2-ZitPGAP localization states ( diffuse , monopolar or bipolar ) in the ΔzitP cells from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus . The total cell count ( n ) is shown above the stacked bar . Value is expressed in percentage of whole cell population . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 01410 . 7554/eLife . 18647 . 015Figure 3—figure supplement 8 . Effect of DUF3426 on ZitP function . ( A ) Relative β-galactosidase activity of lacZ-based promoter-probe reporters to the pilA and CC_1982 promoter in WT and zitP cells expressing Dendra2-ZitP variants from Pxyl at the xylX locus . Xylose was added to the medium . ( B ) Epifluorescence ( Dendra2 ) and phase images depicting the localization profile of Dendra2-CpaM in ΔzitP cells complemented with pMT463-derived plasmids expressing either WT ZitP ( ZitPWT ) or ZitP1-133 . We used empty pMT463 ( - ) as a control . Scale bars: 1 µm . ( C ) Epifluorescence ( Dendra2 ) and phase contrast overlays depicting the localization of Dendra2- ZitPGAP from the xylX locus in ΔzitP cells . Scale bars: 1 µmDOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 015 PALM analysis disclosed differently shaped and sized complexes of Dendra2-ZitP at each Caulobacter pole . Both Dendra2-ZitP clusters appear extended , suggesting that ZitP multimerization along the polar membrane is spatially restricted ( Figure 3A–B; Figure 3—figure supplement 2A and C ) . Quantification of the 2D area and shape-based analyses ( circularity , solidity and eccentricity ) showed that ZitP clusters extending into the base of the stalk are significantly larger and differently shaped than the extended fluorescent foci lining the cap of the opposite ( swarmer ) pole ( Figure 3B and D; Figure 3—figure supplement 2A , C and 5A–D ) . In further support of the existence of two distinct nanostructures of ZitP at each pole , genetic experiments revealed that different pathways govern ZitP polarization: one requiring PopZ and another operating independently of PopZ . Imaging of Dendra2-ZitP in ΔpopZ cells revealed mainly monopolar foci ( Figure 3B and F; Figure 3—figure supplement 1C and 2C ) , resembling those seen at the pole opposite the stalk in WT cells ( Figure 3B and D; Figure 3—figure supplement 2C and 5C–D ) . Quantitative analysis of the polar residence time using stroboscopic single particle tracking PALM ( Gebhardt et al . , 2013 ) revealed a strong reduction in polar binding times of Dendra2-ZitP in ΔpopZ compared to that of WT cells ( Figure 3E; Figure 3—figure supplement 6A–D ) . Thus , PopZ promotes the formation of a large polar ZitP assembly at the stalked pole , whereas a small complex of ZitP sequesters independently of PopZ at the opposite pole . To identify the determinants within ZitP governing the differential polar localization and to test if they support specific functions , we first constructed a mutant variant of ZitP in which all four zinc-coordinating cysteine residues in the zinc-finger domain ( Bergé et al . , 2016 ) are replaced by serine residues ( henceforth ZitPCS , Figure 1C ) . The motility of ΔzitP cells expressing ZitPCS or Dendra2-ZitPCS is reduced compared to those expressing the WT version of ZitP ( ZitP or Dendra2-ZitP; Figure 2G and K ) . While Dendra2-ZitPCS exclusively localizes to the pole opposite the stalk in ΔzitP cells ( Figure 3G; Figure 3—figure supplement 7A ) , it still supports lysis by φCbK ( Figure 2A ) and CpaC assembly ( Figure 2D ) . ZitPCS supports localization of Dendra2-CpaM to the pole opposite the stalk and co-immunoprecipitation experiments show that it interacts with CpaM ( Figure 3—figure supplement 7B–D ) . ZitPCS also confers ( CpaM-dependent ) firing of CtrA-activated promoters with similar efficiency as WT ZitP ( Figure 3—figure supplement 8A ) . Since Dendra2-CpaM is also still monopolar in ∆popZ cells , zinc-binding within the zinc_ribbon_5 domain is necessary for the interaction between PopZ and ZitP ( Bergé et al . , 2016 ) , but not for CpaM localization/interaction . Thus , inactivation of the zinc-coordinating residues in ZitP effectively mimics the monopolar localization of Dendra2-ZitP in ∆popZ cells and functions as unmodified ZitP with respect to the functions that depend on CpaM . By contrast , the opposite effect was seen when ZitP1-133 , a ZitP variant that lacks the periplasmic DUF3426 but retains the cytoplasmic and TM domains ( residues 1-133 , Figure 1C ) , is expressed in ∆zitP cells . ZitP1-133 supports efficient motility and is polarly localized , but no longer supports pilus function ( i . e . plaque formation by φCbK ) , CpaM localization and efficient CtrA-activated transcription ( Figure 2G and J , Figure 3—figure supplement 8A–B ) . Thus , the periplasmic DUF3426 plays a critical role in promoting pilus assembly through the polar recruitment of CpaM . Support for the notion that DUF3426 function is regulated from sequences N-terminal to the DUF3426 came from a forward genetic screen ( see Materials and Methods ) that led to the identification of ZitPGAP ( Figure 1C ) , a mutant variant in which residues Arg93 and Ala94 preceding the TM domain are deleted . ZitPGAP supports motility ( Figure 2K ) , but neither plaque formation by φCbK , nor CpaC* production ( Figure 2A and D ) . As ZitPGAP still localizes to the cell poles , interacts with CpaM and recruits Dendra2-CpaM ( Figure 3—figure supplements 7A–D and 8C ) , ZitP also acts on pilus biogenesis independently of CpaM localization . Taken together our experiments indicate that function and localization of ZitP can be genetically uncoupled . The periplasmic DUF3426 region is required for pilus biogenesis and CtrA-dependent transcription and it implements these functions via the recruitment of CpaM to the pole opposite the stalk . The zinc_ribbon_5 domain promotes PopZ-dependent localization of ZitP to the stalked pole and efficient swarming motility by an unknown mechanism . Interestingly , in a related study , we recently found that ZitP controls PopZ localization independently of the DUF3426 ( Bergé et al . , 2016 ) . Synchronization studies and genetic experiments with cell cycle mutants showed that ZitP and CpaM polarization is temporally and functionally coordinated with cell cycle progression . Immunoblotting revealed the steady-state levels of ZitP and CpaM to fluctuate during the cell cycle ( Figure 4A ) , exhibiting a trough during the G1→S transition and concomitant loss of polar fluorescence at this time ( Figure 4B–C ) . Consistent with the genetic and cytological hierarchy , ChIP-Seq data shows that the early S-phase regulator GcrA directly promotes ZitP and CtrA expression , while the late S-phase regulator CtrA activates expression of CpaM ( Fiebig et al . , 2014; Fioravanti et al . , 2013; Fumeaux et al . , 2014; Murray et al . , 2013 ) . Moreover , ZitP , CtrA and CpaM abundance is reduced when GcrA is depleted or inactivated ( Figure 4D ) . ZitP expression is also strongly reduced in the absence of the CcrM adenine methyltransferase that methylates adenines at the N6-position in the context of 5’-GANTC-3’ sequences . GANTC methylation is required for efficient recruitment of GcrA to its target promoters ( Fioravanti et al . , 2013; Murray et al . , 2013 ) . 10 . 7554/eLife . 18647 . 016Figure 4 . Cell cycle regulation of ZitP and CpaM localization . ( A ) Immunoblots showing the levels of ZitP , CpaM and master cell cycle regulators along the C . crescentus cell cycle in a synchronized WT population . The upper scheme depicts C . crescentus cell cycle stages . ( B ) ( C ) Epifluorescence ( Dendra2 ) and Nomarski ( DIC ) images depicting the localization of Dendra2-ZitP ( B ) and Dendra2-CpaM ( C ) in synchronized ΔzitP or ΔcpaM cells , respectively . We induced expression of Dendra2 fusions expressed from the xylose-inducible Pxyl promoter on a plasmid integrated at the chromosomal xylX locus . Schematic drawings highlight Dendra2 localizations . After synchronization , cells were resuspended in M2G and imaged every 20 minutes . Scale bars: 1 µm . ( D ) Steady-state levels of ZitP , CpaM , CtrA , GcrA , CcrM and MreB ( control ) in WT , gcrA and ccrM mutant cells . Xylose ( 0 . 3% , xyl ) or glucose ( 0 . 2% , glu ) were supplemented to the medium in order to induce/deplete GcrA in ΔgcrA xylX::Pxyl-gcrA cells . ( E ) Schematic representation of the two Caulobacter cell poles . At the stalked pole , the PopZ matrix promotes the recruitment of ZitP . The Zn2+-bound zinc-finger domain of ZitP prevents ZitP/CpaM association and influences CtrA activity and swarming motility . At the opposite pole , the inactive Zn2+-unbound zinc-finger domain allows the formation of the ZitP/CpaM complex and the export and assemblage of CpaC in the outer membrane ( OM ) independently of PopZ . DOI: http://dx . doi . org/10 . 7554/eLife . 18647 . 016 Additionally , we found that the DivJ-PleC-DivK ( kinase-phosphatase-substrate ) system that regulates cell cycle progression and polar development influences the appearance of polar Dendra2-ZitP and Dendra2-CpaM ( Figure 3F , Figure 3—figure supplement 1C–D ) . Specifically examining the localization in mutants where the phosphoflux is shifted towards the accumulation of the phosphorylated form of the DivK cell fate determinant ( Tsokos and Laub , 2012 ) , we found that such a mutation ( inactivation of the PleC phosphatase , ∆pleC ) promotes ZitP/CpaM polarization as indicated by the bipolar localization of Dendra2-CpaM . By contrast , mutations that have the opposite effect on DivK activity or DivK phosphorylation ( caused by the divKCS or ∆divJ mutation ) , disfavour Dendra2-ZitP ( but not Dendra2-CpaM ) polarization ( Figure 3F , Figure 3—figure supplement 1C–D ) . Thus , polar reprogramming of ZitP and CpaM is deeply integrated into the Caulobacter cell cycle through conserved components of the α-proteobacterial cell cycle ( Brilli et al . , 2010 ) . The pole-specific and distinctly shaped assemblies of ZitP are governed via independent localization pathways and linked with functional specialization ( Figure 4E ) . While ZitP acts on pilus assembly by recruiting CpaM and , subsequently , the CpaC pilus channel to the pole opposite the stalk ( 1B and 4E ) , CpaM is also required for efficient activation of CtrA-dependent promoters by an unknown mechanism . A similar reduction in CtrA-dependent transcription occurs in ∆zitP cells that are unable to localize CpaM . While diminished CtrA activity can undermine motility by reducing the number of motile G1-phase cells in the population ( Sanselicio et al . , 2015; Sanselicio and Viollier , 2015 ) , ZitP affects motility in another way , since ∆zitP cells are diminished in motility compared to ∆cpaM cells . Moreover , ectopic induction of the alarmone ( p ) ppGpp reinforces CtrA abundance and activity ( Boutte et al . , 2012; Gonzalez and Collier , 2014; Lesley and Shapiro , 2008; Ronneau et al . , 2016; Sanselicio and Viollier , 2015 ) , but only modestly improves the motility of ∆zitP cells . Such a motility defect also manifests when ZitPCS , a variant that no longer localizes to the stalked pole , is expressed in ∆zitP cells . How ZitP promotes swarming motility from the stalked pole is unclear , but there is precedence of other regulators ( SpmX/Y and CpdR ) that localize exclusively to the stalked pole and affect Caulobacter motility indirectly by regulating cell cycle factors ( Janakiraman et al . , 2016; McGrath et al . , 2006; Radhakrishnan et al . , 2008 ) . Moreover , SpmX and CpdR interact with PopZ directly and their localization is compromised in the absence of PopZ ( Bowman et al . , 2010; Holmes et al . , 2016 ) . It is therefore conceivable that ZitP also affects motility indirectly from the stalked pole , possibly via cell cycle regulation , flagellar performance and/or polarity . The fact that the motility defect of ∆zitP cells can be restored by compensatory mutations in a switch component ( FliG ) of the flagellar motor ( Kojima and Blair , 2004 ) , suggests that flagellar performance , reversals or timing ( i . e . the length of flagellation in the cell cycle ) could be altered by the ∆zitP mutation . Zinc-finger domain proteins other than ZitP may be implicated in linking motility and polarity . The gliding motility protein AgmX confers a flagellum- and pilus-independent form of surface motility in Myxococcus xanthus ( Nan et al . , 2010 ) , a δ-proteobacterium that periodically reverses the polarity of movement . Since AgmX also harbors a related N-terminal Zinc-finger domain , at least two related zinc-finger domains control different types of motility . This is intriguing and hints at a potentially important and conserved role of such zinc-finger domain proteins in developmental processes that rely on protein polarization in bacteria and polar matrix proteins such as PopZ to interact with them . In a complementary study , we additionally show in vitro and in vivo that zinc-bound ZitP binds PopZ directly and regulates PopZ localization without the periplasmic DUF3426 domain ( Bergé et al . , 2016 ) , suggesting that this activity in ZitP may underlie the aforementioned CtrA-independent role in motility . The conservation of ZitP , CpaM ( Figure 1D ) and PopZ orthologs ( Bowman et al . , 2010 ) in distant α-proteobacterial lineages that reside in different ecological niches hints that the functions that these proteins control are not unique to the Caulobacter branch . Indeed , we describe an interaction between ZitP and PopZ in several distinct α-proteobacterial lineages ( Bergé et al . , 2016 ) . On a more general scale , our work suggests that pole-specific functions conferred by bipolar regulators may be commonly used in bacteria and possibly eukaryotes . Such mechanisms could be relevant for toggle proteins , moonlighting/trigger enzymes ( Commichau and Stulke , 2015 ) and other bifunctional regulators ( Radhakrishnan and Viollier , 2012 ) that have more than one biochemical activity and function , for example kinase-phosphatases or synthase-hydrolases of cyclic-di-GMP sequestered to both cell poles ( Boyd , 2000; Kazmierczak et al . , 2006; Tsokos and Laub , 2012 ) . In sum , the functional and topological versatility of ZitP illustrates how a conserved regulator is used to coordinate multiple functions from different locations and structures in the same cell , relying on distinct protein domains and partners to control localization or to implement function . As these functions and polar remodelling events are coordinated with cell cycle progression in Caulobacter via conserved cell cycle proteins , it is likely that superimposed temporal layers similarly act on ZitP and CpaM orthologs in other α-proteobacterial cell cycles . Caulobacter crescentus NA1000 and derivatives were grown at 30°C in PYE or in M2 salts plus 0 . 2% glucose ( M2G ) supplemented with 0 . 4% liquid PYE ( Ely , 1991 ) . Escherichia coli S17-1 , S17-1 λpir and EC100D ( Epicentre Technologies , Madison , WI ) were cultivated at 37°C in LB . We added 1 . 5% agar to PYE plates , and motility was assayed on PYE plates containing 0 . 3% agar . We added D-xylose ( 0 . 3% except if otherwise stated ) , glucose ( 0 . 2% ) , sucrose ( 3% ) , kanamycin ( solid , 20 µg/ml; liquid , 5 µg/mL ) , tetracycline ( 1 µg/mL ) , spectinomycin ( liquid , 25 µg/mL ) , spectinomycin/streptomycin ( solid , 30 and 5 µg/mL , respectively ) , apramycin ( 10 µg/mL ) , gentamycin ( 1 µg/mL ) and nalidixic acid ( 20 µg/mL ) , as required . Swarmer cell isolation , electroporation , biparental mating , and bacteriophage φCr30-mediated generalized transduction were performed as described before ( Chen et al . , 2005; Ely , 1991; Simon et al . , 1983; Viollier and Shapiro , 2003 ) . Bacterial strains , plasmids , and oligonucleotides used in this study are listed and described in supplementary tables . β-Galactosidase assays were performed at 30°C as described previously ( Huitema et al . , 2006; Viollier and Shapiro , 2003 ) . Experimental values represent the averages ( standard error of the mean , SEM ) of at least three independent experiments . To image C . crescentus , overnight cultures were diluted in fresh PYE , xylose was added ( 0 . 3% final concentration ) , and the cells were grown for 3 hours to mid-exponential phase ( OD ( 660 ) ~ 0 . 4 ) . Two uL of culture was placed on a agarose pad containing PYE . The agarose pad was mounted in a silicone gasket ( Grace Biolabs 103280 ) sandwiched between two microscope coverslips to minimize shrinkage of the agarose . The temperature of the microscope enclosure during experiments was 24°C . Images were acquired using a previously described custom built PALM microscope ( Holden et al . , 2014 ) . Fluorescent proteins were excited at 560 nm , and photoactivation was induced at 405 nm at ~ 0–16 W/cm2 . For PALM images of Dendra2-ZitP in C . crescentus , cells were imaged at an exposure time of 10 milliseconds for 10 , 000 frames , and an excitation intensity of ~4 kW/cm2 . For stroboscopic single particle tracking PALM measurement of ZitP binding time , cells were imaged at an exposure time of 30 milliseconds , with a variable interval between each frame , at an excitation intensity of ~1 kW/cm2 . PALM localizations were accumulated in a 2D histogram; the resulting image was blurred with a 2D Gaussian of radius 15 nm to reflect the localization uncertainty of the measurement . The image was gamma adjusted to 0 . 5 to compensate for the large dynamic range of the image , and the ‘Red Hot’ ImageJ colormap was applied . Binding time , τoff , of ZitP to the C . crescentus poles was determined via stroboscopic single particle tracking PALM ( Gebhardt et al . , 2013; Manley et al . , 2008 ) . Under these conditions , Dendra2 bleached under continuous illumination with a photobleaching lifetime , τb , on the order of 50 milliseconds . Since rapid diffusion means that Dendra2-ZitP is only visible when bound to the membrane , and since photobleaching will shorten the observed binding time , the effective on-time of a single Dendra2-ZitP molecule , τeff , will be the convolution of the photobleaching lifetime , τb , and the binding lifetime τoff , ( 1 ) τeff−1=τoff−1+τb−1 , Effective on-time was measured by combining individual Dendra2-ZitP localizations in adjacent frames into tracks ( Crocker and Grier , 1996 ) , and fitting a single exponential model to the observed the track length distribution ( Figure 3—figure supplement 6A ) . In order to measure binding times longer than the photobleaching lifetime , the photobleaching lifetime of the fluorescent protein may be artificially extended by using stroboscopic illumination , introducing large gaps between short periods of illumination . This increases the effective bleaching lifetime to: ( 2 ) τbl′=τblτtlτint , where τtl is duration of the gap ( time lapse/strobe interval ) , τint is camera integration time . By measuring the effective on-time for multiple different stroboscopic illumination times , τtl , and performing a fit of: ( 3 ) τeff= ( τoff−1+τintτblτtl ) −1 , to the data , both the binding time and photobleaching lifetime may be calculated ( Gebhardt et al . , 2013 ) ( Figure 3—figure supplement 6B and C Model 1 ) . We performed non-linear least squares fitting of the raw τeff data directly to Eq . 3 , instead of calculating the quantity τtl/τeff and performing a linear fit as proposed by Gebhardt and coworkers ( Gebhardt et al . , 2013 ) , since the inverse transform proposed results in a non-linear transformation of the sample error distribution incompatible with least squares fitting . We observed that for stroboscopic illumination times significantly greater than the binding time , the data appeared to transition from the hyperbolic relationship predicted by Eq . 3 to a zero-gradient plateau ( Figure 3—figure supplement 6B ) , giving very poor fits between Eq 3 and the data , especially for the ΔpopZ strain which appeared to have a shorter Dendra2-ZitP binding lifetime ( Figure 3—figure supplement 6B ) . We hypothesized that this was due to an inability to accurately estimate effective on-time when molecules bind and unbind in a time significantly less than the duration of a single strobing interval ( since the observed track length will almost always equal 1 frame ) . We confirmed this hypothesis by performing the stroboscopic tracking analysis on simulated data ( Figure 3—figure supplement 6C ) . We simulated timetraces of molecules binding/unbinding with finite bleaching lifetimes , and measured the observed on-time for each simulated molecule by fitting a single exponential to the on-time histogram as above . We observed as hypothesized that the observed off-times showed a sharp plateau for long-strobe intervals due to the finite integration time of the measurement , giving a poor fit of Eq 3 to the data ( Figure 3—figure supplement 6C ) . In order to correct for this , we modified the fitting model to include a minimum measurable on-time plateau: ( 4 ) τeff= ( τoff−1+τintτblτtl ) −1 , τtl>τtlmin , τeff=τtlmin , otherwise . Use of the modified model allowed us to obtain accurate fits to the entire simulated dataset ( Figure 3—figure supplement 6C; Eq 4 ) . We therefore used our updated model to fit the experimental data ( Figure 3E and Figure 3—figure supplement 6B ) and to calculate the observed binding times ( Figure 3—figure supplement 6D ) . This gave a much better fit to the data , both at late and early strobe intervals . Notably , independent fits to the WT and ΔpopZ datasets gave similar observed τtlmin of ~0 . 4 frames , supporting the use of the updated model . In order to estimate the area of Dendra2-ZitP polar complexes , observed localizations were clustered based on local density using DBSCAN ( Endesfelder et al . , 2013; Ester et al . , 1996 ) . Identified clusters were converted to PALM images binarized , and morphologically closed ( Figure 3—figure supplement 5Bi-iii ) . By performing morphological closing on the binary image , we obtained segmented clusters ( Figure 3—figure supplement 5Biii ) which were less sensitive to noise and better reflected the visually estimated extent of the non-segmented cluster . For each identified cluster , the area of the segmented cluster was calculated . For the NA1000 xylX::Pxyl-dendra2-zitP strain , clusters were visually identified as belonging to the stalked or flagellar poles based on the PALM and phase contrast images of the region . For the ΔpopZ::Ω xylX::Pxyl-dendra2-zitP strain , there was no clear difference in pole morphology , so the cluster area for cells was calculated without discriminating poles . Measurement noise means that the measured area of even a zero-area cluster will be larger than zero ( and approximately proportional to the localization uncertainty ) . To test whether Dendra2-ZitP formed an extended polar complex , we compared the area of ZitP clusters to the measured area of simulated zero-area clusters by generating simulated datasets containing localizations coming from a point source , with photon count , background noise and total number of localizations equal to the median values of either the WT or ΔpopZ::Ω datasets ( Figure 3—figure supplement 5D ) . The cluster area of the simulated datasets was then calculated as above . We also calculated the following shape based metrics to further quantify the differences in pole shape: circularity , solidity and eccentricity ( Figure 3—figure supplement 5C ) . Circularity measures similarity of a shape to a circle , C= 4πAp2 , where A is shape area and p is perimeter . Solidity measures the extent to which a shape is convex or concave , S= AH , where A is shape area and H is the convex hull area of the shape . Eccentricity measures how elongated a shape is , E= ab , where a is the length of the minor axis and b is the length of the major axis . Since the observed distributions showed significant non-normality , statistical significance was assessed by the non-parametric test , Mood’s median test . Stars on Figure 3D and Figure 3—figure supplement 5C indicate statistical significance: n . s , p>0 . 05; p<0 . 05; p<0 . 01; **p<0 . 001 . The stalked and the other ( swamer ) pole foci in WT showed statistically significant differences ( p<0 . 001 ) in area , circularity and solidity , supporting the conclusion that ZitP forms distinct polar assemblies . The WT stalked pole showed statistically significant differences ( p<0 . 001 ) to the ΔpopZ::Ω mutant foci for area , circularity , solidity and eccentricity , supporting the conclusion that PopZ specifically promotes the formation of large polar assemblies at the stalked pole . A himar1-based transposon mutagenesis of the NA1000 ( wild-type , WT ) strain was done using the E . coli S17-1 λpir strain containing the himar1-delivery plasmid pHPV414 ( Viollier et al . , 2004 ) . The mutant library was selected on plates containing nalidixic acid and kanamycin embedded in top agar containing φCbK . Colonies emerging from this selection were pooled . We then created generalized transducing lysate from this pool using phage φCr30 and transduced it into strain PV14 ΔpilA-cpaF::Ωaac3 ( conferring resistance to aparamycin ) , selecting for apramycin and kanamycin resistant transductants to eliminate himar1 insertions in the pilA-cpaF locus . The transductants were pooled and a generalized transducing lysate was prepared from this pool using φCr30 . This new lysate was then used to transduce NA1000 to kanamycin resistance and the resulting clones were individually tested for resistance to φCbK . The himar1 insertion site mapping of φCbK–resistant himar1 mutants was done as described before ( Viollier et al . , 2004 ) . To isolate the zitPGAP mutation , we generated a mutant library of zitP alleles by electroporating pMT335-zitP into the mutator E . coli XL1-Red strain . We collected and pooled over 20 , 000 clones for plasmid extraction and we electroporated the plasmid library into the ΔzitP mutant . We incubated the electroporated cells during two hours for regeneration and next added φCbK for one hour in order to eradicate clones that bear a mutated zitP allele restoring effective phage infection . Finally , we plated cells on soft ( 0 . 3% swarming ) agar to evaluate the motility properties . We picked and streaked out motile clones for amplification and plasmid extraction and introduced the plasmids back into a ΔzitP background in the perspective to confirm the motility-proficient and φCbK resistant phenotypes . We isolated a unique plasmid , pMT335-zitPGAP , which bears the zitPGAP allele ( deletion of the Arg93 and Ala94 in the ZitP protein ) . Protein samples were separated by SDS-PAGE and blotted on PVDF ( polyvinylidenfluoride ) membranes ( Merck Millipore ) . Membranes were blocked for 1 hour with Tris-buffered saline , 0 . 05% Tween 20 ( TBST ) , and 5% dry milk and then incubated for an additional 1 hour with the primary antibodies diluted in TBST , 5% dry milk . The membranes were washed 4 times for 5 minutes in TBST and incubated for 1 hour with the secondary antibody diluted in TBST , and 5% dry milk . The membranes were finally washed again 4 times for 5 minutes in TBST and revealed with Immobilon Western Blotting Chemoluminescence HRP substrate ( Merck Millipore ) and Super RX-film ( Fujifilm ) . Rabbit antisera were used at the following dilutions: anti-CtrA ( 1:10 , 000 ) , anti-PilA ( 1:10 , 000 ) , anti-FljK ( 1:50 , 000 ) , anti-CpaC ( 1:5000 ) , anti-ZitP ( 1:5000 ) , anti-CpaM ( 1:5000 ) and anti-GcrA ( 1:2000 ) . HRP-conjugated Anti-rabbit secondary antibody was used at 1:20 , 000 dilution ( Jackson ImmunoResearch , USA ) . PYE or M2G cultivated cells in exponential growth phase were immobilized using a thin layer of 1% agarose . Fluorescence and DIC images were taken with an Alpha Plan-Apochromatic 100x/1 . 46 DIC ( UV ) VIS-IR oil objective on an Axio Imager M2 microscope ( Zeiss ) with 488 nm laser ( Visitron Systems GmbH , Puchheim , Germany ) and a CoolSnap HQ ( Boutte et al . , 2012 ) camera ( Photometrics ) controlled through Metamorph V7 . 5 ( Universal Imaging ) . Images were processed using Image J software . Quantifications were done by manually numbering cells in the diffuse , monopolar or bipolar state . The PCR-amplified zitPCterm and cpaMΔTM genes were cloned into the pET28a vector ( Novagen ) . The His6-ZitPCterm and His6-CpaMΔTM recombinant proteins were overexpressed in E . coli strain Rosetta and purified in standard native conditions on Ni2+-NTA agarose as described previously to raise rabbit polyclonal IgGs in New Zealand White rabbits ( Josman LLC , Napa , CA ) . We followed the TAP procedure as was previously described ( Puig et al . , 2001 ) . When a 1 L-culture reached OD660 between 0 . 4 and 0 . 6 in the presence of 50 mM vanillate , cells were harvested by centrifugation ( 6000xg , 10 min ) . We washed the pellet in 50 mL of buffer I ( 50mM sodium phosphate pH 7 . 4 , 50 mM NaCl , 1 mM EDTA ) and lysed for 15 minutes at room temperature in 10 mL of buffer II ( buffer I + 0 . 5% n-dodecyl-β-D-maltoside , 10mM MgCl2 , two protease inhibitor tablets [Complete EDTA-free , Roche] per 50 mL of buffer II , 1x Ready-Lyse lysozyme [Epicentre] , 500U of DNase I [Roche] ) . Cellular debris was removed by centrifugation ( 7000xg , 20 minutes , 4°C ) . The supernatant was incubated for 2 hours at 4°C with IgG Sepharose beads ( GE Healthcare Biosciences ) that had been washed once with IPP150 buffer ( 10 mM Tris-HCl pH 8 , 150 mM NaCl , 0 . 1% NP40 ) . After incubation , the beads were washed at 4°C three times with 10 mL of IPP150 buffer and once with 10 mL of TEV cleavage buffer ( 10 mM Tris-HCl pH 8 , 150 mM NaCl , 0 . 1% NP40 , 0 . 5 mM EDTA , 1 mM DTT ) . The beads were then incubated overnight at 4°C with 1 mL of TEV solution ( TEV cleavage buffer with 100 U of TEV protease per ml [Promega] ) to release the tagged complex . 3 mM CaCl2 was then added to the solution . The sample with 3 mL of calmodulin-binding buffer ( 10 mM β-mercaptoethanol , 10 mM Tris-HCl pH 8 , 150 mM NaCl , 1 mM magnesium acetate , 1 mM imidazole , 2 mM CaCl2 , 0 . 1% NP40 ) was incubated for 1 hour at 4°C with calmodulin beads ( GE Healthcare Biosciences ) that previously had been washed once with calmodulin-binding buffer . After incubation , the beads were washed three times with 10 mL of calmodulin-binding buffer and eluted five times with 200 µL IPP150 calmodulin elution buffer ( calmodulin-binding buffer substituted with 2 mM EGTA instead of CaCl2 ) . Amicon Ultra-4 spin columns ( Ambion ) were used to concentrate eluates . Eluates were analyzed by SDS-PAGE and stained with silver using the Biorad Silver Stain Plus kit ( Biorad , USA ) . We then cut specific bands and directly sent the gel slices to the Taplin Biological Mass Spectrometry Facility ( Harvard Medical School , Boston , USA ) for mass spectrometric analyses . Cells were harvested from a 50 mL-culture ( OD ( 660 nm ) between 0 . 4–0 . 6 ) by centrifugation at 5000xg for 10 minutes . We washed the cell pellet in 10 mL of buffer I ( 50mM Tris-HCl ( pH 7 . 5 ) ; 50 mM NaCl; 1mM EDTA ) , centrifuged the cell again and resuspended in 1 mL of buffer II ( buffer I plus 0 . 5% n-dodecy-β-D-maltoside; 10 mM MgCl2; EDTA-free protease inhibitors ) . We incubated the mixture for 15 minutes with 5000 units of Ready-Lyse lysozyme ( Epicentre ) , and 30 units of DNase I ( Roche ) . Cellular debris were removed by centrifugation at 10 , 000xg for 3 minutes at 4°C . We cleared the supernatant by incubation for 1 hour at 4°C with Protein-A agarose beads ( Roche ) previously washed three times with 500 µL of buffer II . We removed agarose beads by centrifugation and added to the pre-cleared solution polyclonal IgG rabbit serum for 90 min at 4°C ( dilution 1:500 ) . Next , we trapped for 1 hour at 4°C the antibodies-proteins complexes with the addition of Protein-A agarose beads ( Roche ) previously washed three times with 500 µL of buffer II . The samples were then centrifuged at 3000xg for 1 minute at 4°C and the supernatant was removed . The beads were washed once with buffer I plus 0 . 5% n-dodecy-β-D-maltoside , three times with 500 µL of wash buffer ( 10 mM Tris-HCl at pH 7 . 5; 150 mM NaCl; 0 . 1% n-dodecy-β-D-maltoside ) and finally resuspended in 70 µl SDS sample buffer ( 50 mM Tris–HCl at pH 6 . 8 ) , 2% SDS , 10% glycerol , 1% β-mercaptoethanol , 12 . 5 mM EDTA , 0 . 02% Bromophenol Blue ) , heated to 95°C for 10 minutes and stored at −20°C . Swarming properties were assessed with 5 µl-drops of overnight culture spotted on PYE soft agar plates ( 0 . 3% agar ) and grown for 24 hours . Phage susceptibility assays were conducted by mixing 500 µL of overnight culture in 6 mL soft PYE agar and overlaid on a PYE agar plate . Upon solidification of the soft ( top ) agar , we spotted 5 µL-drops of serial dilution of phages ( φCbK or φCr30 ) and scored for plaques after one day incubation at 4°C . We centrifuged 5 mL mid-log phase cultures of WT or mutant strains and resuspended them in 700 µl of PYE . Then , we pumped in and out ( 10x ) the cells into a syringe endowed with a thin diameter needle . We centrifuged the shear-stressed cells to remove cells debris and collected 200 µL of each supernatant . We added SDS-blue straining and loaded samples on SDS-PAGE gels . To determine the adsorption rate of φCbK , Caulobacter crescentus NA1000 and derivatives were first grown overnight in M2G medium at 30°C and then re-started in fresh M2G at 30°C with shaking until the bacterial cell culture reached an OD660 value of 0 . 4 ( 0 . 4 × 108 CFU/ml ) . Then cell cultures were infected by 0 . 02 multiplicity of φCbK infection ( MOI: ratio of phage to bacteria number ) . The mixtures were incubated at 30°C without shaking for phage adsorption , followed by separation of unbound phages by centrifugation at 13 , 000 rpm in specified time points up to 30 minutes . Supernatants were immediately supplemented by the addition of chloroform ( 1/20 of cell culture volume ) and mixed vigorously to kill remaining bacterial cells . A control tube containing only φCbK ( equivalent to 0 . 02 MOI ) was maintained in parallel for the duration of the experiment and used as reference to control the initial phage plaque-forming units ( pfu ) titer . A 50 µL of the phage supernatant from each tube was mixed with 200 µL of Caulobacter crescentus NA1000 culture at log phase and incubated without shaking at room temperature for 10 minutes to allow adsorption . Infected cells were added to 6 mL of soft PYE agar ( 0 . 5% ) and immediately overlaid on 1 . 5% PYE agar plates . Plates were incubated at 30°C for 24 hours , when pfu were visible . The φCbK adsorption value ( in% of the initial phage pfu titer ) was calculated . Values are the mean of three biological replicates; error bars represent data ranges . Cells in exponential growth phase ( OD660nm=0 . 3–0 . 6 ) cultivated in PYE , were fixed in ice-cold 77% ethanol solution . Fixed cells were re-suspended in FACS staining buffer , pH 7 . 2 ( 10 mM Tris-HCl , 1 mM EDTA , 50 mM NaCitrate , 0 . 01% Triton X-100 ) and then treated with RNase A ( Roche ) at 0 . 1 mg mL−1 for 30 minutes at room temperature . Cells were stained in FACS staining buffer containing 0 . 5 μM of SYTOX Green nucleic acid stain solution ( Invitrogen ) and then analysed using a BD Accuri C6 flow cytometer instrument ( BD Biosciences ) . Flow cytometry data were acquired and analysed using the CFlow Plus V1 . 0 . 264 . 15 software ( Accuri Cytometers Inc . ) . 20 , 000 cells were analysed from each biological sample . Experimental values represent the averages of three independent experiments . We spotted several 5 µL-drops of ∆zitP overnight culture on soft agar plates and waited for flares spreading out the bulk of cells . Flares were peaked out and streaked on fresh agar plates for amplification and subsequently challenged for motility in comparison to WT and ∆zitP strains . Motility-proficient clones were sent for Illumina HiSeQ 2000 sequencing ( Fasteris , www . fasteris . com/ ) . Genomes were compared to NA1000 genome and we identified a single mutation in the fliG gene ( D306G ) . In order to backcross the fliGD306G allele in different backgrounds , the suppressor strain was electrotransformed with the suicide vector pNTPS138-hook and selected on kanamycin-supplemented plates for single crossing-over in close vicinity of the fliG locus . We prepared lysate of this strain , transduced the fliGD306G-linked pNTPS138 into WT and ∆zitP cells and screen by sequencing for clones harbouring the fliGD306G allele . Finally , we grew up the strain without any antibiotic and selected for plasmid excision by plating an overnight culture on sucrose .	Living cells become asymmetric for many different reasons and how they do so has been a long-standing question in biology . In some cells , the asymmetry arises because a given protein accumulates at one side of the cell . In particular , this process happens before some cells divide to produce two non-identical daughter cells that then go on to develop in very different ways – which is vital for the development of almost all multicellular organisms . The single-celled bacterium Caulobacter crescentus also undergoes this type of asymmetric division . The polarized Caulobacter cell produces two very different offsprings – a stationary cell and a nomadic cell that swims using a propeller-like structure , called a flagellum , and has projections called pili on its surface . Before it divides asymmetrically , the Caulobacter cell must accumulate specific proteins at its extremities , or poles . Two such proteins are ZitP and CpaM , which appear to have multiple roles and are thought to interact with other factors that regulate cell division . However , little is known about how ZitP and CpaM become organized at the poles at the right time and how they interact with these regulators of cell division . Mignolet et al . explored how ZitP becomes polarized in Caulobacter crescentus using a combination of approaches including biochemical and genetic analyses and very high-resolution microscopy . This revealed that ZitP accumulated via different pathways at the two poles and that it formed distinct structures at each pole . These structures were associated with different roles for ZitP . While ZitP recruited proteins , including CpaM , required for assembly of pili to one of the poles , it acted differently at the opposite pole . By mutating regions of ZitP , Mignolet et al . went on to show that different regions of the protein carry out these roles . Further experiments demonstrated that regulators of the cell division cycle influenced how ZitP and CpaM accumulated and behaved in cells , ensuring that the proteins carry out their roles at the correct time during division . These findings provide more evidence that proteins can have different roles at distinct sites within a cell , in this case at opposite poles of a cell . Future studies will be needed to determine whether this is seen in cells other than Caulobacter including more complex , non-bacterial cells .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "short", "report", "cell", "biology" ]	2016	Functional dichotomy and distinct nanoscale assemblies of a cell cycle-controlled bipolar zinc-finger regulator
The neuronal microtubule cytoskeleton underlies the polarization and proper functioning of neurons , amongst others by providing tracks for motor proteins that drive intracellular transport . Different subsets of neuronal microtubules , varying in composition , stability , and motor preference , are known to exist , but the high density of microtubules has so far precluded mapping their relative abundance and three-dimensional organization . Here , we use different super-resolution techniques ( STED , Expansion Microscopy ) to explore the nanoscale organization of the neuronal microtubule network in rat hippocampal neurons . This revealed that in dendrites acetylated microtubules are enriched in the core of the dendritic shaft , while tyrosinated microtubules are enriched near the plasma membrane , thus forming a shell around the acetylated microtubules . Moreover , using a novel analysis pipeline we quantified the absolute number of acetylated and tyrosinated microtubules within dendrites and found that they account for 65–75% and ~20–30% of all microtubules , respectively , leaving only few microtubules that do not fall in either category . Because these different microtubule subtypes facilitate different motor proteins , these novel insights help to understand the spatial regulation of intracellular transport . The extended and polarized morphology of neurons is established and maintained by the cytoskeleton ( Stiess and Bradke , 2011; Bentley and Banker , 2016 ) . One of the functions of the microtubule cytoskeleton is to provide a transport network inside the neurons long axon and branched dendrites ( Kapitein and Hoogenraad , 2015; Burute and Kapitein , 2019 ) . Directional transport is enabled by the structural polarity of microtubules , which is recognized by motor proteins that drive transport to either the minus end ( dynein ) or plus end ( most kinesins ) . Using this network , intracellular cargos attached to microtubule-based motor proteins ( kinesins and dynein ) can be shipped between different neuronal compartments ( Hirokawa et al . , 2010; Kapitein and Hoogenraad , 2011 ) . To facilitate proper delivery , transport is regulated at multiple levels . For example , molecular motors , motor adaptor proteins , and cargos themselves undergo tight biochemical regulation in response to changes in metabolic state and extracellular cues ( Hirokawa et al . , 2010; Tempes et al . , 2020 ) . An equally important component is the composition and spatial distribution of microtubule tracks , which is the main subject of this study . Previous work has revealed that the affinity of motors for the microtubule lattice can be modulated by microtubule-associated proteins ( MAPs ) or by post-translational modifications ( PTMs ) of tubulin ( Atherton et al . , 2013; Sirajuddin et al . , 2014; Jurriens et al . , 2021; Monroy et al . , 2020; Park and Roll-Mecak , 2018; Janke and Magiera , 2020 ) . For example , kinesin-1 motors move preferentially on microtubules marked by acetylation and detyrosination ( Cai et al . , 2009; Dunn et al . , 2008 ) , while kinesin-3 prefers tyrosinated microtubules ( Guardia et al . , 2016; Tas et al . , 2017; Lipka et al . , 2016 ) . When tubulin is incorporated into the microtubule lattice , it carries a genetically encoded C-terminal tyrosine , which can subsequently be proteolytically removed to yield detyrosinated microtubules ( Kapitein and Hoogenraad , 2015; Janke and Magiera , 2020 ) . Therefore , tyrosinated tubulin can be regarded as a marker for freshly polymerized microtubules . Such microtubules undergo cycles of growth and shrinkage and are therefore referred to as dynamic microtubules . Following polymerization , tubulins can also acquire new chemical groups through post-translational modifications , such as acetylation and polyglutamylation . Additionally , detyrosinated tubulin can be further proteolytically processed at the C-terminal to yield delta 2-tubulin ( Paturle-Lafanechère et al . , 1991 ) . Such modifications often accumulate on microtubules that are long-lived and resist cold-induced or drug-induced depolymerization , which are therefore termed stable microtubules . Despite many biochemical and physiological studies underpinning the importance of various microtubule modifications ( Sirajuddin et al . , 2014; Janke and Magiera , 2020; Nekooki-Machida and Hagiwara , 2020; Roll-Mecak , 2019 ) , little is known about the relative abundance and spatial organization of different microtubule subsets within neurons . In earlier work , we revealed that stable and dynamic microtubules in dendrites are organized differently and often have opposite orientations , explaining why kinesin-3 can drive efficient anterograde transport in dendrites , unlike kinesin-1 ( Tas et al . , 2017 ) . Nonetheless , many important aspects of the neuronal microtubule array have remained unexplored . First , do tyrosination and acetylation mark two clearly defined subsets or are there also subsets that are both highly tyrosinated and acetylated ? Furthermore , what is the three-dimensional organization of different subsets and their relative abundance ? Finally , do acetylation and tyrosination together mark all microtubules or are there additional subsets that carry neither of these groups ? Although microtubule organization in dendrites has previously been studied using electron microscopy ( Baas et al . , 1988; Kubota et al . , 2011 ) , this method is difficult to combine with selective markers and therefore cannot robustly identify and map microtubule subsets throughout dendrites . Here , we use a variety of super-resolution techniques to explore the quantitative and spatial distribution of microtubule subsets in dendrites . We find that acetylated microtubules accumulate in the core of the dendritic shaft , surrounded by a shell of tyrosinated microtubules . High-resolution microscopy enabled frequent detection of individual microtubule segments , which could be used to carefully quantify the tyrosination and acetylation levels of these segments . This revealed that these two modifications are anti-correlated and define two distinct microtubule subsets . In addition , it enabled us to estimate the absolute number of acetylated and tyrosinated microtubules within dendrites , which revealed that they account for 65–75% and ~20–30% of all microtubules , respectively , leaving only few microtubules that do not fall in either category . Together , these results provide new quantitative insights into the uniquely organized dendritic microtubule network and help to understand the spatial regulation of neuronal transport . We started by mapping the spatial distribution of acetylated and tyrosinated microtubules throughout the dendrite using both 2D and 3D stimulated emission depletion ( STED ) microscopy . Consistent with our earlier observations , this revealed that acetylated microtubules in DIV9 neurons tend to be distributed closer to the central axis of the dendrite , while the tyrosinated microtubules seem to be enriched at the outer surface , close to the membrane ( Figure 1A , B ) . To quantify this observation , we built radial distribution maps of the intensity of acetylated and tyrosinated microtubules along the dendrite , which could be averaged to quantitatively describe the radial distribution of these two subsets ( Figure 1C , D , Video 1 ) . This revealed that the differential spatial organization was maintained along the length of an individual dendrite ( Figure 1D , E ) and for dendrites with various diameters ( Figure 1—figure supplement 1 ) , independent of STED imaging modality ( Figure 1F , G ) . Whereas the intensity of both total tubulin and these two subsets decreased as the dendrite became smaller ( displaying a quadratic dependence on dendrite width ) , the relative intensity of both subsets was nearly constant along the dendrite ( Figure 1—figure supplement 3 ) . Besides acetylation and tyrosination we also tried detecting detyrosinated tubulin . However , whereas the antibody that we used did reveal clear overlap between acetylation and detyrosination in non-neuronal cells , we did not obtain reliable staining in neurons , ( Figure 1—figure supplement 4 ) . Delta 2-tubulin was found on a subset of microtubules that overlapped with the most central part of the acetylated bundles , but was not further explored here ( Figure 1—figure supplement 4 ) . We next attempted to quantify the absolute number of acetylated and tyrosinated microtubules . This cannot be achieved by just comparing fluorescent intensities , because staining efficiencies and fluorophore properties differ for each subset and need to be rescaled using single microtubules of each type as a reference . However , we were unable to distinguish individual microtubules within axons or dendrites , since ( as it is known from electron microscopy studies ) the distance between adjacent microtubules is often smaller than the resolution of STED ( Figure 1—figure supplement 3; Baas et al . , 1988; Baas et al . , 1989 ) . Microtubules in the cell body , however , were more dispersed and here individual microtubules could often be resolved ( Figure 2A , B ) . We therefore set out to develop a workflow to enable the robust quantification of acetylation and tyrosination levels on individual microtubules , which could subsequently be used to determine the number of acetylated and tyrosinated microtubules in dendrites . As a first step , we performed three-color STED microscopy in the soma and dendrites of DIV9 cells to detect total ( alpha- ) tubulin , tyrosinated tubulin , and acetylated tubulin . For the analysis of singe-microtubule intensities , we used a subvolume of the cell body just below the nucleus ( Figure 2A , B ) , where the majority of microtubules were located in the x , y plane and confined to a relatively thin flat layer . We established a custom curvilinear structure detection algorithm to detect filament segments in all three channels and to quantify their background-corrected fluorescence intensity for all channels ( Figure 2B–D , Figure 2—figure supplement 1; Steger , 1998 ) . Next , we focused on the robust estimation of average single filament intensity in the total tubulin channel . We observed that the average intensity of total tubulin was slightly lower for segments detected using acetylated tubulin , compared to segments detected using either total and tyrosinated tubulin ( Figure 2—figure supplement 1 ) . Possibly , the subset characterized by acetylation is highly modified and therefore stained less effectively by the alpha-tubulin antibody . To take this into account , we pooled together the total tubulin intensities of all segments , independent of the detection channel ( Figure 2E ) . This histogram displayed a skewed distribution with a tail in a range of higher intensities ( Figure 2E ) . We reasoned that the peak of the distribution represents the intensities of single microtubules , while the tail corresponds to the presence of microtubule bundles containing two or more overlapping microtubules , as both of these groups could clearly be distinguished in the original images ( Figure 2B , D ) . To obtain a robust estimate for the intensity of individual microtubules , we fitted the histograms with a sum of two Gaussian distributions ( Figure 2E ) . The first Gaussian represents the intensity distribution of a single microtubule , with a standard deviation determined by several different factors ( antibody staining variations , microtubules going in and out of focus ) . The second Gaussian corresponds to bundles of two microtubules and mathematically represents a convolution of first Gaussian with itself ( assuming random independent intensity sampling of two microtubules in a bundle ) . We used the mean value of the first Gaussian as an estimate of average single microtubule intensity in the total channel . We proceeded with an estimation of the average levels of tyrosination and acetylation of an individual microtubule ( per cell ) . To exclude segments corresponding to the bundles of multiple microtubules , we only analyzed segments for which the total tubulin intensity was below the mean intensity of the first Gaussian plus one standard deviation ( Figure 2E ) . Visual inspection of segments below and above the threshold confirmed that this filtering eliminated the majority of thick or bright bundles ( Figure 2F ) . Consistently , including only the single-MT segments determined from total tubulin intensities also resulted in more unimodal and symmetric distributions for the intensities of tyrosination and acetylation levels of segments identified in these respective channels ( Figure 2G , Figure 2—figure supplement 1 ) . The average intensities of single tyrosinated or acetylated microtubule were estimated as the average values of intensities detected and quantified in the same corresponding channel . These values were used for the normalization of intensities shown at Figure 2G and Figure 2—figure supplement 2 . In addition , we also quantified the levels of tyrosination and acetylation of all segments detected in the acetylation and tyrosination channel , respectively . This analysis enabled us to build two two-dimensional histograms that show the levels of both tyrosination and acetylation for microtubule segments detected either in the acetylation channel or the tyrosination channel ( Figure 2G ) . The resulting histograms show that segments detected by acetylation have , on average , lower levels of tyrosination than segments detected by tyrosination , and vice versa ( Figure 2G ) . This quantitatively confirms the general impression that these chemical groups mark two different subsets and that microtubules with high levels of acetylation are mostly detyrosinated . However , despite clearly separating into two subsets , even highly acetylated microtubules display residual tyrosination , whereas many tyrosinated microtubules have some extent of acetylation . The measured relative level of tyrosination for acetylated microtubules , compared to average tyrosination of tyrosinated microtubules , which we termed α , was equal to 0 . 53 ± 0 . 11 ( average ± SD ) and the level of acetylation for tyrosinated microtubules , compared to the acetylation of acetylated microtubules , termed β , was 0 . 45 ± 0 . 07 ( Figure 2H ) . These results demonstrate that microtubules can be divided in two different subsets based on the detection of tyrosination and acetylation . We will refer to microtubules detected in the acetylated tubulin channel , displaying on average 47% lower levels of tyrosination than microtubules detected in the tyrosinated channel , as stable microtubules . Likewise , dynamic microtubules are identified as microtubules detected in the tyrosinated tubulin channel and feature 55% lower levels of acetylation than microtubules detected in the acetylated channel . We next set out to use the intensities of total tubulin , acetylation , and tyrosination on individual microtubules to determine the both the total number of microtubules within dendrites , as well as the number of stable and dynamic microtubules within dendrites . To estimate the total number of microtubules , the dendritic intensity of total tubulin was divided by the single-microtubule intensity ( assuming consistent labeling throughout the neuron ) . However , for the quantification of stable and dynamic microtubules , we needed to correct for the ‘chemical crosstalk’ that we observed , that is the tyrosination and acetylation levels detected for stable and dynamic microtubules , respectively . As a result , the integrated tyrosinated intensity of a dendrite was not just the sum of intensities of dynamic microtubules , but also included the contribution from the residual tyrosination of stable microtubules ( and vice versa ) . This situation was analogous to instances of spectral crosstalk in fluorescence microscopy , where emission from one dye is detected in the spectral channel of another dye ( Zimmermann , 2005 ) , and we therefore used standard formulas for spectral unmixing and our estimates for α and β ( Figure 2H , I ) to take this posttranslational modification crosstalk into account . When we calculated the composition of the dendritic microtubule network , we focused on the proximal 5–10 μm of a dendrite ( Figure 3A , B ) and used the corresponding single-filament intensity and crosstalk estimates from the soma of the same cell . First of all , this showed that the total number of microtubules in a dendrite depends linearly on its cross-sectional area in the range from 1 to 10 μm2 with a slope of 68 microtubules per μm2 . In addition , it revealed that dendrites have over four times more acetylated microtubules than tyrosinated microtubules ( 74 ± 8% versus 16 ± 11% , average ± SD ) and that this factor was largely independent of the diameter of the dendrite ( Figure 3C , D , E ) . We furthermore found that these two subsets did not completely account for the total number of microtubules that we measured , leaving a small fraction of 10 ± 14% of microtubules that were classified as neither acetylated nor tyrosinated . A potential weakness of the analysis that we performed is that it assumes that the measured levels of acetylation and tyrosination on stable and dynamic subsets found in the cell body are comparable to those found within dendrites . To overcome this , single-microtubule levels of acetylation and tyrosination would need to be measured directly in the dendrites , which requires 3D images in which single dendritic microtubules are clearly distinguishable . Because this was not possible using our STED microscopy approach , we switched to expansion microscopy ( ExM ) to improve both lateral and axial resolution ( Jurriens et al . , 2021; Tillberg et al . , 2016 ) . In expansion microscopy , stained samples are embedded in and crosslinked to a swellable hydrogel , followed by proteolytic digestion and physical expansion , which will increase the spacing between the remaining gel-linked protein fractions and fluorophores . Since gels expand in all dimensions , this leads to an isotropic resolution improvement determined by the expansion factor of the gel ( about four times ) . Indeed , expanded samples demonstrated a substantial increase in the clarity with which microtubule organization could be perceived ( Figure 4A , B , Figure 4—figure supplement 1 , Video 2 ) . We therefore repeated our analysis of the spatial distribution of tyrosinated and acetylated microtubules and found that the peripheral enrichment of tyrosinated microtubules was even more pronounced in ExM samples , as shown in y , z cross-section images ( Figure 4B ) and radial distribution plots ( Figure 4C , D , E , F ) . Even though visual tracing of individual filaments remained challenging ( Figure 4—figure supplement 1 ) , we were able to estimate the relative abundance of acetylated and tyrosinated microtubules by decomposing the radial density of total tubulin as a sum of the acetylated and tyrosinated radial densities ( Figure 4G ) . Although this analysis does not take into account the fraction of microtubules that is neither tyrosinated or acetylated , it independently confirms the prevalence of acetylated microtubules ( 65% ) over tyrosinated ( 35% ) . The successful decomposition of total tubulin using only these two subsets , as well as the higher fraction of tyrosinated tubulin in comparison with our estimate from soma-based intensity rescaling ( Figure 3E ) , prompted us repeat our microtubule counting using dendrite-based intensity rescaling . Unfortunately , our ExM data also did not have sufficient resolution to resolve enough individual microtubules to reliably determine the single-microtubule estimates of acetylation and tyrosination required for such analysis . First , we tried to use ExSTED microscopy to improve resolution ( Gao et al . , 2018 ) and found that three-color volumetric STED acquisition of ExM samples resulted in substantial photobleaching , which strongly impaired the integrated intensity analysis ( data not shown ) . We then realized that most dendritic microtubules are organized in bundles that run parallel to the coverslip and thus resolving microtubules would be easier if we could alter the sample orientation such that microtubules are aligned with the optical axis of our microscope . Since in a regular ExM acquisition the axial dimension ( the poorest ) of the PSF is oriented perpendicular to the filaments ( located parallel to the coverslip plane ) , we decided to generate thick gel slices that were rotated by 90 degrees , a procedure we termed FlipExM ( Figure 5A , B ) . In this configuration , we exploit the better lateral resolution to resolve individual microtubules , while PSF blurring along the optical axis happens parallel to filaments ( Figure 5B , Videos 3 and 4 ) . Although we still could not discern individual microtubules within tight bundles , we observed many individual microtubules traversing through the dendritic volume . We therefore set out to quantify the intensities of these microtubules , so that these could be used to quantify the total number of microtubules and the abundance of microtubule subsets within the dendrite . The cross-sections of individual microtubule filaments were automatically detected in each channel in dendrites cross-sections ( Figure 5B , bottom row ) , and we quantified their area and their background-corrected intensity in each channel . To exclude noise and bundles , we then applied area and roundness filters on our detections ( Figure 5C , D ) . After this geometrical filtering , the intensity distribution showed a similar bimodal or skewed shape as found earlier for the filaments in the cell body ( Figure 5E , Figure 2E , Figure 5—figure supplement 1 ) . Therefore , we again used curve fitting ( similar to Figure 2 ) to estimate the average intensity of microtubule cross-sections in each channel . The distributions of tyrosinated and acetylated detections in the tyrosination/acetylation plane ( Figure 5F , Figure 5—figure supplement 2 ) appeared very similar to the data obtained earlier using the cell body ( Figure 2G ) , but with more distinct separation between the two clusters . Compared to the cell body data , we found slightly different values for the average tyrosination level of acetylated microtubules ( 0 . 45 ± 0 . 05 ) , as well as the acetylation level for tyrosinated microtubules ( 0 . 60 ± 0 . 17 ) ( Figure 5G ) . Finally , we used the single-microtubule intensity levels measured directly within dendrites to quantify total microtubule numbers , as well as the number of acetylated and tyrosinated microtubules ( Figure 5G–J ) . First , we divided the integrated cross-section intensity of the total tubulin channel by our single cross-section intensity estimate for dendrites with different diameters ( Figure 5G–J ) . A linear fit through the total number of microtubules as a function of cross-sectional area yields an estimated microtubule density of 68 and 53 microtubules per square micrometer for the cell body and dendrite methods , respectively ( Figure 5J ) . Next , to determine the number of acetylated and tyrosinated microtubules , we employed the ‘modification unmixing’ approach mentioned previously . Consistent with our earlier results , this analysis revealed that stable , acetylated microtubules form the largest population ( 72 ± 6% ) . The fraction of tyrosinated microtubules was larger than our earlier estimate ( 26 ± 8% ) , at the expense of the fraction of microtubules that were neither acetylated nor tyrosinated ( 2 ± 5% , Figure 5H , I , Figure 5—figure supplement 3 ) . These results indicate that acetylated and tyrosinated microtubules together account for 98% of all dendritic microtubules , with acetylated microtubules being almost three times more abundant . The high density of the neuronal microtubule cytoskeleton has so far obscured its exact composition and organization . Earlier work has used electron microscopy to reveal the number and spatial organization in dendrite cross-sections , but this technology is difficult to combine with the robust detection of distinct subsets ( Baas et al . , 1988; Kubota et al . , 2011 ) . While early work on axonal microtubules successfully detected modified microtubules using immunoelectron microscopy , these methods have not been systematically applied to dendrites ( Baas and Black , 1990; Ahmad et al . , 1993 ) . Here , we used super-resolution light microscopy to explore the composition and architecture of the microtubule cytoskeleton in dendrites . In addition to visualizing all microtubules using an antibody against alpha-tubulin , we focused on two microtubule subsets: those labeled using antibodies against acetylation and tyrosination , typically classified as stable and dynamic microtubules ( Janke and Magiera , 2020; Guardia et al . , 2016; Schulze and Kirschner , 1987 ) . Volumetric STED and expansion microscopy revealed a striking spatial organization in which stable , acetylated microtubules are enriched in the core of the dendritic shaft , surrounded by a shell of dynamic , tyrosinated microtubules ( Figures 1 and 4 ) . While our earlier two-dimensional super-resolution imaging already hinted at a spatial separation between different subsets ( Tas et al . , 2017 ) , the current work provides the first quantitative three-dimensional mapping of subset organization throughout a large set of dendrites . The enrichment of dynamic microtubules near the plasma membrane is consistent with the well-established interplay between growing microtubule plus ends and ( sub ) cortical complexes ( van de Willige et al . , 2016; Akhmanova and Steinmetz , 2015 ) . More specifically , dynamic microtubules have been shown to regularly invade into dendritic spines to facilitate intracellular transport or regulate spine morphology in response of specific synaptic stimuli ( Esteves da Silva et al . , 2015; Jaworski et al . , 2009; McVicker et al . , 2016; Hu et al . , 2008; Schätzle et al . , 2018 ) . Next to ensuring the enrichment of dynamic microtubules near the plasma membrane , spatial separation between stable and dynamic microtubules might also promote efficient intracellular transport by separating cargoes driven by subset-specific motors . Moreover , for motors that do not discriminate between microtubule subsets , the spatial separation between mostly minus-end out oriented stable microtubules and mostly plus-end out oriented dynamic microtubules ( Tas et al . , 2017 ) will facilitate directional transport by limiting directional switching induced by cargo-attached motors binding to neighboring microtubules of opposite polarity . In future work , we will explore how the transport patterns of different cargoes depend on the associated motors and the organization of the neuronal microtubule cytoskeleton . The use of three-color super-resolution imaging allowed us to include a marker for total tubulin and , in combination with novel analysis methods , provide two independent estimates for the total number of microtubules in dendrite sections , as well as the number of acetylated and tyrosinated microtubules ( Figure 3 , Figure 5 ) . Our estimates for the total microtubules were obtained by dividing the total intensity of a generic tubulin stain by the intensity measured on individual microtubules in either the soma ( STED , Figure 3 ) or dendrite itself ( Flip-ExM , Figure 5 ) , which revealed an average density of 68 or 53 microtubules per μm2 , respectively . These values are consistent with earlier estimates using electron microscopy ( 66 microtubules per μm2 ) ( Kubota et al . , 2011 ) . Although we used various filtering steps to prevent mistaking small microtubule bundles for individual microtubules , it remains possible that occasional inclusion of such bundles increased our estimate for single microtubules , thereby lowering our estimate for the total number of microtubules . Alternatively , these differences could be caused by local differences in expansion factor or just reflect sample-to-sample differences in the number of microtubules per area . Nonetheless , the close correspondence between our estimates and values obtained using electron microscopy on dendritic cross-sections demonstrates the strength of combining super-resolution microscopy with quantitative analysis . In all our estimates , acetylated microtubules strongly outnumber tyrosinated microtubules . The two single-microtubule calibration methods ( i . e . soma versus dendrite ) yielded strikingly similar estimates for the percentage of acetylated microtubules ( 74 ± 8% versus 72 ± 6% ) , while their estimates for tyrosinated and other ( non-tyrosinated and non-acetylated ) microtubules differed to some extent ( 16% tyrosinated and 10% other microtubules for soma versus 26% and 2% for dendrite estimations , respectively ) . These results suggest that the acetylation level of stable microtubules is similar between soma and dendrites , whereas the tyrosination level of dynamic microtubules could be higher in the soma than the dendrite . For the soma-based method , this would overestimate the dendritic single-microtubule tyrosination levels and result in undercounting dendritic tyrosinated microtubules , leaving a larger fraction of other microtubules . The idea that dendritic dynamic microtubules are on average more detyrosinated is consistent with our finding that these microtubules also have higher levels of acetylation compared to the soma ( Figure 5G ) . We therefore consider the dendrite-intrinsic measurements to be more reliable , that is 72 ± 6% acetylated , 26 ± 8% tyrosinated and 2 ± 5% other microtubules . It is important to note that the two markers that we used , acetylation and tyrosination , represent only a small part of the possibly ways in which the microtubule surface can become differentiated , such as through other modifications like polyglutamylation , phosphorylation , palmitoylation , incorporation of different tubulin isoforms , and adsorption of different MAPs ( Sirajuddin et al . , 2014; Janke and Magiera , 2020 ) . We focused on acetylation and tyrosination because these modifications label clearly distinct subsets and display the most direct correlation with microtubule stability ( e . g . only acetylated microtubules remain after nocodazole treatment ( Tas et al . , 2017 ) and motor selectivity ( e . g . Kinesin-1 binding correlates with acetylation ( Tas et al . , 2017; Jansen et al . , 2021 ) ) . Other modifications , such as glutamylation , are more rheostatic and are known to play different roles at different levels ( Roll-Mecak , 2019; Valenstein and Roll-Mecak , 2016 ) . Importantly , the binary classification scheme used to classify microtubules as either acetylated or tyrosinated is most likely an oversimplification that does not do full justice to the rich modification landscape of microtubules , where also different parts of a microtubule can display different modifications ( Baas and Black , 1990; Ahmad et al . , 1993 ) . Our choice for acetylation and tyrosinations was furthermore prompted by the availability of reliable antibodies , which remains a challenge for many other modifications . Remarkably , our analyses revealed that labeling acetylated and tyrosinated microtubules leaves only a very small fraction ( 2% ) of microtubules unlabeled . This suggests that most detyrosinated microtubules in dendrites are also acetylated and that other modifications or MAPs are found on microtubules that are either tyrosinated or acetylated . In this work , we have introduced innovative imaging and analysis approaches to quantitatively map the neuronal cytoskeleton . In future work , we aim to map how other modifications and various microtubule-associated proteins are distributed over these two subsets of microtubules . In addition , the distribution of modifications and microtubule-associated proteins along the length of individual microtubules should be mapped to better understand how dynamic microtubules may become stabilized . We anticipate that such experiments will benefit from ongoing advances in expansion microscopy , such as iterative expansion or single-step approaches with higher expansion factors . Dissociated hippocampal neuron cultures were prepared from embryonic day 18 rat pups of mixed gender according to the previously published protocol ( Kapitein et al . , 2010 ) . Briefly , cells were plated on 18‐mm glass coverslips coated with poly‐l‐lysine ( 37 . 5 mg/ml ) and laminin ( 1 . 25 mg/ml ) in a 12‐well plate at a density of 50 k/well . Cultures were maintained in Neurobasal medium ( NB ) supplemented with 2% B27 , 0 . 5 mM glutamine , 16 . 6 μM glutamate , and 1% penicillin/streptomycin at 37°C in 5% CO2 . COS7 cells were cultured in DMEM supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin . We followed our recently published immunostaining/expansion protocol , described in details in Jurriens et al . , 2021 . In short , at DIV9 neurons were pre-extracted for 1 min using 500 µl of 0 . 3% Triton X-100 ( Sigma X100 ) , 0 . 1% glutaraldehyde ( Sigma G7526 ) in MRB80 buffer ( 80 mM Pipes ( Sigma P1851 ) , 1 mM EGTA ( Sigma E4378 ) , 4 mM MgCl2 , pH 6 . 8 ) and fixed for 10 min using 4% PFA ( Electron Microscopy Sciences 15 , 710 ) and 4% sucrose in MRB80 buffer ( both solutions were pre-warmed to 37°C ) . Fixed neurons were washed three times in PBS and permeabilized for 10 min in 500 µl of 0 . 25% Triton X-100 in MRB80 buffer . Samples were further incubated in 500 µl of blocking buffer ( 3% w/v BSA in MRB80 buffer ) for at least 45 min at room temperature . Finally , fixed neurons were sequentially incubated for two hours at room temperature ( or overnight at 4°C ) with primary and secondary antibodies diluted in blocking buffer ( 3% w/v BSA in MRB80 buffer ) and washed three times in PBS . The same fixation protocol was used for staining with COS7 cells . We used the following combinations of primary ( dilution 1;500 for STED and confocal; dilution 1:200 for expansion ) and secondary ( dilution 1;500 for STED and confocal; dilution 1:250 ) antibodies: mouse monoclonal anti-acetylated tubulin ( Sigma , [6-11B-1] , T7451 ) with Abberior Star 635P goat anti-mouse IgG ( H + L ) ( Abberior GmbH ST635P-1001–500 UG ) , rat monoclonal anti-tyrosinated tubulin ( Abcam , [YL1/2] , ab6160 ) with Alexa Fluor 594 goat anti-rat IgG ( H + L ) ( Molecular Probes , Life Technologies A11007 ) and rabbit recombinant anti-alpha tubulin antibody ( Abcam , [EP1332Y] , 52866 ) , rabbit polyclonal anti-detyrosinated tubulin antibody ( Merck , AB3210 ) , and rabbit polyclonal anti-delta2 tubulin antibody ( Millipore , AB3203 ) with Alexa Fluor 488 goat anti-rabbit IgG ( H+L ) ( Thermo Fisher Scientific , A-11034 ) . For staining of COS7 cells the antibody combinations were slightly different to ensure optimal signal intensity . We used rabbit polyclonal anti-detyrosinated tubulin antibody with Abberior Star 635P goat anti-rabbit igG ( H+L ) ( Abberior GmbH ST635P-1002–500 UG ) , mouse monoclonal anti-acetylated tubulin with Alexa Fluor 594 goat anti-mouse IgG ( H+L ) ( Molecular Probes , Life Technologies A11032 ) and rat monoclonal anti-tyrosinated tubulin with Alexa Fluor 488 goat anti-rat IgG ( H+L ) ( Molecular Probes , Life Technologies A11006 ) . Expansion microscopy ( ExM ) was performed according to the proExM protocol ( Tillberg et al . , 2016 ) with the detailed description published in Jurriens et al . , 2021 . Briefly , immunostained neurons on 18‐mm glass coverslips were incubated overnight in PBS with 0 . 1 mg/ml Acryloyl‐X ( Thermo Fisher , A20770 ) and afterwards washed three times with PBS . Per coverslip , we made 200 μl of gelation solution by mixing 188 μl of monomer stock solution ( 1 × PBS , 2 M NaCl , 8 . 625% ( w/w ) sodium acrylate ( SA ) ( Sigma Aldrich 408220 ) , 2 . 5% ( w/w ) acrylamide ( AA ) , 0 . 15% ( w/w ) N , N′-methylenebisacrylamide ) , 8 μl of 10% ( w/w ) tetramethylethylenediamine ( TEMED , BioRad 161–0800 ) accelerator and 4 μl of 10% ( w/w ) ammonium persulfate ( APS , Sigma Aldrich A3678 ) initiator ( added at the last step ) . Of the gelation solution , 120 μl was transferred to a gelation chamber , made out of a silicone mold with an inner diameter of 13 mm ( Sigma-Aldrich , GBL664107 ) attached to a parafilm-covered glass slide . The sample was put cells-down on top of the chamber to close it off . After incubation at RT for 1–3 min , the sample was transferred to a humidified 37°C incubator for at least 30 min to fully polymerize the gel . After gelation , the gel was transferred to a 12-well plate with 2 ml of digestion buffer ( 1 × TAE buffer ( 40 mM Tris , 20 mM acetic acid , 1 mM EDTA , pH8 ) , 0 . 5% Triton X-100 , 800 mM NaCl , 8 U/mL proteinase-K ( ThermoFisher , EO0492 ) ) for 4 hr at 37°C for digestion . The gel was transferred to 50 ml deionized water for overnight expansion , and water was refreshed once to ensure the expansion reached plateau . Plasma‐cleaned 24 × 50 mm rectangular coverslips ( VWR 631–0146 ) for gel imaging were incubated with 0 . 1% poly‐l‐lysine to reduce drift of the gel during acquisition . The gel was mounted using custom‐printed imaging chambers ( Jurriens et al . , 2021 ) . The expansion factor was calculated for each sample as a ratio of a gel's diameter to the diameter of the gelation chamber and was in the range of 4 . 14–4 . 16 . For the FlipExM samples , we cut a thin piece of gel ( 1 cm x 3 mm ) using a razor blade and flipped it on its cut edge during transfer to the imaging chamber . Data from non-expanded samples were acquired using a Leica TCS SP8 STED 3X microscope with a pulsed ( 80MHz ) white-light laser , HyD detectors and spectroscopic detection using HC PL APO 100×/1 . 40 Oil STED WHITE ( Leica 15506378 ) oil-immersion objective . For Abberior STAR 635P and Alexa 594 we used 633 nm and 594 nm laser lines for excitation and a 775 nm synchronized pulsed laser for depletion , with a time gating range of 0 . 3–7 ns . For Alexa 488 we used 488 nm excitation , 592 nm continuous depletion laser line and time gate of 1 . 1–7 ns . Emission detection windows were 500–560 nm , 605–630 nm and 640–750 nm for Alexa 488 , Alexa 594 and Abberior STAR 635P , respectively . No bleed-through was observed between the channels . For three-color cell body imaging ( Figure 2 , Figure 3 ) , each fluorescent channel was imaged using the 2D STED configuration ( vortex phase mask ) in sequential z-stack mode from highest to lower wavelength , to prevent photobleaching by the 592 nm depletion laser line . For two-color imaging of dendrites ( Figure 1 ) , we used the Abberior STAR 635P/Alexa 594 combination and a single 775 nm depletion line and therefore acquired images in line-sequential mode . For the 3D STED imaging , we used a combined depletion PSF light path consisting of a mixture 60% Z-donut and 40% vortex phase mask , providing approximately isotropic resolution . For the data shown in Figure 2 , Figure 3 , the size of the field-of-view was in the range of 30–50 µm and it was positioned to include the whole cell body of a neuron ( soma ) and the first 5–10 µm of dendrites emanating from it ( Figure 2A , Figure 3A ) . For the data shown in Figure 1 , the size of the field-of-view was in the range of 50–100 µm and it covered 30–50 µm of the proximal dendrites . The depth of z-stacks varied in the range from 3 to 6 µm for each acquisition and for all cases it was chosen to fully cover the dendrite’s thickness . The lateral pixel size was in the range of 27–30 nm with a distance between z-planes in the range of 150–160 nm . The z-stacks were subjected to a mild deconvolution using Huygens Professional software version 17 . 04 ( Scientific Volume Imaging , The Netherlands ) with CMLE ( classic maximum likelihood estimation ) algorithm with parameters of SNR ( Signal-to-Noise Ratio ) equal to 7 over 10 iterations . After the deconvolution , z-stacks of tyrosinated and acetylated channels were registered in 3D to total tubulin channel using maximum intensity projections in XY and XZ planes using Correlescence plugin v . 0 . 0 . 4 ( https://github . com/ekatrukha/Correlescence archived on Zenodo repository https://doi . org/10 . 5281/zenodo . 4534715 ) for ImageJ . Expanded gels were imaged using the same Leica TCS SP8 STED 3X microscope with a pulsed ( 80 MHz ) white-light laser , HyD detectors and spectroscopic detection using a HC PL APO 86 ×/1 . 20 W motCORR STED ( Leica 15506333 ) water-immersion objective with a correction collar . Each fluorescent channel was imaged in confocal line-sequential mode . For Alexa488 , we used 488 nm excitation and 500–560 nm emission range , for Alexa594 we used 594 nm excitation and 605–630 nm emission and Abberior STAR 635P we used 633 nm excitation and 640–750 nm emission . For ExM samples , the size of the field-of-view was in the range of 50–100 µm and had a thickness in the range of 10–20 µm , chosen to cover the whole volume of a dendrite . The dimensions of FlipExM stacks were 20–30 µm in XY and 30–50 µm in Z . In both cases , the pixel size in XY plane was in the range of 60–80 nm and the distance between z-planes was in the range of 150–180 nm . The z-stacks were subjected to a mild deconvolution using Huygens Professional software version 17 . 10 ( Scientific Volume Imaging , The Netherlands ) with CMLE ( classic maximum likelihood estimation ) algorithm with parameters of SNR ( Signal-to-Noise Ratio ) equal to 15 over 10 iterations . From registered z-stacks , we chose substacks of 4–6 frames ( 0 . 7–1 µm thick ) located at the bottom of the cell under the nucleus ( Figure 2A ) , where the density of microtubule network was low . Using maximum intensity projections of these substacks in each fluorescent channel , we extracted segments of microtubule filaments using CurveTrace plugin ver . 0 . 3 . 5 ( https://github . com/ekatrukha/CurveTrace archived on Zenodo repository https://doi . org/10 . 5281/zenodo . 4534721 ) for ImageJ implementing ( Steger , 1998 ) . The detection parameters used were: line width of 2 . 5 pixels ( 75 nm ) ( standard deviation of line thickness ) and minimum segment’s length of 0 . 6 µm . The detection of filaments was limited to the area of cell body , excluding dendrites ( Figure 2B , Figure 2—figure supplement 1 ) . After detection , each segment of microtubule was stored as a polyline ROI ( region of interest ) file in ImageJ format , essentially represented as a set of ordered XY coordinates . The detection was performed separately for each fluorescent channel , to take advantage of sparser filament’s subnetworks with less overlap , displayed in tyrosinated and acetylated channel ( Figure 2B , bottom row ) . The quantification of filament intensities was performed on the sum of slices of the substacks used for the detection ( SUM projection ) . The intensity of filament segments detected in the different channels was quantified for each fluorescent channel ( total , tyrosinated , acetylated ) , producing nine datasets ( Figure 2C ) . For each detected polyline ROI segment of length Lmid , we first measured the integrated intensity Imid with a line width wmicrotubule of 6 pixels ( ~180 nm ) covering the area of Smid . In this case wmicrotubule corresponds to the width used to normalize intensity of a single microtubule segment . As a second step , we measured the integrated intensity Iwide of the same segment with a wider line’s width of 13 pixels ( ~390 nm ) covering the area of Swide . From these measurements , the average intensity of the background IBG was calculated as: ( 1 ) IBG=Iwide-ImidSwide-Smidand the background-corrected average intensity of a segment Isegm as: ( 2 ) Isegm=Imid-IBGSmidSmid Since the line width was fixed , this value does not depend on the length of the filament and essentially represents average fluorescent intensity along the filament . The described measurements were automated using an ImageJ script ( Katrukha , 2021 ) . The histogram of segment intensities in total channel ( pooled from all three detections ) was fitted with a sum of two Gaussians ( Figure 2E ) expressed as: ( 3 ) ρsegm ( Isegm ) =a1exp⁡ ( − ( Isegm−ITot ) 22σTot2 ) +a2exp⁡ ( − ( Isegm−2ITot ) 24σTot2 ) where a1 , a2 correspond to the amplitudes ( weights ) of first and second Gaussians , ITot and σTot are the average intensity and standard deviation of the Gaussian corresponding to the single microtubule intensity distribution ( for the second Gaussian , after the convolution , average intensity and standard deviation are 2ITot and 2σTot ) . The fitting was performed for each cell individually , to eliminate a difference in imaging conditions and heterogeneity of the sample . The fitted value of average intensity ITot was used later for the estimation of total microtubule numbers in dendrites ( see next section ) . For quantification of the average levels of tyrosination and acetylation per single microtubule , we introduced a threshold of ITot+σTot on the corresponding total tubulin intensity of segments detected in acetylated and tyrosinated channels ( Figure 2E ) . Only the segments which total tubulin intensity was below this threshold were used for the calculation of average intensities of single tyrosinated ITyr or acetylated IAc microtubules . The average values for each channel were used for the normalization of intensities presented at Figure 2G , H and Figure 2—figure supplment 2 . The fitting and threshold filtering was performed using custom written MATLAB scripts ( Katrukha , 2021 ) . To estimate the average number of microtubules in dendrites , we first built summary ( integrated ) XY projection images of the z-stacks containing the whole depth of dendrites . Similar to the quantification of single microtubule intensity , we drew a straight line ROI of 2-3 µm ( Ldendrite ) along a dendrite segment with a width wdendrite ( using ImageJ ) . This width varied depending on the dendrite and was chosen to visually include its whole thickness , covering an area of Smiddendrite . We measured the integrated intensity over this area , denoted as Imiddendrite . In the second step , we measured the integrated intensity Iwidedendrite of the same straight line with a width increased by 10 pixels ( 300 nm ) covering the area of Swidedendrite . From these measurements , the average intensity of the background IBGdendrite was calculated similar to Equation ( 1 ) as: ( 4 ) IBGdendrite=Iwidedendrite-ImiddendriteSwidedendrite-Smiddendriteand the background-corrected average intensity per area of a dendrite segment Idendrite was calculated as: ( 5 ) Idendrite=Imiddendrite-IBGdendriteSmiddendrite/ ( Ldendritewmicrotubule ) wmicrotubule ( 6 ) nTot=ITotdendriteITotwhere ITotdendrite is dendrite’s intensity in total tubulin channel calculated according to Equation ( 5 ) and ITot is average single microtubule intensity in total tubulin channel ( see previous section ) . The specific values of ITot were taken from the same cell/z-stack containing the dendrite . To calculate numbers of microtubules in tyrosinated and acetylated channel we used following formulas ( Figure 2I ) : ( 7 ) ITyrdendrite=nTyr+αnAcITyr ( 8 ) IAcdendrite=βnTyr+nAcIAcwhere ITyrdendrite , IAcdendrite are background corrected dendrite intensities calculated according to Equation ( 5 ) , ITyr and IAc average single microtubule intensities in tyrosinated and acetylated channel , α stands for average level of tyrosination for microtubules detected in the acetylated channel , β corresponds to the average acetylation level of microtubules detected in the tyrosinated channel ( Figure 2H ) and nTyr , nAc are numbers of tyrosinated and acetylated microtubules . The solution of system Equation ( 7 ) - ( 8 ) gives the final formulas: ( 9 ) nTyr=θTyr-αθAc ( 1-αβ ) ( 10 ) nAc=θAc-βnTyrwhere θTyr=ITyrdendrite/ITyr and θAc=IAcdendrite/IAc . Equations ( 9 ) - ( 10 ) were used to report the number of tyrosinated and acetylated microtubules in Figure 3C-D . In addition , these values were used to calculate the tyrosinated and acetylated percent of total microtubules number nTot reported in Figure 3E . The number of 'non-modified' , other microtubules was calculated as nOther=nTot-nTyr-nAc . The dendritic cross-section area ( Figure 3C–D ) was measured by building XZ resliced cross-section along a perpendicular line in the area of intensity measurement . We acquired z-stacks covering the whole thickness of a dendrite ( using 2D or 3D STED in Figure 1 and confocal for ExM samples in Figure 3 ) . Using maximum intensity projection in XY plane , we marked the middle of dendrite with a polyline ROI of appropriate thickness . We used ‘Selection->Straighten’ function of ImageJ on original z-stacks to generate B-spline interpolated stacks , so a dendrite became straight and oriented along X axis . From those stacks , we generated a resliced stack in the plane perpendicular to dendrite’s axis ( YZ ) for the analysis of radial intensity distribution in the cross-section ( Figure 1A–B ) . To find the boundary outline of the dendrite in each slice , we used a custom written set of ImageJ macros allowing semi-automated analysis ( Katrukha , 2021 ) . The process consisted of two stages: finding the bounding rectangle encompassing the dendrite's intensity and building a smooth closed spline ( approximately in the shape of an oval , see below ) . Illustration of full analysis workflow is presented in Video 1 . First , we calculated a center of mass ( based on intensity ) coordinates xc , yc for tyrosinated ( STED data ) or total tubulin ( ExM ) channels . Then we specified a rectangular ROI R of maximum area under conditions that it was still located inside the image and that its center ( intersection of diagonals ) was positioned at the center of mass . In the next step , we progressively downsized the rectangle from each side to find the position where edge's intensity becomes equal to some threshold value ( see below ) . We describe it here for the right side , but the same procedure was applied to all sides . Given an initial rectangle R of width w , height h and top left corner coordinates xR , yR , we built a set of rectangles with the width wi ranging from w/2 to w ( with a step size of one pixel ) with the same height h and same and fixed position of the top left corner . For each rectangle from this set , we measured the integrated intensity , providing the integrated intensity as a function of width Iint ( wi ) . The intensity of the edge Ie ( wi ) was calculated as a derivative of this function , that is Ie ( wi ) =Iint ( wi ) – Iint ( wi+1 ) and normalized by its maximum and minimum value . A typical shape of Ie ( wi ) represents a peak around w/2 ( a center of dendrite/rectangle ) that is gradually decaying toward periphery . For the first image in the resliced stack , by decreasing the value of wi starting from w , we found the first value of Ie ( wi ) that exceeds a threshold normalized intensity value of Ithr and its corresponding width wRB . The coordinate of the right boundary ( RB ) was calculated as xRB=xR +wRB . The threshold intensity value Ithr was in the range of 0 . 2–0 . 4 and chosen for each first image in a stack manually to provide the values of xRB corresponding to the visual boundary of dendrite's intensity . The procedure was repeated for all other sides of the rectangle R , providing coordinates of left xLB , top yTB and bottom yBB boundaries . For horizontal boundaries , the width was kept the same and the position of the opposite edge was kept constant while building Iint ( hi ) . Using the newly found coordinates of the boundaries , we thereby built updated rectangle RB encompassing dendrite's cross-section . This method worked robustly in many cases , but it failed in the presence of axons that were often wrapped around a dendrite . In the YZ plane , those axons produced additional fluorescent spots next to dendrite cross-section that were included into rectangle RB . In the shape of Ie ( wi ) curve they manifest themselves as additional local peaks . Therefore , procedure of finding RB from initial rectangle R for all other images in the resliced YZ stack ( apart from the first ) was modified . In these cases , we scanned Ie ( wi ) by both decreasing and increasing values of wi in the w/2 to w range . During a scan , we recorded all values of wi that corresponded to each threshold Ithr from the set of 0 . 1 to 0 . 5 with a step of 0 . 1 . From these we calculated a set of candidate right boundary positions , from which we chose the one that is closest to the corresponding boundary from the previous slice image in the stack . This value was recorded as the new edge of RB rectangle at the current image . The procedure was repeated for each edge and after finding boundary rectangles for the whole stack , they were inspected and corrected manually . To build a closed smooth spline contour around the irregular shaped dendrite's cross-section , in addition to vertical and horizontal boundaries , we also determined diagonal boundary points . For that we built an intensity profile along the 20–40 pixels wide line ROI connecting left top and right bottom corners of the rectangle RB . After normalization of intensity to minimum and maximum , we found coordinates of two points on the both halves of line where intensity is closest to 0 . 15–0 . 2 of its maximum value , denoted ( xLD1 , yLD1 ) and ( xLD2 , yLD2 ) . The same procedure was performed on the diagonal segment connecting left bottom and top right corners of rectangle RB , providing points ( xRD1 , yLD1 ) and ( xRD2 , yRD2 ) . The ordered set of eight points with coordinates ( xLB , yc ) , ( xLD1 , yLD1 ) , ( xc , yTB ) , ( xRD1 , yLD1 ) , ( xRB , yc ) , ( xLD2 , yLD2 ) , ( xc , yBB ) , ( xRD2 , yRD2 ) was used to construct smooth closed spline boundary C passing through all of them ( ImageJ functions makePolygon and ‘Fit Spline’ ) . The final outlines for each image were inspected visually and if necessary , corrected manually . In addition , for some ExM data profiles with low background we used an alternative , faster algorithm to find the boundary . We detected a set of points representing local fluorescence intensity maxima in each YZ slice ( corresponding to MTs cross-sections ) . Using this set , we built a convex hull and constructed a spline from it . Again , we manually checked and corrected generated outlines . To build the radial intensity distribution , for each image we iteratively reduced the contour C with steps of one pixel using ImageJ function ‘Enlarge ROI’ with negative values ( it uses Euclidean distance map threshold ) , while measuring its area Sk and integrated intensity ICk ( where index k denotes the step ) . We calculated the average intensity MIk of each contour in the shrinking series as derivative: ( 11 ) MIk=ICk-ICk-1Sk-Sk-1 To get the radial distribution , for each k step we recalculated area Sk to radius using the formula Rk=Sk/π and normalized it by maximum value . Finally , to get a probability density function ρR , we normalized MI ( R ) by the area under the curve . For the decomposition of total tubulin radial density ρTotR as a weighted sum of tyrosinated ρTyrR and acetylated ρAcR densities ( Figure 4G ) we minimized the mean square error MSE ( wTyr , wAc ) between two curves: ( 12 ) MSEwTyr , wAc=∑RρTotR-wTyrρTyrR-wAcρAcR2where wTyr and wAc correspond to the weights of tyrosinated and acetylated densities . By taking the derivatives of Equation ( 12 ) and making them equal to zero , the solution can be found in a closed form: ( 13 ) wAc=ρTotρAcρTyr2-ρTotρTyrρTyrρAcρTyr2ρAc2-ρTyrρAc2 ( 14 ) wTyr=ρTotρTyr-wAcρTyrρAcρTyr2where angle brackets denote averaging over the whole radius range . It must be noted , that even without addition of a stronger assumption wTyr+wAc=1 , our analysis provided values that satisfy this relation . The cross-sections of microtubules in YZ FlipExM appeared as a set of fluorescent spots ( Figure 5B ) . For intensity analysis , we used ComDet v . 0 . 5 . 3 plugin for ImageJ ( https://github . com/ekatrukha/ComDet archived on Zenodo repository https://doi . org/10 . 5281/zenodo . 4281064 ) , which reports spot area , width , height in the detected channel and quantifies the background corrected integrated intensity in all three channels . To filter out possible microtubule bundles , we applied a lower bound threshold of 0 . 8 on the spot ‘roundness’ θ , expressed as: ( 15 ) θ=wspothspotmaxwspot , hspot2where wspot and hspot correspond to spot width and height . For spot areas detected per channel and per dendrite , we performed an MLE fit to the normal distribution and obtained estimates for the mean Smean and standard deviation σarea , which were used to filter out spots with areas outside the inclusion range with lower bound max ( Q1 , Smean – σarea ) and upper bound ( Smean+σarea ) , see Figure 5C , D . For the lower bound , Q1 stands for 25th percentile , it was added to robustly remove false positives . After the ‘roundness’ and area filters , an estimation of average single microtubule intensity was performed in a similar way as in Figure 2 , that is by fitting a sum of two Gaussians ( Equation ( 3 ) ) to the histogram of intensity distributions ( Figure 5E , Figure 5—figure supplement 1 ) . For each slice of YZ stack , we calculated normalized integrated intensity in each channel as a sum of all spot's intensities divided by a single microtubule intensity derived from the fit . To calculate absolute MTs numbers per slice , we used the same Equation ( 6 ) - ( 8 ) as in Figure 1 . Since there was little variability in the MTs numbers along the proximal part of the dendrite used for analysis ( Figure 5—figure supplement 4 ) , we calculated average MT number over all slices for each channel per dendrite , reported in Figure 5H . To calculate the average level of tyrosination of microtubules detected in the acetylated channel α and the average level of acetylation of microtubules detected in the tyrosinated channel β from the FlipExM data , we used three different methods . Here , the estimation was more challenging , because their distributions displayed very long tails and therefore the average crosstalk values were still quite large ( Figure 5—figure supplement 1A ) . In the first method , we estimated α and β by pooling all filtered spot detections and calculating their average values ( Figure 5F , G; Figure 5—figure supplement 5 , left panels ) . In the second method , we only included intensities in the acetylated/tyrosinated detection channels that were less than the values of mean + SD in the same channel ( Figure 5—figure supplement 5 , middle panels ) . In the third method , we fitted the distribution of acetylation levels on tyrosinated MTs ( and vice versa ) with a sum of two Gaussian function , to obtain the values of α and β as positions of peaks at the Ace/Tyr plane ( Figure 5—figure supplement 5 , right panels ) . The two last methods provided smaller values for α and β , which resulted in slightly different estimates for the percentage of tyrosinated/acetylated microtubules ( Figure 5—figure supplement 5 ) . The lateral resolution of STED 2D ( Figure 1—figure supplement 2A , x-axis ) was calculated using parameter-free decorrelation method ( Descloux et al . , 2019 ) on the maximum intensity projection of cell body z-stacks ( Figure 2B ) . Same images and MT segments detections were used to calculate an average FWHM of individual MTs segments ( Figure 1—figure supplement 2A , y-axis ) using ‘Fit Gaussian to Curves’ function of CurveTrace ImageJ plugin ( see above ) . In addition , vertical and horizontal Gaussian fits to the YZ-cross-sections images of MTs from ( Figure 1A–B , Figure 2B ) were used to estimate lateral and axial resolution of STED 2D and 3D ( Figure 1—figure supplement 2B ) . For the analysis of MTs modifications along the length of individual dendrites ( Figure 1—figure supplement 3 ) , we performed a series of confocal tilescan acquisitions of large areas ( 100–400 µm in size ) around neuron cell bodies . Each tile was acquired as a z-stack covering the whole thickness of dendrites with the same excitation/emission settings as described in ‘ExM/FlipExM samples imaging’ section . Tiles’ SUM projections were stitched together in ImageJ using ‘Pairwise Stitching’ plugin ( Preibisch et al . , 2009 ) . Individual dendrites were traced manually with polyline ROI in ImageJ , which was subsequently fitted with a spline . We used a custom written ImageJ macro to fit perpendicular intensity profile at equidistantly sampled points of the ROI in each channel to a Gaussian function with a background offset . The fluorescent intensity at each point along a dendrite was estimated as multiplication of amplitude to the standard deviation of fitted Gaussian . Using Matlab script , the intensity profiles at each channel were normalized by the average value of first 5 µm and smoothened with a window of 2 µm . Intensity values above 1 . 25 were excluded from the analysis to remove occasional intensity spikes caused by intersecting neurites . The values of fitted standard deviation in total tubulin channel was used to calculate FWHM ( Figure 1—figure supplement 3 ) .	Cells in the body need to control the position of the molecules and other components inside them . To do this , they use a system of proteins that work a bit like a road network . The ‘roads’ are tubular structures known as microtubules , while ‘vehicles’ are transporters , called motor proteins , that ‘walk’ along the microtubules . Microtubule networks are important in all cells , but especially in neurons , which can grow very large . These cells have tree-like branches called dendrites that receive messages from other neurons . Dendrites contain different types of microtubules with many chemical modifications . These modifications consist of specific molecules or ‘groups’ becoming attached to or removed from the microtubules to change their properties – for example , microtubules can be ‘acetylated’ or ‘detyrosinated’ . Motor proteins prefer different kinds of microtubules , and so understanding transport inside cells involves creating a precise roadmap showing how many of each type of microtubule exist and where they go . Using different super-resolution microscopy techniques , Katrukha et al . created maps of the microtubules in rat neurons . These show that acetylated microtubules form a core in the centre of the dendrites , while tyrosinated microtubules ( which did not undergo detyrosination ) line the cell membrane of the dendrites . Katrukha et al . then used the maps to determine that acetylated microtubules account for 65 to 70% of all microtubules , while tyrosinated microtubules make up 20 to 30% . This means that most microtubules fall into these two categories . The work by Katrukha et al . provides one of the first quantitative estimates of the relative amount of acetylated and tyrosinated microtubules , starting to shed light on how cells control their transport network . This could ultimately allow researchers to explore how transport changes in health and disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology" ]	2021	Quantitative mapping of dense microtubule arrays in mammalian neurons
Cytosine methylation regulates essential genome functions across eukaryotes , but the fundamental question of whether nucleosomal or naked DNA is the preferred substrate of plant and animal methyltransferases remains unresolved . Here , we show that genetic inactivation of a single DDM1/Lsh family nucleosome remodeler biases methylation toward inter-nucleosomal linker DNA in Arabidopsis thaliana and mouse . We find that DDM1 enables methylation of DNA bound to the nucleosome , suggesting that nucleosome-free DNA is the preferred substrate of eukaryotic methyltransferases in vivo . Furthermore , we show that simultaneous mutation of DDM1 and linker histone H1 in Arabidopsis reproduces the strong linker-specific methylation patterns of species that diverged from flowering plants and animals over a billion years ago . Our results indicate that in the absence of remodeling , nucleosomes are strong barriers to DNA methyltransferases . Linker-specific methylation can evolve simply by breaking the connection between nucleosome remodeling and DNA methylation . Cytosine methylation provides a mechanism to heritably alter the genome without permanent modification of the DNA sequence ( Kim and Zilberman , 2014; Du et al . , 2015 ) . In eukaryotes , Dnmt1 family methyltransferases , called MET1 in plants , rely on selective recognition of hemi-methylated symmetrical CG dinucleotides by an obligate cofactor ( Kim and Zilberman , 2014 ) . This mechanism semiconservatively propagates methylation patterns following cell division , and – especially in plants – across generations . Plant and animal genomes also contain cytosine methylation outside CG dinucleotides ( Du et al . , 2015 ) . In plants , the CMT3 methyltransferase family catalyzes methylation of CNG trinucleotides , conventionally described as CHG ( where H is any non-G base ) to avoid overlapping CG sites ( Law and Jacobsen , 2010 ) . The related CMT2 family can methylate cytosines outside CG and CNG contexts ( Zemach et al . , 2013; Stroud et al . , 2014 ) , a pattern of specificity referred to as CHH . CMT2 and CMT3 preferentially methylate heterochromatic transposable elements ( TEs ) and rely on a positive feedback loop with dimethylation of lysine 9 of histone H3 . Plants also possess an RNA-directed DNA methylation ( RdDM ) pathway , in which 24 nucleotide RNA molecules guide DRM methyltransferases ( homologs of animal Dnmt3 ) to initiate DNA methylation in all sequence contexts , and to maintain CHH methylation at relatively euchromatic TEs ( Matzke and Mosher , 2014 ) . As implied by the above description of methylation pathways , DNA methylation represses TE transcription and transposition in plants and vertebrates ( Law and Jacobsen , 2010 ) . Methylation also regulates endogenous genes: methylation close to the transcriptional start site can cause gene silencing , whereas methylation of other genic regions can promote or counteract expression ( Jones , 2012 ) . Genetic defects in the methylation machinery lead to human diseases such as cancer ( Baylin and Jones , 2016 ) and disruption of methylation patterns during clonal propagation of plants causes developmental defects that hamper agriculture ( Ong-Abdullah et al . , 2015; Springer and Schmitz , 2017 ) . Faithful propagation of DNA methylation is clearly essential , yet our understanding of the underlying processes is far from complete . In particular , we do not know how DNA methyltransferases contend with nucleosomes , which comprise the basic repeating units of chromatin . Nucleosomes consist of about 147 bp of DNA tightly wrapped around an octameric complex of histone proteins , with the major and minor grooves of the double helix alternately facing toward and away from the histones roughly every 10 bp ( Luger et al . , 1997 ) . Nucleosomes are separated by short stretches of linker DNA , which are variably bound by histone H1 ( Raghuram et al . , 2009 ) . Flowering plants and mammals show an overall enrichment of methylation over nucleosomes , suggesting that the nucleosome is the preferred target for DNA methylation ( Chodavarapu et al . , 2010; Chen et al . , 2015 ) . Furthermore , a mild but robust 10 bp DNA methylation periodicity in both lineages suggests that methyltransferases directly modify DNA that is wrapped in nucleosomes ( Chodavarapu et al . , 2010 ) . However , mammalian methyltransferase activity is inhibited by nucleosomes and H1 in vitro ( Robertson et al . , 2004; Takeshima et al . , 2006; Takeshima et al . , 2008; Felle et al . , 2011; Schrader et al . , 2015 ) , and regions of the human and mouse genomes with reliably positioned nucleosome arrays organized by CTCF binding exhibit preferential methylation of linker DNA ( Kelly et al . , 2012; Baubec et al . , 2015 ) . Methylation is also almost completely confined to linker DNA in several diverse species of marine algae ( Huff and Zilberman , 2014 ) , suggesting that nucleosomes are generally refractory to DNA methylation . The reasons for the apparent discrepancy between in vitro and in vivo methyltransferase preferences , and for the observed interspecies differences , remain unknown . Adding to the mystery , plants and mammals depend on the DDM1/Lsh family of Snf2-class nucleosome remodelers – proteins that can alter nucleosomes in various ways – to achieve normal levels of DNA methylation ( Jeddeloh et al . , 1999; Zhou et al . , 2016 ) . Arabidopsis plants with mutations in DDM1 suffer major methylation losses in all sequence contexts , primarily in heterochromatic TEs , though much smaller losses are also observed in genes ( Jeddeloh et al . , 1999; Teixeira et al . , 2009; Zemach et al . , 2013; Ito et al . , 2015 ) . Knockout of mouse Lsh likewise causes major depletion of DNA methylation from repetitive heterochromatin ( Dennis et al . , 2001 ) , with more variable effects on methylation in genes and other sequences ( Myant et al . , 2011; Tao et al . , 2011 ) . Genetic inactivation of Arabidopsis histone H1 partially rescues the ddm1 hypomethylation phenotype , suggesting that DDM1 facilitates methyltransferase access to H1-containing chromatin ( Zemach et al . , 2013 ) . However , how DDM1 does this is unknown , and the incomplete rescue of DNA methylation in h1ddm1 compound mutants remains unexplained . More generally , how nucleosome remodelers facilitate methylation in species in which methyltransferases are proposed to act on the nucleosome surface is unclear . Here we explore how DNA methylation is regulated by DDM1/Lsh , H1 and nucleosomes . We find that nucleosomes and H1 are barriers to DNA methylation and that DDM1/Lsh remodelers are required for methylation of nucleosomal DNA . At heterochromatic loci with reliably positioned arrays of nucleosomes , methylation in h1ddm1 plants is strikingly periodic , with nearly wild-type levels in linkers and nearly ddm1 levels in nucleosomes . Our results indicate that methyltransferases generally require remodeling activity to access nucleosomal DNA . Lack of such remodeling biases methylation toward linker DNA , providing a straightforward explanation for the linker-specific methylation patterns found in diverse eukaryotes . To investigate the relationship between nucleosomes and DNA methylation , we deep sequenced unamplified DNA isolated from micrococcal nuclease ( MNase ) -digested chromatin of Arabidopsis plants with inactivating mutations in both canonical linker histone H1 genes ( h1 ) , ddm1 mutants , compound h1ddm1 mutants , and in WT controls . We selected mononucleosomes in silico to avoid inconsistencies stemming from DNA electrophoretic mobility . Using biological replicates , we defined four groups of nucleosomes per genotype , from well-positioned to poorly-positioned ( see Materials and methods ) , and anchored methylation analysis at the presumptive dyad ( nucleosome center ) ( Chen et al . , 2014 ) . We focused our initial analysis on heterochromatic sequences ( see Materials and methods ) , where H1 is most abundant ( Rutowicz et al . , 2015 ) and DDM1 is most important for maintenance of DNA methylation ( Zemach et al . , 2013 ) . CG methylation shows a mild but clear depletion over well-positioned nucleosomes in WT and h1 mutant plants , and an overt depletion in ddm1 and h1ddm1 plants ( Figure 1A , Figure 1—figure supplement 1 , loci with well-positioned nucleosomes on the left ) . Poorly positioned nucleosomes by definition do not have a well-delineated core region , and these loci accordingly exhibit a relatively flat level of methylation in all genotypes ( Figure 1A , Figure 1—figure supplement 1 ) . The difference between core and linker methylation is much greater in h1ddm1 than in ddm1 ( Figure 1B ) , an effect that is observable at individual loci ( Figure 1C , Figure 1—figure supplement 2 ) , where methylation of linker regions – much stronger in h1ddm1 than in ddm1 – alternates with weak or absent methylation of nucleosome cores . For the above analyses , we defined quality of nucleosome positioning separately for each genotype , and therefore the nucleosomes that comprise each of the four positioning groups vary between genotypes . To control for any potential sequence bias across genotypes , we directly compared the best positioned nucleosomes shared by WT and h1ddm1 , or by h1 and h1ddm1 ( Figure 1—figure supplement 3; see Materials and methods ) . This analysis produced the same pattern as in Figure 1 , at nearly the same scale . Interestingly , using only WT well-positioned nucleosomes severely attenuates the core-to-linker CG methylation difference in ddm1 and essentially eliminates any difference in h1ddm1 ( Figure 1—figure supplement 3 ) , illustrating the importance of genotype-specific nucleosome mapping . Our results indicate that nucleosomes and H1 are obstacles to DNA methylation in vivo . Nucleosomes , which form much more stable associations with DNA than H1 ( Misteli et al . , 2000 ) , are more powerful obstacles . In WT an h1 plants , DDM1 activity allows DNA methyltransferases robust access to DNA . In ddm1 mutants , nucleosomes are inaccessible and H1 blocks linker DNA , but the greater accessibility of linkers allows higher methylation levels at these sequences ( Figure 1A , B ) . In the absence of DDM1 and H1 , linker DNA is accessible to methyltransferases , but nucleosome cores are not , producing the robust methylation periodicity in h1ddm1 plants ( Figure 1A–C ) . Interestingly , non-CG methylation shows a pronounced nucleosome core depletion at well-positioned nucleosomes in all genotypes ( Figure 1D , E , Figure 1—figure supplement 3 ) . In the CHG context , h1ddm1 has the most exaggerated core-to-linker methylation differential , followed by h1 ( Figure 1B , D , Figure 1—figure supplement 3 ) . In the CHH context , h1ddm1 and h1 show a similarly strong core-to-linker methylation differential ( Figure 1B , E , Figure 1—figure supplement 3 ) . Thus , whereas DDM1 allows for near-perfect efficiency of nucleosome core CG methylation ( Figure 1A ) , non-CG methylation of nucleosomal DNA is well below that of linkers even in WT ( Figure 1D , E ) . Loss of H1 increases the contrast between linker and core methylation levels ( Figure 1A–E ) . Taken together , the observed CG and non-CG methylation patterns strongly argue that in the absence of nucleosome remodeling by DDM1 , DNA within heterochromatic nucleosomes is unavailable to methyltransferases . DDM1 belongs to an ancient protein family that includes mouse Lsh , inactivation of which causes extensive DNA hypomethylation ( Dennis et al . , 2001; Law and Jacobsen , 2010 ) . To assess whether Lsh , like DDM1 , permits methylation of DNA wrapped in nucleosomes , we examined published methylation data from Lsh null mouse embryonic fibroblasts and WT controls ( Yu et al . , 2014 ) in relation to nucleosome positions ( Teif et al . , 2012 ) ( Figure 1F ) . Because this analysis requires reliably well-positioned nucleosomes , we assessed regions of the mouse genome immediately flanking CTCF binding sites , which are surrounded by positioned nucleosome arrays ( Fu et al . , 2008 ) . As previously reported , we find that WT DNA methylation is periodic around CTCF binding sites , with clear depletion at nucleosome core regions ( Kelly et al . , 2012; Baubec et al . , 2015 ) ( Figure 1F ) . Cells with inactivated Lsh show a much stronger methylation periodicity , with far greater depletion at nucleosome cores ( Figure 1F ) . This contrast is most evident in the linkers that extend from the most reliably positioned nucleosomes on either side of the CTCF binding sites ( emphasized with arrows in Figure 1F ) , which show nearly WT levels of DNA methylation in Lsh mutant cells . Our results indicate that nucleosomes present barriers to mammalian as well as plant DNA methyltransferases , and that the conserved function of the DDM1/Lsh family of remodelers is to enable methylation of nucleosomal DNA . Arabidopsis TEs can be roughly grouped into heterochromatic and euchromatic elements ( Zemach et al . , 2013 ) . Heterochromatic TEs , which we analyzed in Figure 1 , tend to be longer , are methylated at CHH sites primarily by CMT2 and require DDM1 for methylation in all contexts ( Zemach et al . , 2013; Stroud et al . , 2014 ) . Euchromatic TEs are much shorter , less dependent on DDM1 , and CHH methylation at these elements is maintained by DRM1 and DRM2 ( Zemach et al . , 2013; Stroud et al . , 2014 ) . Because DRM1/2-dependent loci are short and often bordered by unmethylated DNA , analysis of methylation anchored to the centers of differentially methylated regions ( DMRs; see Materials and methods ) between drm1drm2 mutants and wild-type produces a sharp peak ( plots in the right column of Figure 2A–C ) . Nonetheless , anchoring methylation analysis to well-positioned nucleosomes shows that CG and CHG methylation at DRM-dependent loci is depleted over nucleosomes much as it is in heterochromatin ( Figure 2A–B , Figure 2—figure supplement 1 ) , demonstrating that nucleosomes in euchromatic TEs are refractory to DNA methylation . CHH methylation is also depleted over nucleosomes , but unlike in heterochromatin , the depletion is not obviously enhanced in any of the mutant genotypes ( Figure 2C , Figure 2—figure supplement 1 ) , probably because the RdDM pathway that guides DRM methyltransferases is largely independent of DDM1 but is associated with other remodelers ( Kanno et al . , 2004; Smith et al . , 2007; Zemach et al . , 2013; Han et al . , 2015 ) . As expected , CMT2-dependent loci recapitulate the behavior of heterochromatic TEs ( Figure 2D–F , Figure 2—figure supplement 1 ) . Due to the strong linker enrichment of methylation in h1ddm1 heterochromatin ( Figure 1A–E ) , even analyses centered on DMRs between cmt2 and wild-type without regard to nucleosome position produce periodic h1ddm1 methylation patterns ( plots in the right column of Figure 2D–E ) . Previous work demonstrated a global enrichment of DNA methylation on Arabidopsis nucleosomes ( Chodavarapu et al . , 2010 ) . We have so far described the opposite pattern , but our analysis has been confined to TEs . To identify regions of the genome that show the reported preferential methylation of nucleosomal DNA , we expanded our analysis to include all nucleosomes . We find that nucleosomes within genes possess a marked enrichment of CG methylation in WT , whereas TEs exhibit no depletion , so that methylation is enriched in WT nucleosomes on average ( Figure 3A; see Materials and methods ) . Separating genic nucleosomes by degree of positioning , as we did for heterochromatin ( Figure 1A ) , reveals an enrichment of DNA methylation in the nucleosome core that diminishes with increased positioning uncertainty ( Figure 3—figure supplement 1 ) . The correspondence between nucleosomes and DNA methylation within genes can be easily seen at individual loci ( Figure 3B ) . Interestingly , nucleosomes in the exons of genes show a clear enrichment of DNA methylation , whereas intronic nucleosomes do not ( Figure 3C ) , indicating that methylation is not intrinsically targeted to genic nucleosomes . The enrichment of methylation over nucleosomes in genes does not preclude the possibility that this methylation requires a remodeling activity . The ddm1 and h1ddm1 mutations do not eliminate methylation of genic nucleosomes ( Figure 3A , Figure 3—figure supplement 1 ) , which is consistent with the preferential requirement of DDM1 for heterochromatic methylation ( Zemach et al . , 2013 ) . However , because DDM1 does contribute to DNA methylation in genes ( Zemach et al . , 2013 ) , we asked whether DDM1 is needed for methylation of a subset of genic nucleosomes . Indeed , we find that lowly-expressed genes depend on DDM1 to maintain nucleosome methylation , whereas more highly expressed genes do not , on average , need DDM1 ( Figure 3D ) . Our definition of genes excludes TEs or sequences with TE-like methylation patterns ( Zemach et al . , 2013 ) , therefore this result is not caused by TE contamination of our gene anotations . Genes with lower levels of expression undergo less transcription-coupled nucleosome displacement ( Workman , 2006 ) , and their methylation may therefore be more dependent on DDM1-mediated remodeling than more highly expressed counterparts . Our results demonstrate that DDM1 facilitates methylation of nucleosome-wrapped DNA in a subset of genes as well as in heterochromatin . Perhaps the most compelling evidence that DNA methylation occurs on the surface of the nucleosome is the reported subtle 10 bp periodicity of methylation in all sequence contexts ( Chodavarapu et al . , 2010 ) – a periodicity that reflects the alternate facing of the major and minor grooves of DNA toward and away from the histone octamer . If the 10 bp periodicity is indeed caused by differential accessibility of DNA on the nucleosome surface , it should not be affected by mutation of a nucleosome remodeler . To assess whether DDM1 is required for this pattern , we examined DNA methylation periodicity in ddm1 plants and sibling controls at heterochromatic TEs . We anchored methylation analysis to the ends of 147 bp MNase fragments isolated in silico from our paired-end libraries , and calculated the strength of periodicity in the average per-base methylation signal with fast Fourier transform ( FFT ) computation ( Chodavarapu et al . , 2010; see Materials and methods ) . As expected , WT plants display a 10 bp methylation periodicity in all sequence contexts ( Figure 4A , Figure 4—figure supplement 1A–C ) . Lack of H1 does not appreciably alter the periodicity in non-CG contexts; CG methylation periodicity is weaker than WT ( Figure 4A ) , probably because the very high levels of heterochromatic CG methylation in h1 plants ( Zemach et al . , 2013 ) leave little room for oscillations . Importantly , the 10 bp periodicity is virtually absent in ddm1 heterochromatin in all contexts ( Figure 4A ) . In contrast , methylation of genic nucleosomes , which in most cases does not require DDM1 , displays a clear 10 bp periodicity in all genotypes examined ( Figure 4B , Figure 4—figure supplement 1D ) . To confirm our Arabidopsis results , we examined whether the 10 bp periodicity also declines in Lsh mutants . Because our analysis of Lsh methylation is based on published nucleosome data from WT cells ( Teif et al . , 2012 ) , and corresponding data from Lsh null cells is not available , we based our calculation of methylation periodicity on nucleosomal fragments mapping around the CTCF sites used in Figure 1F , where nucleosomes positioning is highly reliable ( Fu et al . , 2008 ) . We find a clear 10 bp periodicity in WT CG methylation , but in Lsh-/- the signal is noisy and does not exhibit a maximum at 10 bp ( Figure 4C , Figure 4—figure supplement 2 ) . Although this phenotype may be caused by anchoring the Lsh-/- methylation analysis to WT rather than Lsh-/- nucleosomes , bound CTCF sites generally possess very well positioned flanking nucleosomes ( Fu et al . , 2008; Stadler et al . , 2011 ) , suggesting that methylation periodicity is attenuated in the absence of Lsh . Thus , our results indicate that the 10 bp methylation periodicity observed in Arabidopsis and mouse requires a nucleosome remodeler , providing further support for the hypothesis that DNA methyltransferases cannot directly methylate nucleosomes . The depletion of nucleosome methylation in ddm1 and h1ddm1 plants ( Figure 1A–E , Figure 1—figure supplement 3 ) is reminiscent of the linker-specific DNA methylation patterns observed in three lineages of marine algae ( Huff and Zilberman , 2014 ) that diverged from one another , and from flowering plants and animals , over a billion years ago ( Parfrey et al . , 2011 ) . However , the contrast between the full methylation of linker DNA and the absence of nucleosomal methylation in marine algae ( Huff and Zilberman , 2014 ) is much greater than that observed even in the best positioned group of h1ddm1 nucleosomes ( Figure 1A ) . A potential explanation is that the relevant algal species have exceptionally well-positioned nucleosomes ( Huff and Zilberman , 2014 ) , whereas even the best positioned Arabidopsis nucleosome group we identified is likely quite heterogeneous . To test this hypothesis , we identified arrays of well-positioned Arabidopsis nucleosomes by using an unsupervised clustering algorithm ( Stempor et al . , 2016 ) to organize h1ddm1 CG methylation data centered on heterochromatic nucleosomes in the best-positioned group . One of the resultant clusters ( C2 , 4849 loci ) exhibits highly structured CG methylation that alternates with nucleosome positions ( Figure 5A ) . All genotypes show methylation periodicity at C2 in all sequence contexts ( Figure 5A–D ) . Remarkably , h1ddm1 linker CG methylation reaches near-WT levels in this cluster , whereas nucleosome core methylation is close to ddm1 levels ( Figure 5B ) . The periodicity of h1ddm1 non-CG methylation at C2 is also striking , with ddm1-like core methylation and linker methylation that is even higher than WT ( Figure 5C , D , Figure 5—figure supplement 1 ) . The h1ddm1 methylation patterns at C2 closely resemble those seen in algae over a billion years removed from the Arabidopsis lineage . We have shown that nucleosomes are strong impediments to DNA methylation and that the DDM1/Lsh remodeling proteins facilitate nucleosome methylation in vivo . It remains formally possible that DDM1/Lsh remodelers permit methyltransferases to work directly on the nucleosome surface or promote catalysis without affecting methyltransferase access . However , these interpretations are not parsimonious , and are inconsistent with the accumulating evidence that naked DNA is the preferred template for most DNA modifying enzymes . Such enzymes include cytidine deaminases ( Kodgire et al . , 2012 ) , the TET1 dioxygenase ( Kizaki et al . , 2016 ) , CAS9 ( Horlbeck et al . , 2016 ) , and a bacterial cytosine methyltransferase that has the same catalytic mechanism as eukaryotic enzymes ( Kelly et al . , 2012 ) . In conjunction with these observations , our data strongly support the hypothesis that cytosine methyltransferases generally require naked DNA , and that DDM1/Lsh nucleosome remodelers provide this substrate . We propose that DDM1/Lsh-type remodelers render a region of nucleosomal DNA accessible to DNA methyltransferases in a manner analogous to the remodeling mechanism proposed for Swi2/Snf2 ( Zofall et al . , 2006 ) . These remodelers , which are closely related to DDM1 , bind the nucleosome and translocate DNA from linker toward dyad , generating a nucleosome-free DNA loop ( Figure 6 ) ( Flaus et al . , 2006; Zofall et al . , 2006; Zhou et al . , 2016 ) . Such a loop would allow methyltransferases access to the DNA that would be biased by its orientation to the histone octamer surface , plausibly producing the observed subtle 10 bp methylation periodicity . The 10 bp periodicity could also arise if the formation or stability of nucleosomes is affected by the orientation of 5-methylcytosine toward the histone octamer ( Jimenez-Useche et al . , 2013 ) . Despite the overall importance of DDM1 for DNA methylation , many nucleosomes , particularly those in genes , are methylated normally without DDM1 ( Figure 3 ) . A likely explanation is that other nucleosome remodelers facilitate methylation of euchromatic loci . This hypothesis is consistent with the observations that the RdDM pathway requires remodelers other than DDM1 ( Kanno et al . , 2004; Smith et al . , 2007; Zemach et al . , 2013; Han et al . , 2015 ) , although their specific functions remain unknown . The partial restoration of the 10 bp methylation periodicity in h1ddm1 heterochromatin ( Figure 4A ) also suggests that remodelers other than DDM1 can mediate methylation of heterochromatic nucleosomes in the absence of H1 . This idea is supported by the reported ability of H1 to prevent nucleosome remodeling in vitro for some Swi2/Snf2-type remodelers ( Lusser et al . , 2005 ) and could explain why heterochromatic nucleosomes – and not just linkers – show higher methylation in h1 compared to WT , and in h1ddm1 compared to ddm1 ( Figures 1B and 5B–D ) . The likely involvement of multiple remodelers in euchromatic DNA methylation does not by itself explain the robust enrichment of methylation within exonic nucleosomes ( Figure 3 ) . The origins and biological functions of gene body methylation remain quite mysterious ( Zilberman , 2017 ) , and therefore any explanation of its features is inherently speculative . Given that introns do not exhibit preferential methylation of nucleosomes ( Figure 3C ) , the pathways that target methylation to genes are unlikely to favor nucleosomes , especially considering that methylation in genes and TEs is catalyzed by the same methyltranferases ( Law and Jacobsen , 2010 ) . Preferential methylation of nucleosomes within exons may be related to the function of gene body methylation . DNA methylation is known to affect the properties of nucleosomes , such as positioning and stability ( Davey et al . , 1997; Choy et al . , 2010; Jimenez-Useche et al . , 2013; Huff and Zilberman , 2014 ) . If gene body methylation influences splicing , other aspects of RNA processing , or the stability of transcript elongation ( Zilberman , 2017 ) , it may do so in part by altering the properties of exonic nucleosomes . Enrichment of methylation over exonic nucleosomes may therefore be a product of selection . Despite their preferential methylation , genic nucleosomes – at least those experiencing low levels of transcription – are refractory to MET1 methyltransferase activity , as evidenced by their decreased methylation in ddm1 mutants ( Figure 3D ) . The ability of nucleosomes and H1 to block DNA methylation in the absence of remodeling is dramatically illustrated by a set of heterochromatic loci with arrays of positioned nucleosomes ( Figure 5 , Figure 5—figure supplement 1 ) . Here , linker methylation that is close to or even above WT in h1ddm1 plants is juxtaposed with severe depletion of nucleosomal methylation . Methylation patterns at these loci closely resemble the natural patterns of several diverse species in which methylation is confined to linker DNA throughout the genome ( Huff and Zilberman , 2014 ) . This resemblance is unlikely to be coincidental . Species with linker-specific methylation are interspersed on the tree of life with species that do not limit methylation to linkers ( Figure 5—figure supplement 2; Huff and Zilberman , 2014 ) . This evolutionary history has been difficult to explain , particularly if one assumes that transitions to or from linker-specific methylation require changing the intrinsic ability of DNA methyltransferases to access nucleosomes . Our ability to impose an essentially linker-specific methylation pattern on a substantial fraction of the Arabidopsis genome by inactivating two chromatin proteins provides a plausible solution to this quandary . Species in which linker-specific methylation has been described have well-positioned nucleosomes ( Huff and Zilberman , 2014 ) and all but one of these species ( Micromonas pusilla ) lack the winged-helix motif of canonical H1 ( Figure 5—figure supplement 3 ) ( Kasinsky et al . , 2001; Worden et al . , 2009 ) . A transition to linker-specific methylation in a species with reliably positioned nucleosomes and low or absent linker histones would only require the uncoupling of nucleosome remodeling from DNA methylation maintenance . This could occur through inactivation of the relevant remodeler , or – more likely – by a mutation that severs the spatiotemporal connection between DNA methylation and remodeling . Such a mutation , if sufficiently advantageous , could also be fixed in a species with complex nucleosome positioning and high H1 . The linker-specific methylation patterns almost certainly offer a selective advantage to the species that have them , in part by contributing to nucleosome positioning ( Huff and Zilberman , 2014 ) . Thus , loss of H1 could be advantageous in the presence of a linker-specific methylation system , and reliable nucleosome positioning a consequence of such a system . Considering the many independent instances of complete DNA methylation loss among eukaryotes ( Zemach and Zilberman , 2010 ) , fixation of advantageous mutations that separate methylation from remodeling could plausibly occur multiple times during eukaryotic evolution . Such a model would easily explain the diverse methylation patterns observed among eukaryotes by providing the mechanism by which apparently radically different patterns can repeatedly evolve . ddm1-10 ( Zhang et al . , 2016 ) , h1 . 1 and h1 . 2 ( Zemach et al . , 2013 ) Arabidopsis thaliana ( Columbia ecotype ) were crossed to generate h1ddm1 and siblings for this study . Mutation of DDM1 causes permanent methylation losses at many loci that are not recovered after restoration of DDM1 activity ( Teixeira et al . , 2009 ) . We therefore used the ddm1-10 allele , which , unlike the ddm1-2 allele we used in a previous study ( Zemach et al . , 2013 ) , was not homozygous at any point prior to the generation of the compound h1ddm1 mutant reported here . For MNase-seq , RNA-seq , and bisulfite-seq samples , plants were germinated on agarose plates supplemented with Gamborg’s B-5 growth media ( GBP07 , Caisson Labs ) . Following germination , they were transferred to soil and grown on a 16hr-light/8hr-dark long-day schedule . For all experiments described here , rosette leaves of 1 month old plants were used . Genomic DNA ( gDNA ) was extracted from 1-month-old Arabidopsis thaliana rosette leaves with the DNeasy plant mini kit ( Qiagen , cat . no . 69104 ) per the manufacturer’s instructions . Libraries were prepared from roughly 500 ng of purified gDNA that was sheared to approximately 400 bp on a Misonix water bath sonicator , then purified using 1 . 2X volume of Agencourt Ampure beads ( referred to as ‘beads’ henceforth , cat . no . A63881 ) . Following ligation of methylated Truseq sequencing adapters ( Illumina Hayward , CA ) , bisulfite conversion of DNA was carried out according to manufacturer’s protocol ( Qiagen Epitect Kit , cat . no . 59104 ) except without using carrier RNA . DNA was purified twice with 1 . 2X beads and converted a second time to ensure complete bisulfite conversion of unmethylated cytosine . NEB next indexing primers ( cat . no . E7335S ) were used for generating multiplexed libraries during PCR amplification of libraries . Leaves were flash frozen in liquid nitrogen , pulverized with mortar and pestle on dry ice , and the resulting material was subjected to vortexing in Trizol ( Invitrogen , cat . no . 15596–026 ) . Chloroform was then added at one-fifth the total volume and further vortexing was carried out until the solution appeared homogenous . RNA was subsequently pelleted in ice-cold isopropanol . The resuspended RNA was subjected to rRNA removal with Ribo-zero plant kit ( Illumina , MRZPL1224 ) according to the manufacturer’s protocol . 50 ng of ribo-depleted RNA was used for library preparation with the Scriptseq kit ( Epicentre , cat . no . SSV21124 ) following the manufacturer’s protocol but with the following modifications: the RNA fragmentation step was extended to 10 min , and the temperature was increased to 90°C . Approximately 1 g of pulverized flash frozen rosette leaf tissue from F4 generation plants was resuspended on ice in nuclei isolation buffer ( 0 . 25 M sucrose , 15 mM PIPES pH 6 . 8 , 5 mM MgCl2 , 60 mM KCl , 15 mM NaCl , 0 . 9% Triton X-100 , 1 mM PMSF , 1X protease inhibitor cocktail ) and allowed to thaw for 15 min . The suspension was then strained once through Miracloth ( EMD-Millipore , cat . no . 475855 ) and the nuclear pellet was centrifuged in a swinging-bucket rotor at 2000 g in pre-chilled 5 ml conical tubes . Buffer was carefully aspirated from the loose pellet , which was subsequently resuspended in TM2 ( 10 mM Tris–HCl , pH 8 , 2 mM MgCl2 ) and washed twice in TM2 . The pellet was then resuspended in pre-warmed MNase digestion buffer ( 50 mM Tris-HCl , pH 8 , 5 mM CaCl2 ) and 25 units of MNase were added per 0 . 1 g initial starting material and incubated for 10 min at 37°C with periodic agitation; reactions were stopped with EGTA ( 20 mM final concentration ) . This was followed by the addition of SDS to a final concentration of 0 . 625% , increasing the temperature to 75°C , addition of RNase A for 10 min , and the addition of proteinase-K for an additional 15 min . DNA was then extracted with phenol/chloroform , pelleted in 70% EtOH , then resuspended in TE pH 8 . 0 . Verification of MNase digestion was carried out by fractionating one-tenth of the total reaction on a standard TAE gel and checking for a majority mononucleosomal fraction . Following removal of short DNA with beads , libraries were synthesized using total DNA ( without PCR amplification ) using the Nugen Ovation kit ( Cat . no . 0319 ) following the manufacturer’s protocol . All sequencing was carried out as single-end 100 bp reads , except MNase libraries , which were sequenced in paired-end 100 bp reads on Illumina HiSeq 2500 or −4000 at the QB3 Vincent Coates Genomic Sequencing Lab at UC Berkeley . Bisulfite sequencing reads were mapped with Bowtie ( Langmead et al . , 2009 ) using the bs-sequel pipeline ( available at http://dzlab . pmb . berkeley . edu/tools ) , with Bowtie settings allowing for up to two mismatches in the seed and reporting up to 10 matches for a given read . 100 bp paired-end MNase-seq reads were mapped using Bowtie2 ( Langmead and Salzberg , 2012 ) allowing for up to one mismatch in the seed region and only the best match reported . All alignments were made to the TAIR10 Arabidopsis thaliana genome assembly . The Kallisto ( Bray et al . , 2016 ) quant command was invoked with the following settings: --single --fr-stranded -b 100 l 320 s 30 . Kallisto output was fed into the R environment for processing with the Sleuth software package ( Pimentel et al . , 2017 ) , which was used to perform expression quantification and derive expression deciles . For initial identification of nucleosomes , the Python program iNPS ( Chen et al . , 2014 ) was used on our paired-end MNase-seq reads ( 120 to 180 bp , not inclusive ) , which were filtered using SAMtools ( Li et al . , 2009 ) . iNPS produces a number of nucleosome types depending on the shape of read distribution: only ‘main peaks’ from this output were taken from our biological replicates and used in identification of positioned nucleosomes . The other output from this program , the ‘like_Wig’ files , were used in visualization of MNase-seq data in the genome browser as well as for calculating average nucleosome enrichment across loci . We took iNPS nucleosome peaks and removed all that were wider than 140 bp . Remaining nucleosome overlap was calculated using the built-in ‘-wo’ option of bedtools ( Quinlan and Hall , 2010 ) ‘intersect’ function . The percent overlap of peaks from replicate one with peaks from replicate two was then used to define regions of the genome with reliably positioned nucleosomes . Nucleosome peaks that overlapped reciprocally more than 75% were classified as well-positioned ( left-most panel in Figure 1A , for example ) . Less-well-positioned nucleosomes were further defined in 25% increments of overlap , such that the next category overlapped more than 50% but less than 76% , and so on , generating four groups , plus a group that was comprised of other nucleosomes . To isolate differentially-methylated regions ( DMR ) outside of positioned nucleosomes ( see below for definition of DMRs ) , we used DMRs not overlapping any of the four groups defined above . The presumptive centers of these nucleosomal loci were calculated as the arithmetic means of the combined 5’- and 3’-most ends of the overlapping nucleosomes , allowing for multiple overlaps for a given nucleosome . Nucleosome groups per genotype are provided in GEO ( GSE96994 ) . Shared nucleosomes ( Figure 1—figure supplement 2 ) are defined as those nucleosomes whose calculated dyad is within 20 bp ( inclusive ) of the other genotype . For h1ddm1 , three biological replicates were generated for visualization but only two were used as input to determine nucleosome position because addition of a third replicate did not improve the positioning analysis . The third replicate is available in the GEO accession associated with this work . ‘Heterochromatic sequences’ in Figure 1 refers to merged TAIR10 TEs exhibiting >5% CG methylation that are in the upper two quintiles for H3K9me2 enrichment ( as calculated from Stroud et al . , 2014 ) and are longer than 30 bp ( complete merged TE bed file provided in GEO accession GSE96994 ) . TEs referred to related to Figure 3 are all merged TEs , regardless of H3K9me2 level . The TSS ( Figure 3A ) was defined as ±1 kb from the annotated TAIR10 TSS; while gene body was considered the portion downstream of 1 kb . Genes were defined to exclude annotated TEs or sequences methylated like TEs , as in ( Zemach et al . , 2013 ) . 50 bp windows with at least 10% WT CHH methylation ( mCHH ) and with more than 30% loss of mCHH in drm1drm2 or cmt2 were used to define DRM1/2 and CMT2 dependent DMRs , respectively . These windows also exhibited statistically significant decrease in mCHH from WT ( p<0 . 01 , Fisher’s exact test ) . We further filtered out DMRs that exhibited significant mCHH differences relative to WT in both genotypes , ensuring the DMRs are exclusive to the given genotype . After merging adjacent 50 bp DMR windows , there were 54 , 106 CMT2 DMRs and 8640 DRM1/2 DMRs . These coordinates are available in GEO accession GSE96994 . All DMRs overlap at least one annotated TE . Methylation plots in Figure 2 are centered either on ‘group1’ genotype-specific well positioned nucleosomes that overlap the indicated DMR by at least 1 bp or , for the ‘DMR’ plots in the same figure , we used the arithmetic center of DMRs that do not overlap with any of the nucleosomes identified in groups 1 through 4 , also in a genotype-specific manner ( e . g . h1 DMR plots are anchored to those DMRs which do not overlap any of the h1 group 1–4 nucleosomes ) . We aligned sequences with T-coffee ( Notredame et al . , 2000 ) in normal mode and plotted the resulting alignment file with Boxshade ( http://www . ch . embnet . org/ ) . The following sequences were used to perform multiple sequence alignment shown in Figure 5—figure supplement 3: Aureococcus anophagefferens , protein ID 72830 , scaffold_256:8187–9341; Ostreococcus lucimarinus , protein ID 18993 , Chr_20:191746–192084 ( + ) ; Emiliana huxleyi protein ID 448806/460675 , scaffold_1716:1449–2181 ( + ) /scaffold_11:31908–32637 ( - ) ; Micromonas pusilla , protein ID 1714 , scaffold_11:410445–411786 , Arabidopsis thaliana H1 . 1 ( AT1G06760 ) , Neurospora crassa hH1 ( ORF name NCU06863 ) , and Mus musculus H1a ( MGI:1931523 ) . CTCF sites ( Baubec et al . , 2015 ) were used as anchors for plotting averaged DNA methylation and MNase-seq data . Lsh mutant and WT MEF DNA methylation data ( Yu et al . , 2014 ) and WT MNase-seq data ( Teif et al . , 2012 ) were obtained through GEO ( GSE56151 and GSE40896 , respectively ) . Bin widths for methylation and MNase analysis were both set at 10 bp . For Arabidopsis , 147 bp MNase-seq fragments were isolated from one of the biological replicates for each genotype . The resulting reads were used as anchors to calculate mean DNA methylation using bisulfite data ( Chodavarapu et al . , 2010 ) ; this methylation vector was used as input to the TSA R program ( function ‘periodogram’ ) ( Chan and Ripley , 2012 ) , which generated the power spectrum over a range of frequencies . ‘spec’ output was plotted against the inverse of frequency ( in this case , bp ) , truncated to 30 bp and plotted . For mouse , 147 bp fragments from Teif et al . ( 2012 ) that overlapped CTCF sites ( ±1 kb ) ( Baubec et al . , 2015 ) were used as anchors for methylation averaging , and the strength of periodicity was calculated as above . For MNase-seq plotting , in order to compare reads mapped across different genotypes and multiple replicates , we normalized by multiplying the smoothed output of iNPS by the reads per million per nucleotide ( RPM ) . For instance , a 10 bp bin score of MNase-seq in sample A would be multiplied by 1 × 106/ ( number of mapped reads in sample A ) . Perl scripts ( http://dzlab . pmb . berkeley . edu/tools ) were used to generate enrichment score matrices of mapped data around genomic features of interest . These matrices were imported to R ( R Core Team , 2017; Davey et al . , 1997 ) for further processing and visualization using base plotting functions and the ggplot2 library ( Wickham , 2009 ) . Heatmaps were generated with Seqplots ( Stempor et al . , 2016 ) using the self-organizing map ( SOM ) clustering algorithm . Genome tracks are manual screenshots of our data displayed in IGV ( Robinson et al . , 2011 ) .	Living cells add chemical tags to their DNA to regulate which genes are switched on or off at any given time . These tags include methyl groups added to one of the letters of the DNA code called cytosine . Both plants and mammals need cytosine methylation to develop properly . This methylation also keeps sections of foreign DNA that may have invaded the cell in check . DNA inside the cell is tightly packed , wrapped around proteins to form spool-like structures called nucleosomes . Between each nucleosome is a short DNA segment called a linker region . The DNA wound into nucleosomes is generally inaccessible to other proteins , such as those that add methyl groups . Yet , in flowering plants and mammals , cytosine methylation occurs in both nucleosomes and in linker regions . It was not clear how DNA could be modified in the restrained setting of nucleosomes . Enzymes called nucleosome remodelers can loosen nucleosomes to allow other proteins to reach the DNA . Lyons and Zilberman asked whether cytosine methylation occurs on the nucleosome-bound DNA or if it requires enzymes like these to free the DNA from the constraints of the nucleosome . The experiments involved a plant called Arabidopsis thaliana and mouse cells grown in the laboratory . In mutant plants lacking a nucleosome remodeler called DDM1 , cytosine methylation occurred in the linker regions but not in the nucleosomes . Mouse cells lacking the mouse version of DDM1 also showed less cytosine methylation in the nucleosomes . These results suggest that nucleosomes are barriers to the enzymes that modify DNA . Nucleosome remodeling enzymes like DDM1 can overcome these obstacles to enable cytosine methylation of nucleosome-wrapped DNA . These findings imply that cytosine methylation is more easily established and maintained on nucleosome-free DNA . Abnormal patterns of DNA methylation have been linked to medical conditions – such as neurological disorders and cancers – and to plant defects that hamper agriculture . A better understanding of the process may in the future lead to ways to correct problems with cytosine methylation in these different contexts .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "plant", "biology" ]	2017	DDM1 and Lsh remodelers allow methylation of DNA wrapped in nucleosomes
ISWI-family nucleosome remodeling enzymes need the histone H4 N-terminal tail to mobilize nucleosomes . Here we mapped the H4-tail binding pocket of ISWI . Surprisingly the binding site was adjacent to but not overlapping with the docking site of an auto-regulatory motif , AutoN , in the N-terminal region ( NTR ) of ISWI , indicating that AutoN does not act as a simple pseudosubstrate as suggested previously . Rather , AutoN cooperated with a hitherto uncharacterized motif , termed AcidicN , to confer H4-tail sensitivity and discriminate between DNA and nucleosomes . A third motif in the NTR , ppHSA , was functionally required in vivo and provided structural stability by clamping the NTR to Lobe 2 of the ATPase domain . This configuration is reminiscent of Chd1 even though Chd1 contains an unrelated NTR . Our results shed light on the intricate structural and functional regulation of ISWI by the NTR and uncover surprising parallels with Chd1 . Eukaryotic cells package their DNA into chromatin . Chromatin organization allows cells to compact , protect and regulate their genomes . Nucleosomes are the primary building blocks of chromatin . These particles consist of ~150 bp of DNA that wrap almost twice around an octamer of histones . Nucleosomal DNA , however , is not accessible to most nuclear factors . Nature therefore evolved ATP-dependent nucleosome remodeling complexes that can alter the position or the structure of nucleosomes as necessary . Numerous remodeling complexes with distinct activities are active in any cell . Some move nucleosomes along DNA , eject histones or exchange them for histone variants , and some can even perform several of these activities ( Zhou et al . , 2016 ) . How the various remodelers are regulated in response to cellular needs is not well understood . Several remodelers , for instance , respond to post-translational modifications present on histones ( Swygert and Peterson , 2014 ) . Others are directly regulated by post-translational modifications ( Kim et al . , 2010 ) or react to small signaling molecules ( Zhao et al . , 1998 ) . Cells also adjust the subunit composition of remodeling complexes during development ( Lessard et al . , 2007 ) . All these examples indicate exquisite levels of controls exerted over remodeling complexes . The fact that mutations in subunits of human remodeling factors strongly associate with and in some cases drive cancers underscores the necessity to regulate remodeler activity ( Kadoch and Crabtree , 2015; Garraway and Lander , 2013 ) . Remodelers of the ISWI family – like most other remodelers – can reposition nucleosomes along DNA in a process termed nucleosome sliding . ISWI’s activity is directly regulated by the histone H4 N-terminal tail and by DNA that flanks the nucleosome , so called linker DNA . The regulation imposed by these epitopes has direct consequences for the biological output of ISWI remodelers . By measuring the length of linker DNA , ISWI can generate arrays of evenly spaced nucleosomes ( Lieleg et al . , 2015; Yang et al . , 2006; Yamada et al . , 2011 ) , a characteristic feature of chromatin . Arrays of nucleosomes can further compact . In the compacted state , the histone H4 N-terminal tail of one nucleosome contacts the acidic patch formed by H2A and H2B of a neighboring nucleosome ( Luger et al . , 1997; Dorigo et al . , 2004 ) . This interaction sequesters the H4 tail , which now is no longer available for binding to and stimulating the activity of ISWI . Thus , ISWI’s activity on the compacted chromatin would decrease , ensuring the unidirectionality of the reaction . This process is in line with the importance of some ISWI complexes in heterochromatin biology ( Bozhenok et al . , 2002 ) . How ISWI senses the H4 tail is largely unknown . Evidence points to the ATPase domain of ISWI directly binding the H4 tail ( Racki et al . , 2014; Mueller-Planitz et al . , 2013 ) , consistent with the tail directly influencing catalytic reaction steps ( Clapier et al . , 2001; Dang et al . , 2006 ) . However , a domain at the C-terminal side of ISWI , the HAND-SANT-SLIDE ( HSS ) domain , has been implicated in binding the H4 tail as well ( Boyer et al . , 2004; Grüne et al . , 2003 ) . Another layer of regulation is imposed by the non-catalytic subunit termed ACF1 , which associates with ISWI and sequesters the H4 tail under certain conditions ( Hwang et al . , 2014 ) . ISWI recognizes amino acids R17H18R19 within the H4 tail , which are part of a stretch of amino acids called basic patch ( Fazzio et al . , 2005; Hamiche et al . , 2001; Clapier et al . , 2002 ) . Notably , ISWI contains an identical motif , here called AutoN . Mutation of AutoN’s two arginines to alanines ( referred to as 2RA ) increased the DNA-stimulated ATPase activity and nucleosome sliding , and suppressed the dependence of ISWI’s ATPase and sliding activities on the H4 tail . According to the current model , AutoN directly binds to and blocks the H4-tail binding site , acting as a gatekeeper for the H4 tail . This model necessitates a conformational change of the NTR to allow binding of H4 ( Hwang et al . , 2014; Clapier and Cairns , 2012 ) . Indeed , a conformational change could be traced to AutoN upon nucleic acid binding ( Mueller-Planitz et al . , 2013 ) . Of note , the 2RA mutation diminished but did not abolish the H4-tail dependency , implicating also other regions in the H4 recognition process ( Clapier and Cairns , 2012 ) . The AutoN motif is embedded in a structurally and functionally poorly characterized domain referred to as the N-terminal region ( NTR ) . Besides AutoN , the NTR contains additional motifs: an acidic region that we termed AcidicN , the ‘post-post-helicase-SANT-associated' ( ppHSA ) motif , so named because it follows the post-HSA motif in remodelers of the Snf2 family ( Mueller-Planitz et al . , 2013; Szerlong et al . , 2008 ) , and a weakly conserved AT-hook ( Mueller-Planitz et al . , 2013; Aravind and Landsman , 1998 ) . Their functions remain unknown . Here , we systematically interrogated the functions of all conserved motifs within the NTR by mutagenesis and a series of quantitative biochemical assays in vitro and in vivo . We paid particular attention to probe for possible crosstalk between these motifs and the H4 tail to understand its recognition process . Using protein crosslinking followed by mass spectrometry and protein structural modeling we obtained information about the general structural architecture of the NTR-ATPase module . With similar approaches , we mapped the H4-tail binding site . We interpret our results within a unified structural and functional framework for the combined inhibition of ISWI by the NTR and recognition of the histone H4 tail . Contrary to current models , we propose that AutoN does not occlude the binding pocket of the H4 tail and that inhibition by AutoN involves a more elaborate mechanism than simple mimicry of the H4 basic patch . Multiple sequence alignment of ISWI homologs revealed several sequence motifs in the NTR of ISWI ( Mueller-Planitz et al . , 2013 ) . To assess their degree of conservation we queried the UniProt database for ISWI homologs ( Figure 1—figure supplement 1 ) . Sequence alignment of these candidates showed conservation of AutoN ( Clapier and Cairns , 2012 ) but also indicated that two other motifs , termed ppHSA and AcidicN , were at least as conserved ( Figure 1 ) . In contrast , an AT-hook ( Aravind and Landsman , 1998 ) was poorly conserved . Of note , a separate PSI-BLAST of the NTR of ISWI revealed conservation of ppHSA across multiple families of remodelers , including Snf2 , Lsh and Ino80 , suggesting shared function ( Figure 1F ) . ppHSA and AcidicN have not been characterized so far . 10 . 7554/eLife . 21477 . 003Figure 1 . The NTR of ISWI contains several conserved sequence motifs . ( A ) Schematic representation of the ISWI domain composition . The grey inset shows the sequence and motifs of the NTR . Arrows indicate amino acids within the NTR of Drosophila ISWI that crosslinked to Lobe 2 of the ATPase domain ( Table 1 ) . HSS , HAND-SANT-SLIDE domain . ( B–E ) Sequence logos showing the sequence conservation of ( B ) ppHSA , ( C ) AT-hook , ( D ) AutoN , and ( E ) AcidicN . X-Axis values are amino acid positions in D . melanogaster ISWI . See Figure 1—figure supplement 1 for full alignment . ( F ) Alignment of the ppHSA motif of Drosophila ( Dm ) ISWI with the human ( Hs ) ISWI homologs SNF2H and SNF2L and representatives of unrelated remodeler families . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 00310 . 7554/eLife . 21477 . 004Figure 1—figure supplement 1 . Alignment of ISWI homologs from various organisms . Search for homologous proteins and alignment was done using HHblits ( toolkit . tuebingen . mpg . de/hhblits ) . 26 Sequences lacking an NTR were manually deleted from the alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 004 To study its physiological role , we serially truncated the NTR of Isw1 in Saccharomyces cerevisiae ( Figure 2A ) and tested whether these truncation variants complemented a previously characterized growth defect of a yeast triple knockout ( TKO ) strain lacking three remodelers ( ΔISW1 , ΔISW2 , ΔCHD1 ) at elevated temperatures ( Tsukiyama et al . , 1999 ) . To assess whether complementation was dependent on the expression level , the alleles were placed under the control of synthetic promoters of varying strengths ( Blazeck et al . , 2012 ) . Protein expression levels were measured by Western blot analysis ( Figure 2—figure supplement 1E ) . 10 . 7554/eLife . 21477 . 005Figure 2 . Functional importance of the NTR of yeast Isw1 in vivo . ( A ) Successive N-terminal truncation mutants of Isw1 . Note that Isw1ΔNTR lacked the entire N-terminus up to the first seven residues of AcidicN ( Figure 1E ) . ( B ) Complementation assay with Isw1ΔppHSA . A yeast strain lacking ISW1 , ISW2 and CHD1 ( TKO ) was transformed with Isw1 derivatives under control of promoters of varying strengths . In comparison to a strain lacking only ISW2 and CHD1 ( DKO ) , Isw1WT fully complemented the growth phenotype at elevated temperatures ( 37°C ) . In contrast , Isw1ΔppHSA did not complement at any expression level . Results for other Isw1 variants can be found in Figure 2—figure supplement 1 . Growth was assessed by spotting tenfold serial dilutions of liquid cultures . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 00510 . 7554/eLife . 21477 . 006Figure 2—figure supplement 1 . Complementation assay with N-terminal truncation variants of Isw1 . ( A–D ) Growth assays as in Figure 2B . Expression levels were estimated by Western analysis ( see panel E ) . Results in D ( 30°C and 37°C ) are replotted from Figure 2 . ( E ) Exemplary Western blot to determine relative expression levels ( tabulated in A–D ) using an Anti-TAP antibody . TAP-tagged Isw1 variants under control of the indicated promoter were expressed . Their expression level was normalized against genomically TAP-tagged wild-type Isw1 . Errors are minimal and maximal values of two technical replicates . H3 served as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 006 Expression of none of the N-terminal truncation variants fully complemented the growth phenotype , indicating functional relevance of the NTR in vivo . In contrast , the TKO strain that was complemented with full-length Isw1 grew essentially as well as the ΔISW2 , ΔChd1 double knockout strain ( DKO; Figure 2B , Figure 2—figure supplement 1A ) . Isw1 variants that lacked the AutoN-AcidicN region in addition to ppHSA grew modestly better than Isw1ΔppHSA , in line with the general inhibitory nature of AcidicN and AutoN ( compare rows 1 and 2 of Figure 2—figure supplement 1B , C to the same rows in D; see also below ) . We noted a pronounced toxicity of all Isw1 mutants as indicated by slow growth at elevated expression levels ( for instance , compare row four with row five in Figure 2—figure supplement 1B , C , D ) . Full-length Isw1 , on the other hand , was not toxic at comparable expression levels ( Figure 2—figure supplement 1A ) . Toxicity at high expression levels could be caused by structural instability of the N-terminally truncated Isw1 variants . Indeed , analogous ISWI derivatives from Drosophila melanogaster proved difficult to purify ( see below ) , supporting the notion that mutations in the NTR destabilize ISWI structure . Toxicity of the Isw1 NTR deletions precluded a detailed analysis in vivo . Importantly , the in vivo results left open the possibility that NTR-deleted Isw1 was catalytically inactive . We therefore continued to study the function of the NTR motifs in vitro using purified Drosophila ISWI proteins . Although ISWI variants carrying mutations or deletions in the NTR generally expressed well , we failed to purify them using standard protocols . For each ISWI variant , we screened through a variety of expression and purification strategies to improve the yield of soluble protein . The strategies that we employed included fusion to solubility tags ( Z2 , GB1 , NusA , TrxA ) , fusion to or co-expression of chaperones ( trigger factor , GroES/GroEL , DnaK/DnaJ/GrpE ) and inclusion of protease sites ( 3C ) at three locations in the NTR to cleave off parts of the N-terminus after purification . The strategies that proved successful are summarized schematically in Figure 3—figure supplement 1 and Figure 6—figure supplement 1 . We first benchmarked the DNA- and chromatin-stimulated ATPase activities of ISWI that lacked ppHSA ( ISWIΔppHSA ) or both ppHSA and AT-hook ( ISWIΔppHSA; ΔAT-hook ) against the activity of wild-type ISWI ( ISWIWT ) . We used saturating ATP and nucleic acid concentrations as indicated by control experiments with varying levels of ligands ( Figure 3—figure supplement 2 ) . DNA- and chromatin-stimulated ATPase rates of the truncation mutants varied by no more than 1 . 8-fold from ISWIWT ( Figure 3A , B ) indicating that ppHSA and AT-hook were largely dispensable for ATP hydrolysis and for proper recognition of chromatin . 10 . 7554/eLife . 21477 . 007Figure 3 . The ppHSA motif is largely dispensable for catalysis . ( A ) N-terminal truncation mutants of Drosophila ISWI . ( B ) DNA- and nucleosome-stimulated ATP turnover . ATPase rates were measured in the presence of saturating concentrations of ATP ( 1 mM ) , DNA ( 0 . 2 g/l ) or nucleosomes ( 0 . 1 g/l ) . Errors for nucleosome-stimulated rates of ISWI deletion mutants are minimal and maximal values of two independent measurements , and s . d . for all other measurements ( n ≥ 4 ) . ATPase rates in absence of nucleic acids were <0 . 022 s−1 for all ISWI variants ( data not shown ) . ( C ) Remodeling activity was determined by measuring the accessibility changes of a unique KpnI restriction site in a 25-mer nucleosomal array ( 100 nM nucleosomes , 300 nM enzyme ) . Errors are s . d . ( n ≥ 3 ) except for ISWIΔppHSA; ΔAT-hook for which minimal and maximal values of two independent measurements are shown . Raw data of the remodeling assay can be found in Figure 3—figure supplement 3 . Color code as in panel B . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 00710 . 7554/eLife . 21477 . 008Figure 3—figure supplement 1 . Cloning and purification of N-terminal truncation variants of Drosophila ISWI . ( A ) Construct design . Only the NTR region and N-terminally fused affinity ( His6 ) and solubility tags ( Trigger factor ) are shown ( not to scale ) . Blue arrowheads indicate a TEV cleavage site , orange arrowheads indicate a 3C protease cleavage site that were used to cleave off the tag . ( B ) Coomassie stained SDS-PAGE of purified recombinant ISWI variants . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 00810 . 7554/eLife . 21477 . 009Figure 3—figure supplement 2 . Saturation controls for ISWIWT and ISWIΔppHSA in ATPase assays . ( A , B ) Linearized pT7blue DNA was titrated over a 16-fold range . 0 . 2 mg/ml were saturating for ISWIWT . ( A ) and ISWIΔppHSA ( B ) . ( C , D ) Titration of chromatin assembled on linearized pT7blue DNA . 0 . 1 mg/ml were close to saturation for ISWIWT ( C ) and ISWIΔppHSA ( D ) . Errors are s . d . ( n ≥ 7 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 00910 . 7554/eLife . 21477 . 010Figure 3—figure supplement 3 . Determination of the rate constants for remodeling ( kobs; Figure 3C ) for ISWIWT and N-terminal truncation mutants of ISWI . ( A ) Exemplary remodeling time courses for ISWIWT , ISWIΔppHSA and ISWIΔppHSA; ΔAT-hook . Asterisks mark a contaminating non-nucleosomal DNA ( competitor DNA ) that was not completely removed during preparation of nucleosomal arrays . Mock: Sample lacking ISWI . ( B ) Quantification of time courses shown in ( A ) . Data were fit to a single exponential function to extract the rate constant kobs . The reactions progressed similarly fast when 100 nM and 300 nM enzyme were employed , suggesting saturation of chromatin . Because ISWIΔppHSA; ΔAT-hook at 50 nM was substoichiometric to nucleosomes ( 100 nM ) , it remodeled noticeably more slowly than at 300 nM . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 010 To evaluate whether ppHSA and AT-hook were required to efficiently couple ATP hydrolysis to nucleosome remodeling , we employed a quantitative remodeling assay . This assay monitors remodeling of a single nucleosome in the context of a 25-mer nucleosomal array by measuring the remodeling-dependent exposure of a unique restriction enzyme site originally occluded by the nucleosome ( Mueller-Planitz et al . , 2013 ) . Time courses of the remodeling reaction were fit to single exponential functions to extract the observed remodeling rate constant kobs ( Figure 3C; Figure 3—figure supplement 3 ) , which provided us with a quantitative measure to compare the remodeling activities of ISWI and its derivatives . Remodeling was affected only modestly by deletion of parts of the NTR ( 3 . 3- and 1 . 4-fold for ISWIΔppHSA and ISWIΔppHSA , ΔAT-hook , respectively; Figure 3C ) . In conclusion , ATPase and remodeling data suggested that both ppHSA and AT-hook are not absolutely required for catalysis in vitro . The modest decreases in remodeling activities could be due to lower stability of these enzymes ( see above ) . We speculated that the NTR might stabilize the structure of ISWI by adopting a similar configuration as the two chromo domains of the related remodeler Chd1 . Like the NTR , the chromo domains are located directly N-terminal to the ATPase module . Notably , they bridge over and pack against the second ATPase lobe , presumably locking the ATPase in an inactive state ( Figure 4A ) ( Hauk et al . , 2010 ) . 10 . 7554/eLife . 21477 . 011Figure 4 . The NTR contacts Lobe 2 of the ATPase domain . ( A ) Surface representation of the Chd1 crystal structure ( PDB code 3MWY ) ( Hauk et al . , 2010 ) . ATPase Lobe 1 and 2 are colored dark and light grey , respectively , and the N-terminal chromo domains cyan . ( B ) Homology model of the ISWI ATPase domain ( Forné et al . , 2012 ) . Cyan: hypothetical binding interface of the ISWI NTR ( see main text ) , red: position of Bpa substitution ( H483 ) . ( C–E ) Mass spectrometric validation of the crosslink XL1 ( Table 1 ) formed between Bpa at position 483 and an NTR peptide . ( C ) Isotopic distribution of the crosslinked peptide . ( D ) UV-dependent increase of the signal for the crosslinked peptide . Extracted ion chromatograms of the ions were used for the quantification . ( E ) High resolution , high accuracy MS2 fragmentation spectrum . Top right: summary of observed product ions mapped onto the sequence of the crosslinked peptide . B: Bpa . ( F ) Predicted docking interface of AcidicN ( blue and dark blue ) , AutoN ( cyan and dark blue ) and overlapping regions ( dark blue ) in the structural model of ISWI . The predicted interface for AcidicN overlaps with the interface for the acidic helix of the N-terminal chromo domains of Chd1 ( orange ) ( Hauk et al . , 2010 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 01110 . 7554/eLife . 21477 . 012Figure 4—figure supplement 1 . The effect of the H483B mutation on chromatin remodeling . With 0 . 3 µM enzyme , the observed rate constant for remodeling ( kobs ) was ~threefold affected by the mutation . Note , however , that ISWIH483B , in contrast to ISWIWT , was not fully saturating at this concentration , as suggested by the saturation control ( 0 . 1 µM enzyme ) . Error bars are s . d . ( n ≥ 3 ) for ISWIWT and minimal and maximal values of two independent measurements for ISWIH483B . Data marked with an asterisk ( * ) was replotted from Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 01210 . 7554/eLife . 21477 . 013Figure 4—figure supplement 2 . Validation of additional crosslinks detected in the ISWIH483B dataset . ( A–C ) Crosslink XL7 ( Table 1 ) to P75 . ( D–F ) Crosslink XL6 to P71K72 . For description of data , see Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 01310 . 7554/eLife . 21477 . 014Figure 4—figure supplement 3 . Structural predictions of NTR elements . ( A ) Predicted docking interface of AcidicN . The cyan to blue color scale denotes low to high contact probabilities . The location of AcidicN in the MtISWI crystal structure ( PDB 5JXR ) , which became available during the revision of this study , is shown in red . ( B ) Predicted docking interface ( blue ) of the AutoN-AcidicN peptide in the structural model of ISWI . During docking , both Lobe 1 and Lobe 2 were present ( cf . the docking interface shown in Figure 4F ) . The acidic helix of the N-terminal chromo domains of Chd1 is shown in orange for reference . ( C ) Structure prediction of a peptide comprising AutoN and AcidicN ( DHRHRKTEQEEDEELL ) by PEP-FOLD and I-TASSER . ( D ) Helical Propensity of amino acids 23–112 of ISWI predicted by four different algorithms ( see legend ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 014 To explore , we first determined the binding interface of the chromo domains ( amino acids 239–284 ) on Lobe 2 of the ATPase module using the PISA algorithm ( www . ebi . ac . uk/pdbe/pisa/ ) and visualized the analogous surface on a homology model of ISWI ( Figure 4B; cyan ) . We then site-specifically inserted the UV-crosslinking amino acid p-benzoyl-p-phenylalanine ( abbreviated Bpa or B ) into this hypothetical binding interface in ISWI ( H483B; Figure 4B , red ) using established strategies ( Forné et al . , 2012; Chin et al . , 2002 ) . We first tested whether mutagenesis of H483 impacted catalysis . The H483B mutation diminished the DNA- and chromatin-stimulated ATPase activity of full-length ISWI by fourfold each , a result that may not be surprising given that the mutation is located in the conserved ‘block D’ of Snf2 ATPases ( Flaus et al . , 2006 ) ( Figure 6—figure supplement 4 ) . Importantly , the remodeling activity of ISWIH483B was reduced to a similar degree ( threefold ) , indicating that the efficiency of remodeling per hydrolyzed ATP was unchanged ( Figure 4—figure supplement 1 ) . We conclude that ISWIH483B , albeit hydrolyzing ATP more slowly than ISWIWT , efficiently coupled ATP hydrolysis to chromatin remodeling , which suggested that the mutant remained structurally largely intact . Also , auto-regulation of ISWIH483B by its NTR was unperturbed because mutagenesis of the NTR had analogous effects on ISWIWT and ISWIH483B ( see below ) , further justifying the use of ISWIH483B for crosslinking experiments . Crosslinking of full-length ISWIH483B was induced by UV irradiation , and the crosslinks were mapped by high accuracy mass spectrometry ( MS ) ( Forné et al . , 2012; Mueller-Planitz , 2015 ) . Remarkably , ISWIH483B crosslinked to several positions in the NTR within or adjacent to the ppHSA motif ( Figure 1A , arrows; Figure 4C–E; Figure 4—figure supplement 2; Table 1 ) . We independently replicated these crosslinking results with a truncated form of ISWI ( ISWI26-648 ) , which lacked the HSS domain and non-conserved N-terminal amino acids ( data not shown ) . 10 . 7554/eLife . 21477 . 015Table 1 . Overview of crosslinks formed by ISWIH483B . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 015IDMass ( D a ) Error ( ppm ) Bpa peptideTarget peptideSequence* , †SiteSequence†SiteXL13220 . 4946−1 . 9LDGQTPBEDRNR483QTEIFTHFMoxTNSAK‡59–60XL23204 . 5056−3 . 7LDGQTPBEDRNR483QTEIFTHFMTNSAK‡60–61XL32950 . 3474−1 . 0LDGQTPBEDR483QTEIFTHFMoxTNSAK‡55–59XL42934 . 3571−2 . 6LDGQTPBEDR483QTEIFTHFMTNSAK‡59–61XL52207 . 0968+0 . 2LDGQTPBEDRNR483SPTKPK‡69–72XL61936 . 9645−6 . 1LDGQTPBEDR483SPTKPK‡71–72XL71736 . 8594−6 . 1LDGQTPBEDR483GRPK75B symbolizes Bpa . †Crosslinked amino acids are underlined; oxindicates oxidized methionine ( +15 . 9949 Da ) . ‡Precise attachment sites not distinguishable from data . In our previous work , we incorporated Bpa in a variety of places on Lobes 1 and 2 of the ATPase domain but never observed crosslinks to the NTR ( Forné et al . , 2012; Mueller-Planitz , 2015 ) . We therefore suggest that the ppHSA motif specifically docked to a location in proximity of amino acid 483 in Lobe 2 . Docking of the NTR against Lobe 2 may be necessary for the structural integrity of ISWI-type remodelers ( see above ) . The presence of ppHSA in other remodelers ( Snf2 , Lsh and Ino80; Figure 1F ) predicts similar functions beyond the ISWI family . If the NTR is structurally close to Lobe 2 of the ATPase module , AutoN and the neighboring AcidicN motif may also be able to contact Lobe 2 . To explore this idea , we performed in silico docking studies to predict the binding site of AutoN and AcidicN . We carried out three independent docking runs to model the interaction of Lobe 2 with AutoN , AcidicN and AutoN-AcidicN , respectively ( see Materials and methods for details ) . All three ab initio docking runs yielded a large cluster of models that identified the preferred binding site for AutoN and AcidicN ( Figure 4F; Figure 4—figure supplement 3A ) . Docking of scrambled peptides as a control partially diminished the preference for this binding pocket ( data not shown ) . Docking of AutoN-AcidicN against a homology model comprising both ATPase lobes gave very similar results , suggesting specificity of the motifs for binding to Lobe 2 ( Figure 4—figure supplement 3B ) . We validated the docking results by mutagenesis further below . Strikingly , AcidicN , which is predicted to be α-helical ( Figure 4—figure supplement 3C , D ) , contacted Lobe 2 precisely where an acidic helix of the chromo domains of Chd1 bound ( Hauk et al . , 2010 ) , which suggested conservation of this binding mode . Based on our results , we propose the NTR to adopt a structural architecture akin to the chromo domains of Chd1 ( Figure 4A ) despite complete lack of sequence conservation between the two . Due to sequence similarity , the H4 tail and AutoN may compete for the same binding site ( Hwang et al . , 2014; Clapier and Cairns , 2012 ) . We thus set out to identify the H4-tail binding pocket within ISWI and compare it to the predicted AutoN interaction surface . We adopted two complementary crosslinking approaches . First , we used two different H4-tail peptides , which carried a Bpa moiety either at amino acid 1 or 10 ( T1B and L10B peptides , respectively ) , and bound these peptides to ISWI26-648 in the presence of DNA ( Mueller-Planitz et al . , 2013 ) . After irradiation , a lower-mobility band was detected by SDS-PAGE , which suggested successful crosslinking ( Figure 5—figure supplement 1A , E ) . We mapped several crosslinks of the H4 peptides to Lobe 2 by MS ( Figure 5—figure supplement 1A–F; Table 2 ) . Control experiments showed that the T1B H4 peptide stimulated the ATPase activity like a wild-type H4 peptide ( Figure 5—figure supplement 2A ) . 10 . 7554/eLife . 21477 . 016Table 2 . Overview of H4-tail mediated crosslinks . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 016IDReliabilityH4Remodeler constructMass ( Da ) Error ( ppm ) H4 peptideRemodeler peptideSequenceSiteSequence†SiteXL11highnucleosomalISWIWT2034 . 8571−0 . 2XGR1QIQEFNMDNSAK495XL12highnucleosomalSNF2H2251 . 9753−0 . 4GXGK10VLDILEDYCMWR520XL13ahighpeptideISWI26-6481648 . 7601−0 . 4BGR1LDGQTPHEDR482XL13bhighpeptideISWI26-6481918 . 9052−0 . 9BGR1LDGQTPHEDRNR482XL13chighpeptideISWI26-6483340 . 5374−2 . 8BGR1LDGQTPHEDRNRQIQEFNMDNSAK482XL14mediumnucleosomalSNF2H2222 . 9624−1 . 5XGR1VLDILEDYCMWR‡519–22XL15mediumpeptideISWI26-6481257 . 6261+2 . 3BGR1MVIQGGR578XL16mediumpeptideISWI26-6481424 . 7832−3 . 9BGR1IVERAEVK568XL17mediumpeptideISWI26-6481453 . 7998−4 . 6GBGK10IVERAEVK568B symbolizes Bpa; X symbolizes Benzophenone-labeled cysteine . †Crosslinked amino acids are underlined . ‡Precise attachment sites not distinguishable from data . Because the peptides may not exclusively bind ISWI in the physiological binding pocket , we pursued a second approach . We reconstituted entire nucleosomes bearing a photo-reactive benzophenone on the N-terminal tail of H4 . Benzophenone labeling was achieved by chemical modification of single cysteine mutants of H4 ( T1C and L10C ) . These nucleosomes bound to full-length ISWI and stimulated its ATPase activity like wild-type nucleosomes ( Figure 5—figure supplement 2B , C ) suggesting that they were properly recognized by the remodeler . UV-irradiation of full-length ISWI bound to benzophenone-labeled T1C nucleosomes retarded the mobility of the remodeler during SDS-PAGE , indicative of successful crosslinking ( Figure 5A ) . MS analysis mapped a crosslink to Lobe 2 of ISWI ( Figure 5B–D ) . We repeated these crosslinking experiments with the human ISWI homolog SNF2H . Both T1C- and L10C-labeled nucleosomes crosslinked to Lobe 2 of SNF2H ( Figure 5—figure supplement 1G–J ) . In summary , two very different crosslinking approaches , one employing Bpa-containing peptides and one using benzophenone-derivatized nucleosomes , consistently yielded crosslinks between the H4 tail and Lobe 2 of the ATPase domain . Table 2 lists all crosslink candidates , classified in terms of their reliability ( see Materials and methods ) . Notably , methionine residues were overrepresented as targets of the photo-crosslinking approach , consistent with the known preference of benzophenones for methionine ( Wittelsberger et al . , 2006 ) . In summary , our data strongly indicated the H4-tail binding site to reside on or close to Lobe 2 . 10 . 7554/eLife . 21477 . 017Figure 5 . The binding sites of the NTR and the H4-tail on Lobe 2 are proximal . ( A–D ) Crosslinking of nucleosomes containing benzophenone-labeled H4 to ISWI . ( A ) Crosslinking time course analyzed by SDS-PAGE and Coomassie staining . The asterisk marks a UV-irradiation dependent band of lower mobility containing the crosslink mapped in B–D . ( B–D ) Mapping and validation of a crosslink ( XL11; Table 2 ) formed in the upshifted band in A . Isotopic distribution of the crosslinked peptide , MS2 spectrum and quantification as in Figure 4 . ( E ) Crosslink-guided in silico docking of an H4 peptide ( amino acids 1–20 ) to ISWI . The predicted docking interface of the H4 tail on Lobe 2 is illustrated in a yellow and red color scale , which indicates low to high contact probabilities between the docked H4 tail and Lobe 2 . The contact probabilities were calculated from a family of 383 docked structures ( see Materials and methods ) . For comparison , the predicted docking interface of the NTR is shown in shades of blue ( see Figure 4F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 01710 . 7554/eLife . 21477 . 018Figure 5—figure supplement 1 . Additional crosslinks between the H4 tail and ISWI or SNF2H . ( A–D ) SDS-PAGE analysis and MS mapping of crosslinks formed between a synthetic H4 tail peptide containing Bpa at amino acid one to the following amino acids in ISWI26-648: H482 ( A; XL13a in Table 2 ) ; M578 ( B; XL15 ) ; R568 ( C; XL16 ) . ( E–F ) MS mapping and SDS-PAGE analysis of crosslinks formed between a synthetic H4 tail peptide containing Bpa at amino acid 10 to R568 of ISWI26-648 ( XL17 ) . Asterisks next to SDS gels mark upshifted bands indicative of successful crosslinking . ( G , H ) SDS-PAGE analysis and MS mapping of crosslinks formed by benzophenone-labeled T1C nucleosomes to C519M520W521R522 of SNF2H ( XL14 ) . ( I , J ) SDS-PAGE analysis and MS mapping of crosslinks formed by benzophenone-labeled L10C nucleosomes to position 520 of SNF2H ( XL12 ) . B: Bpa , X: Benzophenone-labeled cysteine , 0N40: Mononucleosomes with 40 bp of DNA flanking one side of the nucleosome . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 01810 . 7554/eLife . 21477 . 019Figure 5—figure supplement 2 . Controls for possible adversary effects of covalent modifications of the H4 tail . ( A ) H4 tail peptides ( amino acid 1–24 ) carrying a T1B substitution stimulated the ATPase of ISWIWT ( 0 . 5 µM ) activity similarly well as WT tail peptides in the presence of 1 . 2 g/l salmon sperm DNA . In contrast , a scrambled sequence with a Bpa moiety at position one did not noticeably stimulate the ATPase . The ATP concentration was 1 mM . Scrambled and WT peptide data were replotted from ref . 15 . ( B ) T1C and L10C mononucleosomes labeled with 4- ( N-Maleimido ) benzophenone ( 4MBP ) stimulated the ATPase activity of ISWIWT similarly well as WT nucleosomes . A reaction without nucleosomes ( − ) served as a control . ( C ) WT and 4MBP-labeled T1C and L10C mononucleosomes ( 200 nM ) bound ISWIWT ( 0 to 400 nM ) similarly well in an electrophoretic mobility shift assay . Samples were resolved on a 5% native polyacrylamide gel . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 01910 . 7554/eLife . 21477 . 020Figure 5—figure supplement 3 . Surfaces on Lobe 2 that were sampled by selected amino acids in the H4 tail during crosslink-guided structural docking . ( A ) Interaction surface of histone H4 T1 on Lobe 2 of ISWI . Crosslinked amino acids ( Table 2 ) are shown as spheres . Blue , high confidence crosslink positions used for modeling; black and grey , lower confidence crosslink positions . Precise attachment sites are not available for XL14 ( grey ) . ( B ) Interaction surface of H4 L10 . Coloring of spheres as in ( A ) . ( C ) Predicted interaction surface of H4 K16 . H4 K16 from the crystal structure of MtISWI Lobe 2 in complex with an H4 peptide ( PDB 5JXT ) is shown as stick representation for reference . The color scales indicate contact probabilities between individual amino acids in the H4 tail and Lobe 2 across a family of 383 structural models . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 020 To identify the H4-tail binding pocket we turned to crosslink-guided in silico docking of the H4-tail peptide . We only used the five crosslinks for this analysis that passed stringent quality controls ( Table 2 , high reliability; see also Materials and methods ) . The predicted docking interface is visualized in Figure 5E and Figure 5—figure supplement 3 . Note that not all lower quality crosslinks were compatible with this binding mode , possibly because the H4-tail peptide bound flexibly or in multiple binding modes ( Racki et al . , 2014 ) . Some of these modes may not be strongly populated or functionally active as crosslinking can in principle trap fleeting intermediates . We also cannot rule out false positives among the lower quality candidates . Interestingly , the predicted docking interface of AcidicN was in close proximity to the H4-tail interface . This prompted us to investigate the function of AcidicN and – in the following section – its potential involvement in the H4-tail recognition process . To study its function , we replaced three or six negatively charged amino acids in AcidicN by uncharged ones using conservative E to Q and D to N mutations . These mutants were denoted ISWI+3 and ISWI+6 respectively ( Figure 6A ) . To improve solubility , ISWI+6 was fused to a solubility tag ( Z2-tag; Figure 6—figure supplement 1 ) . Control experiments ruled out interference of the Z2-tag on catalytic properties of ISWI ( Figure 6—figure supplement 2A , B ) . 10 . 7554/eLife . 21477 . 021Figure 6 . AcidicN is a strong negative regulator of the ATPase . ( A ) Design of AcidicN derivatives of ISWI ( see also Figure 6—figure supplement 1A ) . ( B ) Effects of AcidicN mutation on ATP hydrolysis in absence or presence of saturating concentrations of DNA and chromatin . In absence of DNA , ATPase activities of ISWIWT ( # ) and ISWI+3 ( § ) were ≤0 . 06 s−1 . Errors are s . d . ( n ≥ 4 ) . ( C ) Effects of AcidicN mutation on the remodeling activities . Nucleosomal arrays containing wild-type H4 were used . Errors are s . d . ( n ≥ 3 ) except for ISWI+3 for which minimal and maximal values of two independent measurements are shown . Color code as in panel ( B ) . Raw data of the remodeling assay can be found in Figure 8—figure supplement 1 . Results for ISWIWT ( ) are replotted for comparison from Figure 3B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 02110 . 7554/eLife . 21477 . 022Figure 6—figure supplement 1 . AcidicN and AutoN mutants . ( A ) Construct design . Only the NTR region of ISWI including affinity and solubility tags are shown ( not to scale ) . Blue arrowheads indicate a TEV cleavage site . Tags were removed by protease cleavage as indicated . ( B ) Coomassie-stained SDS-PAGE of purified recombinant proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 02210 . 7554/eLife . 21477 . 023Figure 6—figure supplement 2 . Comparison of ATPase and remodeling activities of ISWI control variants used in this study . ( A ) The Z2 solubility tag did not interfere with DNA- and chromatin-stimulated ATPase activities . Saturating amounts of nucleic acid ligands ( 0 . 2 mg/ml of linearized pT7blue and 0 . 1 mg/ml of chromatin assembled on the same DNA , respectively ) and ATP ( 1 mM ) were used . The unstimulated basal activity was ≤0 . 05 s−1 . Errors are s . d . ( n ≥ 3 ) . ( B ) The Z2 solubility tag did not interfere with remodeling rates on wild-type H4 containing chromatin and tail-less H4 chromatin . Z2-tagged ISWI+6; E257Q , which contained a point mutation in the ATPase domain rendering it catalytically inactive , was included as a control . Its activity on tail-less H4 arrays was undetectable ( § ) . Errors are s . d . ( n ≥ 3 ) except for the ATPase-dead construct ( ISWI+6; E257Q ) , which was tested once . ( C ) DNA- and chromatin-stimulated ATP hydrolysis rates of the ATPase dead double mutant ISWI+6; E257Q were negligible ( ≤0 . 04 s−1 ) . Errors are s . d . ( n ≥ 3 ) for ISWIWT and minimal and maximal values of two independent measurements for ISWI+6; E257Q . The asterisks ( * ) mark data that were replotted for comparison from Figure 3B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 02310 . 7554/eLife . 21477 . 024Figure 6—figure supplement 3 . Saturation controls for ISWI+6 in ATPase assays . ( A ) Linearized pT7blue DNA was titrated over a 16-fold range . 0 . 2 mg/ml were saturating . ( B ) Titration of chromatin assembled on linearized pT7blue DNA . 0 . 1 mg/ml were close to saturation . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 02410 . 7554/eLife . 21477 . 025Figure 6—figure supplement 4 . AcidicN mutations upregulate the ATPase activity of ISWIH483B . Relative to ISWIWT , ISWIH483B had a ~fourfold diminished DNA- and chromatin-stimulated ATPase activity . Additional mutation of AcidicN ( +3; +6 ) strongly activated both DNA- and chromatin-stimulated ATP turnover . Errors are s . d . for ISWIWT and minimal and maximal values of two independent measurements for all other constructs . Data for ISWIWT ( * ) were replotted for comparison from Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 025 Of note , the +3 and+6 mutants had a strongly deregulated ATPase , hydrolyzing ATP markedly faster than ISWIWT when presented with saturating amounts of naked DNA . In fact , DNA-stimulated ATPase rates of ISWI+6 reached values of nucleosome-stimulated ISWIWT rates . Also its basal ATPase activity was strongly ( 20-fold ) upregulated compared to ISWIWT ( Figure 6B; Figure 6—figure supplement 3A , B ) . To rule out that co-purifying contaminating ATPases overwhelm the ATPase signal , we combined the +6 mutation with a point mutation in the ATPase that abrogates ATPase activity ( E257Q ) . ATP hydrolysis and remodeling were negligible for ISWI+6; E257Q , providing strong evidence against this possibility ( Figure 6—figure supplement 2B , C ) . In contrast to the DNA-stimulated reaction , nucleosome-stimulated ATPase and remodeling activities were comparable between the AcidicN mutants and ISWIWT ( Figure 6B , C ) . Taken together , these results indicated that the AcidicN mutants were not simply hyperactive , but misregulated instead . More specifically , mutation of AcidicN prevented ISWI from properly recognizing whether chromatin was bound and led to futile ATP hydrolysis in the absence of chromatin . To independently test this conclusion and to further validate the usefulness of the H483B mutant used further above , we combined the H483B and AcidicN mutations ( Figure 6—figure supplement 4 ) . DNA-stimulated ATP hydrolysis was strongly upregulated in the ISWI+3; H483B and ISWI+6; H483B double mutants relative to the ISWIH483B single mutant and reached levels of the chromatin-stimulated reaction . These data closely paralleled and therefore independently validated our results obtained with ISWI+3 and ISWI+6 . We conclude that AcidicN regulates ISWIWT and ISWIH483B in a very similar fashion , further justifying the use of ISWIH483B for crosslinking experiments above . To validate the predicted binding interface of AcidicN on Lobe 2 and to further probe the functionality of this interaction , we introduced mutations in Lobe 2 . We selected three positively charged residues for mutagenesis , K403 , R458 and R508 , which are predicted to participate in docking to the negatively charged AcidicN motif ( Figure 7A; Figure 7—figure supplement 1 ) . Charge-reversal of these residues would be expected to weaken docking of AcidicN to Lobe2 and – in the simplest case – phenocopy the effects of the mutation of AcidicN . Indeed , the interface mutants had a strongly upregulated DNA-stimulated ATPase activity whereas chromatin-stimulated ATP turnover and nucleosome remodeling were largely unaffected ( Figure 7B , C; Figure 7—figure supplement 2 ) . The interface mutants therefore behaved just like the AcidicN mutants discussed above . A control mutant ( ISWIR486; 488D ) , carrying amino acid substitutions just outside of the predicted AcidicN binding interface , however , retained its ability to discriminate chromatin over DNA in the ATP hydrolysis assay ( Figure 7B ) . These data support the notion that AcidicN interacts with Lobe 2 at the predicted interface and that this interaction is functionally important to discriminate whether chromatin is bound to the enzyme . 10 . 7554/eLife . 21477 . 026Figure 7 . Validation of the predicted binding interface of AcidicN on Lobe 2 . ( A ) Homology model of the ISWI ATPase domain . Dark and light grey , ATPase lobes 1 and 2 , respectively; blue , hypothetical binding interface of AcidicN as in Figure 4—figure supplement 3A . Positively charged residues selected for mutagenesis are shown in red ( AcidicN interface mutant ) and orange ( control mutant ) . ( B ) Mutation of the AcidicN interface ( K403D , R458D and R508D ) strongly upregulated DNA-stimulated ATP hydrolysis relative to ISWIWT , whereas the nucleosome-stimulated ATP turnover was similar . In contrast , a control mutation ( R486; 488D ) had little effect on ATP hydrolysis . Saturating concentrations of DNA and chromatin were used . Errors are s . d . for ISWIWT and minimal and maximal values of two independent measurements for all other constructs . ( C ) AcidicN interface variants of ISWI robustly remodeled nucleosomes within twofold of ISWIWT . Nucleosomal arrays containing wild-type H4 were used . Errors are s . d . ( n ≥ 3 ) for ISWIWT and minimal and maximal values of two independent measurements for all other constructs . Raw data of the remodeling assay can be found in Figure 7—figure supplement 2 . Color code as in ( B ) . Results for ISWIWT ( * ) were replotted for comparison from Figure 3B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 02610 . 7554/eLife . 21477 . 027Figure 7—figure supplement 1 . Coomassie-stained SDS-PAGE of purified recombinant ISWI constructs analyzed in Figure 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 02710 . 7554/eLife . 21477 . 028Figure 7—figure supplement 2 . Determination of rate constants for remodeling of AcidicN interface mutants . ( A ) Exemplary remodeling time courses on WT H4-arrays for interface mutants . Asterisks mark a contaminating non-nucleosomal DNA ( competitor DNA ) that was not completely removed during preparation of nucleosomal arrays . Mock: Sample lacking ISWI . ( B ) Quantification of time courses shown in ( A ) . Data were fit to a single exponential function to extract kobs ( see Figure 7C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 028 To explore whether AcidicN takes part in H4-tail recognition , we measured the dependence of AcidicN mutants on the H4 tail in remodeling assays . Strikingly , the +3 and+6 ISWI derivatives lost most of their reliance on the H4 tail during remodeling ( Figure 8A ) . In contrast , ISWIΔppHSA and ISWIΔppHSA; ΔAT-hook retained a strong H4-tail dependence , which indicated that ppHSA had little involvement in H4-tail recognition . 10 . 7554/eLife . 21477 . 029Figure 8 . Mutation of AcidicN , the AcidicN binding interface or AutoN suppresses dependence on the H4-tail . ( A ) H4-tail dependence of the remodeling activities of ISWI variants . Values were calculated from the observed remodeling rate constants obtained for WT and tail-less H4 chromatin ( Figure 8—figure supplement 1E ) . ( B ) ATP hydrolysis measurements of ISWT+6 , ISWI2RA and ISWI+6; 2RA in absence or presence of saturating concentrations of DNA and chromatin . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 02910 . 7554/eLife . 21477 . 030Figure 8—figure supplement 1 . Raw data of the remodeling assays . ( A–D ) Determination of rate constants ( kobs ) from remodeling assays for ISWI+6 , ISWI2RA and ISWI+6; 2RA . Shown are exemplary time courses on nucleosomal arrays containing wild-type ( A , C ) and tail-less H4 ( B , D ) . Data were fit to a single exponential function to extract the rate constant kobs . ( E ) Rate constants for remodeling of ISWI variants used in this study ( all 300 nM ) . Errors are s . d . ( n ≥ 3 ) for ISWIWT , ISWIΔppHSA , and ISWI+6 and minimal and maximal values of two independent measurements for all other variants . Samples , in which the enzyme concentration was not saturating , are indicated ( § ) . Data marked with asterisks ( * ) were replotted from previous figures for better overview . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 030 Two AcidicN interface mutants described above ( K403D and R458D ) also depended less on the H4 tail during remodeling than ISWIWT ( Figure 8A ) . The third mutant ( R508D ) and the control mutant ( R485; 488D ) were apparently still sensitive towards loss of the H4 tail . These two mutants , however , were not saturated with tail-less chromatin so that the calculated values represented upper limits for the H4-tail dependence ( <24 fold and <15 fold , respectively; Figure 8—figure supplement 1E and data not shown ) . Lack of the H4-tail dependence of AcidicN mutants was reminiscent of the phenotype previously described for the ISWI2RA mutation in AutoN ( Clapier and Cairns , 2012 ) . The 2RA mutation ( Figure 6—figure supplement 1 ) suppressed the dependence on the H4 tail also in our experiments , albeit our quantitative analysis showed an even more robust reduction than previously seen ( Figure 8A ) . ISWI2RA was catalytically fully active , as was an AcidicN and AutoN double mutant ( ISWI+6; 2RA; Figure 8—figure supplement 1 ) . Like the respective single mutants , ISWI+6; 2RA barely relied on the presence of the H4 tail ( Figure 8A; Figure 8—figure supplement 1 ) . Compared to the respective single mutants , the ISWI+6; 2RA double mutant hydrolyzed ATP even faster in the absence of any ligand ( Figure 8B ) . This result suggested that both motifs contributed to repression of the basal ATPase activity . In contrast , DNA- and chromatin-stimulated ATP turnover rates were not further perturbed by the double mutation ( Figure 8B ) , consistent with both motifs cooperating during discrimination of chromatin from DNA . Dozens of ATP-dependent chromatin remodeling factors are at work in any eukaryotic cell . Their activities impact every process that involves the cell’s genetic material , including transcription , replication , DNA repair and recombination . Dysfunction and improper regulation of these complexes may have dire consequences for human health ( Kadoch and Crabtree , 2015; Garraway and Lander , 2013 ) . Perhaps as a consequence , remodelers across many families independently evolved intricate mechanisms for autoregulation ( Clapier and Cairns , 2012; Hauk et al . , 2010; Wang et al . , 2014; Clapier et al . , 2016; Gottschalk et al . , 2009 ) . It has been known for many years that the activity of ISWI remodelers is regulated by the H4 tail ( Clapier et al . , 2001 ) . Regulation by the H4 tail was later also discovered for remodelers of the Chd1 ( Ferreira et al . , 2007 ) and Alc1 families ( Ahel et al . , 2009 ) . The molecular mechanism of H4-tail recognition and regulation has remained elusive , not least because the tail’s binding site had not been mapped . Using crosslinking-MS , we found the H4 tail to bind to the conserved Lobe 2 of the ATPase module . Direct binding to the ATPase domain explains regulation of otherwise divergent remodeler families and explains the influence of the H4 tail on catalytic , as opposed to purely binding steps ( Clapier et al . , 2001; Dang et al . , 2006 ) . Our data do not rule out additional binding sites on other domains and on ISWI’s partner subunit ACF1 as proposed earlier ( Boyer et al . , 2004; Grüne et al . , 2003; Hwang et al . , 2014 ) . ISWI and Chd1 proteins have evolved a complex autoregulatory mechanism . This mechanism involves an autoinhibitory domain N-terminal to the ATPase . Inhibition by this domain is countered in an unknown fashion by H4-tail binding . Two limiting scenarios can explain the data ( Figure 9 ) . 10 . 7554/eLife . 21477 . 031Figure 9 . Proposed models for autoregulation imposed by the NTR and the recognition process of the H4 tail . The ppHSA motif , AcidicN and AutoN dock against Lobe 2 of the ATPase domain , promoting an overall structural architecture of the ATPase module that is reminiscent of Chd1 ( Figure 4A ) . AcidicN and AutoN functionally collaborate in the H4 tail recogniton process . The docking site of AutoN-AcidicN is adjacent to the H4 tail potentially allowing simultaneous binding ( top ) . Alternatively , the H4 tail may displace the NTR as suggested previously ( bottom ) ( Clapier and Cairns , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 031 The first model has been proposed earlier ( Clapier and Cairns , 2012 ) and posits that AutoN acts as a pseudosubstrate by mimicking part of the basic patch of the H4 tail . In fact , AutoN ( amino acids ‘RHRK’ , which are present in many but not all ISWI proteins; Figure 1D and Figure 1—figure supplement 1 ) was initially discovered by way of its resemblance to the amino acids ‘R17H18R19K20’ on histone H4 ( Clapier and Cairns , 2012 ) . In this model , the basic patch of the H4 tail must compete with AutoN for the same binding site on the ATPase domain , such that AutoN and possibly the entire NTR is displaced upon tail binding ( Clapier and Cairns , 2012; Hauk et al . , 2010 ) . This model is supported by the observations that the NTR can in principle undergo conformational changes ( Mueller-Planitz et al . , 2013 ) and that the chromo domains of Chd1 must rearrange before the ATPase domain assumes a catalytically active conformation ( Hauk et al . , 2010 ) . Direct experimental support for a shared binding site of AutoN and H4 basic patch has been lacking , however , and the resemblance of the two motifs may be purely coincidental in principle . Of the four amino acids that resemble the H4 tail , only three ( R17H18R19 ) were found to be functionally important for ISWI enzymes ( Fazzio et al . , 2005; Clapier et al . , 2002; Clapier and Cairns , 2012 ) . Recent crystallographic evidence also did not support the molecular mimicry hypothesis ( see below ) ( Yan et al . , 2016 ) . We favor a second , simpler model , which does not invoke molecular mimicry ( Figure 9 ) . In this model , the AutoN and the H4-tail binding sites are not identical , possibly allowing simultaneous binding of both to Lobe 2 at least temporarily . This scenario is fully compatible with our suggestion that the docking sites for the H4 tail and AutoN-AcidicN are adjacent to each other but not overlapping ( Figure 5E ) . Conceivably , the negatively charged AcidicN motif may even promote binding of the basic H4 tail to a neighboring site . A structural rearrangement of the NTR upon H4 tail binding is compatible with but not required in this model . Similarly , conformational changes of the NTR upon DNA binding ( Mueller-Planitz et al . , 2013; Hauk et al . , 2010 ) or during other steps of the reaction cycle are also fully consistent with it . Intriguing parallels between ISWI’s NTR and Chd1’s chromo domains become apparent . Our crosslinking results indicate that the NTR of ISWI docks against Lobe 2 of the ATPase domain in a very similar fashion as the chromo domains of Chd1 , and docking appears to involve an acidic motif in both cases ( Hauk et al . , 2010 ) . Thus , the overall conformational architecture of ISWI’s ATPase module may be shared with Chd1 . Moreover , both domains are known to inhibit the ATPase , both are predicted to undergo conformational changes upon substrate binding ( Mueller-Planitz et al . , 2013; Clapier and Cairns , 2012 ) and both confer sensitivity towards the histone H4 tail ( Clapier and Cairns , 2012; Hauk et al . , 2010 ) . Thus , despite complete lack of sequence conservation between both domains , they appear to have evolved very similar functionalities . The NTR of ISWI contains several conserved motifs whose functions have mostly remained unexplored so far . Because the ppHSA motif and adjacent regions crosslinked to the ATPase lobe 2 , we suggest that it is important for docking the NTR against the ATPase domain . Consistent with such a structural role of this motif , we found that ISWIΔppHSA is destabilized in vitro and in vivo . Of note , the ppHSA motif is present in a wide variety of unrelated remodelers , including Ino80 , Lsh , and Snf2 , suggesting that their ATPases , too , might bind the ppHSA motif and assemble into a structurally analogous architecture . In this study , we functionally characterized AcidicN , a novel motif in the NTR . ISWI with a mutated AcidicN had a deregulated , hyperactive ATPase activity . Notably , this mutant hydrolyzed ATP with comparable velocities when bound to either DNA or nucleosomes , indicating that it lost its ability to discriminate between them . In particular , it lost its H4-tail dependence . This phenotype is reminiscent of mutations in the acidic helix in Chd1 ( Hauk et al . , 2010 ) , underscoring the functional parallels between the NTR and chromo domains discussed above . The effects of AcidicN mutations were also remarkably similar to AutoN mutations ( Clapier and Cairns , 2012 ) , which suggested that they work together . The mechanism of autoinhibition by the NTR therefore may involve more than simple mimicry of H4’s basic patch by AutoN ( Clapier and Cairns , 2012 ) . Supporting its functional importance , AcidicN is at least as conserved as AutoN in our alignments . During the revision of this manuscript , a crystal structure of the ATPase module of ISWI from a thermophilic fungus became available ( Yan et al . , 2016 ) . Even though both studies relied on different approaches , they arrived at very similar conclusions . As suggested by our crosslinking and modeling data , the NTR packed against the ATPase domain in the structure of the thermostable ISWI . We correctly predicted the AcidicN binding pocket on Lobe 2 ( Figure 4—figure supplement 3A ) , and our crosslinks between Lobe 2 and the NTR were fully supported by the structure as well . Finally , the authors succeeded in co-crystallizing a histone H4-tail peptide with Lobe 2 of the ATPase . Even though only the basic patch of the tail peptide was visible in the structure , its location overlapped well with the position of the modeled H4 basic patch ( Figure 5—figure supplement 3C ) . AutoN crystallized in closer proximity to the interaction site of the basic patch than suggested by modeling , but molecular mimicry of AutoN with the basic patch was not supported by the structure . Sensitive biophysical assays will be instrumental in the future for resolving conformational changes that may occur during H4-tail recognition and for understanding their functional importance in ISWI complexes . Moreover , ascertaining the predicted role of H4-tail recognition for the formation or maintenance of compact heterochromatic regions remains an important goal . Search for homologous proteins of full-length Drosophila ISWI and alignment of sequences were done using HHblits with standard settings . Sequence logos of conserved NTR motifs were derived with WebLogo three from this alignment ( Schneider and Stephens , 1990 ) . Proteins containing the ppHSA motif were identified by PSI-BLAST against the 120 N-terminal amino acids of ISWI . The alignment was done using T-Coffee . Sequence alignments were visualized using Jalview 2 . 9 . S . cerevisiae Isw1 alleles were cloned into selected destination vectors of a galactose-inducible hybrid promoter library ( generously provided by Dr . Hal Alper , UT Austin , USA ) ( Blazeck et al . , 2012 ) . The following destination vectors were used , sorted according to increasing promoter strength: Gal4pBS2-Pleum ( denoted ‘+’ in Figure 2B; Figure 2—figure supplement 1 ) , Gal4pBS4-Pleum ( ‘++’ ) , UASgal-A9-Pcyc ( ‘+++’ ) , and UASgal-Pgal ( ‘++++’ ) . Destination vectors were XbaI and ClaI digested and gel purified . Isw1 derivatives were PCR-amplified from yeast genomic DNA , gel purified and ligated into the destination vectors by Gibson assembly . All spotting assays employed untagged Isw1 variants . All constructs were sequence verified before transformation . As an empty vector control , the UAS promoter , coding and terminator sequences were removed from the Gal4pBS2-Pleum plasmid by AscI and MluI digest and subsequent self-ligation . YTT227 ( TKO ) , YTT225 ( DKO ) and W1588-4c ( wild-type; Table 3 ) were transformed with indicated plasmids via a standard transformation protocol . Single colonies were picked and grown overnight in Synthetic complete ( SC ) -Ura + Glucose ( 2% ) media . The culture was then diluted to OD 0 . 05 in SC-Ura Galactose media ( 2% ) and grown for 24 hr . Cells were diluted again to OD 0 . 1 in Galactose media and grown for another 24 hr before spotting . Cells were diluted to OD 1 . 0 , and tenfold serial dilutions were spotted on galactose media and incubated at 30°C , 37°C or 38 . 5°C for 72 hr . At least two replicates were performed on different days with a single transformant of a sequence-verified clone . 10 . 7554/eLife . 21477 . 032Table 3 . Yeast strains used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 21477 . 032StrainGenotypeReferenceW1588-4CMATa ade2-1 his3-11 , 15 leu2-3 , 112 trp1-1 ura3-1 can1-100 but RAD5 Tsukiyama et al . ( 1999 ) YTT227MATa ade2-1 his3-11 , 15 leu2-3 , 112 trp1-1 ura3-1 can1-100 but RAD5 isw1::ADE2 isw2::LEU2 chd1::TRP1 Tsukiyama et al . ( 1999 ) YTT225MATa ade2-1 his3-11 , 15 leu2-3 , 112 trp1-1 ura3-1 can1-100 but RAD5 isw2::LEU2 chd1::TRP1 Tsukiyama et al . ( 1999 ) YFMP047MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 ISW1-TAP::HIS3MX6 Open Biosystems For Western analysis , Isw1 variants were C-terminally tagged by fusion to a cassette containing a ( GGS ) 2 linker , a 3C cleavage site , a ( GGS ) 5 linker and a TAP tag . YTT227 that expressed TAP-tagged Isw1 variants was induced with galactose as above , diluted to OD 0 . 1 and grown to OD 1 . 0 in 10 ml SC-Ura + Galactose media . YFMP047 ( Table 3 ) , containing a genomically TAP-tagged Isw1 allele , was grown as a control in YPAD media . Cells were harvested , washed twice with cold water and dissolved in 200 μl Extraction buffer ( 40 mM Hepes-KOH pH 7 . 5 , 10% Glycerol , 350 mM NaCl , 0 . 1% Tween-20 , 1 μg/ml Pepstatin , 2 μg/ml Leupeptin , 2 μg/ml Aprotinin , 1 mM PMSF . Glass beads ( 200 μl ) were added , and the suspension was vortexed for 10 min with a 30 s on/off cycle on ice . After centrifugation ( 13 , 000 rpm , 20 min 4°C ) , supernatants were harvested , aliquoted ( 50 µl ) , flash frozen , and stored at −80°C for subsequent use . Supernatants were thawed on ice and 50 μg of each extract was loaded on a 10% SDS gel . Anti-TAP antibody ( CAB1001 , ThermoFisher; 1:5000 dilution ) was used to detect TAP-tagged ISW1 mutants and anti-H3 antibody ( ab1791 , Abcam; 1:20 , 000 dilution ) was used as a loading control . Membranes were scanned using the LI-COR Odyssey IR imaging system ( ODY-0853 ) and bands were quantified using Image Studio Lite v5 . 2 . 5 . Expression levels were normalized to the signal of genomically integrated TAP-tagged Isw1 . Two technical replicates were performed . A pPROEX-HTb–based expression plasmid with the gene encoding Drosophila ISWIWT ( kindly provided by C . Müller; EMBL , Heidelberg , Germany ) served as the template for all ISWI variants . An overview over cloned ISWI variants is presented in Figure 3—figure supplement 1A and Figure 6—figure supplement 1A . All ISWI genes were fused N-terminally to a His6-tag . To generate ISWIΔppHSA and ISWIΔppHSA; ΔAT-hook , a 3C cleavage site was introduced at the desired site by QuikChange mutagenesis or polymerase incomplete primer extension . The trigger factor gene was amplified from pTf16 ( Takara Bio Inc . ) and fused to the ISWIWT gene by Gibson assembly . ISWI+6 was subcloned into the pET-Z2 plasmid ( kindly provided by Dr . Arie Geerlof , Helmholtz Zentrum , Munich , Germany ) . Expression and purification of His6-tagged ISWIWT and its derivatives was performed essentially as described ( Forné et al . , 2012 ) with the following variations . Tags or parts of the NTR were cleaved off by specific proteases ( TEV and 3C , respectively ) as indicated ( Figure 3—figure supplement 1A; Figure 6—figure supplement 1A ) . ISWIH483B was expressed and purified as described ( Forné et al . , 2012 ) . During its purification , the UV light of the FPLC remained switched off to protect the Bpa residue . All ISWI variants were purified once except ISWI+3 and Z2-ISWI+6 , which were purified twice . The independent preparations were indistinguishable in ATPase assays ( ISWI+3 and Z2-ISWI+6 ) . Whereas ISWI+3 preparations were not directly compared , independent Z2-ISWI+6 preparations also yielded same results in remodeling assays . A pBH4-based expression plasmid encoding full-length human SNF2H ( kindly provided by G . Narlikar; UCSF , San Francisco ) was transformed into Rosetta competent E . coli cells . Protein expression was performed in 2x YT medium ( 20 g/l tryptone , 10 g/l yeast extract , 10 g/l NaCl ) supplemented with 34 mg/l chloramphenicol and 100 mg/l ampicillin . Expression of SNF2H was induced by addition of 0 . 4 mM IPTG at 18°C for approximately 18 hr . Bacteria cells were resuspended in 20 ml lysis buffer per 1 l culture ( 25 mM HEPES pH 8 . 0 , 300 mM KCl , 7 . 5 mM imidazole , 10% glycerol , 1 mM DTT ) supplemented with protease inhibitors ( 1 mM PMSF , 1 mg/l Aprotinin , 1 mg/l Leupeptin , 0 . 7 mg/l Pepstatin ) per 1 l culture , and lysed by French Press ( Thermo Spectronic ) and ultrasonication ( Branson ) . Per 1 l lysed bacteria culture , 1000 U Benzonase ( Merck Millipore ) were added . The lysate was clarified by centrifugation ( 30 min , SS34 rotor ) . The N-terminal His6-tagged SNF2H was purified by nickel affinity chromatography ( HisTrap HP , 5 ml; GE Healthcare ) . An elution gradient was applied with 25 mM HEPES pH 7 . 0 , 300 mM KCl and 400 mM Imidazole and enzyme-containing fractions were pooled . Contaminating DNA was removed by passing the sample over an anion exchange column ( Mono Q 5/50 GL ion exchange column; GE Healthcare ) that was pre-equilibrated in SEC buffer ( 25 mM HEPES pH 7 . 5 , 300 mM KCl , 1 mM DTT ) . The flow-through of the column was collected . The protein sample was concentrated to 0 . 5–1 ml per 1 l of original E . coli culture in centrifugal filters ( Amicon Ultra-4 , 30 kDa MWCO; Millipore ) . TEV protease ( prepared in-house ) was added to a final concentration of 0 . 075–0 . 15 mg/ml and the concentrated protein sample was dialyzed against 1 l SEC buffer overnight in dialysis tubing ( 6000–8000 Da MWCO; Sectra/Por ) . The protein sample was loaded onto a size exclusion chromatography column ( Superdex 200 HiLoad 16/60 , 120 ml; GE Healthcare ) pre-equilibrated in SEC buffer . Elution fractions were pooled according to purity and , as necessary , concentrated and dialyzed into storage buffer ( 25 mM HEPES pH 7 . 5 , 210 mM KCl , 15% glycerol , 1 mM DTT ) for at least 16 hr . Drosophila histones were purified as described ( Klinker et al . , 2014; Luger et al . , 1999 ) . The 187 bp long Widom-601 derivative used for end-positioned mononucleosomes ( 0N40 ) was excised from pFMP151 with SmaI ( NEB ) and PAGE purified . DNA for 25-mer nucleosomal arrays used in remodeling assays was excised from pFMP233 with EcoRI HF , HincII and AseI ( NEB ) and purified by phenol/chloroform extraction and ethanol precipitation . Polynucleosomes used in ATP-hydrolysis assays were assembled on linearized plasmid DNA ( pT7 blue derivative ) . Histone octamers , mononucleosomes and polynucleosomes , including 25-mer nucleosomal arrays , were prepared by salt-gradient dialysis as described ( Mueller-Planitz et al . , 2013; Luger et al . , 1999 ) . Mononucleosomes were further purified by glycerol gradient ultracentrigation . Nucleosomal arrays were purified further by Mg2+ precipitation ( 3 . 5 mM for WT-H4 arrays , 8 . 5 mM for H4-tail deleted arrays ) ( Mueller-Planitz et al . , 2013 ) . The concentration of nucleosomal DNA was determined by measuring its UV absorption at 260 nm . For nucleosomal arrays , concentrations refer to the concentration of individual nucleosomes . Remodeling and ATPase assays were performed in 25 mM HEPES-KOH , pH 7 . 6 , 50 mM NaCl , 1 mM MgCl2 , 0 . 1 mM EDTA , 10% glycerol , 0 . 2 g/l BSA and 1 mM DTT at 26°C in the presence of a ATP regenerating system as described ( Mueller-Planitz et al . , 2013 ) . ATP hydrolysis was monitored by an NADH-coupled ATP hydrolysis assay ( Mueller-Planitz et al . , 2013; Forné et al . , 2012 ) . Saturating concentrations of ATP-Mg2+ ( 1 mM ) and of nucleic acids ligands were used ( 0 . 2 mg/ml of linearized pT7blue and 0 . 1 mg/ml of chromatin assembled on the same DNA , respectively ) . Saturation of DNA and chromatin was controlled by varying the concentration of the ligands at least 16-fold ( Figure 3—figure supplement 2; Figure 6—figure supplement 3 ) . Occasional occurrence of air bubbles in ATPase experiments precluded accurate measurements; affected samples were excluded from the analysis . In no other assays were outliers excluded . Remodeling activity was probed by a restriction enzyme accessibility assay ( Mueller-Planitz et al . , 2013 ) . A 25-mer nucleosomal array with a 197 bp nucleosomal repeat length was used . The 19th nucleosome of this array occluded a unique KpnI site at position −32 relative to its dyad ( Mueller-Planitz et al . , 2013 ) . Arrays ( 100 nM ) were incubated with ISWI derivatives at the indicated concentrations , ATP-Mg2+ ( 1 mM ) and KpnI ( 2 U/ml ) . Reactions were quenched with SDS ( 0 . 4% ) and EDTA ( 20 mM ) before the samples were deproteinized , ethanol precipitated and resolved by agarose gel electrophoresis ( Mueller-Planitz et al . , 2013 ) . kobs for remodeling was obtained by fitting time courses to a single exponential function . When the enzyme concentration was varied ≥threefold , typically between 100 nM and 300 nM , similar values for kobs were obtained with a few exceptions , suggesting that arrays were generally saturated ( Figure 3—figure supplement 3; Figure 4—figure supplement 1; Figure 7—figure supplement 2; Figure 8—figure supplement 1A–D and data not shown ) . The exceptions comprised ISWIΔppHSA; ΔAT-hook , ISWI+3 and ISWIH483B on WT-arrays and ISWI+3 , ISWIR508D and ISWIR486; 488D on tail-less H4 arrays . To site-specifically attach a UV-reactive benzophenone residue to full-length histone H4 , single cysteines were introduced into the histone H4 tail by site directed mutagenesis at the indicated positions . 4- ( N-Maleimido ) benzophenone ( Sigma ) was dissolved to 100 mM in N , N-Dimethylformamide ( DMF ) and added to a final concentration of 3 mM to denatured single cysteine variants of H4 ( 1 mg/ml ) in 20 mM Tris/HCl pH 7 . 1 , 7 M Guanidine-HCl , 5 mM EDTA , 2 mM TCEP for 2 hr at room temperature . After a 3 hr incubation in the dark , the labeling reaction was stopped by adding 20 mM DTT for 20 min . UV-Crosslinking was performed in uncoated 384-well plates or 96-well plates ( Greiner ) on ice using the 365 nm irradiation of a BioLink UV-Crosslinker ( Peqlab ) for the indicated durations . Crosslinking between benzophenone-labeled nucleosomes ( 0N40; 1 µM ) and stoichiometric amounts of ISWI or SNF2H was performed in 20 mM Tris/HCl , pH 7 . 7 , 100 mM KCl , 0 . 1 mM EDTA , 3 mM DTT . Crosslinking between ISWI26-648 ( 0 . 1 mg/ml ) and a histone H4 peptide comprising the 24 N-terminal amino acids of H4 carrying a Bpa substitution at position 1 or 10 was carried out in the presence of 13 µM 59 bp DNA duplex in 25 mM HEPES-KOH , pH 7 . 6 , 50 mM NaCl , 1 mM MgCl2 , 0 . 1 mM EDTA , 10% glycerol and 1 mM DTT for 3 hr as above . Samples were subsequently digested with benzonase before further processing . Crosslinking between Bpa variants of ISWI ( H483B ) was carried out as described ( Forné et al . , 2012 ) . UV-irradiated samples and unirradiated control samples were separated by SDS-PAGE and Coomassie stained . Protein bands were excised and trypsin digested for subsequent mass spectrometry as described ( Forné et al . , 2012; Wilm et al . , 1996 ) . For LC-MS/MS , 5 µl were injected in either an Ultimate 3000 system ( Thermo ) and desalted on-line in a C18 micro column ( 75 µm i . d . x 15 cm , packed with C18 PepMap , 3 µm , 100 Å by LC Packings ) or desalted offline using C18 Stagetip and injected in an Ultimate 3000 RSLCnano system ( Thermo ) . Desalted sample was then separated in a 15 cm analytical C18 micro column ( 75 µm i . d . packed with C18 PepMap , 3 µm , 100 Å by LC Packings or homepacked 75 μm ID with ReproSil-Pur C18-AQ 2 . 4 μm from Dr . Maisch ) with a 40 to 60 min gradient from 5% to 60% acetonitrile in 0 . 1% formic acid . The effluent from the HPLC was directly electrosprayed into an LTQ-Orbitrap XL as described before ( Forné et al . , 2012 ) or a Q Exactive HF MS ( Thermo ) . The Q Exactive HF MS was operated in a data-dependent mode . Survey full scan MS spectra ( from m/z 375–1600 ) were acquired with resolution R = 60 , 000 at m/z 400 ( AGC target of 3 × 106 ) . The ten most intense peptide ions with charge states between 3 and 5 were sequentially isolated to a target value of 1 × 105 , and fragmented at 27% normalized collision energy . Typical mass spectrometric conditions were: spray voltage , 1 . 5 kV; no sheath and auxiliary gas flow; heated capillary temperature , 250°C; ion selection threshold , 33 . 000 counts . Each Thermo binary raw file was converted to a dta file using Decon2LS ( Zimmer et al . , 2006 ) or to an mgf file using Proteome Discoverer 1 . 4 ( Thermo ) and -as needed- recalibrated with the Post-Search Recalibrator Node . Crosslinks were mapped by Crossfinder ( Forné et al . , 2012; Mueller-Planitz , 2015 ) . Typical error windows were ±10 ppm for MS1 searches and ±15 ppm for MS2 searches . All amino acid residues were regarded as potential sites of crosslinking . Crosslink candidates were independently validated by the authors J . L . , S . P . , N . H . and F . M . -P . and rated as high , medium and low confidence . The validation comprised a general assessment of the spectrum quality , removal of wrong product ion assignments , and evaluation of the actual evidence for the presence of the two peptides within the crosslink . The mass spectrometry data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD005831 . Interactions between NTR motifs and the histone H4 tail with Lobe 2 were modeled using the fully blind peptide-protein docking protocol pepATTRACT ( Schindler et al . , 2015a ) in the ATTRACT docking engine ( de Vries et al . , 2015 ) ( www . attract . ph . tum . de/peptide . html ) . The termini of the motifs ( ‘peptides’ ) were left uncharged , other parameters were set to the default values as described ( Schindler et al . , 2015a ) . Briefly , for each motif three idealized peptide conformations ( extended , α-helical and poly-proline ) were generated from sequence and this peptide ensemble was docked rigidly against the protein domain using the ATTRACT coarse-grained force field ( Zacharias , 2003 ) . The top-ranked 1000 structures were subjected to two stages of atomistic refinement using the flexible interface refinement method iATTRACT ( Schindler et al . , 2015b ) and a short molecular dynamics simulation in implicit solvent with the AMBER program ( Case et al . , 2014 ) . The pepATTRACT protocol requires neither knowledge about the peptide binding site nor of the bound peptide conformation and is therefore suitable for predicting complexes between proteins and motifs from intrinsically unstructured regions . The structure of ISWI ATPase Lobe 2 ( residues 352–637 ) was modeled by homology from the structure of Chd1 ( PDB 3MWY ) using MODELLER ( Webb and Sali , 2016 ) . We performed three docking runs modeling the potential binding site for the AutoN motif ( residues 89–97; DHRHRKTEQ ) , the AcidicN motif ( residues 96–104; EQEEDEELL ) and the full module AutoN+AcidicN ( residues 89–104; DHRHRKTEQEEDEELL ) separately . During the first two runs , we modeled the peptide ensemble from the sequence as described above . For the third run , we used PEP-FOLD2 ( Shen et al . , 2014 ) , PEP-FOLD3 ( Lamiable et al . , 2016 ) and I-TASSER ( Yang et al . , 2015 ) servers to predict the structure of the module . We used the resulting 13 conformations for ensemble docking to Lobe 2 . To test the specificity of the docking solutions , we also modeled Lobe 2 binding to scrambled sequences of AutoN ( HRQHKDERT ) , AcidicN ( LEDELQEEE ) and AutoN-AcidicN ( HLREQLDTHEREDEKE ) . Docking of AutoN-AcidicN against the homology model comprising both ATPase lobes was done as described above . During docking of the histone H4 tail ( residues 1–20 ) to Lobe 2 , we used the five high confidence crosslinks ( Table 3 ) , which – due to redundancy – provided three unique amino acid linkages . These linkages were used as upper harmonic distance restraints with a maximum distance of 20 Å to guide the modeling ( pepATTRACT-local protocol ) ( Schindler et al . , 2015a ) . After molecular dynamics refinement ( see above ) , which did not apply crosslinking restraints for technical reasons , models were filtered for those that still satisfied the distance restraints provided by the crosslinks , yielding 383 models . All figures were created using PyMOL ( www . pymol . org ) .	In the cells of animals , plants and other eukaryotes , DNA wraps tightly around proteins called histones to form structures known as nucleosomes that resemble beads on a string . When nucleosomes are sufficiently close to each other they interact and clump together , which compacts the DNA and prevents the genes in that stretch of DNA being activated . But how do cells mobilize their nucleosomes ? A nucleosome remodeling enzyme called ISWI can slide nucleosomes along DNA . ISWI becomes active when it interacts with a ‘tail’ region of a histone protein called H4 . However , the H4 tail prefers to interact with neighboring nucleosomes instead of with ISWI . Therefore when ISWI slides a nucleosome close to another one , the H4 tail of the nucleosome binds instead to its new neighbor so that ISWI cannot continue to slide . By this mechanism , ISWI is proposed to pile up nucleosomes , which subsequently compact , leading to the inactivation of this part of the genome . To investigate how ISWI recognizes the H4 tail , Ludwigsen et al . mapped where the H4 tail binds to ISWI by combining the biochemical methods of cross-linking and mass spectrometry . In addition , mutagenesis experiments identified a new motif in the enzyme that is essential for recognizing the H4 tail . In the absence of the nucleosome , this motif – called AcidicN – works with a neighboring motif called AutoN to keep ISWI in an inactive state . The two motifs also work together to enable ISWI to distinguish between nucleosomes and DNA . Further evidence suggests that other remodeling enzymes have similar regulation mechanisms; therefore this method of controlling nucleosome remodeling may have been conserved throughout evolution . Further studies are now needed to detect the shape changes that occur in ISWI as it recognizes the histone tail and work out how this leads to nucleosome remodeling . Inside cells , ISWI is usually found within large complexes that consist of many proteins . It therefore also remains to be discovered whether the proteins in these complexes impose additional layers of regulation and complexity on the activity of ISWI .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "biochemistry", "and", "chemical", "biology" ]	2017	Concerted regulation of ISWI by an autoinhibitory domain and the H4 N-terminal tail
The glymphatics system describes a CSF-mediated clearance pathway for the removal of potentially harmful molecules , such as amyloid beta , from the brain . As such , its components may represent new therapeutic targets to alleviate aberrant protein accumulation that defines the most prevalent neurodegenerative conditions . Currently , however , the absence of any non-invasive measurement technique prohibits detailed understanding of glymphatic function in the human brain and in turn , it’s role in pathology . Here , we present the first non-invasive technique for the assessment of glymphatic inflow by using an ultra-long echo time , low b-value , multi-direction diffusion weighted MRI sequence to assess perivascular fluid movement ( which represents a critical component of the glymphatic pathway ) in the rat brain . This novel , quantitative and non-invasive approach may represent a valuable biomarker of CSF-mediated brain clearance , working towards the clinical need for reliable and early diagnostic indicators of neurodegenerative conditions such as Alzheimer’s disease . The recent identification of the glymphatic system and the dural lymphatic network provide exciting new perspectives on waste clearance mechanisms within the central nervous system ( CNS ) ( Louveau et al . , 2015; Iliff et al . , 2012 ) . According to the glymphatics hypothesis , cerebrospinal fluid ( CSF ) crosses from the subarachnoid space into the periarterial space where it swiftly flows towards the brain tissue . Fluid then passes into the parenchyma from the perivascular space , a transition mediated by aquaporin-4 ( AQP4 ) channels that reside on the end feet of astrocytes . This periarterial inflow creates a convective flux of fluid across the parenchyma that exits via perivenous channels , carrying with it ‘waste products’ of brain metabolism . As such , the glymphatic pathway has been proposed to function as a ‘cleaning system’ of the brain . The exchange of CSF with interstitial fluid ( ISF ) is an established mechanism underlying the clearance of amyloid beta ( Aβ ) , recognised as a leading molecular candidate to initiate Alzheimer’s disease ( AD ) ( Iliff et al . , 2012; Weller et al . , 2008; Bakker et al . , 2016; Xu et al . , 2015; Tarasoff-Conway et al . , 2015; Hawkes et al . , 2014 ) . Despite evidence that aspects of the glymphatic pathway are preserved across species ( Goulay et al . , 2017; Ringstad et al . , 2017; Dobson et al . , 2017 ) , key questions remain on the anatomy and function in the human brain and to what extent it contributes to pathology . Currently , however , these questions cannot be answered because there are no non-invasive techniques for assessment . The development of non-invasive methods to image CSF-mediated brain clearance pathways , such as the glymphatic system , would enable repeated and practical measurement to investigate this system in the human brain and the intact animal skull . This , in turn , may help fully characterise impairment of CSF-mediated clearance pathways with age ( Kress et al . , 2014 ) , as well as the contribution to Aβ accumulation in AD . Ultimately , such methods could address the pressing clinical need for reliable and early biomarkers of AD , by identifying patients at risk of Aβ accumulation due to failing clearance mechanisms . The perivascular space is a fluid filled compartment that surrounds selected blood vessels in the brain ( Huffman et al . , 2016 ) . Perivascular channels form a central component of the glymphatic pathway that is said to drive rapid CSF-ISF exchange . Although the precise routes and fluid dynamics that underlie CSF-ISF exchange remain controversial ( Holter et al . , 2017; Hladky and Barrand , 2014; Brinker et al . , 2014; Smith et al . , 2017 ) , several independent groups have identified perivascular channels as central to this pathway ( Iliff et al . , 2012; Bedussi et al . , 2015; Lochhead et al . , 2015; Rennels et al . , 1985 ) . As such , the perivascular space represents a promising target for non-invasive imaging biomarkers of CSF-ISF exchange . To date , perivascular function has been studied using only invasive methods: ex-vivo microscopy ( Bedussi et al . , 2015 ) , two-photon imaging ( Iliff et al . , 2012 ) and contrast-enhanced MRI following intra-cranial/lumbar injection ( Iliff et al . , 2013a; Yang et al . , 2013 ) . In this work we introduce the first non-invasive method for the assessment of perivascular function using contrast-free MRI , and demonstrate use of the method in the rodent brain . Despite recognition that the perivascular space facilitates CSF-ISF exchange , the nature of fluid movement within this channel is yet to be unambiguously determined . Broadly , the glymphatics hypothesis describes perivascular fluid movement as possessing coherent , bulk flow ( Iliff et al . , 2012 ) . However , this has been questioned by other studies which propose that the fast distribution of CSF-tracers along the perivascular space can be explained by rapid dispersion of fluid/tracers via mechanical pulsations , with little bulk flow ( Hladky and Barrand , 2014; Asgari et al . , 2016 ) . Given the current uncertainty , when considering non-invasive MRI techniques for assessment , diffusion MRI represents a prime candidate for initial application owing to its established sensitivity to water dispersion , together with evidence of sensitivity to bulk flow ( non-plug e . g . laminar flow ) from prior studies of the cerebral vasculature ( Wells et al . , 2017 ) . That is , irrespective of whether perivascular fluid movement is dominated by bulk flow or rapid dispersion with little bulk flow , diffusion MRI sequences , if appropriately tuned , should yield sensitive and quantitative correlates of fluid movement , albeit non-specific to flow coherence . In this study , we apply ultra-long echo time ( TE ) , diffusion weighted MRI sequences to assess fluid movement within perivascular channels surrounding the middle cerebral artery ( MCA ) of the healthy rat brain . In addition , given evidence that cerebral arterial pulsation is a key mechanism that drives PVS fluid movement ( Rennels et al . , 1985; Iliff et al . , 2013b ) , we investigate the dependence of the technique on vascular pulsatility through cardiac gating and modulation by the adrenoceptor agonist , dobutamine . This technique represents the first non-invasive biomarker of perivascular action , working towards new translational techniques to assess CSF mediated brain clearance pathways and their role in disease . The ultra-long TE MRI sequence presented here is designed to attenuate the measured signal from the blood and parenchyma that immediately surround the perivascular space in order to minimise partial volume effects , which represent a potential confounder for assessment by MRI given the small size of this compartment . Figure 1A shows a b0 image of the axial slice through the ventral aspect of the rat brain . The subarachnoid CSF that bathes the Circle of Willis ( CoW ) can be clearly observed , with marked contrast between the blood vessels within the CoW and surrounding CSF . Bright tracts appear either side of both MCA branches ( Figure 1A ) which , due to the ultra-long echo time , must derive from fluid filled compartments of similar composition to the CSF in the subarachnoid space . This observation , together with the characteristic morphology that runs alongside and parallel to the MCA , is consistent with the description of the perivascular space as a fluid filled compartment that surrounds major blood vessels feeding the brain ( Huffman et al . , 2016 ) . Indeed , the location of this compartment is highly consistent with direct assessment from a previous study ( Figure 1B , adapted from Lochhead et al . , [Lochhead et al . , 2015] ) . The precise definition of the perivascular ( and ‘paravascular’ ) space is somewhat unclear , as highlighted in a number of recent articles ( Hladky and Barrand , 2014; Brinker et al . , 2014; Bedussi et al . , 2017 ) . Whether the fluid filled tracts around the MCA that we observe ( Figure 1A ) occupy a physically and functionally distinct ‘paravascular’ space as described by Iliff et al . , ( Iliff et al . , 2012 ) forms a more continuous pathway with subarachnoid CSF as described by Bedussi et al ( Bedussi et al . , 2017 ) . , or are well described by a perivascular space as proposed by Lochhead et al . , ( Lochhead et al . , 2015 ) remains unknown . Irrespective of the precise anatomical bordering of the fluid filled tracts identified in this work , and despite these semantic differences , all the aforementioned studies have highlighted the movement of fluid that surrounds subarachnoid arteries as a key site of CSF-tracer inflow towards the parenchyma . Hence non-invasive assessment of fluid movement within this compartment represents a meaningful measure of CSF-ISF exchange pathway function . Application of a motion probing gradient ( MPG ) along the principle direction of the perivascular tracts located around the MCA was observed to markedly attenuate the signal from these tracts relative to when the MPG was applied perpendicular to their principle orientation ( Figure 2A ) . Accordingly , across the 10 subjects , within the right perivascular space , the pseudo-diffusion coefficient ( D* ) parallel to PVS orientation was significantly greater than D* in either perpendicular direction ( p<0 . 01 respectively ) . In a similar fashion , D* ( parallel to principle direction of left PVS ) was significantly greater than D* in either perpendicular direction [p<0 . 01] . ( Figure 2B ) . These data demonstrate that the MRI sequence employed here can detect the directional dependence of fluid movement within the perivascular space ( the principal directionality of which is parallel to their orientation ) , which verifies that they are sensitised to the movement of fluid within this compartment . Within the CSF in the subarachnoid space , it was observed that D* when the MPGs were applied in the in-plane orientation ( i . e . parallel to the left or right branch of the MCA ) were both significantly greater than D* in the through plane orientation [p<0 . 01] . This is consistent with the known direction of CSF movement in the rostral-caudal direction within this region from prior invasive studies ( Iliff et al . , 2013a; Lee et al . , 2018 ) . Having verified the sensitivity of the MRI sequence to fluid movement within the perivascular and subarachnoid space , MPGs were then applied in six different directions to generate a pseudo diffusion tensor image that reflects the directionality and magnitude of subarachnoid CSF and perivascular fluid movement . Figure 3 illustrates that , for the subarachnoid space ROI , the mean D* tensor ellipsoid ( n = 6 ) was well aligned with the known principle direction of CSF movement ( caudal-rostral , observed in several invasive studies of the rodent brain [Iliff et al . , 2013a; Lee et al . , 2018] ) . Likewise , Figure 3 illustrates that the principle direction of the mean D* tensor of the left and right perivascular space , respectively , was aligned with the orientation of the respective branch of the MCA . The D* tensors for each of the individual animals are shown in Figure 2—figure supplement 1 , which show reasonable consistency with the directionality of the mean tensors shown in Figure 3 . The magnitude of the D* tensors within this region were markedly reduced post-mortem , which demonstrates that a large component of the D* measurements reflects fluid movement driven by physiological perturbations such as cardiac and respiratory pulsation and secretion from the choroid plexus ( Figure 3—figure supplement 1 ) . This may also partially reflect the reported collapse of the PVS post mortem [1] ( indeed visual inspection of the b0 images indicates a reduction in signal intensity within this region [data not shown] ) . Fractional anisotropy ( SEM ) within the right and left perivascular space and the subarachnoid space was 0 . 44 ( ±0 . 04 ) , 0 . 36 ( ±0 . 04 ) and 0 . 6 ( ±0 . 02 ) respectively with mean diffusivity ( SEM ) calculated to be 0 . 0042 ( ±0 . 0003 ) , 0 . 0052 ( ±0 . 0003 ) , 0 . 0065 ( ±0 . 0007 ) mm2/s . Figure 3E shows a map of pseudo diffusion tensors for a single subject . The principal direction of the D* tensors in the perivascular tracts that surround the left and right MCA respectively can be seen to run parallel to the orientation of the MCA . Likewise , the principal orientation of the individual voxel D* tensors can be seen to run rostral-caudal in the mid-section of the CoW . Previous studies have identified cerebral vascular pulsation to play a prominent role in perivascular fluid propulsion . To investigate this mechanism , MRI data were captured during both cerebral arterial pulsation and diastole using ECG gating with variable delays to image capture ( 36 ms and 116 ms from the r-wave to the centre of ‘diffusion’ weighting respectively ) . The results are shown in Figure 4 , where a striking and highly directional dependence of D* on cerebral vascular pulsation was observed in the PVS [Figure 4] . D* in the PVS was ~300% greater during arterial pulsation relative to diastole when motion probing gradients were applied parallel to the principle orientation ( p<0 . 01 ) . We recorded a more moderate dependence ( p=0 . 1 ) on the r-wave delay within the CSF ROI at the mid-section of the CoW ( although visual inspection of the D* maps suggests that other regions within the subarachnoid CSF appeared to show greater changes with the r-wave delay ) . Minimal dependence of the D* measures on the r-wave delay was observed in the third ventricle ( p=0 . 2 ) . Administration of the adrenoceptor agonist , dobutamine , increased heart rate from ( 354 ± 8 to 519 ± 17 bpm ) . A 65% increase in D* along PV channels was recorded ( p<0 . 01 ) following dobutamine with comparatively little change after vehicle ( Figure 4C ) . No significant changes were observed in the subarachnoid space ROI at the mid-section of the CoW following dobutamine ( p=0 . 39 , although visual inspection of the data suggests other regions within the subarachnoid CSF did show marked increases in D* ) . Dobutamine had minimal effect on D* within the third ventricle ( p=0 . 30 ) . Together these data are concordant with previous invasive measures demonstrating that perivascular fluid movement is driven by cerebral vascular pulsation and that we are now able to capture this mechanistic dependence non-invasively using the techniques introduced here . In this study , we introduce a novel MRI method to measure a distinct feature of brain physiology that , to date , has only be assessed using invasive methods – the movement of fluid in the perivascular space . The perivascular space serves as a preferential pathway for CSF-ISF exchange , an important mechanism supporting the clearance of potentially harmful molecules , such as Aβ , from the CNS . This non-invasive and translational method may have utility in AD research given evidence that Aβ accumulation ( in late stage , sporadic AD ) occurs not because of increased Aβ production but because of decreased rates of Aβ clearance ( Mawuenyega et al . , 2010 ) . Thus , this technique may expedite greater understanding of how Aβ clearance mechanisms become impaired with ageing ( Kress et al . , 2014 ) and in turn reveal a new window in early AD pathogenesis in which to target future diagnostic and treatment strategies . The technique may have broader utility to a range of neurological conditions given reported associations between glymphatic function in , for example , stroke ( Gaberel et al . , 2014 ) and traumatic brain injury ( Iliff et al . , 2014 ) . The precise mechanisms that underlie CSF-ISF exchange are yet to be fully defined and this remains an active area of research . Accumulative evidence , however , has established cerebral vascular pulsation as an important mechanism underlying perivascular fluid movement ( Rennels et al . , 1985; Iliff et al . , 2013b ) . Here , we have captured the action of cerebral arterial pulsation to drive perivascular fluid movement using non-invasive techniques ( Figure 4 ) . The measured D* showed a remarkable dependence on vascular pulsation with a ~300% increase recorded during arterial pulsation relative to diastole ( Figure 4 ) . Moreover , D* ( non-gated ) was found to markedly increase following adrenoceptor agonist , dobutamine . The non-invasive nature of this technique may enable future studies to investigate the mechanistic link between vascular pulsatility and PVS fluid movement in the healthy human brain , and its modulation by pathology as well as novel therapy . In this study , D* estimates were captured using a b0 image and then with motion probing gradients applied at a single b-value , in different directions . Future studies may wish to examine the behaviour of the PVS signal over a greater range of b-values ( and different values of δ and Δ ) to examine whether , in combination with more advanced signal modelling , this may reveal more detailed insight into PVS fluid movement . Of note , a previous study aimed to correlate MRI measures of water diffusivity from the PVS to AD severity ( Taoka et al . , 2017 ) . However , this earlier work presents limited evidence as to the contribution of the perivascular space to the measured MRI signal and hence that the parameters extracted from their measurements provide meaningful correlates of PVS fluid movement . The expression of AQP4 appears to be mechanistically important in CSF-ISF exchange ( Smith et al . , 2017; Mestre , 2017 ) . However , although genetic deletion of AQP4 was found to markedly decrease rates of small molecular weight tracer inflow from the CSF into the brain , it did not appear to affect the movement of tracers along para-arterial channels ( Iliff et al . , 2012 ) . Thus , by extension , as the technique here is targeted to PVS fluid movement , it may not be sensitive to AQP4 related modulation of CSF-ISF exchange through genetic deletion of AQP4 in the rodent brain . Hence , future studies are required to fully elucidate the relationship between para/perivascular fluid movement , CSF-ISF exchange and AQP4 expression ( Brinker et al . , 2014 ) . Furthermore , rates of glymphatic inflow have been linked to changes in extracellular space volume ( Xie et al . , 2013 ) and central noradrenaline activity ( Benveniste et al . , 2017 ) and how these factors may modulate measures of D* captured using the techniques presented here would be an interesting avenue of further study . Moreover , how the technique introduced here may be influenced by pathology is an important consideration . For example , the composition of the CSF and PV fluid may change in disease , in turn altering the relaxation times of this compartment ( for example the presence of iron could reduce PVS T2 ) . Whilst this may not confound measures of D* , as relaxation time changes will be accounted for by the acquisition of a b0 image at identical TR and TE , this may change contrast between the PVS and surrounding tissue and potentially lessen the SNR of the measurements . MR relaxometry studies targeted to the normal and enlarged PVS may be an interesting avenue of future investigation leading to novel biomarkers of PVS composition . Efforts are ongoing to investigate the sensitivity of the method to detect dysfunction of perivascular fluid movement associated with ageing and models of pathological conditions , with the knowledge that clinical translation of this non-invasive approach may be practically achievable in the near future . All imaging was performed using a 9 . 4T VNMRS horizontal bore scanner ( Agilent Inc . , Palo Alto , CA ) . A 72 mm inner diameter volume coil was used for RF transmission and signal was received using a 4 channel array head coil ( Rapid Biomedical ) . The imaging gradient hardware was calibrated using a custom designed structural phantom , as previously described ( O'Callaghan et al . , 2014 ) . A key aspect of the MRI sequence was the use of a long TE to attenuate the signal from the surrounding arterial blood and tissue ( T2 ~30 and 38 ms respectively at 9 . 4T [Wells et al . , 2013] ) relative to the MRI signal from CSF in the subarachnoid space and fluid in perivascular channels ( T2 ~111 ms [Kuo et al . , 2005] ) . In order to achieve this , a fast spin echo ( FSE ) sequence was employed ( 180° refocusing pulses ) with an echo train length of 16 giving an effective echo time of 142 ms ( thus the ultra-long TE is compatible with a multiple echo train FSE readout for SNR efficiency ) . Therefore , at this echo time , the signal from the grey matter tissue , blood and CSF will have decayed to ~2 , 1 and 28% of the theoretical signal at TE = 0 respectively . In addition , the use of an ultra-long TE permits a long echo train per excitation ( 16 echoes ) to increase the SNR efficiency of the acquisition ( i . e . SNR per unit time ) . Finally , the use of a relatively long TR ( 5000 ms ) , further weights the measured MRI signal from CSF/interstitial fluid relative to surrounding blood/tissue . It should be noted that , as part of the FSE readout , phase encoding lines will be acquired at a range of different TEs and thus the eventual contrast in the image may deviate from that predicted by assuming a constant TEeff across all phase-encoding steps . Simulations ( data not shown ) indicate that this effect was minimal in the current study but future applications should consider this aspect of MRI image capture . In this study , four separate sets of experiments were performed , that can be divided into ‘multiple direction diffusion weighted imaging’ ( n = 10 ) , ‘diffusion tensor imaging’ ( n = 6 ) , ‘ECG-gating ( n = 5 ) ’ and’ Dobutamine ( n = 6 ) ’ . An axial slice was positioned at the ventral aspect of the brain at the level of the Circle of Willis ( CoW - see Figure 1 ) . A series of scout images were acquired with the slice orientation and position manually altered in an iterative manner in order that the perivascular space around the MCA could be optimally visualised . The angular orientation of the image was then changed so that the animals right perivascular tracts ( surrounding the MCA in the axial slice ) was aligned with the orientation of the frequency encoding ( FE ) imaging gradients . In doing so , the animals left perivascular tracts then become approximately aligned with the phase encoding ( PE ) imaging gradients ( see Figure 2 ) . This ensured that , when applying diffusion ( or motion probing ) gradients along the FE direction , the direction of diffusion weighting was parallel to the right perivascular tract and perpendicular to the left tract; and vice versa when applying diffusion gradients along the PE direction . As a result , the sensitivity for measuring differences in fluid movement along and across both tracts was maximised . A fast-spin echo imaging sequence was acquired with the following sequence parameters: TR = 5 s , Echo Train Length = 16 , effective TE = 142 ms , echo spacing = 10 ms , FOV = 25 × 25 mm , matrix size = 128 × 128 , slice thickness = 0 . 8 mm or 1 mm , number of averages = 12 . A b = 0 image was acquired with minimal diffusion weighting ( b0 ) and then with separate acquisitions with the motion probing gradients applied in three principle directions ( X , Y , Z ) with a b-value of 107 s/mm2 ( δ = 5 ms , Δ = 26 ms , G = 4 . 2 G/cm ) . Regions of interest were manually drawn around the perivascular tracts surrounding the left and right MCA , as well as within the CSF of the subarachnoid space in the mid-section of the CoW from the b0 images . The subarachnoid space ROI was chosen because previous invasive measures have demonstrated rapid caudal-rostral CSF-tracer movement in this region ( Gaberel et al . , 2014; Mesquita et al . , 2015 ) . As such , data from this ROI can provide a degree of validation for the technique if the directionality of fluid movement is found to be consistent with the established caudal-rostral fluid movement . The pseudo-diffusion coefficient ( D* ) was then calculated for each direction of the applied motion probing gradients using the following equation:S=S0exp ( −bD∗ ) where S is the measured signal at b = 107 s/mm2 , S0 is the signal taken from the b0 image . In this work we choose to report the exponential decay coefficient as the pseudo diffusion coefficient ( D* ) since this is analogous to the Intra-voxel Incoherent Motion ( IVIM ) literature where in-vivo D* estimates reflect an unknown contribution from relatively coherent flow in large and/or directionally ordered vessels and isotropic fluid motion derived from randomly orientated vessels within a MRI voxel . A paired t-test was applied to investigate ( i ) if D* was greater when the motion probing gradient was applied parallel to the principle direction of the perivascular tracts , relative to application in each of the orthogonal planes for the left and right perivascular channels respectively; ( ii ) if D* in the subarachnoid space ROI was significantly greater in the FE and PE directions than in the through plane slice selection direction . Images were acquired with no ‘diffusion weighting’ ( b0 ) and then using motion probing gradients applied in 6 different directions ( δ = 7 . 5 ms , Δ = 52 ms , G = 1 . 5 G/cm , b value ~100 s/mm2 ) respectively with the following sequence parameters: TR = 5 s , Echo Train Length = 16 , effective TE = 142 ms , FOV = 30×15 mm , matrix size = 128×64 , slice thickness = 1 mm , number of averages = 24 . Pseudo-Diffusion tensors were generated using a calculated b-matrix that incorporated the ‘diffusion’ weighting introduced by the imaging gradients . As described above , ROIs were drawn around the perivascular tracts that surround the animal’s left and right MCA , as well as the CSF in the subarachnoid space that resides in the mid-section of the Circle of Willis . For visualisation purposes , pseudo-diffusion tensor ellipsoids were generated using the fanDTasia routines in Matlab ( Barmpoutis et al . , 2007 ) . For pseudo-diffusion tensor mapping , images were smoothed using an edge preserving filter and thresholded based on signal intensity , to remove signals that did not principally derive from fluid filled compartments and images were generated using the Explore DTI toolbox ( Leemans et al . , 2009 ) . Maps were colour coded according to their principle orientation . In one animal , the diffusion tensor sequence was applied to the brain immediately post-mortem . In these experiments , a three lead electrode was used to measure ECG signals in the bore of the magnet . The diffusion weighted sequence was acquired with the following parameters: TR = 5 s , Echo Train Length = 16 , effective TE = 142 ms , echo spacing = 10 ms , FOV = 25 × 25 mm , matrix size = 128 × 128 , slice thickness = 1 mm , number of averages = 12 , δ = 5 ms , Δ = 26 ms , diffusion gradient amplitude = 2 . 3 G/cm , b value ~45 s/mm2 , diffusion gradients applied in two directions ( in plane , parallel to the PVS around the left and right MCA respectively ) . Image capture was gated to the ECG signal and image acquisition began either directly after the r-wave or with an 80 ms delay . Given that the diffusion weighting is applied during the first echo time at 72 ms , this results in a delay of 36 ms from the r-wave to the centre of diffusion weighting ( i . e . the first 180° refocusing pulse ) or 116 ms with the additional 80 ms delay . As Δ was 26 ms in these acquisitions , the ‘diffusion weighting’ was therefore applied between 23 and 49 ms from the r-wave and 103 and 129 ms from the r-wave respectively . Given previous recordings of pulse wave velocity in the mouse brain of 2 . 69 m/s ( Di Lascio et al . , 2014 ) and given an approximate distance from the heart to the MCA of 10 cm in ~400 g rats ( together with the separation between adjacent r-waves to be ~150 ms ) we define the separate acquisitions to therefore take place during cerebral arterial pulsation or diastole . It should be noted that due to the ECG gating employed in these experiments , the TR will vary slightly between successive echo trains , but given the minimum TR was 5 s and that the r-r interval in the rat is ~150 ms , this should introduce relatively little variation into the measured MRI signal . ROIs were drawn around the left and right PVS and within the mid-section of the subarachnoid space as before . In addition , ROIS were drawn within the third ventricle to examine the r-wave delay dependence on measures of D* within ventricular CSF . The average D* in the PVS ( MPGs applied parallel and perpendicular to PVS orientation respectively ) was taken for each rat and a paired t-test was used to investigate if D* ( MPGs parallel to PVS orientation ) was greater during arterial pulsation relative to diastole for each region . Data were acquired in 6 male Sprague Dawley rats using the identical MRI sequence approach described above ( ‘ECG gating’ ) but with no ECG gating . Dobutamine ( n = 3 subcutaneous bolus , 2 mg/kg ( Buttrick et al . , 1988 ) in saline ~0 . 6–0 . 8 ml ) or saline vehicle ( n = 3 ) was then delivered and the same acquisitions were performed after bolus infusion .	Our brain is bathed in cerebrospinal fluid , a clear liquid that ‘cushions’ the fragile organ . This liquid travels into the brain along special channels – the perivascular space – that surround certain blood vessels . As the fluid washes in and out of the brain , it takes with it potentially harmful molecules , such as the aggregates that build up to cause Alzheimer’s disease . If this brain-cleaning system becomes faulty , it could lead to neurodegenerative diseases . However , it is extremely difficult to measure the activity of this intricate and delicate system , and most studies so far have had to use invasive techniques that usually require brain surgery . Now , Harrison et al . adapt a technique , called diffusion tensor magnetic resonance imaging ( MRI ) , to visualise how the cerebrospinal fluid moves in the perivascular space in healthy rats . The non-invasive MRI method captures how the cerebrospinal fluid is driven into the brain when the blood vessels nearby expand and contract; as the vessels pulsate with each heartbeat , there is a 300% increase in the movement of the fluid in the perivascular space . This approach could be applied to understand exactly how neurodegenerative diseases emerge when the cerebrospinal fluid stops to properly clean the brain . Ultimately , the method could be used to detect when the cleansing system starts to fail in people , which could help to treat patients before their brains accumulate too many harmful substances .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "short", "report", "neuroscience" ]	2018	Non-invasive imaging of CSF-mediated brain clearance pathways via assessment of perivascular fluid movement with diffusion tensor MRI
PPP-family phosphatases such as PP1 have little intrinsic specificity . Cofactors can target PP1 to substrates or subcellular locations , but it remains unclear how they might confer sequence-specificity on PP1 . The cytoskeletal regulator Phactr1 is a neuronally enriched PP1 cofactor that is controlled by G-actin . Structural analysis showed that Phactr1 binding remodels PP1's hydrophobic groove , creating a new composite surface adjacent to the catalytic site . Using phosphoproteomics , we identified mouse fibroblast and neuronal Phactr1/PP1 substrates , which include cytoskeletal components and regulators . We determined high-resolution structures of Phactr1/PP1 bound to the dephosphorylated forms of its substrates IRSp53 and spectrin αII . Inversion of the phosphate in these holoenzyme-product complexes supports the proposed PPP-family catalytic mechanism . Substrate sequences C-terminal to the dephosphorylation site make intimate contacts with the composite Phactr1/PP1 surface , which are required for efficient dephosphorylation . Sequence specificity explains why Phactr1/PP1 exhibits orders-of-magnitude enhanced reactivity towards its substrates , compared to apo-PP1 or other PP1 holoenzymes . PPP family phosphatases are metalloenzymes that carry out the majority of protein serine/threonine dephosphorylation ( Brautigan and Shenolikar , 2018 ) . The three Protein Phosphatase 1 ( PP1 ) isoforms regulate diverse cellular processes , acting in partnership with over 200 different PP1-interacting proteins ( PIPs ) . Some PIPs are PP1 substrates , but others are PP1 cofactors , which variously determine substrate specificity , subcellular targeting and/or coupling to regulatory pathways ( Bollen et al . , 2010; Cohen , 2002 ) . The PP1 catalytic site lies at the intersection of three putative substrate-binding grooves ( Egloff et al . , 1995; Goldberg et al . , 1995 ) , and PIPs can interact both with these grooves and with other PP1 surface features . To do this they use a variety of short sequence elements , of which the best understood is the RVxF motif ( Choy et al . , 2014; Egloff et al . , 1997; Hendrickx et al . , 2009; Hurley et al . , 2007; O'Connell et al . , 2012; Ragusa et al . , 2010; Terrak et al . , 2004 ) . Unlike protein Ser/Thr kinases , PP1 exhibits little sequence-specificity by itself ( Brautigan and Shenolikar , 2018; Miller and Turk , 2018 ) . Moreover , no instances of PIP-induced sequence-specificity are known , although it is well established that PIPs can enhance or inhibit PP1 activity towards particular substrates ( Ichikawa et al . , 1996; Johnson et al . , 1996 ) . Some PIPs contain autonomous substrate-binding domains , which facilitate substrate recruitment ( Boudrez et al . , 2000; Choy et al . , 2015 ) , while others constrain substrate specificity by occluding PP1 surfaces such as the RVxF-binding pocket and/or substrate-binding grooves ( Hirschi et al . , 2010; Ragusa et al . , 2010 ) . Interestingly , several PIPs extend the PP1 substrate-binding grooves and/or significantly alter PP1 surface electrostatics , without altering the conformation of the catalytic site ( O'Connell et al . , 2012; Ragusa et al . , 2010; Terrak et al . , 2004 ) . How this might affect substrate selection remains unclear , as the sequence-specificities of these PIP/PP1 holoenzymes , and how they bind their substrates , have not been characterised . The four Phosphatase and actin regulator ( Phactr ) proteins ( Allen et al . , 2004; Sagara et al . , 2003 ) are novel PIPs that are implicated in cytoskeletal regulation in animal models ( Hamada et al . , 2018; Kim et al . , 2007; Zhang et al . , 2012 ) and cell culture settings ( Huet et al . , 2013; Wiezlak et al . , 2012 ) . The Phactrs bind G-actin via multiple RPEL motifs present in their conserved N- and C-terminal regions ( Huet et al . , 2013; Mouilleron et al . , 2012; Sagara et al . , 2009; Wiezlak et al . , 2012 ) . Their C-terminal RPEL domain overlaps the PP1 binding sequence ( Allen et al . , 2004; Larson et al . , 2008; Sagara et al . , 2003 ) , and G-actin competes with PP1 for Phactr binding ( Huet et al . , 2013; Wiezlak et al . , 2012 ) . As a result , like other RPEL proteins , G-actin/Phactr interactions respond to fluctuations in actin dynamics ( Diring et al . , 2019; Miralles et al . , 2003; Vartiainen et al . , 2007 ) . Extracellular signals , acting through the ‘rho-actin’ signal pathway , thus control Phactr/PP1 complex formation by inducing changes in cellular G-actin concentration ( Wiezlak et al . , 2012 ) . The biochemical function of Phactr/PP1 complexes has been unclear . Phactr1 and Phactr3 inhibit dephosphorylation of phosphorylase a by PP1 in vitro ( Allen et al . , 2004; Sagara et al . , 2003 ) , but Phactr4/PP1 complex formation is associated with cofilin dephosphorylation in vivo ( Huet et al . , 2013; Zhang et al . , 2012 ) . Here , we show that Phactr1 confers sequence-specificity on the Phactr1/PP1 holoenzyme . We identify substrates for Phactr1/PP1 , and determine the structures of Phactr1/PP1-substrate complexes . We show that efficient catalysis requires interactions between conserved hydrophobic residues in the substrate and a novel Phactr1-PP1 composite surface , which comprises a hydrophobic pocket and associated amphipathic cavity with a surrounding basic rim . These interactions allow Phactr1/PP1 to recognise its substrates 100-fold more efficiently compared with PP1 alone or the spinophilin/PP1 complex , in which the hydrophobic groove is remodelled differently . Previous studies have shown that Phactr1 C-terminal sequences are necessary and sufficient for interaction with PP1 ( Allen et al . , 2004; Wiezlak et al . , 2012 ) : they contain an RVxF-like sequence , LIRF , and can functionally substitute for the related PP1-binding domain of yeast Bni4 ( Larson et al . , 2008; Figure 1A ) . We synthesised peptides corresponding to Phactr1 ( 517-580 ) from all four Phactr family members , and measured their affinity for recombinant PP1α ( 7-300 ) using Bio-layer interferometry ( BLI ) . Phactr1 ( 517-580 ) bound with an affinity of 10 . 4 nM , comparable to Phactr3 , while the other Phactrs bound more weakly ( Figure 1B and C , Figure 1—figure supplement 1A ) . This Kd is similar to PIPs such as spinophilin ( 8 . 7 nM ) and PNUTS ( 9 . 3 nM ) ( Choy et al . , 2014; Ragusa et al . , 2010 ) , and somewhat stronger than PPP1R15A and NIPP1 ( Choy et al . , 2015; O'Connell et al . , 2012 ) . We determined the structures of purified Phactr1 ( 507-580 ) /PP1α ( 7-300 ) ( 1 . 9 Å ) and Phactr1 ( 516-580 ) /PP1α ( 7-300 ) ( 1 . 94 Å ) , which crystallised at pH 8 . 5 and pH 5 . 25 , respectively , and exhibit differing conformations of Phactr1 residues C-terminal to residue 567 ( Figure 1D , Figure 1—figure supplement 1B ) . In both structures , the PP1 catalytic site contains two presumed manganese ions and a phosphate anion , as previously reported ( Figure 1—figure supplement 1C; Egloff et al . , 1995 ) , and its conformation is virtually identical to that seen in other PP1 complexes , such as spinophilin/PP1 ( RMSD 0 . 23 Å over 255 Cα atoms ) ( Choy et al . , 2014; Ragusa et al . , 2010 ) . The Phactr1 C-terminal sequences are well resolved in the pH 8 . 5 structure , which is also adopted by Phactr1 ( 516-580 ) /PP1α ( 7-300 ) when substrates occupy the PP1 active site , and which is supported by BLI data ( discussed below ) . It is therefore likely be representative of the structure at physiological pH . In contrast , at pH 5 . 25 Phactr1 C-terminal sequences adopted a distinct and poorly resolved conformation , perhaps induced by protonation of the three C-terminal histidines ( Figure 1—figure supplement 1B ) . In the discussion that follows , we focus on the pH8 . 5 structure . Phactr1 residues 516–542 wrap around PP1 , occluding its C-terminal groove , and covering 2260 Å2 of solvent-accessible surface ( Figure 1D , Figure 1—figure supplement 1D ) , contacting PP1 in a strikingly similar way to spinophilin , PNUTS and PPP1R15B ( Chen et al . , 2015; Choy et al . , 2014; Ragusa et al . , 2010 ) . These contacts include a non-canonical RVxF motif ( Egloff et al . , 1997; Hendrickx et al . , 2009 ) , a ϕϕ motif ( O'Connell et al . , 2012 ) , an Arg motif ( Choy et al . , 2014 ) , and a previously unrecognised Trp motif ( Figure 1E–H; Figure 1—figure supplement 2A ) . The RVxF sequence LIRF ( 519-522 ) is critical for Phactr1/PP1 complex formation ( Figure 1E , Figure 1—figure supplement 2A ) . Its deletion decreased binding affinity ~104 fold , while the I520A and F522A mutations reduced it 4-fold and ~650 fold respectively ( Figure 1C ) . The RVxF residue L519 also makes contacts with G-actin in the trivalent G-actin•Phactr1 RPEL domain complex ( Mouilleron et al . , 2012 ) , explaining why PP1 and G-actin binding to Phactr proteins is mutually exclusive ( Figure 1—figure supplement 2B; Wiezlak et al . , 2012 ) . The other contacts also contribute to PP1-binding affinity . The ϕϕ motif , EVAD ( 527-530 ) , adds a β-strand to PP1 β-strand β14 , extending one of PP1’s two central β-sheets ( Figure 1F ) . The Phactr1 Arg motif contacts the PP1 C-terminal groove , with R536 forming a bidentate salt bridge with PP1 D71 , and a hydrogen bond with N271 . This is stabilised by an intrachain cation-π interaction with Phactr1 Y534 , which also hydrogen bonds with PP1 D71 , and makes hydrophobic contacts with P24 and Y70 ( Figure 1G , Figure 1—figure supplement 2A ) . Substitution of R536 by proline , as in the mouse Phactr4 ‘humdy’ mutation ( Kim et al . , 2007 ) , reduced Phactr1-binding affinity >300 fold , while Y534A reduced it > 10 fold ( Figure 1C ) . The Trp motif , W542 , which is constrained by a salt bridge between R544 and D539 , makes hydrophobic contacts with PP1 I133 and Y13 ( Figure 1H ) , similar to those in the PPP1R15B and spinophilin PP1 complexes ( Chen et al . , 2015; Ragusa et al . , 2010; Figure 1—figure supplement 2A ) . Although W542A reduced PP1-binding affinity ~40-fold , R544A reduced it ~ threefold ( Figure 1C ) . The Phactr1 sequences C-terminal to the RVxF-ϕϕ-R-W string form a novel structure specific to the Phactr1/PP1 complex . Residues 545–565 include a five-turn amphipathic α-helix that abuts the PP1 hydrophobic groove: Phactr1 residues L545 , I553 , L557 , F560 make hydrophobic contacts with PP1 I133 , R132 , W149 , and K150 , while E556 and E564 make salt bridges with PP1 R132 and K150 , respectively , and K561 hydrogen bonds with PP1 S129 and D194 ( Figure 2A ) . Phactr1 then turns , contacting the PP1 α7-α8 loop: Phactr1 H568 hydrogen bonds to PP1 P192 and Phactr1 S571 , making hydrophobic contacts with PP1 M190 , while Phactr1 L570 and L574 make hydrophobic contacts with PP1 M190 and I189/P196/L201 respectively . The Phactr1 C-terminus then folds back onto the amphipathic helix , stabilised by multiple intrachain interactions . These include hydrophobic contacts between Phactr1 T575 , F577 and K561; hydrogen bonds between the R576 carbonyl and K561 , between the H578 amide and N558 , and between the R579 carbonyl and R554 , and a salt bridge between the P580 carboxylate and the R554 guanidinium . The Phactr1 C-terminus also contacts PP1 through hydrogen bonds between R576 and PP1 D194 , and H578 and PP1 S129 ( Figure 2A ) . Mutagenesis experiments confirmed the importance of these interactions for Phactr1-PP1 binding ( Figure 1C ) . A triple alanine substitution of Phactr1 L557 , F560 and L574 ( LFL-3A ) resulted in a ~ 900 fold drop in binding affinity , while the acidic substitution L574D reduced binding affinity by 17-fold . Mutations F577A and H578A reduced binding activity by 16-fold and 7-fold , respectively , corroborating previous co-immunoprecipitation experiments ( Allen et al . , 2004; Sagara et al . , 2009 ) , and mutation of all three C-terminal histidines ( HHH-A ) reduced affinity >50 fold ( Figure 1C ) . The docking of Phactr1 C-terminal sequences across the PP1 hydrophobic groove substantially modifies its topography ( Figure 2B ) . The positioning of the Phactr1 amphipathic helix creates a deep hydrophobic pocket comprising Phactr1 W542 , L545 , K550 , I553 , R554 , L557 and H578 , and PP1 I130 I133 and Y134 . Adjacent to the pocket , a narrow amphipathic cavity is formed by the positioning of Phactr1 K561 and R576 across the PP1 hydrophobic groove: its polar side comprises Phactr1 K561 and R576 , and PP1 D194 and S129 , while its hydrophobic side is formed from PP1 residues C127 , I130 , V195 , W206 and V223 . Three water molecules and a glycerol are resolved within the cavity , whose hydrophobic side forms part of the binding site for the PP1 inhibitors tautomycetin and tautomycin ( Choy et al . , 2017 ) . The new composite surface is crowned by a basic rim , formed from Phactr1 residues K550 , R554 , R576 , H578 and R579 ( Figure 2B ) , which radically alters the surface charge distribution ( Figure 2C ) . Thus , Phactr1 binding profoundly transforms the surface of PP1 adjacent to its catalytic site . This transformation is distinct from that seen in the spinophilin/PP1 complex ( Ragusa et al . , 2010 ) , which modifies the hydrophobic groove in a different way and leaves the PP1 surface electrostatics unchanged ( Figure 2—figure supplement 1 ) . We previously showed that expression of an activated Phactr1 derivative that constitutively forms the Phactr1/PP1 complex , Phactr1XXX , induces F-actin rearrangements in NIH3T3 fibroblasts , provided it can bind PP1 ( Wiezlak et al . , 2012 ) . Indeed , overexpression of the Phactr1 PP1-binding domain alone can also induce such changes , suggesting that the Phactr1 C-terminal sequences are sufficient to allow recognition of at least some substrates ( Figure 3—figure supplement 1A; see Discussion ) . These observations suggest that the Phactr1/PP1 complex might dephosphorylate target proteins involved in cytoskeletal dynamics . To identify potential Phactr1/PP1 substrates , we used differential SILAC phosphoproteomics in NIH3T3 cells expressing Phactr1XXX . Over 3000 phosphorylation sites were quantified , and each assigned a dephosphorylation score comparing their phosphorylation with that observed in cells inducibly expressing Phactr1XXXΔC , which lacks the PP1 binding sequences , or vector alone ( Figure 3A; Figure 3—figure supplement 1B , Supplementary file 1 , Table A ) . Annotation enrichment analysis of the whole dataset ( Cox and Mann , 2012 ) identified two Gene Ontology Biological Process categories that exhibited a significantly higher mean dephosphorylation score upon expression of Phactr1XXX: ‘regulation of actin filament-based process’ and ‘regulation of actin cytoskeleton organisation’ ( Figure 3B , Supplementary file 1 , Table B ) . In keeping with this , proteins with a high dephosphorylation score included many cytoskeletal components and regulators ( Figure 3C , Supplementary file 1 , Table A ) . Proline-directed sites predominated in the dataset as a whole , but those sites with dephosphorylation scores > 2 . 5 were enriched in acidic residues at positions +2 to +7 relative to the phosphorylation site , with small hydrophobic residues enriched at positions +4 and +5 , and basic residues N-terminal to it ( Figure 3D ) . Since this sequence bias was not observed at all sites , we tested directly whether sites were Phactr1/PP1 substrates using an in vitro peptide dephosphorylation assay . Synthetic phosphopeptides containing the candidate sites exhibited a range of KM and kcat/KM values: for example , IRSp53 pS455 exhibited a low KM and high kcat/KM; spectrin αII pS1031 had a similar KM but lower kcat/KM; and afadin pS1275 was as reactive as spectrin αII pS1031 , but with much poorer KM ( Figure 3C; Figure 3—figure supplement 1C; Supplementary file 2 ) . Interestingly , the four substrates with the lowest KM - IRSp53 pS455 , Plekho2 pS395 , spectrin αII pS1031 , and Tbc1d15 pS205 - all contained a leucine doublet at the +4 and +5 positions ( see Discussion ) . We generated phosphospecific antisera against IRSp53 pS455 , spectrin αII pS1031 , and afadin pS1275 . Immunoblot analysis showed that in NIH3T3 cells , phosphorylation of IRSp53 S455 and afadin S1275 was substantially decreased upon expression of Phactr1XXX , and by serum stimulation , which activates rho-actin signalling and Phactr-family/PP1 complex formation ( Miralles et al . , 2003; Vartiainen et al . , 2007; Wiezlak et al . , 2012; Figure 3—figure supplement 1D; Supplementary file 3 ) . Moreover , treatment of cells with cytochalasin D ( CD ) , which binds G-actin and disrupts its interaction with RPEL proteins ( Vartiainen et al . , 2007; Wiezlak et al . , 2012 ) significantly decreased phosphorylation of IRSp53 S455 and afadin S1275 ( Figure 3—figure supplement 1E; Supplementary file 3 ) . In contrast , treatment of cells with the latrunculin B ( LB ) , which increases G-actin concentration through blockade of F-actin assembly , but whose binding to G-actin does not affect RPEL-actin interaction , had the opposite effect ( Figure 3—figure supplement 1E; Supplementary file 3 ) . These data suggest that endogenous RPEL protein ( s ) , presumably Phactr-family members , control IRSp53 S455 and afadin S1275 phosphorylation in NIH3T3 cells . To explore directly the specific involvement of Phactr1 in protein dephosphorylation we turned to neurons , which express Phactr1 at high level ( Allen et al . , 2004 ) . Phactr1 mutations are associated with morphological and functional developmental defects in cortical neurons ( Hamada et al . , 2018 ) , and expression of Phactr1XXX induced morphological defects upon expression in cultured hippocampal neurons ( Figure 3—figure supplement 2A ) . To assess whether Phactr1 controls phosphorylation in neurons , we analysed hippocampal and cortical neurons from wildtype and Phactr1-null animals . Phactr1-null mice are viable; they do not show any obvious developmental abnormalities , and expression of the other Phactr proteins is apparently unaffected ( Figure 3—figure supplement 2B ) . Neurons were treated with LB or CD to inhibit or activate RPEL proteins , and phosphorylation profiles analysed using TMT phosphoproteomics . Among ~9000 phosphorylation sites quantified , we found 44 sites on 37 proteins that differed significantly in their response to these stimuli in Phactr1-null cells ( Figure 3E , Supplementary file 1 , Table E ) . The sequence context of these sites was similar to that of those seen in NIH3T3 cells ( Figure 3—figure supplement 2C ) , and seven , including IRSp53 pS455 , spectrin αII pS1031 , and afadin pS1275 , were observed in both cell types ( Figure 3F , Figure 3—figure supplement 2D; Supplementary file 3 ) . In sum , these results identify multiple Phactr1 substrates in NIH3T3 cells and neurons , many of which are cytoskeletal components or regulators . To understand the molecular basis for substrate recognition by Phactr1/PP1 , we sought to determine the structures of Phactr1/PP1-substrate complexes . We were unable to co-crystallise Phactr1/PP1 with unphosphorylated or glutamate-phosphomimetic peptides encompassing IRSp53 S455 , afadin S1275 or spectrin αII S0131 , so we used a PP1-substrate fusion strategy similar to that used for PP5 ( Oberoi et al . , 2016 ) . Fusion proteins comprising PP1 ( 7–304 ) joined via a ( SG ) 9 linker to unphosphorylated substrate sequences were coexpressed with Phactr1 ( 516-580 ) for structural analysis . This approach allowed determination of the structures of the IRSp53 ( 449-465 ) and spectrin αII ( 1025–1039 ) complexes at 1 . 09 Å and 1 . 30 Å resolution , respectively ( Figure 4A–C; Table 1; Figure 4—figure supplement 1; referred to hereafter as the IRSp53 and spectrin complexes ) , but was unsuccessful with Tbc1d15 , Plekho2 , afadin , and cofilin . The Phactr1/PP1-IRSp53 and Phactr1/PP1-spectrin complexes crystallised in the same spacegroup . Each asymmetric unit contains two complexes ( RMSD 0 . 22 Å and 0 . 17 Å over 288 Cα , respectively ) , in which substrate-Phactr1/PP1 interactions are mostly conserved , albeit with some minor differences ( Figure 4—figure supplement 1A and B ) . The substrate sequences , whose N-termini are largely unresolved , make extensive contacts across the PP1 catalytic site , then extend in a sinuous trajectory across the composite Phactr1/PP1 surface , making numerous hydrophobic and ionic contacts ( Figure 4C ) . Like the Phactr1/PP1 holoenzyme , the complexes contain a presumed phosphate anion at the catalytic site , and thus appear to represent putative enzyme/product complexes ( see below ) . The structure of Phactr1/PP1 in both complexes is identical to that of the isolated Phactr1/PP1 complex ( RMSD 0 . 31 Å over 313 Cα and 0 . 32 Å over 307 Cα , respectively ) . IRSp53 and spectrin make virtually identical contacts with the PP1 catalytic cleft , predominantly via their mainchains ( Figure 4C; Figure 4—figure supplement 1C ) . In IRSp53 complex 2 , K452IRSp53 ( −3 relative to the phosphorylation site ) makes a salt bridge with PP1 D220 carboxyl , in the acidic groove , while water-bridged hydrogen bonds link the mainchain carbonyl and the sidechain hydroxyl of S453IRSp53/S1029spectrin ( −2 ) to PP1 Y272 and the phosphate , and to V250 , respectively . The dephosphorylated S455IRSp53/S1031spectrin ( 0 ) hydroxyl interacts with PP1 R96 , H125 and the phosphate , its mainchain amide and carbonyl contacting the phosphate and PP1 R221 , respectively ( Figure 4A–C , Figure 4—figure supplement 1A and B ) . The phosphate is inverted compared with the holoenzyme complex , losing its contact with PP1 R96 and H125 , but making contact with the S455IRSp53/S1031spectrin hydroxyl ( Figure 4D; Figure 4—figure supplement 1A–C ) . C-terminal to the dephosphorylated serine , T456IRSp53/R1032spectrin ( +1 ) hydrogen bonds with PP1 Y134 via its mainchain amide , with R1032spectrin making two additional hydrogen bonds , with PP1 Y134 via its carbonyl , and the D220 carbonyl via its sidechain . The majority of the PP1 catalytic cleft residues that contact IRSp53 and spectrin are conserved among PPP family members , including PP5 , which has been crystallised in complex with a phosphomimetic derivative of its substrate , Cdc37 ( S13E ) ( Oberoi et al . , 2016 ) . Strikingly , in that complex , the PP5 residues corresponding to PP1 R96 , H125 , Y134 , R221 , and Y272 make contacts with Cdc37 ( S13E ) analogous to those seen in the Phactr1/PP1-substrate complexes , despite the fact that Cdc37 ( S13E ) docks in the PP5 catalytic cleft in the opposite orientation to IRSp53 and spectrin ( Figure 4—figure supplement 2; see Discussion ) . It is generally accepted that the phosphate seen in many PPP family protein structures binds the active site in a similar way to the substrate phosphate ( Egloff et al . , 1995; Griffith et al . , 1995; Mueller et al . , 1993; Swingle et al . , 2004 ) , and we therefore assume that the phosphate in our Phactr1/PP1 holoenzyme structure is positioned similarly to the phosphorylated S455IRSp53/S1031spectrin of the bound substrate ( Figure 4D ) . Consistent with this idea , in the PP5/Cdc37 ( S13E ) complex , the phosphomimetic glutamate sidechain carbonyl is virtually superposable on the phosphate in the Phactr1/PP1 complex ( Figure 4—figure supplement 3 ) . The Phactr1/PP1 complex also contains a bound water , W1 , presumably activated by the metal ions and PP1 D64 and D92 ( Figure 4D ) . W1 is oriented appropriately for in-line nucleophilic attack on the substrate phosphate: this would be facilitated by protonation of the phosphoserine oxygen by the PP1 catalytic dyad H125-D95 , and would result in inversion of the phosphate ( Figure 4E , Figure 4—figure supplement 3 ) . The structures of our Phactr1/PP1-IRSp53/spectrin complexes are consistent with their representing the resulting enzyme-product complexes , stablised through the tethering of the substrate sequences to the holoenzyme . The structures thus provide direct evidence for the in-line nucleophilic hydrolysis mechanism for PPP family phosphatases , as outlined by the Barford and Ciszak groups ( Egloff et al . , 1995; Swingle et al . , 2004 ) . C-terminal to their interaction with the catalytic cleft , both IRSp53 and spectrin follow similar trajectories from residues +2 to +7 ( Figure 4A–C , Figure 4—figure supplement 1C ) . The most striking feature of the complexes is the multiplicity of contacts made between the substrate and the novel composite Phactr1/PP1 surface created by extension of the PP1 hydrophobic groove , most of which are conserved between the two substrates ( Figure 4C and F ) . In both structures , the residues +3 to +6 form a β-turn , allowing the hydrophobic doublet L459-L460IRSp53/L1035-L1036spectrin ( +4/+5 ) to make intimate contact with the Phactr1/PP1 hydrophobic pocket , entirely burying the +5 leucine sidechain ( Figure 4C and F ) . These interactions are stabilised by hydrogen bonds between the mainchain carbonyls of substrate residues +2 , +4 , +5 and Phactr1 R576 K550 , and R554 , and between the sidechain of the substrate acidic +6 residue ( D461IRSp53/E1037spectrin ) and Phactr1 H578 ( Figure 4A–C ) ; a hydrogen bond between the D461IRSp53 mainchain carbonyl and Phactr1 R579 is substituted by a hydrogen bond between the E1037spectrin carboxylate and Phactr1 R579 mainchain amide . At position +2 ( G457IRSp53/E1033spectrin ) , the mainchain amide and carbonyl make hydrogen bonds with the mainchain carbonyls of PP1 R221 and Phactr1 R576 respectively; in addition , the E1033spectrin side chain spans the top of the amphipathic cavity to make an additional salt bridge with the Phactr1 R576 side chain ( Figure 4B and C ) . To investigate the functional significance of the Phactr1/PP1-substrate interactions seen in the structures , we tested mutated IRSp53 peptide substrates in the in vitro dephosphorylation assay ( Figure 5A ) . Alanine substitutions at positions S453 ( −2 ) , S454 ( −1 ) , and T456 ( +1 ) decreased catalytic efficiency somewhat , apparently reflecting an increased KM . In contrast , alanine substitution of residues L459 , L460 , and D461 ( +4 to +6 ) individually increased KM , but had variable effects on catalytic efficiency , with L459A and D461A increasing it , and L460A reducing it ( see Discussion ) . Pairwise combination of substitutions at positions +4 through +6 suppressed reactivity even more , with the LL459/460AA mutant being essentially unreactive . Alanine substitution at K462 ( +7 ) and D463 ( +8 ) either decreased or increased KM with corresponding effects on catalytic efficiency , consistent with a preponderance of acidic residues at positions +6 to +8 ( see Figure 3D ) . Upon expression in NIH3T3 cells , IRSp53 ( L460A ) exhibited enhanced basal phosphorylation , and was less susceptible to CD-induced dephosphorylation than wildtype IRSp53 , as assessed using the phospho-S455 antibody ( Figure 5B ) . These data show that substrate interactions with the Phactr1/PP1 hydrophobic pocket are important for efficient dephosphorylation both in vitro and in vivo . The effects of substrate alanine mutations on KM suggests that they affect the affinity of Phactr1/PP1-substrate interactions . To investigate substrate binding directly , we used a peptide overlay assay in which unphosphorylated substrate peptides were immobilised on membranes , and tested for their ability to recruit recombinant GST-Phactr1 ( 516-580 ) /PP1 ( 7–300 ) complex from solution . Since these peptides lack phosphoserine , it is likely that their binding will predominantly be determined by interactions outside the catalytic site . Only peptides representing low KMsubstrates - IRSp53 , spectrin αII , Tbc1d15 and Plekho2 - exhibited detectable binding under the conditions of the assay ( Figure 5—figure supplement 1; see Figure 3C ) . To assess the contribution of individual residues to binding , we used IRSp53 and spectrin arrays in which each substrate residue was systematically changed to every other amino acid ( Figure 5C and D ) . Tryptophan and cysteine substitutions at non-critical residues led to a general increase in binding affinity , and we therefore did not attempt to interpret these substitutions . Analysis of both arrays implicated residues −2 to +7 in substrate-binding affinity . Positions +4 and +5 displayed the strongest selectivity , for hydrophobic residues and leucine respectively: indeed , phenylalanine substitution of IRSp53 L459 ( +4 ) , increased binding affinity and catalytic efficiency ( Figure 5A ) . In contrast , the sequence dependence at other positions was strongly context-dependent . In the vicinity of the target residue , IRSp53 binding was relatively unaffected by N-terminal variation , while T456 ( +1 ) was suboptimal , with basic or hydrophobic residues being preferred; in contrast , spectrin binding was strongly selective at N-terminal positions , with a preference for basic residues at −1 and +1 , and serine at −2 . Similarly , IRSp53 exhibited a strong preference for acidic residues at position +6 , perhaps reflecting the presence of the suboptimal neighbouring basic residue , K462 ( +7 ) , which was not the case for spectrin . Taken together with the kinetics data , these results show that interactions with the Phactr1/PP1 hydrophobic pocket are a critical determinant of Phactr1/PP1 substrate recognition . In IRSp53 , the dephosphorylated residue S455 is flanked by three other potential phosphoacceptors , S453 , S454 , and T456 . We considered the possibility that strong interactions with the Phactr1/PP1 hydrophobic pocket might also allow efficient dephosphorylation of these residues . That this might be the case was suggested by our structural analysis of a phosphomimetic fusion construct , in which the S455 is substituted by glutamate , at 1 . 39 Å resolution ( Phactr1/PP1-IRSp53 ( S455E ) ) . Strikingly , in this complex , the phospomimetic glutamate ( S455E ) does not replace phosphate at the catalytic site . Instead , IRSp53 T456 occupies the position corresponding to S455 in the wildtype complex , with a recruited phosphate positioned as in the wild-type complex , and the glutamate ( S455E ) in effect occupies position −1 , pointing away from the catalytic site ( Figure 6A ) . Nevertheless , the critical contacts between IRSp53 ( S455E ) L459/L460 and the Phactr1/PP1 hydrophobic pocket are maintained , extending the intervening sequence G457-N458-L459 ( Figure 6B ) . Consistent with this , both wildtype IRSp53 pT456 and IRSp53 E455/pT456 were effective substrates for Phactr1/PP1 , exhibiting three- to fivefold higher KM , but only twofold lower catalytic efficiency , while IRSp53 pS453 and pS454 were very poor substrates , exhibiting 10–30 fold lower catalytic efficiency ( Figure 6C ) . Thus , Phactr1/PP1 substrates can tolerate either three or four residues between the target residue and the hydrophobic residue that engages the Phactr1/PP1 hydrophobic pocket ( see Discussion ) . Since the composite hydrophobic surface of the Phactr1/PP1 holoenzyme plays an important role in substrate binding and catalytic efficiency , we next investigated to what extent interaction with Phactr1 confers substrate specificity on PP1 . We were particularly interested to test whether the high KM and lower binding affinity of Phactr1/PP1 substrates such as afadin and cofilin might be associated with an increased promiscuity in their reactivity with different PP1 complexes . In addition to comparing Phactr1/PP1 with apo-PP1 , we therefore also compared it with the spinophilin/PP1α complex ( Figure 1—figure supplement 2A , Figure 6—figure supplement 1A; hereafter spinophilin/PP1 ) ( Ragusa et al . , 2010 ) . As discussed above , spinophilin also interacts with PP1 through an extended RVxF-ϕϕ-R-W string , but remodels the hydrophobic groove differently , and does not change PP1 surface electrostatics ( Figure 2—figure supplement 1; Ragusa et al . , 2010 ) . We also assessed the substrate specificity of a fusion protein in which the Phactr1 residues 526–580 were fused via a short SG linker to PP1 residue 304 , solved at 1 . 78 Å resolution ( Figure 6D ) . This fusion protein lacks the RVxF-PP1 interaction critical for formation of the authentic Phactr1/PP1 complex , but nevertheless generates a composite hydrophobic surface virtually identical to that seen in the Phactr1/PP1 holoenzyme ( RMSD 0 . 25 Å over 2395 atoms; Figure 6—figure supplement 1B ) . We tested a series of Phactr1/PP1 substrate phosphopeptides with decreasing catalytic efficiency and increasing KM , and analogous phosphopeptides from glycogen phosphorylase and GluR1 , substrates of the GM/PP1 and spinophilin/PP1 complexes ( Hu et al . , 2007; Ragusa et al . , 2010 ) . Compared with apo-PP1 and spinohilin/PP1 , Phactr1/PP1 exhibited 100- to 400-fold greater catalytic efficiency against its substrates IRSp53 pS455 , CD2ap pS458 , and afadin pS1275 ( Figure 6E ) . However , while Phactr1/PP1 dephosphorylated cofilin pS3 with a low catalytic efficiency comparable to afadin pS1275 , this was only twofold enhanced compared with PP1 or spinophilin/PP1 ( Figure 6E; see Discussion ) . Thus , at least some Phactr1 substrates with high KM , are likely to be substates for multiple different PP1 complexes ( see Discussion ) . Phactr1 did not enhance the activity of PP1 against the GM/PP1 substrate glycogen phosphorylase pS15 , or the spinophilin/PP1 target GluR1 pS863 ( Figure 6E , Figure 6—figure supplement 1C ) . The PP1-Phactr1 fusion protein behaved similarly to the Phactr1/PP1 complex , indicating that RVxF interactions are not essential for specificity ( Figure 6E ) . We found that spinophilin did not enhance PP1 activity against the short GluR1 pS863 peptide ( Figure 6—figure supplement 1C ) , although it is active against a longer GluR1 fragment ( Ragusa et al . , 2010 ) , suggesting that substrate specificity is not directed by its modified hydrophobic groove . Taken together with the results in the preceding sections , these results show that the composite surface formed by interaction with Phactr1 is responsible for the substrate specificity of the Phactr1/PP1 holoenzyme . We used proteomic approaches coupled with Phactr1 overexpression and inactivation studies to show that Phactr1/PP1 dephosphorylates multiple target proteins involved in cytoskeletal structures or regulation . This supports previous studies showing that cytoskeletal phenotypes result from Phactr1 and Phactr4 mutations ( Hamada et al . , 2018; Kim et al . , 2007; Zhang et al . , 2012 ) , and are induced by overexpression of Phactr1 and Phactr4 derivatives that constitutively associate with PP1 ( Huet et al . , 2013; Wiezlak et al . , 2012 ) . Inspection of the top-scoring Phactr1/PP1 substrates shows that acidic residues are over-represented C-terminal to the target phosphorylation site , and that hydrophobic residues are preferred at positions +4/+5 . Signal-regulated dephosphorylation of actin regulators by Phactr/PP1 complexes provides a new perspective on cytoskeletal regulation . For example IRSp53 was previously characterised as a Cdc42/Rac effector that controls F-actin assembly at membrane protrusions ( Scita et al . , 2008 ) . The Phactr1/PP1 target residue pS455 characterised here is one of four negatively-acting putative AMPK sites that recruit 14-3-3 proteins ( Cohen et al . , 2011; Kast and Dominguez , 2019; Robens et al . , 2010 ) and our data indicate another of these sites , S367 , is also a Phactr1/PP1 target ( Supplementary file 1A ) . Rho-actin signalling thus provides an additional positively acting signal input to IRSp53 , probably controlling interaction of the neighbouring SH3 domain with its effectors . Interestingly , the Phactr1/PP1 sites in spectrin αII and girdin are also adjacent to SH3 and SH2 domains , respectively ( Lin et al . , 2014; Figure 3C; Supplementary file 1 ) . Our data also confirm that the actin depolymerising factors cofilin and destrin are also dephosphorylated ( and activated ) by Phactr1 , in agreement with previous studies of Phactr4 ( Huet et al . , 2013 ) . Since Phactr proteins are inactivated by G-actin , their activation of cofilin potentially provides a feedback loop that couples F-actin severing to decreased G-actin level , but more work is required to establish this . Phactr1 is enriched at the post-synaptic density ( PSD ) ( Allen et al . , 2004 ) , and is required for neuronal migration and arborisation ( Hamada et al . , 2018 ) . We identified multiple Phactr1-dependent substrates in hippocampal neurons . Several of these , including IRSp53 and spectrin αII , are dephosphorylated during long-term potentiation ( LTP ) ( Li et al . , 2016 ) , and IRSp53-null mice exhibit deficits in hippocampal learning and memory ( Bobsin and Kreienkamp , 2016; Kang et al . , 2016 ) . Moreover , dephosphorylation of cofilin is implicated in early-stage dendritic spine remodelling , along with G-actin itself ( Bosch et al . , 2014; Lei et al . , 2017 ) . These data suggest that rho-actin signalling to Phactr1 and the resulting protein dephosphorylations may contribute to synaptic plasticity , and indeed in humans Phactr1 mutations cause the infantile seizure condition West syndrome ( Hamada et al . , 2018 ) . There may be multiple targets for rho-actin signalling to RPEL proteins in this setting , as the RPEL protein ArhGAP12 ( Diring et al . , 2019 ) also influences dendritic spine morphology ( Ba et al . , 2016 ) . Future work will focus on the functional significance of neuronal rho-actin signalling to Phactr/PP1 substrates . We used a PP1-substrate peptide fusion strategy to characterise Phactr1/PP1-substrate interactions . In our structures , the substrate sequences are not phosphorylated , but a phosphate is recruited to the catalytic cleft: they thus appear to represent enzyme/product complexes , presumably stabilised by virtue of the protein fusion . The inversion of the phosphate relative to its orientation in the Phactr1/PP1 holoenzyme provides persuasive support for the in-line nucleophilic attack model for PPP phosphatases proposed by others ( Egloff et al . , 1995; Swingle et al . , 2004 ) . Substrate interaction with the catalytic cleft , which does not change its conformation , is predominantly mediated by mainchain interactions with residues conserved among the PPP family . We were surprised to see that Phactr1/PP1 substrates dock with the catalytic cleft in the opposite orientation to that previously seen in a complex between PP5 and a phosphomimetic substrate derivative ( Oberoi et al . , 2016 ) . However , the clear sequence bias observed in Phactr1/PP1 substrate sequences suggests that the orientation seen in our IRSp53 and spectrin αII complexes is strongly preferred . The most striking aspect of substrate recognition is the role played by the composite surface generated by the close apposition of the Phactr1 extreme C-terminal sequences and the PP1 hydrophobic groove . This creates a new hydrophobic pocket , into which the substrate +4/+5 hydrophobic residues are inserted , and an adjacent amphipathic cavity , which appears less important , making no specific contacts with IRSp53 , and being only partially occupied by spectrin αII . The preference for acidic residues C-terminal to the dephosphorylation site presumably reflects the basic electrostatics associated with the novel composite Phactr1/PP1 surface . Biochemical studies suggest that the additional binding energy provided by substrate interactions with the hydrophobic pocket is responsible for the enhanced catalytic rate of Phactr1/PP1 compared with apo-PP1 or other PP1 complexes . According to this view , it might be expected that substrates with higher KM might be more promiscuous in their interaction with PP1 complexes . Cofilin pS3 might be such a substrate: it exhibits similar catalytic efficiency to afadin pS1275 , but has little preference for Phactr1/PP1 over apo-PP1 or spinophilin/PP1 . However , high binding affinities might limit catalytic efficiency by slowing product dissociation , a situation perhaps exemplified by spectrin . PP1 inhibitors generally lack specificity because they target the catalytic site ( Zhang et al . , 2013 ) , and our findings suggest that targeting the composite Phactr1/PP1 surface may be a good strategy to create specific Phactr/PP1 inhibitors . An interesting consequence of the winding trajectory of the substrate following the catalytic site , and its strong interaction with the hydrophobic pocket , is that variation in the number of residues between the two can be tolerated . For example , IRSp53 pT456 , with only four residues to the pocket-bound L460 , has comparable kinetic properties to the bona fide IRSp53 pS455 . We speculate that Phactr1/PP1 substrates such as afadin pS1275 , where the sole hydrophobic residue is at +4 relative to the target residue , may well represent such ‘stretched’ substrates . This flexibility complicates the unambiguous definition of the dephosphorylation consensus site , and further studies will be required to produce a structure-based alignment of the substrates that we have identified . We propose that Phactr/PP1 substrates can be defined by a hydrophobic doublet at position +4/+5 or +3/+4 relative to the dephosphorylation site , with leucine preferred at the distal position , embedded within acidic sequences . The simple Phactr1/PP1 core recognition sequence , S/T-x ( 2-3 ) -ϕ-L , is reminiscent of those seen for protein kinases ( Miller and Turk , 2018 ) and is perhaps the first identified for PP1 . Mutation of positions +4 and/or +5 in candidate Phactr/PP1 substrates will be a useful way to generate authentic constitutively phosphorylated mutants for functional studies , as an alternative to ‘phosphomimetic’ glutamate substitution of the phosphorylated residue . The lack of structural information for PP1-substrate complexes has hitherto precluded demonstration of how PIP-dependent remodelling of the substrate-binding grooves of PP1 can influence its ability to recognise substrate . In spinophilin , which also binds PP1 via a RVxF-ϕϕ-R-W string , sequences C-terminal to the W-motif form a helix that remodels the PP1 hydrophobic groove differently ( Ragusa et al . , 2010 ) . Spinophilin therefore does not detectably enhance PP1 activity towards Phactr/PP1 substrates , although the relative contributions of the spinophilin-remodelled PP1 surface , and the adjacent PDZ domain , to substrate specificity remain to be determined ( Ragusa et al . , 2010 ) . In contrast , MYPT interacts with PP1 in a different way to Phactr1 and spinophilin , extending both the C-terminal and acidic grooves ( Terrak et al . , 2004 ) , and it will be interesting to see how this facilitates substrate recognition . In all these complexes , PIP-PP1 interaction also acts indirectly to constrain substrate binding specificity , by occluding potential substrate-binding sites on the PP1 surface such as the RVxF binding pocket ( Ragusa et al . , 2010 ) . Our results underscore the importance of the composite Phactr1/PP1 surface in substrate recognition and specificity . Nevertheless , as with protein kinases , these interactions alone are unlikely to define substrate selection completely ( Miller and Turk , 2018 ) . Phactr-family members do not appear to contain any conserved domains that might represent autonomous substrate-binding domains , as seen in NIPP1 ( Boudrez et al . , 2000 ) and PPP1R15A ( Chen et al . , 2015; Choy et al . , 2015; Crespillo-Casado et al . , 2018 ) . On the other hand , many PIPs act to target PP1 complexes to specific subcellular compartments , proteins or macromolecules ( Cohen , 2002 ) . Phactr-family proteins exhibit differential subcellular localisations , including the nucleus and cell membrane ( Huet et al . , 2013; Wiezlak et al . , 2012 ) , and Phactr1 has been shown to interact directly with the KCNT1 potassium channel ( Ali et al . , 2020; Fleming et al . , 2016 ) . At least for Phactr3 and Phactr4 , membrane targeting involves conserved N-terminal sequences ( Huet et al . , 2013; Itoh et al . , 2014 ) which overlap a G-actin controlled nuclear import signal in Phactr1 . The genetic dominance of West syndrome Phactr1 mutations and Phactr4 R536P/humdy mutation also is consistent with Phactr-family proteins interacting with other cellular components in addition to PP1 ( Hamada et al . , 2018; Kim et al . , 2007; Zhang et al . , 2012 ) . These considerations , and the coupling of Phactr1/PP1 holoenzyme formation to rho-actin signalling , make it likely that the range of substrates controlled by Phactr1/PP1 in a particular setting will reflect the state of cellular actin dynamics and subcellular localisation of the complexes as well as the direct recognition of substrate primary sequence by the holoenzyme . pET28 PP1 ( 7–330 ) and pcDNA3 . 1 IRSp53 were gifts from Wolfgang Peti and Eunjoon Kim ( KAIST , S . Korea ) respectively . Other plasmids were: modified pTRIPZ ( Diring et al . , 2019 ) ; pEF Phactr1 and derivatives ( Wiezlak et al . , 2012 ) ; pGEX 6p2 ( GE Healthcare ) ; and pGRO7 ( Takara ) . For protein expression , Phactr1 ( 507-580 ) and Phactr1 ( 516-580 ) sequences were expressed using pGEX-6P2 . PP1-substrate chimeras were derivatives of pET28 PP1 ( 7–330 ) in which PP1 sequences 7–304 were joined by a ( Ser-Gly ) 9 linker to either IRSp53 ( 449-465 ) ( QQGKSSSTGNLLDKDDL ) IRSp53 ( 449-465 ) -S455E ( QQGKSSETGNLLDKDDL ) , or spectrin αII ( 1025–1039 ) ( DPAQSASRENLLEEQ ) . pET28 PP1-Phactr1 ( 526-580 ) , derived from pET28 PP1 ( 7–330 ) , encodes PP1 ( 7–304 ) -SGSGS-Phactr1 ( 526-580 ) . Plasmid construction and mutagenesis used standard methods , the NEB NEBuilder HiFi DNA Assembly Cloning Kit , or the NEB Q5 Site-Directed Mutagenesis Kit . Primers are listed in Supplementary file 4 . Protein expression in BL21 ( DE3 ) E . coli cells ( Invitrogen ) was with pGRO7 coexpression as described ( Choy et al . , 2014 ) . Overnight pre-cultures ( 400 ml ) were grown in LB medium supplemented 1 mM MnCl2 and used to inoculate a 100L fermenter . After growth to OD600 of ~0 . 5 , 2 g/L of arabinose was added to induce GroEL/GroES expression . At OD600 ~1 , the temperature was lowered to 10°C and protein expression induced with 0 . 1 mM IPTG for ~18 hr . Cells were harvested , re-suspended in fresh LB medium/1 mM MnCl2/200 μg/ml chloramphenicol and agitated for 2 hr at 10°C . Harvested cells were resuspended in lysis buffer ( 50 mM Tris-HCl , pH 8 . 5 , 5 mM imidazole , 700 mM NaCl , 1 mM MnCl2 , 0 . 1% v/v TX-100 , 0 . 5 mM TCEP , 0 . 5 mM AEBSF , 15 μg/ml benzamidine and complete EDTA-free protease inhibitor tablets ) , lysed by French press , clarified , and stored at −80°C . Phactr1/PP1 complexes were batch-adsorbed onto glutathione-sepharose affinity matrix , and recovered by cleavage with 3C protease at 4°C overnight in 50 mM Tris-HCl , pH 8 . 5 , 500 mM NaCl , 0 . 5 mM TCEP . Eluted complex was further purified via adsorption on Ni-NTA IMAC , and elution with 50 mM Tris pH 8 . 0 , 200 mM Imidazole , 700 mM NaCl and 1 mM MnCl2 at 4°C . Finally , proteins were purified using size exclusion chromatography on a Superdex 200 26/60 column equilibrated in complex buffer ( 20 mM Tris-HCl pH 8 . 5 , 0 . 25 M NaCl and 0 . 4 mM TCEP ) . Bio-layer interferometry ( BLI ) was as described ( Bertran et al . , 2019 ) , using the Octet Red 96 ( ForteBio ) . 50 μg/ml His-tagged PP1α was immobilised on Nickel-coated biosensor ( Ni-NTA , ForteBio ) , and the loaded biosensors then incubated with 0 . 1–10 μM Phactr1 peptides in Octet buffer ( 50 mM Tris pH 7 . 5 , 500 mM NaCl , 0 . 5 mM TCEP , 0 . 1% Tween 20 , 500 mg BSA/100 ml ) . Curve fitting , steady state analysis , and calculation of kinetic parameters were done using Octet software version 7 . 1 ( ForteBio ) . For peptides used , see Supplementary file 4 . Phactr1/PP1 in complex buffer was concentrated to 10 mg/ml and crystallised at 20°C using sitting-drop vapour diffusion . Sitting drops of 1 μl consisted of a 1:1 ( vol:vol ) mixture of protein and well solution . Well solutions were as follows . Phactr ( 507-580 ) /PP1α ( 7-300 ) : 7 . 5% PEG 3350 , 0 . 2 M MgSO4; Phactr1 ( 516-580 ) /PP1α ( 7-300 ) : 1M LiCl , 0 . 1 M tri-sodium citrate pH 5 . 25 , 10% PEG 6000; Phactr1/PP1-IRSp53 ( S455E ) : 20% PEG 3350 , 0 . 2 M NaBr; PP1-Phactr1 ( 526-580 ) , 20% PEG 3350 , 0 . 2M Potassium Citrate . Crystals appeared within 24–48 hr and reached their maximum size after 4 to 7 days , apart from Phactr ( 507-580 ) /PP1α ( 7-300 ) , for which the best crystals appeared after 3–7 weeks , reaching their maximum size after 8 weeks . For Phactr1/PP1-IRSp53 and Phactr1/PP1-spectrin α II complexes , crystallisation was achieved by microseed matrix screening ( D'Arcy et al . , 2014 ) using Phactr1/PP1-IRSp53 ( S455E ) crystals . Phactr1/PP1-IRSp53 crystallised in 20% PEG 3350 , 0 . 2 M KSCN , 0 . 1 M BIS-Tris propane pH 8 . 5; Phactr1/PP1-spectrin αII crystallised in 20% PEG 3350 , 0 . 2 M NaI , 0 . 1 M BIS-Tris propane pH 8 . 5 . Crystals appeared within a day and reached maximum size within 4 to 5 days . All crystals were cryoprotected in well solution supplemented with 15% glycerol + 15% ethylene glycol and flash-frozen in liquid nitrogen . X-ray data were collected at 100 K at beamlines I04-1 ( mx9826-17 ) , I02 ( mx9826-26 ) , I03 ( mx18566-37 ) , I24 ( mx18566-38 ) , I04 ( mx18566-29 ) , and I04-1 ( mx18566-55 ) of the Diamond Light Source Synchrotron ( Oxford , UK ) . Data collection and refinement statistics are summarised in Table 1 . Data sets were indexed , scaled , and merged with xia2 ( Winter et al . , 2013 ) . Molecular replacement used the atomic coordinates of human PP1 from PDB 4M0V ( Choy et al . , 2014 ) in PHASER ( McCoy et al . , 2007 ) . Refinement used Phenix ( Adams et al . , 2010 ) . Model building used COOT ( Emsley et al . , 2010 ) with validation by PROCHECK ( Vaguine et al . , 1999 ) . Residues modelled in the different structures are summarised in Supplementary file 5 . Figures were prepared using PYMOL ( Schrodinger LLC , 2020 ) . Atomic coordinates and crystallographic structure factors have been deposited in the Protein Data Bank under accession codes PDB 6ZEE , Phactr ( 507-580 ) /PP1α ( 7-300 ) ; PDB 6ZEF , Phactr1 ( 516-580 ) /PP1α ( 7-300 ) ; PDB 6ZEG , Phactr1/PP1-IRSp53; PDB 6ZEH , Phactr1/PP1-spectrin; PDB 6ZEI , Phactr1/PP1-IRSp53 ( S455E ) ; PDB 6ZEJ , PP1 ( 7–304 ) -SGSGS-Phactr1 ( 526-580 ) . Phosphatase assays ( 50 µl ) were performed in 96-well plates . Peptides ( 40 µl ) were serially diluted 1 . 5-fold from 1 mM in complex buffer ( 20 mM Tris-HCl pH 8 . 5 , 250 mM NaCl , 0 . 4 mM TCEP ) . Phactr1/PP1 ( 0 . 2–3U , 10 µl ) was added and after 15 min at room temperature , reactions were quenched by addition of 100 µl of Biomol Green reagent ( Enzo Life Sciences ) for 30 min , Absorbances at 620 nm were measured using the SpectraMax Plus 384 microplate reader and converted into phosphate concentrations using standard curves . Rate constants were estimated in GraphPad Prizm 8 by fitting product concentration readouts to modified Michaelis-Menten equation:PtE=kcatKMC ( CKM+1 ) ( P , product released at time t; E , Phactr1/PP1 concentration; kcat/KM , catalytic efficiency; C , initial substrate concentration; KM , Michaelis constant ) . The activity of the Phactr1/PP1 preparation was established on the day of each experiment , using 125 µM IRSp53 ( 447-465 ) pS455 peptide as standard , with Phactr1/PP1 at various concentrations . One unit of phosphatase activity was defined as the concentration of Phactr1/PP1 complex that generates 15 µM of phosphate in 15 min . To normalise the activity of different phosphatases , a para-nitrophenylphosphate ( pNPP ) -based assay was used . 10 µl phosphatase ( 200 nM ) was added to 40 µl pNPP ( 2-fold serial dilution from 100 mM ) , incubated at room temperature for 15 min , then quenched with 25 µl 3 M NaOH . Para-nitrophenol product was measured by absorbance at 405 nm . Full assay data can be found in Supplementary file 2 . Peptides were synthesised by the Francis Crick Institute Peptide Chemistry Science Technology Platform using standard techniques . Peptide arrays ( 5–10 nmol/spot ) were synthesised on a derivatised cellulose membrane Amino-PEG500-UC540 using Intavis ResPep SLi automated synthesiser ( Intavis Bioanalytical Instruments ) . Dry membranes were blocked for 1 hr in 5% milk in Tris-buffered saline supplemented with 0 . 1% Tween-20 ( TBST ) with agitation , rinsed with TBST , and incubated overnight at 4°C with 10 µg/ml GST-Phactr1 ( 516-580 ) /PP1 ( 7–300 ) complex in TBST . After three 10-min washes with TBST , membranes were incubated with 1:5000 HRP-conjugated anti-GST in 5% milk/TBST at room temperature for 1 hr , washed three times with TBST , and binding revealed using SuperSignal West Pico Plus reagent ( ThermoFisher ) as chemiluminescent substrate . Images were taken using Amersham Imager 600 ( GE ) . Mouse NIH3T3 fibroblasts ( mycoplasma-free ) were maintained in DMEM ( Gibco ) with 10% fetal calf serum ( FCS ) and penicillin-streptomycin at 37°C and 10% CO2 . Cells were transfected using Lipofectamine 2000 ( Invitrogen ) ( 150 , 000 cells per well , six-well dish ) . For SILAC proteomics , NIH3T3 cell line pools were generated stably carrying doxycyline-inducible pTRIPZ-Phactr1 derivatives , using puromycin selection . Phactr1 expression was induced by doxycycline addition as indicated in the figure legends . Cells were maintained overnight in DMEM/0 . 3% FCS , and then stimulated with 15% FCS for 1 hr or cytochalasin D ( CD ) or latrunculin B ( LatB ) for 30 min . Primary rat hippocampal neurons ( DIV 14 ) were cultured as described ( Baltussen et al . , 2018 ) and transfected using Lipofectamine 2000 . Total RNA was purified using the GenElute mammalian total RNA kit ( Sigma ) and cDNA synthesised using the Transcriptor First Strand cDNA Synthesis kit ( Roche ) with random hexamer primers . Real‐time‐qPCR was performed using the 7500 Fast Real-Time PCR System ( Thermo Fisher Scientific ) with the SYBR green reaction mix ( Life Technologies ) . Primers are listed in Supplementary file 4 . The relative abundance of target cDNA was normalised against rps16 cDNA abundance in each sample . Immunofluorescence microscopy in fibroblasts was performed as described ( Vartiainen et al . , 2007; Wiezlak et al . , 2012 ) . F-actin was detected with FITC-phalloidin ( Invitrogen ) and nuclei were visualised using DAPI . For immunofluorescence microscopy primary rat neurons were transfected with expression plasmids and fixed 24 hr later in 4% paraformaldehyde/4% sucrose in PBS before staining with Flag and GFP antibodies . Images were taken with Leica SP5 confocal microscope with 63x ( NA 1 . 4 ) oil objective . GFP was used to monitor the shape of dendritic spines , and Flag for transfected cells . An image stack with 0 . 5 μm z-intervals was obtained to capture a part dendritic arbour at high magnification . In each experiment , at least 10 cells per mutant were imaged and the morphology of dendritic spines was quantified blindly . Statistical analysis was performed in GraphPad Prizm eight using Welch’s t-test function . SDS-PAGE analysis of cell lysates and immunoblotting was performed using standard techniques; the signal was visualised and quantified using Odyssey CLx instrument ( LI-COR ) and the Image Studio ( LI-COR ) Odyssey Analysis Software . Primary antibodies used were anti-Flag ( Sigma , F7425 ) , anti-IRSp53 ( Santa Cruz , sc-50011 and Abcam , ab15697 ) , anti-afadin ( Santa Cruz , sc-74433 ) , anti-GST ( VWR , RPN1236 ) . Secondary antibodies labelled with IRDye 800CW and IRDye 680LT were from LICOR . Rabbit anti-IRSp53 pS455 , anti-afadin pS1275 and anti-spectrin pS1031 antibodies were custom-made ( Covalab ) ; for antigen sequences used see Supplementary file 4 . Cells were maintained for at least six passages in ‘heavy’ ( R10K8 ) or ‘light’ ( R0K0 ) DMEM medium supplemented with 10% SILAC dialysed fetal calf serum ( 3 kDa cutoff ) . Phactr1 expression was induced by doxycycline addition to 1 μg/ml for 5 hr . Cells were harvested processed for mass spectrometry essentially as described ( Pattison et al . , 2016; Touati et al . , 2018 ) with minor modifications . 1 mg each of ‘light’ and ‘heavy’ labelled lysates were mixed and dried . Protease digestion and phosphopeptide enrichment was done as described with minor modifications ( Pattison et al . , 2016; Touati et al . , 2018 ) . Peptides were dissolved in 35 μl of 1% TFA , with sonication , and fractionated on a 50 cm , 75 μm I . D . Pepmap column with elution directly into the LTQ-Orbitrap Velos . Xcalibur software was used to setup data-dependent acquisition in top10 mode . Raw mass spectrometric data were processed in MaxQuant ( version 1 . 3 . 0 . 5 ) for peptide and protein identification; database search was with the Andromeda search engine against the Mus musculus canonical sequences from UniProtKB . Fixed modifications were set as Carbamidomethyl ( C ) and variable modifications set as Oxidation ( M ) , Acetyl ( Protein N-term ) and Phospho ( STY ) . The estimated false discovery rate was set to 1% at the peptide , protein , and site levels . A maximum of two missed cleavages were allowed . Phosphorylation site tables were imported into Perseus ( v1 . 4 . 0 . 2 ) for analysis . Contaminants and reverse peptides were cleaned up from the Phosphosites ( STY ) . Dephosphorylation score was defined as log2 ( dephosphorylation score ) =0 . 25*[log2 ( 2H/1L ) + log2 ( 3H/1L ) + log2 ( 2L/1H ) + log2 ( 3L/1H ) ] where 1 , 2 and 3 denote Phactr1XXX , Phactr1XXXΔC , and empty vector cells , and H and L denote samples from cells grown in R10K8 and R0K0 media , respectively . Sequence logos were generated using WebLogo ( https://weblogo . berkeley . edu/logo . cgi/ ) . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium ( http://proteomecentral . proteomexchange . org ) via the PRIDE partner repository with the dataset identifier PXD019977 . The Phactr1-null ( tm1d ) allele was derived from the CSD79794 Phactr1 Tm1a allele ( KOMP; https://www . komp . org/geneinfo . php ? geneid=74176 ) by sequential action of Flp and Cre . Heterozygous animals were crossed . Hippocampal and cortical tissue was extracted from E16 . 5 embryos , and cultured in 12-well plate dishes ( 500 , 000 cells per well ) as described ( Baltussen et al . , 2018 ) before genotyping . Two biological replicates were processed for wildtype or Phactr1-null neurons . On DIV10 , neurons were treated for 30 min with CD ( 10 μM ) , LB ( 1 μM ) or vehicle ( DMSO ) . Preparation of lysates , protease digestion and phosphopeptide enrichment was done as described with minor modifications ( Eder et al . , 2020 ) . Phosphopeptides were eluted directly into the Orbitrap Fusion Lumos , operated with Xcalibur software , with measurement in MS2 and MS3 modes . The instrument was set up in data-dependent acquisition mode , with top 10 most abundant peptides selected for MS/MS by HCD fragmentation . Raw mass spectrometric data were processed in MaxQuant ( version 1 . 6 . 2 . 10 ) ; database search against the Mus musculus canonical sequences from UniProtKB was performed using the Andromeda search engine . Fixed modifications were set as Carbamidomethyl ( C ) and variable modifications set as Oxidation ( M ) , Acetyl ( Protein N-term ) and Phospho ( STY ) . The estimated false discovery rate was set to 1% at the peptide , protein , and site levels , with a maximum of two missed cleavages allowed . Reporter ion MS2 or Reporter ion MS3 was appropriately selected for each raw file . Phosphorylation site tables were imported into Perseus ( v1 . 6 . 1 . 2 ) for analysis . Contaminants and reverse peptides were cleaned up from the Phosphosites ( STY ) and the values normalised using Z-score function across columns . Cortical and hippocampal as well as MS2/MS3 data across two biological replicates were pooled . ( DMSO-CD ) differences were calculated and compared between Phactr1 WT and KO neurons using the two-sample t-test . Phosphorylation sites exhibiting significantly different dephosphorylation in WT neurons compared with KO neurons were considered to be Phactr1-dependent . Sequence logos were generated using WebLogo ( https://weblogo . berkeley . edu/logo . cgi/ ) . The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium ( http://proteomecentral . proteomexchange . org ) via the PRIDE partner repository with the dataset identifier PXD019882 .	Specific arrangements of atoms such as bulky phosphate groups can change the activity of a protein and how it interacts with other molecules . Enzymes called kinases are responsible for adding these groups onto a protein , while phosphatases remove them . Kinases are generally specific for a small number of proteins , adding phosphate groups only at sites embedded in a particular sequence in the target protein . Phosphatases , however , are generalists: only a few different types exist , which exhibit little target sequence specificity . Partner proteins can attach to phosphatases to bring the enzymes to specific locations in the cell , or to deliver target proteins to them; yet , it is unclear whether partner binding could also change the structure of the enzyme so the phosphatase can recognise only a restricted set of targets . To investigate this , Fedoryshchak , Přechová et al . studied a phosphatase called PP1 and its partner , Phactr1 . First , the structure of the Phactr1/PP1 complex was examined using biochemistry approaches and X-ray crystallography . This showed that binding of Phactr1 to PP1 creates a new surface pocket , which comprised elements of both proteins . In particular , this composite pocket is located next to the part of the PP1 enzyme responsible for phosphate removal . Next , mass spectrometry and genetics methods were harnessed to identify and characterise the targets of the Phactr1/PP1 complex . Structural analysis of the proteins most susceptible to Phactr1/PP1 activity showed that they had particular sequences that could interact with Phactr1/PP1’s composite pocket . Further experiments revealed that , compared to PP1 acting alone , the pocket increased the binding efficiency and reactivity of the complex 100-fold . This work demonstrates that a partner protein can make phosphatases more sequence-specific , suggesting that future studies could adopt a similar approach to examine how other enzymes in this family perform their role . In addition , the results suggest that it will be possible to design Phactr1/PP1-specific drugs that act on the composite pocket . This would represent an important proof of principle , since current phosphatase-specific drugs do not target particular phosphatase complexes .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2020	Molecular basis for substrate specificity of the Phactr1/PP1 phosphatase holoenzyme
Defects in flagella growth are related to a number of human diseases . Central to flagellar growth is the organization of microtubules that polymerize from basal bodies to form the axoneme , which consists of hundreds of proteins . Flagella exist in all eukaryotic phyla , but neither the mechanism by which flagella grow nor the conservation of this process in evolution are known . Here , we study how protein complexes assemble onto the growing axoneme tip using ( cryo ) electron tomography . In Chlamydomonas reinhardtii microtubules and associated proteins are added simultaneously . However , in Trypanosoma brucei , disorganized arrays of microtubules are arranged into the axoneme structure by the later addition of preformed protein complexes . Post assembly , the T . brucei transition zone alters structure and its association with the central pair loosens . We conclude that there are multiple ways to form a flagellum and that species-specific structural knowledge is critical before evaluating flagellar defects . Most vertebrate cells have a cilium or flagellum ( the terms are used here interchangeable ) . It is a thin , microtubule-containing , membrane-covered extension on the surface of the cell , which may generate cell motility and/or act as a sensory and signaling organelle . Defects in cilia and flagella can cause severe human diseases for example skeletal deformations , polycystic kidney disease , infertility , situs invertus , blindness , obesity , and even cancer ( Escudier et al . , 2009; Chandok , 2012; Huber and Cormier-Daire , 2012; Li et al . , 2012 ) . A collective name for these conditions is the ‘ciliopathies’ ( Fliegauf et al . , 2007 ) . Some ciliopathies , such as primary ciliary dyskinesia , are diagnosed by ultrastructural investigations of cilia that should be motile ( Chandok , 2012; Chilvers et al . , 2003; Escudier et al . , 2009; Huber and Cormier-Daire , 2012; Li et al . , 2012; O’Toole et al . , 2012 ) . However , the ultrastructural pathology of many ciliopathies remains unknown . Each flagellum consists of ∼1000 different proteins ( Pazour , 2005; Gherman et al . , 2006; Fliegauf et al . , 2007; Ishikawa et al . , 2012 ) , many of which contribute to its microtubule-based core , called the axoneme . Axonemes originate inside the cell at basal bodies ( Figure 1A , B ) . From the basal body , nine doublet microtubules ( dMTs ) extend into the next section along the flagellum called the transition zone . The transition zone ends at the basal plate , an electron-dense structure found in the region , where the two central pair microtubules ( CPs ) are nucleated and the canonical 9+2 axoneme arrangement starts . The dMTs consist of a complete A-tubule containing 13 protofilaments and an incomplete B-tubule containing 10 protofilaments ( Warner and Satir , 1973; Amos and Klug , 1974; Sui and Downing , 2006; Nicastro et al . , 2011 ) . Dynein arms are bound to the A-tubule of the dMTs and walk on the neighboring B-tubule . This causes dMTs to slide along each other , introducing sheer , which is converted into flagellar bending by several classes of static links . Flagellar bending is controlled so as to induce motility in some cells ( e . g . , spermatozoa , Giardia spp . and Trypanosoma spp . ) and propel the surrounding media in other cells ( e . g . , respiratory tract epithelia , fallopian tube epithelium ) . In addition to the MT components of the axoneme , partially assembled protein modules such as radial spokes ( Qin , 2004; Diener et al . , 2011 ) , nexin links and central pair projections are important for axonemal function through their roles as cross-linkers and regulators of dMT sliding and bending . 10 . 7554/eLife . 01479 . 003Figure 1 . Microtubule organization at the distal tip of the flagellum varies in different cell cycle stages and species . ( A ) A 20 nm tomographic slice showing a new T . brucei flagellum attached to the old flagellum . ( B ) A schematic of the picture in A describing the relevant regions of the flagellum . ( C–G ) Tomographic slices ( 20 nm thick ) showing distal tips of flagella in different stages of the cell cycle . ( C ) T . brucei old flagellum , ( D ) T . brucei short flagellum , ( E ) T . brucei long flagellum , ( F ) C . reinhardtii short flagellum , and ( G ) C . reinhardtii long flagellum . ( C′–G′ ) 3D models of flagella in the same cell cycle stage as in C–G , showing the axoneme with A-tubules in pink , B-tubules in dark blue and central pair in green . Flagellar membrane is shown in transparent pink . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 00310 . 7554/eLife . 01479 . 004Figure 1—figure supplement 1 . Gallery of long growing T . brucei tips , all showing disordered axonemes ( 20 nm thick tomography slices ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 004 In most multicellular organisms , the cilium is produced after the cell has exited the cell cycle , but in many protozoan flagellates , new flagella must be built to maintain motility in daughter cells ( Ginger et al . , 2008; Dawson and House , 2010 ) . Flagellar elongation occurs by addition of protein subunits at the axoneme’s distal end ( Rosenbaum and Child , 1967; Marshall , 2001 ) . Large protein complexes containing the precursor axoneme building blocks are delivered to this site via an evolutionary conserved process called intraflagellar transport ( IFT; [Kozminski et al . , 1993] ) . The roles of IFT in ciliary function are well studied ( Pedersen and Rosenbaum , 2008 ) , and the molecular mechanisms that mediate IFT of axonemal proteins are beginning to be characterized ( Bhogaraju et al . , 2013 ) . The structure of the flagellar tip has been characterized; the B-tubule ends before the A-tubule creating a distal ‘singlet region’ in the flagellum tip of most species ( Ringo , 1967; Satir , 1968; Sale and Satir , 1976; Woolley and Nickels , 1985 ) ; the CPs extend further into the distal tip than the dMTs ( Ringo , 1967 ) ; the dMTs and CPs are linked to the membrane through capping structures ( Dentler , 1980; Woolley et al . , 2006 ) . Yet , we know very little about how the flagellar components , once delivered to the distal tip , are assembled to form the beating flagellum ( Ishikawa and Marshall , 2011; Fisch and Dupuis-Williams , 2012 ) . For example , does the CP extend beyond the dMTs during tip growth , like in the mature flagellum , or is the growth of all MTs synchronized ? Alternatively do the dMTs extend beyond the CP during flagellar extension ? When do other structural modules such as radial spokes , dynein arms , and central pair projections get incorporated ? Clearly , there are multiple possibilities for how a flagellum might extend . We have examined two evolutionary distant organisms , the green algae Chlamydomonas reinhardtii and the parasitic protozoa Trypanosoma brucei , to determine if a consistent pattern of flagellar extension exists . By studying the tips of their growing flagella and their basal plate region , we reveal two separate assembly pathways of flagella extension and maturation . To elucidate the pathways for axoneme elongation , we used electron tomography to examine the tips of actively growing flagella in two organisms , just when the flagella have started growing ( at ∼0 . 7–1 . 5 μm ) , and after a period of flagellar growth ( at 4–10 μm; Table 1 ) . C . reinhardtii has two flagella that are reabsorbed down to their transition zones , which are then expelled prior to mitosis ( Rasi et al . , 2009; Parker et al . , 2010 ) . After mitosis , the small daughter cells remain within the wall of the mother cell where they regrow their flagella; a cell stage we easily identified in the electron microscope . In T . brucei the new flagellum starts growing midway through the cell cycle ( Sherwin and Gull , 1989 ) . Its tip is attached to the side of the old flagellum by a structure called the flagella connector ( Moreira-Leite , 2001; Briggs , 2004; Davidge , 2006 ) . The relative positions of a cell’s two flagella allow an unambiguous identification of which flagellum is old and which is new ( Figure 1A , B ) . Throughout the rest of the cell cycle , the length of the new flagellum correlates with cell cycle progression . 10 . 7554/eLife . 01479 . 005Table 1 . Sample size of each species/cell cycle stageDOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 005Flagellum typeN ( flagella tips reconstructed ) T . bruceiOld6Short5Long HPF10Long chemical fixed2C . reinhardtiiShort4Long4 3D electron tomographic reconstructions of flagellar tips were made of old T . brucei flagella ( slowly growing or of constant length; Video 1 ) ( Davidge , 2006; Farr and Gull , 2009 ) , and growing new flagella that were short or long ( Videos 2 and 3 ) . Comparable images of new growing flagella that were short or long were also obtained from C . reinhardtii ( Videos 4 and 5 ) . The axoneme in all flagellar tips studied , except the T . brucei growing long flagella , displayed the regular spacing of dMTs found in the rest of the flagellum ( Figure 1C , D , F , G , Figure 1—figure supplement 1 ) . In contrast , the growing long T . brucei flagellar tips showed a disorganized array of dMTs ( Figure 1E ) ; some dMTs lay very close to the CPs . The axoneme structure is revealed in 3D models of each of the tips ( Figure 1C′–G′ , Videos 6–10 ) . The disorganized microtubules at the tips of growing flagella in T . brucei indicate that the mechanism of axonemal growth in long flagella is different in this species from that seen in C . reinhardtii , so we investigated the structure of growing flagella further . 10 . 7554/eLife . 01479 . 006Video 1 . Old T . brucei flagellum tip ( related to Figure 1 ) . 1-nm thick sections of a tomogram reconstruction containing the distal tip . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 00610 . 7554/eLife . 01479 . 007Video 2 . Growing short T . brucei flagellum tip ( related to Figure 1 ) . 1-nm thick sections of a tomogram reconstruction containing the distal tip . Please also note the just formed basal plate . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 00710 . 7554/eLife . 01479 . 008Video 3 . Disordered long growing T . brucei flagellum tip ( related to Figure 1 ) . 1-nm thick sections of a tomogram reconstruction containing the distal tip . Note the bent doublet microtubules and the absence of electron-dense structures associated with the central pair . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 00810 . 7554/eLife . 01479 . 009Video 4 . Short growing Chlamydomonas reinhardtii flagellum tip ( related to Figure 1 ) . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 00910 . 7554/eLife . 01479 . 010Video 5 . Long growing Chlamydomonas reinhardtii flagellum tip ( related to Figure 1 ) . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 01010 . 7554/eLife . 01479 . 011Video 6 . 3D model of the microtubules found in an old T . brucei flagellum tip ( related to Figure 1 ) . Model was made by drawing lines in the microtubules and around membranes as seen in a tomogram reconstruction of the old T . brucei flagellum tip ( Video 1 ) . The lines were then provided with a skin through a meshing process . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 01110 . 7554/eLife . 01479 . 012Video 7 . 3D model of the microtubules found in a short growing T . brucei flagellum tip ( related to Figure 1 ) . Model is a segmentation from the tomogram , Video 2 . The tomogram reconstruction contained the most of the flagellum tip , but cropped off some the axoneme shortly proximal to it and one dMT was not found within the reconstruction ( missing in the model ) . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 01210 . 7554/eLife . 01479 . 013Video 8 . 3D model of the disordered microtubules found in a growing T . brucei flagellum tip ( related to Figure 1 ) . Note that some doublet microtubules are in contact with the flagella membrane . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 01310 . 7554/eLife . 01479 . 014Video 9 . 3D model of a short growing C . reinhardtii flagellum ( related to Figure 1 ) . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 01410 . 7554/eLife . 01479 . 015Video 10 . 3D model of a long growing C . reinhardtii flagellum ( related to Figure 1 ) . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 015 The structural disorder in growing long new flagella of T . brucei is most obvious in 3D models generated by tracing the doublet MTs through tomographic reconstructions ( Figure 2A , B ) . ∼300 nm from the tip of a T . brucei flagellum , the nine doublet MTs make an almost perfect circle around the central pair MTs ( Figure 2B3 ) . However , ∼150 nm from the tip , the MT doublets change orientation compared to the central MT pair ( Figure 2B2 ) . From this point toward the axoneme tip , the MT doublets lose their circular arrangement ( Figure 2A–B1 ) . In this disordered region , some MTs lie closer to the CPs , others are further away from them ( Figure 2C , Figure 1—figure supplement 1 ) , indicating that the tip is not just a tapered end . 10 . 7554/eLife . 01479 . 016Figure 2 . Structural disorganization in the growing tips of long T . brucei flagella . ( A ) 3D model of a tomographic reconstruction of a growing , long T . brucei axoneme . The doublets are color coded in a gradient from yellow ( dMT1 ) to red ( dMT9 ) . ( B ) A cut through of the 3D model shows that ( 1 ) at the end of the flagellum , the circular arrangement of the MTs is completely lost , ( 2 ) <0 . 5 μm before the flagella tip , the MTs start losing their circular organization , and ( 3 ) >0 . 5 μm from its tip the axoneme is well organized . ( C ) Individual traces of the nine dMTs in A reveals them bending both toward and away from the CP . dMTs not touching the membrane also showed this random bending ( e . g . , dMTs 2 and 5 ) ( D ) CP projections are arranged as an electron-dense ladder ( arrows ) extending from the central pair . ( E ) In the tip of the T . brucei old flagellum the associated structural proteins are visible all the way to the tip . In the longitudinal view , we see the distal end that then curves and the final 200 nm is shown in cross-sectional view ( insert ) . ( F ) Central pair projections are clearly seen all the way to the end of the central pair in the short growing T . brucei flagellum . ( G ) In the T . brucei long growing flagellum such associated proteins are not visible ( red arrows ) . ( H ) The structural proteins are clearly visible along the length of a mature C . reinhardtii flagellum . ( I and J ) In the growing short C . reinhardtii flagellum , these proteins were also present at the axoneme’s tip , indicating that microtubules and associated structures are simultaneously assembled . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 016 The disorder seen in the growing T . brucei axoneme could be explained by a transient lack of MT-associated structures that are present in the mature axoneme , for example the radial spoke complex , CP projections and/or nexin links . Central pair projections display a ladder-like arrangement along the length of the mature T . brucei flagellum ( Figure 2D ) , and in all the distal flagella tips ( Figure 2E , F , H–J ) except for in the growing T . brucei long flagellum tip ( Figure 2G ) , where the associated complexes were not visible . Therefore , although we have a relatively small sample size ( Table 2 ) , we suggest that during the elongation of a T . brucei flagellum , the microtubules extend beyond the position of the associated proteins , and hence the final structures are added to the already formed but disorganized axonemal tip . 10 . 7554/eLife . 01479 . 017Table 2 . Distances between the axonemal microtubules and the flagellar membraneDOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 017Flagellum typeN ( A/B-tubule and CPs ) Average MT-membrane distance ( nm ) Range: MT-membrane distance ( nm ) T . bruceiOld4826 ± 99–54Short5494 ± 2735–160Long4750 ± 370–157C . reinhardtiiShort6077 ± 3233–214Long6097 ± 4742–230 We investigated whether the plus ends of axonemal MTs pushed against the flagellar membrane as they grew . The 3D nature of our data allowed the quantification of microtubule end distances to the nearest flagellar tip membrane . The reconstructions of old T . brucei flagella showed dMTs as well as the CP plus ends neatly arranged ∼30 nm from the tip membrane ( Figure 3A ) . Only in the disorganized growing long T . brucei flagella did dMTs sometimes touch the flagellar membrane ( distance 0–125 nm; average 44 ± 32 nm; n=36 dMTs; Table 2 ) . In both short and long growing C . reinhardtii flagellum tips , all axonemal MT ends were found within 230 nm of the nearest flagellar tip membrane ( Table 2 ) . In no case did the CPs touch the flagellar tip membrane . Thus , in all cases except growing long T . brucei flagella , flagellar extension does not require close contact between dMTs and flagellar membrane . 10 . 7554/eLife . 01479 . 018Figure 3 . No singlet zone in any of the samples examined . ( A ) The distance between the dMTs/CP MT plus ends and the closest flagellar tip membrane was measured in flagella of each cell cycle stage/species . These properties are represented here as box plots in which the mean value is identified by a dotted horizontal line , the median value by a line , the 25–75% percentile by the height of the box , and the horizontal bars outside the box include the 5th and the 95th percentile . Only in T . brucei long new flagellum ( LNF ) were the microtubules touching the membrane , but in all samples all MT ends were closer than 230 nm to the membrane . ( number above box = n ) ( B ) To determine if dMTs or CP microtubules extended the furthest within the axoneme , all microtubule plus end’s distance to the longest microtubule in the axoneme was measured . ( * ) In T . brucei LNF , the CPs was lagging behind the dMT extension , and in C . reinhardtii CPs extended further than the dMTs . ( C ) The extension of the A- or B-tubule within the doublet was measured . The A-tubule did not always extend the furthest . ( D ) The difference of extension between the A- and B-tubule in a dMT is merely 10–20 nm in both species . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 01810 . 7554/eLife . 01479 . 019Figure 3—figure supplement 1 . Chemical fixation increased the distance between the flagellar tip membrane and the microtubule plus ends . ( A ) A tomographic slice and the belonging 3D model of a high pressure frozen and freeze substituted long new flagellum tip in T . brucei . ( B ) A tomographic slice and the corresponding 3D model of a chemically fixed long new flagellum tip in T . brucei . Note that the membrane is in a similar place in comparison to the flagellar connector as seen in A ( white arrows ) . Black arrows show the distal microtubule ends . The stable position of the flagellum tip membrane in relation to the FC in both sample preparations indicates that the distance increase between the axoneme and the membrane is due to microtubule depolymerization . ( C ) Quantification of distances between CPs and dMTs to the tip membrane . Numbers over the boxes are the amount of MTs measured . ( D ) The distribution of microtubule plus-ends within the axoneme show a much wider range in chemically fixed samples . This explains why the T . brucei new flagellum tip has been shown to lack some dMTs in previous studies e . g . , Briggs ( 2004 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 019 We also investigated whether there was a difference between the elongation of CPs and dMTs . To show microtubule extension within the axoneme , we measured the distance from the ends of the A- , B-tubule and CP to the end of the furthest reaching microtubule of the axoneme . The ends of dMTs and CPs of flagellar ends did not significantly differ in their extension in most samples ( Figure 3B; Table 3 ) . However , we found a significant difference between the extensions of dMTs and CPs in the growing long T . brucei flagella , where the CP lagged behind the dMTs extension by approximately 50 nm ( p<0 . 01 ) . Also in C . reinhardtii , a significant difference was revealed , but here the CP extended ∼50 nm beyond most dMTs ( p<0 . 01 ) . 10 . 7554/eLife . 01479 . 020Table 3 . Statistics on the microtubule extension within the axonemeDOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 020Flagellum typedMT/Cp extension differencePaired t testn ( axonemes ) A-B tubule extension DifferencePaired t testN ( MTs ) T . bruceiOldNo0 . 242No0 . 6323ShortNo0 . 813No0 . 7324LongYes0 . 016No0 . 1236C . reinhardtiiShortNo0 . 283No0 . 1428LongYes0 . 014No0 . 5136 The small distance from the tip in which all the MT tips were found in both T . brucei and C . reinhardtii contradicts the existence of a long singlet region , previously published to be ∼1 μm in C . reinhardtii ( Ringo , 1967; Satir , 1968; Sale and Satir , 1976; Woolley and Nickels , 1985 ) . Furthermore , the extension of the A- and B-tubule within the dMTs revealed that the A-tubule commonly extended the furthest ( Figure 3C ) . However , the extension of the longest sub-fiber did not protrude more than 10–20 nm beyond its partner tubule . No samples showed a significant difference between A- and B-tubule extension ( Figure 3D; Table 3 ) . We also compared the distances from the MT ends to the tip membrane in high pressure frozen vs chemically fixed growing T . brucei flagella ( Figure 3—figure supplement 1 ) . In chemically fixed cells , the dMT and CP ends were found in a wide range of distances , up to 350 nm from the membrane ( average 203 ± 75 nm; n = 40 ) , in contrast to the high pressure frozen T . brucei flagella where the same distance was a maximum of 125 nm ( 45 ± 37 nm; n = 75 ) . The microtubule plus ends within the same axoneme were spread over a larger distance in chemically fixed samples than in high pressure frozen ones . Indeed , a partial axoneme was previously used to localize an image as having been acquired close to the distal flagellum end ( Briggs , 2004 ) , when such an area of a partial axoneme was rarely to be found in this study . These findings , in combination with the absence of a singlet region in high-pressure frozen samples , indicate that the distal flagellar arrangements previously published in C . reinhardtii tips were likely disturbed by the chemical fixation used . We conclude that the singlet region is not found in either species examined , but that the two species examined display two different assembly pathways . Axonemal microtubules have previously been shown to be linked to the flagellar membrane ( Dentler , 1980; Woolley et al . , 2006 ) . We examined the presence of such linkages in all 3D reconstructions available ( Figure 4A–H ) . The distal ends of the flagella were often very electron dense , particularly in the growing long flagella ( Figure 4C , G ) , making visualization of the MTs and their ends difficult . Most MT with a clear end morphology appeared flared , but in some cases caps were visible on CPs ( e . g . , C . reinhardtii short new flagellum; Figure 4E , middle ) . We saw some filamentous material between the CPs and dMTs extending to the flagellar membrane , and also from the CP to the dMTs ( e . g . , Figure 2J ) but it is not a clear structural link . We saw no evidence of a bead and plate structure , as previously described in the C . reinhardtii CP cap found in demembranated mature flagella ( Dentler and Rosenbaum , 1977 ) . In cryo-electron tomography of an intact T . brucei old flagellum tip , we see clear electron densities in the distal end of dMT A-tubules as well as in CPs , but no bead or plate structure ( Figure 3I , J ) . Thus , the distal ends of axonemal microtubules are likely linked to the membrane by fibrous structures in the mature flagellum , but our results are inconclusive about the growing axoneme . 10 . 7554/eLife . 01479 . 021Figure 4 . All axoneme microtubule plus ends are found close to the tip membrane . 4-nm thick slices of tomograms from ( A and B ) T . brucei short new flagellum , ( C and D ) T . brucei long new flagellum , ( E and F ) C . reinhardtii short new flagellum , and ( G and H ) C . reinhardtii long new flagellum . In the left column CP distal ends are shown and the right column , dMT plus ends . dMTs are always displayed with the A-tubule to the left . Examples of flared ends are marked with turquoise arrowhead and capped ends with white arrowheads . Electron-dense structures associated with the C . reinhardtii and T . brucei CPs in long new flagella are marked by green arrows . ( I and J ) 15-nm thick cryo-electron tomography sections of ( I ) CP and ( J ) dMTs at the old T . brucei flagellum tip . Note the electron density extending into the lumen of both CP MTs and into the A-tubule lumen in the dMTs . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 021 We investigated whether the proximal end of the flagellum becomes altered as the flagellum grows and matures . Electron tomographic reconstructions of five cells revealed the ultrastructure of the basal plate and the anchoring of the central pair microtubules throughout the cell cycle in ten flagella ( Figure 5A; five new and old flagellum pairs ) . 10 . 7554/eLife . 01479 . 022Figure 5 . The T . brucei basal plate matures and alters its association with the CP minus ends during the cell cycle . ( A ) A 20 nm thick tomographic slice showing the new flagellum ( NF ) basal plate region . ( B ) The exact thickness of the basal plates and the locations and structure of the CP minus ends within/around them . The new flagellum in Cell1 is so short that no CP microtubules have started to grow yet . ( C ) The length of the new flagellum ( marked with * ) reveals the cell’s position in the cell cycle; The cell earliest in the cell cycle displaying a short new flagellum ( top of Figure ) and the most mature cell with two flagella in individual flagellar pockets at the bottom . The cartoons to the left show outlines of the flagellar pocket region in the cells studied , the cartoons are oriented so that the anterior end of the cell points to the right . ( D ) 10-nm thick tomographic slices of the new flagella showing the bilayered electron-dense material forming the basal plates , with the capped minus end ( e . g . , turquoise arrow ) of one of the CP MTs nucleated on the proximal surface . Cell 1 has such a short flagellum that no CPs has grown yet , but an early basal plate is evident . ( E ) The CP MTs were modeled and the basal plates visualized using electron density thresholding . The bi-layered structure of the basal plate is clearly visible until cell 4 , where the two flagella have separated into two separate flagellar pockets in preparation for cell division . ( F ) A selection of the models displayed in D and F show how the basal plate is formed of two stacked rings early in the cell cycle , which is then lost as the flagellum matures . ( G ) The tomographic slices of basal plates in the old flagella ( OF ) of the same cell as the new flagellum shown to the left . The bilayered structure is mostly lost , the basal plate is longer and the CP microtubule now has an open end ( e . g . , red arrow ) i . e . inserted further into the basal plate . ( H ) The 3D models of the basal plate region in the old flagella reveal amorphous structures that sometimes only anchor one CP MT ( cell 3 and 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 02210 . 7554/eLife . 01479 . 023Figure 5—figure supplement 1 . The ring structure of the new flagellum basal plate in T . brucei . ( A ) A cartoon of the cell that is imaged in B . It has two flagella in separated pockets , meaning that the new flagellum ( marked with an * ) already is long . ( B ) Cross-sectional views ( slices from tomograms ) of the basal plate region in both the old and the new flagellum . The number in the bottom right corner is the distance from the first image . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 023 Several differences were found over the cell cycle in T . brucei: first , the minus ends of the central pair microtubules were capped in all four growing new flagella long enough to have a CP ( Figure 5B and arrow in D; Video 11 ) , whereas in old flagella of the same cells , CPs showed some open ends ( 6 out of 9 CP ends of known morphology; Figure 5B and arrow in G; Video 12 ) . Second , in new flagella , the ∼30-nm thick basal plate consisted of two stacked electron-dense rings , and the minus ends were found within the more distal ring ( Figure 5C , D , Figure 5—figure supplement 1 ) . In old flagella , this defined basal plate structure consisting of two rings was gradually lost and replaced with a more diffuse electron dense mass that became up to 90 nm thick with the progression of the flagellum maturation ( Figure 5B , G , H ) . Third , the central pair minus ends were commonly found within this electron dense mass in old flagella , but in 2 out of the 10 CPs , one of the ends was found around 100 nm distal to the basal plate ( Figure 5B , G; cells 3 and 5 ) , showing a loosened CP anchoring in more mature flagella . 10 . 7554/eLife . 01479 . 024Video 11 . The basal plate in a new T . brucei flagellum ( related to Figure 5 ) . The basal plate has been modeled by density thresholding and is displayed in brown . The CP is modeled as all MTs but also show their capped minus ends . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 02410 . 7554/eLife . 01479 . 025Video 12 . The basal plate of an old T . brucei flagellum ( related to Figure 5 ) . The basal plate has been modeled by density thresholding and is displayed in brown . The CP is modeled and show one capped and one open minus end . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 025 Thus , the T . brucei basal plate structure and its association with the central pair minus ends changes as the flagellum matures . We wondered if this is a conserved feature of the CP nucleating region of flagella ( Euteneuer and McIntosh , 1981; Song and Mandelkow , 1995 ) . We therefore reconstructed nine transition zone regions of C . reinhardtii cells found within different mother cell walls ( Figure 6A ) . These cells had flagella of various lengths , all growing except for two flagella that were found in a mature cell ( shown as ‘long’ ) . We first plotted the thickness of the central cylinder of the transition zone ( previously described as an electron dense H and the core of the 9-pointed star [Ringo , 1967] ) against the length of the flagellum to see if there were any obvious changes to this structure as the flagellum grows . The nine central cylinder structures were all between 120 and 200 nm ( average 158 ± 25 nm ) , but there was no detectable increase in their thickness with flagella length . The assumed CP minus ends were in close proximity to the distal end of the cylinder , and 14 out of 17 minus ends were capped . The open CP ends were found in rather short flagella ( 1 . 2 and 2 μm; Figure 6B; Video 13 ) . 10 . 7554/eLife . 01479 . 026Figure 6 . The transition zone in C . reinhardtii is structurally uniform during the cell cycle . ( A ) A 20-nm thick tomographic slice showing the two flagella extending from the C . reinhardtii cell . ( B ) The exact thickness of the transition zone central tube structure and the locations and structure of the CP minus ends within/around them . Note that the thickness of the central tube does not correlate with the flagellar length . This graph includes both flagella that we know are growing , and flagella of unknown dynamic state . The three colored columns represent the measurements of central tubes found in cell 1 , 2 and 3 in C–E . ( C ) The cartoons to the left show outlines of the cells visualized in 10-nm thick tomographic slices of transition zone ( D ) and ( E ) where their 3D models show that the central tube structure consists of two baskets . The top basket is almost complete and the bottom one is partial . ( F ) The nine central tubes are here arranged by size , and the contribution of the upper and lower basket to the thickness of the structure is displayed . Note that the proportions within the structure remain similar through out , with the upper basket contributing 60 ± 5% , and the lower basket 28 ± 5% of the total central tube thickness . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 02610 . 7554/eLife . 01479 . 027Video 13 . The central cylinder in the C . reinhardtii transition zone ( related to Figure 6 ) . Central cylinder is modeled in brown , the CP in green . Both minus ends are capped . Scale bar = 50 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 027 Three cells were arranged by the lengths of their flagella and thus , their cell cycle stages ( Figure 6C–E ) . The central cylinder looks like two stacked U’s in tomographic slices of all cells ( Figure 6D ) . Density thresholding of our tomograms reveals that the U’s resemble two stacked baskets in 3D ( Figure 6E ) . As the central cylinder grows , the proportions of the baskets remain surprisingly similar ( Figure 6D , E ) , with the longer upper basket constituting 60 ± 5% of the total basal plate thickness and the more partial lower basket constituting 28 ± 5% ( Figure 6F ) . Interestingly , in the one cell with long flagella where we reconstructed both transition zones , the two central cylinder structures were similar in thickness and in their anchoring of the CPs ( Figure 6A , B; long flagellum ) . We conclude that the structural alterations in the region of CP nucleation during flagellar maturation in T . brucei are not conserved between these two species , as the C . reinhardtii structure merely grows lengthwise . Since the flagellum extends at the distal end , one has a timeline of flagellar extension with the most recently built piece at the distal tip and the oldest region at the proximal end ( Figure 7A , B ) . We used this property to dissect the assembly process in the disorganized growing long T . brucei flagellum and found the central pair projections ( shown to be missing in these tips in Figure 2F ) ∼0 . 5 μm from the distal tip ( Figure 7C ) . However , already after 0 . 25 μm the axoneme had its normal circular arrangement , which was also where the radial spokes were incorporated into the axoneme , confirming the crucial role of the radial spokes in the circular arrangement of dMTs of these flagella . Interestingly , it was also at ∼250 nm from the distal tip that the microtubules were found to start in chemically fixed cells ( Figure 1—figure supplement 1 ) , suggesting that the presence of radial spokes stabilized the dMTs and prevented further shrinkage . The paraflagellar rod ( PFR ) , an extra-axonemal para-crystalline structure found in T . brucei and many other kinetoplastids ( Vickerman , 1962; Bastin et al . , 1998; Portman and Gull , 2010; Höög et al . , 2012 ) , was the last component added to the flagellum , ∼800 nm from the tip . 10 . 7554/eLife . 01479 . 028Figure 7 . The axoneme is a timeline of flagellar assembly . ( A ) Because the new flagellum ( * ) grows at its distal tip , the axoneme is older closer to the basal body . This spatial assembly progress is used to reveal the order of structural additions to the growing axoneme . ( B ) Cross-sectional views from a tomogram of T . brucei flagella ( old flagellum to the left and the growing new flagellum to the right ) . In ( A ) the slice is taken close to the tip of the growing flagellum . Only a few doublet microtubules have extended to here . ( B ) 100 nm in to the flagellum , dMTs and CPs are present but arranged in abnormal angles to each other . ( C ) The axoneme starts appearing normal ∼250 nm into the axoneme . However , note the almost complete ring around the CPs in the old flagellum ( arrow ) , most likely formed by central pair projections and/or heads of radial spokes . This ring is still not to be found in this most proximal tomographic slice taken . ( C ) Measurements of axonemal component incooperation was done in a multitude of cross-sectional and longitudinal flagella tip tomography reconstructions . ( D ) T . brucei flagellum grows in a disorganized manner probably due to a lack of associated axonemal proteins in the distal tip , this model is made to approximate where the components are added in comparison to the distal tip . As the flagellum matures ( right ) , the axoneme is organized all the way to the tip and the basal plate has altered structure from the two rings to an electron-dense cloud . Electron-dense structures are visible from the lumen of the CP and A-tubule extending towards the membrane . ( E ) In C . reinhardtii , growth is organized with the CP protruding slightly at the tip of the axoneme . As the flagellum matures , the transition zone central tube extends but no morphological changes were seen . DOI: http://dx . doi . org/10 . 7554/eLife . 01479 . 028 Finally , we used this information to build a summarizing model of the two ways to grow and mature a flagellum in the protozoa T . brucei and C . reinhardtii . In T . brucei the growing axoneme is disorganized , with shorter CPs than dMTs , all of which are possibly not anchored to the flagellar membrane ( Figure 7D ) . The associated axonemal protein modules such as radial spokes and CP projections are added in later assembly steps . In the proximal part of the growing flagellum , the basal plate is first arranged into two electron-dense rings close to which capped CP ends can be found . In the corresponding area of a mature flagellum , the basal plate is a loose electron-dense area in which one , sometimes two , open or closed CP minus ends are found . In C . reinhardtii , the growing axoneme is not very different ( on this ultrastructural level ) from the mature axoneme as the circular arrangement , with the associated axonemal proteins , is maintained all the way to the growing tip ( Figure 7E ) . This holds true also for the central structure of the transition zone , the only maturation of which seems to be a lengthwise extension . Despite the remarkable conservation of the axonemal structure , we have shown two ways to build a flagellum; flagellar growth occurs in both species and length-dependent manner . Whereas , we see no difference in growth in short T . brucei and C . reinhardtii flagella , T . brucei long flagella appear to first elongate their axonemal MTs in a disordered manner , which is then stabilized to a circular arrangement by the addition of associated protein complexes such as the radial spokes . In contrast , our study suggests that the axonemal dMTs and structural proteins are assembling simultaneously in long C . reinhardtii . What regulates such differences in the assembly pathways ? Dentler ( 1980 ) showed that the growing axoneme in C . reinhardtii had structures linking from the tips of the microtubules to the membrane . Woolley et al . ( 2006 ) failed to see such linking structures in growing flagella of Leishmania major ( but did see them in mature flagella ) , a close relative to T . brucei . We speculate that the presence of such anchorage in C . reinhardtii could provide structural clues for the growing axoneme , holding it into a close-to-circular arrangement . This arrangement could then facilitate the incorporation of the associated axonemal proteins such as radial spokes . Another explanation for the apparent difference in axonemal growth mechanism between T . brucei and C . reinhardtii could be that these flagella are elongating at different rates: snapshots of a fast-growing axoneme might capture disordered intermediates that are not seen in a more slowly elongating structure . However , the initial growth rate of regrowing C . reinhardtii flagella ( shed by ionic shock ) is approximately 12–24 µm/hr , which then declines to 9 µm/hr as the flagellum gets longer ( Rosenbaum et al . , 1969; Flavin and Slaughter , 1974 ) . The growth rate of their flagella after mitosis remains unknown , but we assume equal or faster assembly because of the pre-mitotic reabsorbtion of the flagellar components that could then be utilized in the construction of the next flagellum . In T . brucei , we deduced the flagellum growth rate from the growth rate of the PFR to be a constant ∼4 µm/hr ( Bastin et al . , 1999 ) . These growth rates indicate that the structural differences we see are not caused by a slower flagellar growth rate , and therefore more organized growth , in C . reinhardtii . The differences in growth mechanics may not be so surprising , since the two species grow their flagella in different circumstances . In T . brucei , the new flagellum is not necessary for cell motility since the old flagellum is still present and active . In C . reinhardtii on the other hand , both flagella have been shed as a preparation for mitosis , so the cell is completely dependent on the reappearance of the two new flagella for its motility , making the rapid establishment of function important . There are numerous further differences between the flagella of these two organisms: ( 1 ) The beat form of the flagella ( Schmidt and Eckert , 1976; Heddergott et al . , 2012 ) . ( 2 ) The presence of a PFR in T . brucei ( Vickerman , 1962; Bastin et al . , 1998; Portman and Gull , 2010; Höög et al . , 2012 ) . ( 3 ) A rotational CP in C . reinhardtii; ( Mitchell , 2003 ) vs a stationary CP T . brucei ( Gadelha et al . , 2006 ) . Based on the findings presented in this paper , we can now add to this list a difference in the pathways for establishing axoneme organization during MT elongation , and maturation of the proximal region during the cell cycle . It is important to note that we did not use deflagellation to create a situation of regenerating flagella in C . reinhardtii , as this regrowth might be different from the natural situation occurring when flagella regrow after mitosis . Pre-mitotic reabsorption of the flagellum probably allows for storage of the flagellar protein pool , a pool that would have been lost in the event of ionic shock deflagellation . Furthermore , in deflagellation before mitosis the basal plate is lost in a final event of shedding ( Parker et al . , 2010 ) , whereas the stress induced deflagellation triggers a severing event distal to the basal plate , which can then form the transition zone for the next flagellum ( Rosenbaum et al . , 1969 ) . We also revealed that the flagellum tip structure might be different to what was previously described . Most notably , we found no evidence of a singlet region , which has been shown to be over 2 μm long in Tetrahymena sp ( Sale and Satir , 1976 ) and ∼1 μm in C . reinhardtii . We also found no obvious CP cap , not disproving its existence ( we did see evidence for filamentous proteins extending into the CP lumen ) , but showing that in well preserved cells it does not appear as previously described ( Dentler and Rosenbaum , 1977; Dentler , 1980; Fisch and Dupuis-Williams , 2012 ) . Even though we found the CPs to extend further than the dMTs , it was only within ∼50 nm , in contrast to the published flagellar tip arrangement in C . reinhardtii ( of unknown dynamic state ) , that shows the central pair protruding ∼400 nm beyond the doublet microtubules ( Ringo , 1967 ) . These differences in assembly order , speed , function , and structure , in two species both with the conventional 9+2 motile flagella structure , pose the question if the internal environments of flagella of different species are more different than presently assumed . Ciliogenesis will most likely involve a whole subset of flagellar proteins , with functions that are distinct from the IFT of axonemal proteins to the assembling tip . Flagellar growth rates after ionic shock in C . reinhardtii are length dependent , with shorter flagella growing faster than longer ones . Together with a steady disassembly rate , this forms the basis of the balance-point-model of flagellar length regulation ( Marshall et al . , 2005 ) . The faster growth of short C . reinhardtii was then showed to depend on longer IFT trans in the short flagellum ( Engel et al . , 2009 ) . This would fit well with our observations that the modes of assembly in the short and long post-mitotic C . reinhardtii tips do not differ beyond the point of central pair extension . Interestingly , we did detect differences in short vs long growing flagellar tips in T . brucei , an organism that most likely has linear flagellum growth . The absence of several axonemal components at the long growing tip might indicate that IFT is rate limiting in this growth . One would then predict that flagellar length is independent of size of IFT trains in these cells . However , it has recently been shown that the IFT traffic in T . brucei also differs considerably from that of C . reinhardtii ( Buisson et al . , 2013 ) . Thus , to further understand the normal structure and function of flagella , and the pathology of various ciliopathies , it is crucial to further understand ciliogenesis . With this paper we have revealed two modes of flagella growth , an area of flagellum biology that still remained mostly in the dark , and complementing our extensive knowledge on IFT of building material to the site of flagellar growth . Procyclic T . brucei strain 427 ( high pressure freezing; HPF ) or 29–13 ( chemical fixation ) were grown in SDM-79 media supplemented with 10% fetal bovine serum ( for chemical fixation; PAA Laboratories Ltd , UK ) and 20% fetal bovine serum ( for HPF ) ( Brun and Schönenberger , 1979 ) . Cell growth was monitored by using a CASY DT cell counter ( Sedna Scientific , UK ) , and cultures were diluted on a daily basis to maintain a density between 5 × 105 and 1 × 107 cells per milliliter . Cells were prepared for electron tomography by high pressure freezing or chemical fixation followed by epon embedding as described previously ( Höög et al . , 2010 ) . Unperturbed procyclic T . brucei strain 427 was plunge frozen and imaged intact as in ( Höög et al . , 2012 ) . Wild-type Chlamydomonas reinhardtii strain 137C mt+ was grown at room temperature in liquid culture ( Sagar and Granick medium ) using a 10/14 dark/light cycle . After the cultures were shifted to the dark cycle , the cells were prepared for electron microscopy by HPF followed by freeze substitution as essentially described in O’Toole et al . ( 2003 , 2007 ) . Briefly , the liquid culture was spun at 500×g for 5 min and the loose pellet was resuspended in a medium containing 150 mM mannitol for 1 hr . The samples were then spun at 500×g , the supernatant decanted and the loose pellet frozen using a BAL-TEC HPM-010 high-pressure freezer . The frozen samples were freeze substituted in 1% OsO4 and 0 . 1% uranyl acetate in acetone for 3 days then embedded in epon/araldite resin . Semi-thick ( 300–400 nm ) serial sections of the samples were cut using an ultracut UCT ultramicrotome ( Leica Microsystems Ltd , UK ) . The sections were flattened by chloroform gas exposure whilst floating on the water surface . Ribbons of serial sections were put centered on 2 × 1 mm copper palladium slot grids ( Agar Scientific Ltd , UK ) . The sections were stained for 5 min on 2% uranyl acetate followed by 30 s Reynold’s lead citrate . 15-nm colloidal gold particles were applied to both sides of the grid , to be used for image alignment . Tilt series of serial sections from flagella distal tips were acquired using the serialEM software ( Mastronarde , 2005 ) operating a Tecnai TF30 300 kV IVEM microscope ( FEI Co . , The Netherlands ) . Images were collected about two orthogonal axes in 1° increments ( ±60° ) using a Gatan CCD camera ( pixel size 0 . 76–1 . 3 nm ) . Tomographic reconstructions were calculated , the two axis combined , serial sections stitched together and models were created using the IMOD software package ( Kremer et al . , 1996; Mastronarde , 1997 ) . Note that not all flagella tips are complete within the reconstructed volume in one tomogram . Many are reconstructed over serial sections . Measurements were only done within the same tomogram to ensure no error was introduced in the joining process . All flagella visualized were followed through the serial sections , taking lower magnification images of the entire cell to be able to reconstruct the flagellum length . Microtubule ends were classified as in Höög et al . ( 2007 , 2011 , 2013 ) .	Some cells have a whip-like appendage called a flagellum . This is most often used to propel the cell , notably in sperm cells , but it can also be involved in sensing cues in the surrounding environment . Flagella are found in all three domains of life—the eukaryotes ( which include the animals ) , bacteria and ancient , single-celled organisms called Archaea—and they perform similar functions in each domain . However , they also differ significantly in their protein composition , overall structure , and mechanism of propulsion . The core of the flagellum in eukaryotes is made up of 20 hollow filaments called ‘microtubules’ arranged so that nine pairs of microtubules form a ring around two central microtubules . The core also contains many other proteins , but it is not clear how all these components come together to make a working flagellum . Moreover , it is not known if the flagella of different groups of eukaryotes are all assembled in the same way . Now , Höög et al . have discovered that although the core structure of the eukaryote flagellum is highly conserved , it can be assembled in markedly different ways . Some species of eukaryote—such as Chlamydomonas reinhardtii , a single-celled green alga , and Trypanosoma brucei , the protist parasite that causes African sleeping sickness—must grow new flagella when their cells divide , so that each new cell can swim . Using a form of electron microscopy called electron tomography , Höög et al . could see the detailed structure of the growing flagella in three dimensions . At first the cores of the flagella in these two distantly related species grow in the same way . However as the flagella get longer their cores grow in completely different ways . The microtubule filaments in longer flagella grow in a synchronized manner in the alga , but in a disorganized way in the protist . The results of Höög et al . illustrate that it is not advisable to draw generalised conclusions based on studies of a few model species . However , since defects in flagella are known to cause several diseases in humans , this knowledge might inform future studies aimed at developing treatments for infertility , respiratory problems , and certain kinds of cancer .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "microbiology", "and", "infectious", "disease" ]	2014	Modes of flagellar assembly in Chlamydomonas reinhardtii and Trypanosoma brucei
Animals have developed mechanisms to reconstruct lost or damaged tissues . To regenerate those tissues the cells implicated have to undergo developmental reprogramming . The imaginal discs of Drosophila are subdivided into distinct compartments , which derive from different genetic programs . This feature makes them a convenient system to study reprogramming during regeneration . We find that massive damage inflicted to the posterior or the dorsal compartment of the wing disc causes a transient breakdown of compartment boundaries , which are quickly reconstructed . The cells involved in the reconstruction often modify their original identity , visualized by changes in the expression of developmental genes like engrailed or cubitus interruptus . This reprogramming is mediated by up regulation of the JNK pathway and transient debilitation of the epigenetic control mechanism . Our results also show that the local developmental context plays a role in the acquisition of new cell identities: cells expressing engrailed induce engrailed expression in neighbor cells . Regeneration is a classical problem in developmental biology . The cells involved have to undergo a developmental reprogramming so that they acquire the new cell identities necessary to regenerate a lost organ or a damaged tissue . In the classical example of the regeneration of an amputated amphibian limb ( Kragl et al . , 2009 ) , proximal cells have to generate new cells with different—distal—identity . Very little is known about the phenomena and mechanisms behind this developmental reprogramming . We are investigating this problem using the compartments of Drosophila imaginal discs as a model system for regeneration . The subdivision into lineage blocks , termed compartments , is a principal feature of the organization of the body of Drosophila ( Garcia-Bellido et al . , 1973; Morata & Lawrence , 1977 ) . The earliest compartmentalization event is established during embryogenesis ( Lawrence & Morata , 1977 ) and separates anterior ( A ) and posterior ( P ) compartments in each segment . The A/P boundaries in the different body segments not only separate the original lineage blocks of the fly; they also are topological landmarks to delimit Hox genes expression ( Lawrence , 1992 ) . Since the initial establishment of the A/P boundary , the P compartments acquire expression of the engrailed ( en ) gene , which determines anterior or posterior identity within each segment: the ‘on’ state specifies posterior whereas the ‘off’ state specifies anterior identity . This early lineage segregation is preserved for the rest of the development in all larval and adult structures . In the imaginal discs ( the precursors of the adult cuticular structures ) en remains expressed in the P compartments ( Brower , 1986 ) and is also required to maintain the A/P border ( Morata and Lawrence , 1975 ) . The absence of en function in the A compartment allows anterior cells to activate the Hh pathway and subsequently the cubitus interruptus ( ci ) gene , a marker of A compartment identity ( Orenic et al . , 1990 ) . The expression of en is regulated epigenetically . It is kept in off ( silenced ) state in the A compartments through the activity of the Polycomb-G genes ( Pc-G ) ( Busturia and Morata , 1988 ) , whereas the trithorax-G genes ( trx-G ) maintain en in on state in P compartments ( Breen et al . , 1995 ) . In addition to the A/P boundary , there is in the wing disc another lineage border , separating the dorsal ( D ) and the ventral ( V ) compartments . This boundary appears during larval development ( Garcia-Bellido et al . , 1973 ) and is dependent on the activity of the gene apterous ( ap ) , which confers dorsal identity ( Diaz-Benjumea and Cohen , 1993 ) . The A/P and the D/V borders not only separate cells with different identities in the wing disc; they also function as developmental organizers of the imaginal discs . The posterior compartment cells secrete the Hedgehog ( Hh ) morphogen that activates the decapentaplegic ( dpp ) gene in the anterior cells close to the border , from where the Dpp signal diffuses to the two compartments and controls their growth and pattern ( Lawrence and Struhl , 1996 ) . The Wg signal emanating from the D/V border also has a major patterning function in the wing disc ( Irvine and Vogt , 1997 ) . It follows that the establishment and maintenance of the A/P and D/V boundaries are critical factors in the development of the wing disc and that major alterations of these boundaries would grossly interfere with normal pattern and growth . Therefore , it is expected that regenerative processes incorporate mechanisms to ensure the restoration of these borders in damaged discs . The possibility of collapse and subsequent restoration of the A/P border during regeneration in the wing disc was suggested by Szabad et al . ( 1979 ) , although recent reports ( Bergantinos et al . , 2010; Smith-Bolton et al . , 2009 ) have failed to observe lineage transgressions after massive damage to the disc . We are investigating the regenerative response of imaginal discs to the ablation of specific compartments , and especially how the stability of compartment boundaries is maintained . Our results indicate that ablation of the P or the D compartment causes a transient collapse of the A/P or the D/V boundaries , which are very quickly reconstructed . During the reconstruction process some cells are reprogrammed and change their original compartmental identity . We provide evidence that the identity changes are associated with gain of activity of the JNK pathway and with relaxation of the epigenetic control by the Pc-G and trx-G genes . In addition , we suggest that the identity changes observed are induced by a novel mechanism by which isolated cells acquire the identity of their neighbours . In the experiments to induce massive damage to specific compartments , equivalent to ablation , we have utilized the Gal4/UAS/Gal80TS method ( see ‘Materials and methods’ for details ) to force high levels of apoptosis in specific domains ( Bergantinos et al . , 2010; Smith-Bolton et al . , 2009; Herrera et al . , 2013 ) . The temperature-sensitive Gal4 suppressor Gal80 allows temporal manipulation of Gal4 activity . In these experiments a temperature shift from 17 to 29°C causes a loss of Gal80 function , which now permits the activity of the Gal4 driver . In turn , Gal4 activates the UAS-hid and UAS-Flp transgenes ( Figure 1—figure supplement 1 ) . The activation of UAS-hid with the hh-Gal4 driver causes massive apoptosis in the P compartment . In addition , the high levels of Flipase generated by the UAS-flp transgene induce recombination in the act>stop>lacZ cassette , with the result that the cells of the original P compartment become indelibly marked with the activity of the lacZ gene , which encodes the ßGal protein . Under our experimental conditions ( 48 hr of continuous Flp activity ) all the cells of the P compartment acquire the ßGal label ( Figure 1A , B ) . We refer to these cells as the ‘Hh lineage’ . This is an important point because we consider the ßGal label as the indicator of the compartmental provenance of the cells . 10 . 7554/eLife . 01831 . 003Figure 1 . Ablation of the posterior compartment . Wing imaginal discs after 48 hr of induction of GFP ( A , C and E , controls ) or hid ( B , D and F ) . Note that in both cases the βgal lineage label is acquired by all the posterior cells ( A and B ) . The disc in B shows high levels of apoptosis , indicated by Caspase3 activity . ( C and D ) Cross-sections of wing discs perpendicular to the A/P border at the level of the wing pouch . The peripodial membrane is on top , the columnar epithelium on the bottom . Anterior compartment at the left , posterior at the right . There is a marked reduction of the size of P compartment in which hid is expressed ( D ) . The epithelium is much thinner than in control disc ( C ) , indicating a big reduction of cell number due to the ablation ( compare the number of nuclei in the right part of C and D ) . ( E–G ) Shape of the A/P border after 48 hr of GFP ( E ) or hid ( F ) induction . In the later this border becomes wiggly and inter-digitized , a feature quantified in panel G . By allowing 72 hr of recovery this effect is partially recovered ( third bar in panel G ) . Bars represent S . E . M . , n>15 in each genotype and time point , p<0 . 01 , p<0 . 05 . See also Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 00310 . 7554/eLife . 01831 . 004Figure 1—figure supplement 1 . Genetic ablation and lineage labeling system . ( A ) In discs from larvae of the genotype indicated maintained at 17°C the Gal80TS repressor blocks Gal4 activity , thus the UAS vectors cannot be activated . ( B ) By rising the temperature to 29°C the Gal80TS is inactivated , thus provoking two effects: ( i ) over-expression of the pro-apoptotic vector UAS-hid , causing massive cell death in the Hh domain ( P compartment ) , ( ii ) over-expression of the UAS-Flipase construct , which in turn induces recombination in the act>stop>LacZ cassette . This recombination labels indelibly all the cells of the posterior compartment ( green label in the diagram ) . ( C ) By taking back the larvae to 17°C the Gal80TS activity is recovered , thus blocking the expression of both hid and Flp . This allows the recovery of the ablated domain . The lineage labeling also allows recognizing if cells originated outside the ablated domain ( not labeled ) are participating in regeneration . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 004 To ablate the dorsal compartment and to track the lineage of the dorsal compartment cells , we have followed a similar strategy , but using the ap-Gal4 line , which is expressed specifically in the dorsal wing compartment . We describe in detail the results obtained ablating the P compartment . After 48 hr of hid activity , the great majority of P compartment cells in the disc are dead or are dying ( Figure 1B ) . We estimate that about 70% of the cells die and as a result the size of the compartment is greatly reduced ( compare Figure 1C , D ) . However , in spite of the massive cell death in the P compartment , there is clear , although highly irregular ( Figure 1E–G ) , boundary separating anterior and posterior cells , outlined by the anterior compartment marker Ci . After allowing 72 hr of recovery at 17°C , the damaged P compartment is completely regenerated: it exhibits normal size and shape and an almost normal A/P border . We have studied by clonal analysis the growth of the A and P compartments during and after the ablation period . The size of clones in the A and P compartment during the ablation period is similar ( Figure 2A–D ) , although the P compartment clones have probably grown more since some of their cells must have died due to Hid activity . The proliferative response to the ablation becomes evident during the recovery time ( Figure 2E–H ) : clones in the P compartment are about 3 times bigger than those in the A compartment . These additional divisions restore the full size of the compartment . This result also illustrates the independent size regulation of the two compartments ( Martin and Morata , 2006 ) ; once the A/P border is re-established the two compartments grow autonomously . 10 . 7554/eLife . 01831 . 005Figure 2 . Clonal analysis of growth in discs in which the P compartment has been ablated . ( A–D ) GFP-labeled clones only allowed growing during the ablation period . The clones were induced at the beginning of the ablation and the discs fixed at the end of it . Control disc ( B ) a hid-induced disc ( C ) and clone size quantification ( D ) are shown . Note the absence of major differences in clone size between control discs and hid-induced in both compartments . ( E–H ) Clones induced at the end of the ablation period and scored after 72 hr of recovery period at 17°C ( see diagram E ) . Again , a control disc ( F ) , a hid-induced disc ( G ) , and quantification ( H ) are shown . Note in G that the clones in the P compartment are much bigger than those in the A compartment of the ablated discs and those of the P compartment in controls ( F ) . Bars represent S . E . M . , n>25 in each genotype and time point , p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 005 Next , we examined if the ablation of the P or the D compartment affect the stability of the A/P or the D/V border . Our experimental system permits distinguishing the compartmental origin of the cells . Those of the P or D compartment are labeled by lacZ ( ßgal ) activity , whereas the A or V compartment cells are labeled by the lack of it . In the majority of discs in which the P compartment is ablated ( 22 out of 31 ) , we find that cells from the A compartment have contributed to the regenerated P compartment . The transgressing cells are of anterior origin , as indicated by the lack of ßGal staining , but now belong to the P compartment , as showed by the lost the Ci anterior marker ( Figure 3A ) and the gain of en expression ( Figure 3—figure supplement 1A ) . These lineage violations are not observed in control discs . There was the possibility , however , that in the experimental discs the ßGal label of the P compartment might not be as precise as in the controls , leaving some cells unmarked that could be interpreted as of anterior origin . But under that hypothesis these unmarked cells should appear anywhere in the P compartment . We find that the transgressions localize to the proximity of the A/P border ( Figure 3—figure supplement 2 ) , which argues strongly against that possibility . 10 . 7554/eLife . 01831 . 006Figure 3 . Transgressions of the A/P and the D/V border caused by ablation of the P or the D compartment . ( A and B ) Wing discs after 48 hr of Hid treatment followed by 72 hr of recovery . The original Hh lineage is labeled by ßgal ( green ) , the A compartment is marked with an anti-Ci antibody ( red ) and the nuclei with Topro ( blue ) . ( A ) Note the presence of groups of cells ( asterisk ) originated in the anterior compartment ( they lack the βgal label ) but that have lost Ci activity , indicating that they have lost anterior identity . ( B ) The unexpected finding that cells from the P compartment–they are part of the Hh lineage–can penetrate in the A compartment and to acquire anterior identity , as demonstrated by the Ci marker ( arrows ) ( C ) Portion of a disc in which the P compartment has been ablated , containing two sets of ‘twin’ clones labeled with GFP-green/LacZ ßgal ( see ‘Materials and methods’ for details ) . The clones were initiated at the beginning of the ablation of the P compartment and fixed 24 hr after the end of ablation . In those panels the A/P borderline is blue and the twin clones delineated in red or green . Note that the two cases the clones cross over the A/P line . ( D ) Wing disc after 48 hr of Hid induction in the dorsal compartment followed by 72 additional hours of recovery . Note the presence of groups of cells ( asterisks ) that in spite of their ventral origin ( lack of dorsal lineage βgal label ) now present dorsal markers as the mew Integrin ( PS1α ) . See also Figure 3—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 00610 . 7554/eLife . 01831 . 007Figure 3—figure supplement 1 . Transgressions detected during the ablation period and transgressions using p53 as apoptotic inducer . ( A ) Wing disc fixed at the end of 48 hours hid induction . Note that the trespassing cells ( with lack of βgal label , indicated by an asterisk ) , have activity of the posterior marker engrailed . ( B ) Wing disc after 48 hr of p53 over-expression in the posterior compartment and 72 additional hours of recovery . Note the presence of compartment transgressions ( red arrows , cells that lack the βgal label ) . Occasionally , part of the cells of a transgression retain its original identity , as the example pointed with a white arrow , in which a group of Ci positive cells remain trapped in the posterior compartment . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 00710 . 7554/eLife . 01831 . 008Figure 3—figure supplement 2 . Localization on the wing disc of the different transgressions of the A/P border . We have plotted all the transgressions found in after overexpressing the pro-apoptotic gene hid a disc containing the normal doses of Polycomb ( A ) or only one dose ( B ) . The transgressions from A to P are in light blue and the P to A in light red . The darker color indicates the superimposition of several transgressions in some regions . Note that virtually all the transgressions reach the A/P border . Those from A to P exhibit some preference for the region between hinge and pouch , whereas those from P to A localize to the wing pouch . The overall preferential location of the transgressions near the A/P border rules out the possibility of incomplete label of the hh lineage . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 008 In addition , we unexpectedly find that surviving cells from the ablated P compartment may also contribute to the A compartment: cells of posterior provenance , labeled with ßGal , frequently penetrate ( 13 transgressions in a sample of 31 discs ) into the A compartment and acquire anterior identity , visualized by ci activity ( Figure 3B and Figure 3—figure supplement 2 ) . Since the A compartment was not in need of reconstruction , this observation indicates that the transgressions are caused by a transient collapse of the A/P boundary , which affects the two compartments . It is worth mentioning that the A/P transgressions have occurred during the ablation period , as they are found in discs fixed at the end of it , without recovery time ( Figure 3—figure supplement 1A , see also Figure 4E–E′′ ) . This is a significant observation because it suggests that the restoration of the A/P border is immediate and concomitant with the ablation . The lineage violations are not specific to the UAS-hid transgene; we have made similar observations using other pro-apoptotic genes ( Figure 3—figure supplement 1B ) . 10 . 7554/eLife . 01831 . 009Figure 4 . Involvement of the JNK pathway in the transgression of the A/P boundary during regeneration . ( A–A′′ ) Wing disc and magnification showing activation of puc-LacZ ( green ) after ablation of the P compartment . The A/P boundary is outlined by Ci ( red ) label . Note that some anterior cells contain LacZ expression . The puc–lacZ staining is not particularly strong in the P compartment because it is located predominantly on the basal side of the disc , where the dying cells accumulate . ( B and B′ ) Control non-ablated disc showing absence of puc-LacZ activity , except in the stalk region where it is normally expressed . The panel C illustrates the experiment designed to test the requirement for JNK activity for crossing the A/P boundary during regeneration . After removing repression by Gal80TS the dpp-LHG line forces hid activity in the dpp domain ( red ) , located just anterior to the A/P line . At the same time the posterior compartment cells lose the potential to gain JNK activity due to puc over-expression . The results are illustrated in D–F′ . Cells of posterior origin ( green ) can penetrate in the A compartment if they can activate JNK ( E–E′′ ) , but are unable to do so if JNK activation is prevented by puc over-expression . Quantitative data are shown in D . Bars represent S . E . M . , n = 15 in each genotype , p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 009 To reinforce the conclusion for the preceding experiments that the A/P border is transgressed during regeneration of the P compartment , we also performed an experiment of clonal analysis . The idea was that if the A/P border had collapsed during ablation some marked clones induced before or at the time of collapse should include anterior and posterior cells . We generated ‘twin clones’ such that we labeled the progeny of the two cells resulting from mitotic recombination events ( see ‘Materials and methods’ ) . The clones were generated at the beginning of the ablation period and fixed the discs at different times after the end . As illustrated in Figure 3C , we found numerous cases of clones crossing over the A/P border ( 10 cases of crossing in 16 discs examined ) . These transgressions do not appear in discs of the same genotype in which the P compartment is not ablated . A loss of lineage restriction was also observed in the D/V compartment boundary after ablation of the D compartment . As in the experiment to assay the A/P boundary , we can follow the lineage of the D compartments by labeling the cells with the UAS-Flipase system ( see ‘Materials and methods’ and Figure 1—figure supplement 1 ) . We find that cells of ventral provenance appear integrated in the regenerated D compartment ( Figure 3D ) . The change of identity is visualized by the acquisition by these cells of PS1α integrin , a marker of dorsal cells ( Gotwals et al . , 1994 ) . There is evidence that JNK signaling is involved in in situ regeneration in imaginal discs ( Bergantinos et al . , 2010; Herrera et al . , 2013; Worley et al . , 2012 ) and also in transdetermination of transplanted discs ( Lee et al . , 2005 ) . Moreover , a recent report ( Gettings et al . , 2010 ) indicates that JNK activity is required for the change of en expression and subsequent crossing of the A/P border by some cells during dorsal closure in Drosophila embryos . Therefore , we have examined whether it plays a role in the lineage transgressions we observe . We had previously noticed that JNK is up regulated in the P compartments of ablated hh > hid discs , as shown by the activation of the puc-LacZ marker , which is not normally expressed in the wing pouch ( Figure 4B , B′ ) . This was expected because Hid induces JNK ( Shlevkov and Morata , 2012 ) , but we also find anterior compartment cells near the A/P border that express the puc-LacZ marker ( Figure 4A–A′′ ) . We designed a special experiment combining the lexA/lexO ( Yagi et al . , 2010 ) and the Gal4/UAS binary systems to examine the involvement of JNK in the A/P lineage transgressions . The rationale was to induce massive cell death in the region of the A compartment close to the A/P border , and at the same time to prevent activation of JNK in the P compartment ( Figure 4C , details in ‘Materials and methods’ ) . In discs of genotype dpp-LHG>LexO-hid hh-Gal4>UAS-puc Gal80TS the 17 to 29ºC temperature shift causes ablation of the Dpp domain in the A compartment and at the same time prevents JNK activation in the P compartment due to the over-expression of the negative regulator puc ( Martin-Blanco et al . , 1998 ) . In control dpp-LHG>LexO-hid Gal80TS discs the shift causes ablation of the Dpp domain but JNK activity is not prevented . The results are illustrated in Figure 4D–F . In the experimental discs , the transgressions are greatly reduced , clearly indicating a requirement for JNK function . This experiment also shows that A/P transgressions also occur after damage to the A compartment , even if only a portion of the compartment is affected . The preceding experiments establish that after massive damage to the A , P or the D compartment the lineage boundaries collapse but are quickly restored . During the process some cells change their original identity , indicating that those cells have undergone a reprogramming process that modifies the expression of identity genes , en , ci , ap , etc , activating some and repressing others . In support of this , we observe that after ablation of the dorsal compartment ( in ap>hid discs , Figure 5—figure supplement 2 ) the expression of engrailed in the dorsal part of the posterior compartment is lower than in the corresponding region in the ventral compartment . The functional state of Drosophila identity genes , i . e . , their active or silencing mode , is controlled by the genes of the Polycomb ( Pc-G ) or trithorax groups ( trx-G ) ( Kennison and Tamkun , 1988; Lewis , 1978 ) . The Pc-G genes are responsible for keeping in off state ( silencing ) the genes inappropriate for a particular identity , whereas the trx-G genes maintain the activity of identity-specific genes . Therefore , we checked whether the identity changes observed during the reconstruction of compartment boundaries are associated with alterations of the activity of the Pc-G and the trx-G genes . It is known that the Pc-G and trx-G genes control en expression in the A and P compartments ( Busturia and Morata , 1988; Breen et al . , 1995 ) . We have measured the boundary transgressions after damage to P compartments of discs containing only one dose of Polycomb ( Pc3/+ ) or trithorax ( trxE2/+ ) in comparison with controls containing normal doses of the genes . The results indicate an involvement of the Pc-G and trx-G genes ( Figure 5A–G ) : in Pc3/+ discs , the number and size of lineage transgressions is greatly increased with respect to controls , whereas in trxE2/+ discs there are very few lineage transgressions . 10 . 7554/eLife . 01831 . 010Figure 5 . Changes in epigenetic regulation during A/P boundary reconstruction . Comparison of the size of transgressions after Hid treatment found in discs containing normal doses of Polycomb ( Pc ) or of trithorax ( trx ) ( A , control with GFP treatment , B experimental ) , with those in discs containing one dose of Pc ( C , control , D experimental ) or of trx ( E , control , F experimental ) . Note the significant increase in transgressions size in Pc3/+ background and the reduction in trxE2/+ background ( a small transgression is indicated with an arrow in F ) . Panel G shows the quantification of the transgressions size in the three genotypes , each with respect to its own control . Bars represent S . E . M . , n>15 in each genotype , p<0 . 01 , **p<0 . 05 . See also Figure 5—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 01010 . 7554/eLife . 01831 . 011Figure 5—figure supplement 1 . Alterations in the expression of Posterior sex comb ( psc ) and of H3K4 levels in the proximity of the A/P border after Hid administration . ( A and C ) . Psc expression ( red , mtdTomato ) in three control discs is uniform across the border ( A ) , defined by the limit of the lacZ expression under hh control . Note in the three experimental discs ( C ) a lowering of Psc levels , indicated by the bracket . ( B and D ) Similar comparison of H3K4 levels between three control ( B ) and three experimental discs ( D ) . H3K4 is increased in cells in the P compartment and in some anterior cells close to the border , indicated by brackets . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 01110 . 7554/eLife . 01831 . 012Figure 5—figure supplement 2 . Engrailed protein levels are reduced in posterior cells . Wing discs after 48 hr of GFP ( A–A′ , control ) or hid ( B–B′ ) induction in the dorsal compartment ( apterous-Gal4 domain ) . A′ and B′ panels show a heat map representation of the Engrailed protein levels . The D/V border is indicated with a discontinuous line . In the hid-expressing disc ( B–B′ ) , note the reduction of En levels in the dorsal compared to the ventral compartment . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 012 We have screened the expression during regeneration of several Pc-G and trx-G genes and of some epigenetic markers ( Figure 5—figure supplement 1 ) . The QF line ET40 ( Potter et al . , 2010 ) is an insertion of the QF transcription factor in the Posterior sex combs ( Psc ) gene , a member of the Pc-G . In Psc-QF>QUAS-mtdTomato discs in which the P compartment is ablated there is a clear down regulation of Psc activity ( Figure 5—figure supplement 1A , B ) . We also find an increase in the levels of tri-methylation of H3K4 , a mark of active chromatin ( Schubeler et al . , 2004 ) ( Figure 5—figure supplement 1C , D ) . In both cases , the modifications affect the entire posterior compartment and the region of the A compartment close to the A/P border . Taking everything together the preceding results strongly suggest that the changes of identity during regeneration are associated with temporal relaxation of the epigenetic control of compartment identity . We conjecture that as a result cells in the neighborhood of the A/P ( or D/V ) boundary may lose some of their developmental constraints , thus descending to a kind of ‘naive’ ( undetermined ) state . During the reconstruction of the boundaries some cells acquire a new compartmental identity , different from the original one . This implies the novo activation of identity genes . We speculated about the possibility of an induction mechanism by neighbor cells . That is , a cell in a naive state could be induced by its neighbors to express the same identity gene . The reason behind this speculation was the realization that identity changes observed appear to be influenced by the local developmental context , that is , the cells may change from anterior to posterior , but keep the overall wing identity . This suggested that the local developmental environment could play a role in determining the acquisition of the new identity . There is a recent observation ( Garaulet et al . , 2008 ) that dorsal compartment cells expressing en appear to induce the en-LacZ reporter in ventral cells located in their vicinity . We designed two experiments to check whether A compartment cells might be induced to express the posterior identity gene en by the influence of neighbor cells expressing the gene . The first experiment consisted of filling the anterior wing compartment almost entirely with GFP-marked clones of cells expressing en , but leaving a few islands of unmarked anterior cells without exogenous en activity . We find that these unmarked cells often activate the endogenous en gene , as indicated by the expression of the en-lacZ insert ( Figure 6C–E′′ ) and also by the presence of the En protein ( Figure 6—figure supplement 1A–B′′ ) . The gain of en activity is associated with down regulation of the A compartment markers Ci and Ptc ( Figure 6D′′ , E′′ , Figure 6—figure supplement 1C–D′′ ) . Significantly , this activation only occurs in discs densely populated by en-expressing clones; it does not appear if there are few clones ( Figure 6A–B′′ ) . It suggests that the induction process requires the receiving cells to be largely surrounded by inducing cells . This experiment was also performed at 17°C , a temperature at which the Gal4 system is less active and therefore the levels of en over-expression in the clones are lower . The result ( Figure 6—figure supplement 2 ) is that en is also non-autonomously activated under these conditions . 10 . 7554/eLife . 01831 . 013Figure 6 . Induction of en activity by en-expressing neighbors . ( A–E′ ) Clones of cells over-expressing engrailed and GFP were generated 48 hr before puparium formation . The clones of interest are those located in the A compartment . In A compartments with low density of en-expressing clones ( A , B–B′′ ) there is no induction of the en-lacZ reporter in anterior cells , nor there is alteration of Ci levels . However , when the disc is filled with en-expressing clones ( C and magnifications in D–E′′ ) anterior cells in contact with en-expressing clones acquire en-LacZ activity , while Ci protein levels decrease ( D′′ and E′′ ) . ( F–G′ ) Disc with Gal80-expressing clones ( refractory to Gal4 ) in the anterior compartment surrounded by cells over-expressing en-LacZ ( red ) and GFP driven by the act-Gal4 line ( see ‘Materials and methods’ for details ) . The clones show en activity , visualized by en-LacZ ( G ) . ( H and H′ ) High magnification of the A/P border in a control disc doubly stained for Patch , a marker of the A compartment that delineates the A/P border , and en-LacZ . Note that there is no extension of en activity beyond the boundary . ( I–L ) Scheme of our proposal of the ‘induced by neighbors’ model of en activation during the regeneration of the A/P boundary . Before cell death induction ( I ) the A and P compartments are separated by a normal straight A/P border . ( J ) During the cell killing the border collapses due to changes of identity of cells . This allows some intermingling before it is reconstructed; cells of anterior provenance may be surrounded by cells of posterior identity , which induce activity of the endogenous en gene of the anterior cells ( K ) . Once new the identities are established , the differential affinities of the A and P cells contribute to form a new A/P boundary ( L ) . See also Figure 6—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 01310 . 7554/eLife . 01831 . 014Figure 6—figure supplement 1 . Presence of En protein induced by en-expressing neighbors . ( A–B′′ ) Clones of en-expressing cells ( green ) in the A compartment inducing en activity ( red ) in anterior cells visualized with the anti-En antibody ( red ) . The region of the inset in A is magnified in B–B′′ . Notice the En protein in cells close to the border of the clones . ( C–D′′ ) disc containing en-expressing clones ( green ) stained for en-lacZ ( blue ) and Patched ( red ) . The inset area is magnified in D–D′′ . Notice that the expression of en in anterior cells close to the en-expressing clones ( D′ ) is associated with a diminution of Ptc levels ( D′′ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 01410 . 7554/eLife . 01831 . 015Figure 6—figure supplement 2 . Experiment of induction of clones over-expressing en ( green ) by the Gal4 system at 17°C . The genotype of the discs is the same as those in Figure 6; the only difference is the temperature at which the act-Gal4 transgene functions . In the disc in A , the region of the A compartment outlined by the white square is magnified in B . Note the en-lacZ expression ( red ) in cells outside but close to the borders of en-expressing clones . The A/P border s delimited by the high en-lacZ levels in the P compartment . DOI: http://dx . doi . org/10 . 7554/eLife . 01831 . 015 The second experiment consisted of generating clones of cells that are refractory to Gal4 in a field of Gal4-driven en-expressing cells ( see ‘Materials and methods’ ) . Clones in the anterior compartment expressing the Gal4-suppressor Gal80 are surrounded by cells with high levels of en activity driven by the act-Gal4 line . In these clones , we also observe induction of endogenous en activity ( Figure 6F–G′ ) . We believe that the phenomenon of en induction it is not due to the Hedgehog -mediated late activation of en in the anterior cells close to the A/P border ( Blair , 1992 ) , because the en-lacZ insert line we use in these experiments does not show this effect ( Figure 6H , H′ ) . Furthermore , as noted above , we only observe en induction in discs densely populated by en-expressing clones; in discs with fewer clones it does not occur ( compare A–B′′ and C–D′′ panels in Figure 6 ) . The Hh-mediated activation would not depend on clone density . The results reported bear on the regenerative response of the wing disc to ablation of the posterior or the dorsal compartment . Compartments are independent lineage blocks and express different genetic/developmental programs ( Lawrence , 1992 ) . Since the majority of our experiments concern the response of the P compartment , we restrict the discussion to those , although we are confident the conclusions also apply to the D compartment . We find that the P compartment fully regenerates after a massive damage that kills the majority of the cells . There is a proliferative response aimed to restore the normal size of the compartment ( Figure 2 ) . This response is easily visualized during post-ablation time ( Figure 2E–H ) , although it may start during the ablation: the fact that clones in the P compartment , which is under ablation , grow as much as those in the control A compartment ( Figure 2A–D ) suggests that they perform more cell divisions . The autonomous proliferative response of the A and P compartment provides further evidence to the proposal that A and P compartments are units of growth control ( Martin and Morata , 2006 ) . One unexpected result is that the ablation of one compartment , be it the P or the D compartment , results in a transient collapse of the A/P or the D/V boundary . Cells of anterior origin can penetrate into the P compartment ( Figure 3A ) and vice versa , cells of posterior origin can penetrate into the A compartment ( Figure 3B ) . The latter result is significant because the A compartment was undamaged and therefore the penetration of cells of posterior provenance cannot be due to a repair mechanism . Possibly , the collapse of the boundary results from a transient loss ( or debilitation ) of the identity of cells in the P compartment ( Figure 5—figure supplement 2 ) and likely some anterior cells close to the border ( Figure 5—figure supplement 1 ) . As a consequence posterior cells can mix with juxtaposing anterior cells , in effect eliminating the boundary . We believe the behavior of the posterior cells derives from a diminution of en activity , responsible for posterior identity and for maintaining the A/P border ( Morata and Lawrence , 1975 ) . In turn , it appears to be caused by two factors: attenuation of the control exerted by the Pc-G and trx-G genes ( Figure 5 ) , and up regulation of the JNK pathway ( Figure 4A ) . These two pathways are known to be involved in controlling en activity ( Busturia and Morata , 1988; Breen et al . , 1995; Gettings et al . , 2010 ) . We observe that the collapse of the boundary is followed by a very rapid reconstruction . The fact that there is a functional A/P border after 48 hr of ablation and without allowing time for recovery ( Figure 3—figure supplement 1A , also Figure 4E ) , strongly suggests an immediate re-establishment . The reconstruction must result from the acquisition of new anterior or posterior identities by the cells close to the border . We believe that the mechanism behind this reprogramming is the phenomenon of gene induction by neighbors that we report here ( Figure 6A–H′ ) : islands of anterior cells in which en is originally inactive are induced to activate the gene when surrounded by en-expressing cells . This phenomenon provides an explanation for the changes of identity observed in the proximity of the A/P border of regenerating discs that we outline in Figure 6I–L . In the developmental context of the wing disc , the cells around the A/P border only have the option of being en-on or en-off , depending on the active genotype of surrounding cells . This non-autonomous induction may also have general implications about how major genetic/developmental decisions occur during development . In Drosophila many developmental decisions are taken collectively by small groups of cells: the compartmental subdivisions occur in groups of cells ( Garcia-Bellido et al . , 1973 ) . Similarly , the phenomenon of ‘transdetermination’ in which cells change their original segmental determination in transplantation experiments ( Gehring , 1967 ) , is not a clonal but a collective decision . These examples are cases of group decisions , in which the local context plays a role . These group interactions to generate a coherent pattern of developmental resemble the ‘community effect’ described by John Gurdon years ago ( Gurdon , 1988 ) . One could speculate that induction by neighbors may be part of a group decision process during normal development . The initial activation of en during embryogenesis is induced by the transcription factors eve and ftz ( DiNardo et al . , 1988 ) . It is possible that not all the cells respond similarly to Eve or Ftz—it is known that the initial en expression is inhomogeneous ( DiNardo et al . , 1985 ) . The secondary induction via the local context would ensure the coherent activation of engrailed in the entire group . There is a recent report suggesting the possibility that the En protein may act as a short-range signal ( Layalle et al . , 2011 ) . It raises the possibility that the En product itself may activate en non-autonomously . The Drosophila stocks used were hh-Gal4 ( Tanimoto et al . , 2000 ) ; ap-Gal4 ( Calleja et al . , 1996 ) ; ET-40 ( named in the text as Psc-QF ) and QUAS-mtdTomato ( Potter et al . , 2010 ) ; dpp-LHG-86Fb ( Yagi et al . , 2010 ) ; lexO-hid ( a kind gift of Ainhoa Pérez-Garijo and Hermann Steller ) ; puc-lacZ line ( pucE69 ) and UAS-puc14C line ( Martin-Blanco et al . , 1998 ) ; act-Gal4 , UAS-GFP , UAS-hid7 ( named in the text as UAS-hid ) , UAS-Flp , UAS-LacZ , tub-Gal80TS , arm-LacZ FRT80B and ubi-GFP FRT80B ( Bloomington Drosophila Stock Center ) ; UAS-en ( Guillen et al . , 1995 ) ; hs-Flp112 and act>stop>lacZ ( Struhl and Basler , 1993 ) ; ubiP63E>stop>GFP ( Evans et al . , 2009 ) ; tubP>stop>Gal80 ( Bohm et al . , 2010 ) ; mew-YFP ( named in the text as PS1α-YFP , Kyoto Stock Center CPTI-001678 ) ; Pc3 ( Lewis , 1978 ) ; trxE2 ( Kennison and Tamkun , 1988 ) . In the experiments using the Gal4/UAS system the general rule was to use only two UAS transgenes . This was to avoid the possibility of titrating the amount of Gal4 protein available for the UAS vectors . We have used a cell killing method very similar to that previously described ( Bergantinos et al . , 2010; Smith-Bolton et al . , 2009; Herrera et al . , 2013 ) . For ablation of posterior compartments , we used the genotypes UAS-hid tub-Gal80TS/+; hh-Gal4/act>stop>lacZ UAS-Flp ( experimental ) and UAS-GFP tub-Gal80TS/+; hh-Gal4/act>stop>lacZ UAS-Flp ( control ) . To ablate dorsal compartment the hh-Gal4 insertion was substituted by an ap-gal4 line . The tub-Gal80TS vector suppresses Gal4 activity at 17°C , but it is ineffective at 29°C thus allowing to manipulate Gal4 activity by shifting temperature between 17°C and 29°C . Larvae were cultured at 17°C by 7 days ( Figure 1—figure supplement 1A ) . At this point ( approximately the transition between second and third instar ) , the temperature of the cultures was raised to 29°C , thus inactivating the Gal80TS . There are two consequences of this: ( 1 ) an induction of hid overexpression , which induces massive apoptosis; ( 2 ) induction of Flipase recombinase , which in turn drives recombination of the act>stop>lacZ cassette , thus labeling all the cells of the Gal4 domain with indelible activity of the LacZ marker , which encodes the ßgal protein ( Figure 1—figure supplement 1B ) . After 48 hr of ablation treatment , the cultures were returned to 17°C , activating again the Gal80TS repressor thus allowing the recovery of the tissue . In the regenerated structure , we can distinguish the cells that belonged to the original domain ( ßgal positive ) from those cells originated in other compartment ( ßgal negative ) ( Figure 1—figure supplement 1C ) . We used larvae of genotypes ubiP63E>stop>GFP , tub-Gal80TS/ UAS-Flp , UAS-puc; dpp-LHG , hh-Gal4/ lexO-hid ( experimental ) and ubiP63E>stop>GFP , tub-Gal80TS/ UAS-Flp; dpp-LHG , hh-Gal4/ lexO-hid ( control ) . In these genotypes hid is overexpressed in the Dpp domain , the anterior region close to the A/P border , while puc ( only present in the experimental genotype ) and Flipase are overexpressed in the hh domain ( posterior compartment ) . The Flipase in turn induces recombination of the ubiP63E>stop>GFP casette , thus labeling the posterior compartment lineage . The tub-Gal80TS vector suppresses both LHG and Gal4 activity at 17°C . Larvae were cultured at 17°C for 7 days and then the temperature was raised to 29°C to inactivate Gal80TS for 2 days . In this experiment , the transgressions were scored at the end of this ablation period . We have measured the size of transgression size as the total surface , measured in µm2 , of a patch of cells with a compartmental identity ( activity of genes as en or ci ) different from its lineage origin ( e . g . Hh lineage-negative/Ci-negative or Hh lineage-positive/Ci-positive ) . To induce neutral clones in posterior compartment-ablated discs , we heat-shocked ( 12 min 37°C ) larvae of genotype hs-Flp; UAS-hid tub-Gal80TS/+; hh-Gal4/ ubiP63E>stop>GFP . As controls we induced clones in hs-Flp; UAS-hid tub-Gal80TS/+; ubiP63E>stop>GFP/+ larvae in which there is no Hid activation . To analyze clonal growth during the ablation period , the heat-shock was delivered at the beginning of the ablation period and the discs were fixed after 48 hr , at the end of ablation . To analyze the growth during the recovery period the heat shock was delivered at the end of Hid treatment and the discs were fixed after 72 hr . To analyze in detail the transgression of the A/P boundary by clones generated either in the A or the P compartment we used a ‘twin clones’ method , marking the two cellular progenies of a mitotic recombination event . We generated the double LacZ clones and double GFP twins in the genotype hs-Flp112; UAS-hid tub-Gal80TS/+; ubi-GFP FRT80B hh-Gal4/arm-LacZ FRT80B . The clones were induced by heat-shock ( 10 min 37°C ) at the beginning of the Hid treatment . Discs were fixed after 48 hr of Hid induction and 24 additional hours of recovery period . To generate marked clones over-expressing engrailed we used the construction act>CD2 y+>Gal4 ( in the following act>stop>Gal4 ) ( Ito et al . , 1997 ) . In discs of genotype yw hs-Flp; act>stop>Gal4 UAS-GFP/en-LacZ; UAS-en/+ a long heat shock of 30 min gave rise to a very large number of en-expressing clones that are labeled with GFP . For low density clones induction , we used 10 min heat shocks . We co-expressed these genes with UAS-GFP for labeling the clones . We detected the endogenous engrailed gene activity thanks to the insertion en-LacZ ( ryxho25 ) ( Hama et al . , 1990 ) . The genotype used was yw hs-Flp act-Gal4 UAS-GFP; tub-Gal80TS/en-LacZ; tubP>stop>Gal80/UAS-en . In this genotype , a temperature shift from 17°C to 29°C removes the activity of the thermo-sensitive form of Gal80 thus allowing activation of the UAS-en transgene in all the cells of the disc ( the act-Gal4 line confers ubiquitous expression ) , except in the clones resulting from recombination of the tubP>stop>Gal80 cassette . These clones were induced by a heat shock prior to the temperature shift and contain the normal form of the Gal80 protein . The larvae were cultured at 17°C and the heat shock ( 12 min at 37°C ) was administrated 7 days after egg lying . After 24 additional hours at 17°C , the larvae were shifted to 29°C and the discs were fixed 2 days later . Immunostaining was performed as described previously ( Shlevkov and Morata , 2012 ) . Images were captured with Leica TCS SPE and Zeiss LSM510 Vertical confocal microscopes . Images were processed with Fiji-ImageJ and Adobe Photoshop CS4 software using Student’s t for significance tests . Statistical analyses were performed with Microsoft Excel software . The following primary antibodies were used: rabbit anti-Caspase 3 ( Roche , Basilea , Switzerland ) 1:50; mouse anti-ß-Galactosidase ( Promega , Madison , WI ) ; rat anti-Cubitus interruptus ( 2A1; Hybridoma Bank , Iowa City , IA ) 1:25; mouse anti-Engrailed ( 4D9; Hybridoma Bank ) 1:50; mouse anti-Patched ( Apa-1; Hybridoma Bank ) 1:50; rabbit anti-H3K4me3 ( 07-473; Millipore , Billerica , MA ) 1:50 . Fluorescently labeled secondary antibodies ( Molecular Probes , Eugene , OR ) were used in a 1:200 dilution . TO-PRO3 ( Invitrogen , Carlsbad , CA ) was used in a 1:600 dilution to label nuclei; Phalloidin-Cy5 ( 65906; Sigma , St . Louis , MO ) was used in a 1:200 dilution to label the F-actin network ( thus the cell membranes ) .	When cells or tissues are damaged , animals can often regenerate the affected tissues . In an effort to identify the genes and mechanisms that are involved in this regeneration process , researchers often perform experiments on relatively simple organisms or systems . These experiments frequently involve the amputation of specific cells or organs so the researchers can observe and manipulate the events that occur during the subsequent regeneration . One such model organism is the fruit-fly Drosophila . Inside the Drosophila larva are structures called imaginal discs , which are precursors to parts of the outer cuticle of the adult fly . Each imaginal disc contains two main boundaries , dividing it into anterior/posterior and dorsal/ventral compartments: posterior cells , for example , express a gene called engrailed to produce the relevant protein , whereas anterior cells do not; likewise , the gene apterous is expressed in dorsal cells but not ventral cells . These genes , engrailed and apterous , are the key factors that determine the developmental features–and hence the identity—of the posterior and the dorsal cells respectively . Herrera and Morata investigated how cells regenerate when parts of the imaginal disc are destroyed , using a genetic technique that causes high levels of programmed cell death in either the posterior or the dorsal compartments of the disc . Destroying most of the cells in either of these compartments in the imaginal disc leads to a temporary breakdown of the corresponding boundary , which is then rapidly reconstructed . During this reconstruction process , some of the imaginal disc cells are reprogrammed: for example , if the cells in the posterior compartment are destroyed , some anterior cells take on a posterior identity . This reprogramming occurs because the cell destruction disrupts the way that the expression of genes such as engrailed and apterous is controlled by other genes . Neighboring cells can also control gene expression , and therefore cell identity . Cells that express engrailed , for example , cause their neighbors to express engrailed too . This process is likely to involve group interactions that might help the distinct compartments in the imaginal disc to form by ensuring that adjacent cells develop in the same way . Similar processes may also occur as part of the normal development of organisms .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine" ]	2014	Transgressions of compartment boundaries and cell reprogramming during regeneration in Drosophila
Drosophila syncytial nuclear divisions limit transcription unit size of early zygotic genes . As mitosis inhibits not only transcription , but also pre-mRNA splicing , we reasoned that constraints on splicing were likely to exist in the early embryo , being splicing avoidance a possible explanation why most early zygotic genes are intronless . We isolated two mutant alleles for a subunit of the NTC/Prp19 complexes , which specifically impaired pre-mRNA splicing of early zygotic but not maternally encoded transcripts . We hypothesized that the requirements for pre-mRNA splicing efficiency were likely to vary during development . Ectopic maternal expression of an early zygotic pre-mRNA was sufficient to suppress its splicing defects in the mutant background . Furthermore , a small early zygotic transcript with multiple introns was poorly spliced in wild-type embryos . Our findings demonstrate for the first time the existence of a developmental pre-requisite for highly efficient splicing during Drosophila early embryonic development and suggest in highly proliferative tissues a need for coordination between cell cycle and gene architecture to ensure correct gene expression and avoid abnormally processed transcripts . Timing and coordination of biological processes is crucial for cellular homeostasis and normal development . Drosophila melanogaster embryonic development starts with thirteen nuclear divisions without cytokinesis ( syncytial blastoderm ) , these divisions being among the fastest known for any animal embryonic system ( Foe and Alberts , 1983 ) . Drosophila syncytial blastoderm formation relies on the maternally encoded gene products loaded into the egg during oogenesis ( Tadros and Lipshitz , 2009 ) . After fertilization , as nuclei enter interphase 14 , the maternal to zygotic transition ( MZT ) occurs in which the soma suddenly becomes transcriptionally active and many of the maternally encoded gene products are rapidly degraded ( Anderson and Lengyel , 1979; McKnight and Miller , 1976; Yasuda et al . , 1991 ) . While the major burst of zygotic transcription occurs once the nuclei arrest in interphase 14 , there is an initial wave of zygotic gene expression during the syncytial nuclear divisions 8–13 ( Pritchard and Schubiger , 1996; ten Bosch et al . , 2006 ) . Due to the extreme speed of syncytial nuclear divisions , a limitation to the size of early zygotic transcriptional units has been suggested ( McKnight and Miller , 1976; Rothe et al . , 1992; Shermoen and O’Farrell , 1991 ) . Consistently , approximately 70% of early zygotic genes are small in size and intronless ( De Renzis et al . , 2007 ) . As only 20% of Drosophila genes are intronless , it has been proposed that small intronless genes have an important selective advantage for transcription during the syncytial blastoderm formation ( De Renzis et al . , 2007 ) . In yeast , Drosophila , and human cells , pre-mRNA splicing is mostly co-transcriptional ( Ameur et al . , 2011; Khodor et al . , 2011 ) , with in vivo splicing rates being in the order of 30 s to approximately 3 min once the intron is transcribed ( Alexander et al . , 2010; Huranova et al . , 2010; Schmidt et al . , 2011 ) . As most early zygotic transcripts are intronless ( De Renzis et al . , 2007 ) , syncytial blastoderm interphases can be as short as 4 to 5 min , and given the fact that mitosis inhibits splicing ( Shin and Manley , 2002 ) , we hypothesized that further to the selective pressure for small transcriptional units , there is also a pressure to avoid pre-mRNA splicing during early zygotic expression . We isolated two mutant alleles for a subunit of the NTC/Prp19 complexes , known to be important for efficient spliceosome activation , which specifically impaired pre-mRNA splicing of early zygotic but not maternally encoded transcripts . We showed that the differential splicing defects were not related to any particular structure/sequence of the early zygotic transcripts or differential association of spliceosomal components to the NTC/Prp19 complexes . Ectopic maternal expression of an early zygotic transcript in a mutant background was sufficient to suppress its splicing defects , suggesting that they were dependent on the developmental context of gene expression . We reasoned that constraints on pre-mRNA splicing are present during Drosophila early embryonic development . Consistently , a small early zygotic transcript with four introns was poorly spliced in wild-type embryos . Such constraints on pre-mRNA splicing are a likely explanation why most early zygotic genes are intronless and suggest that highly proliferative tissues need coordination between cell cycle and gene architecture for correct gene expression and avoidance of abnormally processed transcripts . Our results strongly argue in favor of a developmental pre-requisite for highly efficient splicing during fast development , therefore we propose that the requirement for overall splicing efficiency is likely to vary during development . Previously we isolated a collection of maternal mutants defective in blastoderm cellularization and/or germ-band extension ( Pimenta-Marques et al . , 2008 ) . Complementation group 7 contained two different mutant alleles with similar defects in blastoderm cellularization . Through deficiency mapping and a candidate gene approach we concluded that both were allelic to the uncharacterized coding gene CG6197 ( Flybase ) . To confirm the mutants’ identity , we rescued their zygotic lethality , female sterility ( germ-line clones ) , and blastoderm cellularization defects ( maternal mutant embryos ) using a genomic fragment construct that contained a wild-type copy of CG6197 ( Figure 1—figure supplement 1A , data not shown ) . Both isolated alleles of CG6197 showed identical phenotypes: maternal mutant embryos ( hereafter referred to as mutant embryos ) showed normal syncytial nuclear divisions ( Figure 1A , B ) but subsequently failed to elongate the cortical nuclei , which became mislocalized during blastoderm cellularization ( Figure 1C–F , quantification in Figure 1G ) . The blastoderm cellularization phenotype was remarkably similar to that described for kugelkern/charleston mutant embryos ( Brandt et al . , 2006; Pilot et al . , 2006 ) . Based on the observed phenotypes , we named the corresponding gene fandango , after the Iberian folk dance . 10 . 7554/eLife . 02181 . 003Figure 1 . Drosophila Fandango/Xab2 is required for blastoderm cellularization . ( A and B ) Panels show embryos with normal syncytial blastoderm nuclear divisions in control embryos ( hs-FLP; FRT42B ) ( A ) and fand1 germ-line clone embryos ( hs-FLP; FRT42B fand1 , maternal mutant ) ( B ) . Embryos were stained for DNA ( green ) and p-Tyrosine ( red ) . ( C–F ) Panels show blastoderm cellularized embryos . Control embryos showed normal epithelial architecture with elongated nuclei and columnar cell shape ( C ) . fand1 germ-line clone mutant embryos showed abnormal epithelial architecture , the cortical nuclei failed to elongate and became mislocalized ( D ) . ( E and F ) Magnification of C and D , respectively . Embryos were stained for Slam ( green ) , Neurotactin ( red ) , and DNA ( blue ) . ( G ) Quantification of fandango maternal mutant embryo phenotype during blastoderm cellularization . Early cellularization: control: 100% normal ( n = 44 ) , fand1: 100% normal ( n = 49 ) ; mid cellularization: control: 100% normal ( n = 25 ) , fand1: 28% normal ( n = 21 ) ; late cellularization: control: 100% normal ( n = 42 ) , fand1: 0% normal ( n = 38 ) . ( H ) Maternally controlled oogenesis was normal in fandango mutant clones . Absence of endogenous nGFP ( green ) indicated that the cells were homozygous for fand1 mutation . Ovaries were stained for F-actin ( red ) and WGA ( blue ) . ( I ) Western blot of whole protein extracts from embryos and ovaries mutant for fand1 and fand2 alleles ( germ-line clones ) showed a clear reduction in Fandango protein levels compared to control tissues . It should be noticed that due to experimental constraints the total protein extracts from mutant ovaries included not only signal from mutant germ-line cells ( homozygous for fand1 ) , but also the tightly associated heterozygote somatic follicle cells . α-Tubulin was used as a loading control . ( J ) Real-time qPCR analysis showed no significant differences in fandango mRNA levels between control and fand1 embryos during development ( Two-way ANOVA p>0 . 05 ns . ) . fandango mRNA levels were normalized with β-actin mRNA levels . ( K–N ) in situ hybridization for nanos RNA ( maternal ) and even-skipped RNA ( early zygotic ) in blastoderm cellularized embryos . Both control ( K ) and fand1 mutant ( L ) embryos showed normal nos localization pattern in the pole cells . fand1 embryos ( N ) showed A–P patterning defects of eve compared to control embryos ( M ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 00310 . 7554/eLife . 02181 . 004Figure 1—figure supplement 1 . fandango mutant alleles contain changes in highly conserved amino acids . ( A ) Panels show the rescue of the transheterozygous zygotic lethality of fand1 and fand2 alleles . Only transheterozygous mutant females carrying the wild-type copy of fandango ( wt-fandango; FRT42B , fand1/fand2 ) were viable ( dark gray box ) . Transheterozygous mutant males without receiving the genomic fragment ( FRT42B , fand1/fand2 ) died ( light gray box ) . In the reciprocal cross , both females and males transheterozygous mutants carrying the wild-type copy of fandango , were viable ( dark gray box ) . ( B ) Fandango has multiple copies of a tetratricopeptide repeat ( TPR ) motif , a protein–protein interaction module found in a number of functionally different proteins . A scheme displaying the distribution of conserved TPR protein domains in Fandango ( top ) . Mutations of fand1 and fand2 alleles affect highly conserved amino acids of the TPR domains 7 and 6 , respectively . fand1 contained a missense point mutation changing an alanine to a valine at aminoacid position 401 ( A401V ) and fand2 contained a microdeletion which resulted in loss of six conserved amino acids from position 355 to 360 ( Δ355–360 ) . Partial alignment of Fandango ( Drosophila melanogaster CG6197 , ref . NP_610891 . 1 ) with orthologous Xab2 ( Homo sapiens , ref . NP_064581 . 2 ) , Xab2 ( Mus musculus , ref . NP_080432 . 1 ) , Xab2 ( Danio rerio , ref . NP_001038248 . 1 ) , C50F2 . 3 ( Caenorhabditis elegans , ref . NP_491250 . 1 ) , cwf3 ( Schizosaccharomyces pombe , ref . NP_596612 . 1 ) , and SYF1 ( Saccharomyces cerevisiae , ref . NP_010704 . 1 ) ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 004 fandango encodes the Drosophila ortholog of yeast SYF1 ( synthetic lethal with cdc41 ) ( Russell et al . , 2000 ) and human XAB2 ( XPA binding protein 2 ) ( Nakatsu et al . , 2000; Kuraoka et al . , 2008 ) . These proteins were described as subunits of the NTC/Prp19 complexes , which are important for spliceosome stabilization and activation ( Chan et al . , 2003; Chang et al . , 2009; Hogg et al . , 2010 ) . Fandango protein has multiple tetratricopeptide repeat ( TPR ) motifs , which is a protein–protein interaction module ( Zeytuni and Zarivach , 2012 ) . Sequencing both alleles of fandango ( fand1 and fand2 ) revealed distinct mutations within the fandango open reading frame ( ORF ) . fand1 contained a missense point mutation in a highly conserved residue within TPR domain VII ( from an alanine to a valine; A401V ) , whereas fand2 contained a microdeletion of 18 nucleotides within TPR domain VI , which deleted six conserved amino acids from position 355 to 360 ( Figure 1—figure supplement 1B ) . In total protein extracts , both fand1 and fand2 mutant embryos showed a significant reduction in Fandango protein levels compared to control ( Figure 1I ) . fandango mRNA levels , analyzed by real-time qPCR , were similar between control and fand1 mutant embryos ( Figure 1J ) , suggesting that the mutation did not alter the stability of the encoding pre-mRNA . As noted above fandango maternal mutant embryos and kugelkern ( kuk ) mutant embryos showed remarkably similar blastoderm cellularization defects ( Brandt et al . , 2006; Pilot et al . , 2006 ) . Since fandango encodes a protein whose yeast and human orthologs are required for efficient spliceosome activity , we hypothesized that Fandango was required for splicing of kuk transcripts . kuk encodes two different transcripts , which vary in intron size ( Figure 2A ) . Both transcripts are predicted to encode the same protein . Analysis of publicly available modENCODE transcriptome datasets ( Graveley et al . , 2011 ) suggested that the large kuk transcript was maternally expressed , whereas the small kuk transcript was only expressed zygotically . Through RT-PCR analysis we confirmed that similarly to control maternal genes ( nanos and oskar ) the large kuk transcript was maternally expressed ( being present in unfertilized eggs ) , whereas the small kuk transcript was exclusively zygotically expressed ( being present only in fertilized eggs ) as the case of well-known early zygotic genes ( even-skipped and krüppel ) ( Figure 2—figure supplement 1A ) . 10 . 7554/eLife . 02181 . 005Figure 2 . Splicing of early zygotic but not maternally encoded pre-mRNAs is affected in fandango mutants . ( A ) The kugelkern ( kuk ) locus encodes two transcripts of different size , kuk-lt containing a large intron and kuk-st with a short intron . Orientation and position of primers used for splicing analysis is indicated ( arrows ) . ( B ) RT-PCR analysis of kuk transcripts . Control embryos yielded PCR products in the size predicted for the properly spliced forms of both kuk transcripts using exon–exon ( e–e ) primers ( green dots , kuk-lt: 431 bp and kuk-st: 437 bp ) . fandango maternal mutant embryos ( fand1 and fand2 alleles ) showed splicing defects only in the kuk-st transcript; PCR products were detected by e–e primers in the size expected for intron retention ( red dots , kuk-st: 596 bp ) and by intron–exon ( i–e ) primers ( kuk-st: 474 bp ) . Splicing of the kuk-lt was not affected in fandango mutant background; PCR products were only detected with e–e primers in the predicted size for the correctly spliced pre-mRNA ( green dots , kuk-lt: 431 bp ) . ‘No RT’ controls ( only total RNA as template ) yielded no amplification , meaning there was no contamination with genomic DNA in the samples tested . ( C ) RT-PCR analysis of maternal and early zygotic genes . Maternal transcripts were properly spliced , in both , control and fand1 mutant embryos; PCR products were only detected using e–e primers ( green dots , grk: 527 , nos: 581 , osk: 762 bp ) . In contrast , early zygotic transcripts were correctly spliced only in control embryos ( green dots , kr: 559 , eve: 828 , ftz: 753 bp ) . fand1 mutant embryos yielded PCR products in the size predicted for intron retention with e–e primers ( red dots , kr: 932 , eve: 899 , ftz: 900 bp ) and with i–e primers ( kr: 629 , eve: 720 , ftz: 595 bp ) . All PCR bands showed in the panels were cloned and sequenced to confirm their identity . Green dots indicate correctly spliced transcripts , red dots indicate unspliced transcripts ( intron retention ) . ( D ) RNA-Seq data confirmed that zygotic but not maternally encoded transcripts displayed a large fraction of splicing defects ( intron retention ) in fand1 mutant embryos . The panel shows box plot of the distribution of numbers of reads per bp relative to the total number of reads falling inside a 100 bp window centered around the 5′splice sites of zygotic ( n = 408 splice sites from 270 genes ) or maternal genes ( n = 5876 splice sites from 2048 genes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 00510 . 7554/eLife . 02181 . 006Figure 2—figure supplement 1 . Splicing of early zygotic but not maternally encoded pre-mRNAs is affected in fandango mutants . ( A ) RT-PCR analysis of kuk transcripts from unfertilized and fertilized eggs . kuk-lt is maternal and kuk-st only zygotically expressed . kuk-st ( 437 bp ) transcripts are only detected in fertilized eggs , as are other well-known early zygotic transcripts ( kr: 559 , eve: 828 bp ) . kuk-lt ( 431 bp ) is amplified from both fertilized and unfertilized eggs , as are other known maternal transcripts ( nos: 581 , osk: 762 bp ) . ( B ) RT-PCR analysis of kuk transcripts with a specific reverse primer for kuk-st ( kuk-st4 ) . Control embryos yielded PCR products with the size expected for properly spliced kuk transcripts using e–e primers ( green dots , kuk-lt: 431 bp and kuk-st: 2257 bp ) . fand1 maternal mutant embryos showed splicing defects in kuk-st; PCR products were detected in the size expected for intron retention with e–e primers ( red dot , kuk-st: 2413 bp ) and by i–e primers ( kuk-st: 2293 bp ) . Splicing of the kuk-lt was not affected in fandango mutant background; a PCR product was only detected with e–e primers in the expected size for correctly spliced pre-mRNAs ( green dot , kuk-lt: 431 bp ) . ( C ) RT-PCR analysis of kuk-st showed the rescue of splicing defects observed in both fandango alleles by a genomic fragment construct derived from the third chromosome that contained a wild-type copy of fandango . Embryos analyzed were laid by GLC females FRT42B fand1/CyO; wt-fandango or FRT42B fand2/CyO; wt-fandango . GLC FRT42B and mutant GLC fand1 and fand2 embryos were used as controls . ( D ) RT-PCR analysis of maternally encoded transcripts from wild-type and fandango mutant ovaries ( germ-line clones ) failed to detect any splicing defects . The PCR products detected in both samples were of the size predicted for properly spliced pre-mRNAs ( green dots , nos: 581 , osk: 762 , stg: 614 bp ) . All PCR products shown in these panels were cloned and sequenced to confirm their identity . Green dots indicate correctly spliced transcripts and red dots indicate unspliced transcripts ( intron retention ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 00610 . 7554/eLife . 02181 . 007Figure 2—figure supplement 2 . Early zygotic but not maternally encoded pre-mRNAs shows significant intron retention in fandango mutants . ( A ) Zygotic genes display intron retention in the fand1 mutant , while maternal genes do not . This panel shows Integrative Genomics Viewer ( IGV ) screenshots of RNA-Seq read coverage from tophat alignments of a selected set of known early zygotic and maternally expressing genes . Within each gene , the image scale is identical for both fand1 and control . ( B ) Zygotic genes in fand1 mutant present clear evidence of intron retention , independent of intron size . The panel displays box plot distributions of percentage intron retention of exon–intron boundaries of early zygotic and maternal genes in fand1 and control . Exon–intron boundaries were divided in bins , by intron size ( less than 100 bp; from 100 to 500 bp; and greater than 500 bp ) . Sizes were selected empirically to have comparable dataset sizes in each bin ( 150–200 boundaries for zygotic genes , 1000–3000 for maternal genes ) . The frequency of intron retention was determined by comparing the number of unsplit reads overlapping the splice site with the number of reads that showed an exon–exon split ( see ‘Materials and methods’ for more details ) . ( C ) Splice sites in zygotic and maternal genes presented the same characteristic sequence pattern ( 5′ GT; 3′ AG ) . ( D ) Zygotic genes and maternal genes ( see ‘Materials and methods’ for more details ) did not reveal any distinguishing features in terms of exon–intron structure . The exon frequency is close to that expected from a random distribution , roughly 50% ( with the obvious exception of gene endings—TSS and TTS ) . For comparison , the 59 early zygotic genes described by ( De Renzis et al . , 2007 ) , 70% of which are intronless , display a very distinct , non-random , pattern . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 007 To investigate by RT-PCR whether Fandango was required for splicing of kuk pre-mRNAs , specific sets of primers ( exon–exon , e–e; intron–exon , i–e ) were designed for each kuk transcript , taking advantage of a longer 3′UTR in the small kuk transcript ( Figure 2A ) . Surprisingly , whereas fandango embryos showed significant splicing defects of the small zygotic kuk transcript , the large maternal kuk transcript was correctly spliced ( Figure 2B; Figure 2—figure supplement 1B ) . Splicing defects were fully rescued by a genomic fragment construct that contained a wild type copy of fandango ( Figure 2—figure supplement 1C ) . The differential requirement of Fandango for splicing of kuk transcripts prompted us to investigate more than 20 other maternal and early zygotic genes . RT-PCR analysis of fandango embryos invariably showed splicing defects of early zygotic but not maternally encoded transcripts ( Figure 2C , data not shown ) . High-throughput transcriptome sequencing ( RNAseq ) confirmed that splicing of early zygotic but not maternally encoded gene products was affected in fandango embryos ( Figure 2D , Figure 2—figure supplement 2A ) . Maternal transcripts , whose intron size was equivalent to those observed in early zygotic transcripts , were unaffected ( Figure 2—figure supplement 2B ) , which showed that Fandango was not specifically rate limiting for splicing of small introns . Comparison analysis of 5′ and 3′ splice site consensus sequences between maternal and zygotic pre-mRNA transcripts showed no significant differences ( Figure 2—figure supplement 2C ) and the two populations of transcripts displayed a similarly heterogeneous exon–intron structure ( Figure 2—figure supplement 2D ) . RT-PCR analysis of maternally encoded transcripts from wild-type and fandango mutant ovaries ( germ-line clones ) also failed to detect splicing defects ( Figure 2—figure supplement 1D ) . This suggested that the absence of splicing defects of maternally encoded transcripts in fandango embryos was not due to specific degradation of unspliced transcripts during oogenesis . The differential requirement of Fandango for splicing of early zygotic encoded transcripts is fully consistent with the observation that maternally controlled oogenesis , primordial germ-cell formation , and syncytial nuclear divisions were normal in fandango mutants ( Figure 1A , B , H , K , L ) , whereas the first detectable phenotype only occurred during zygotically controlled blastoderm cellularization ( Figure 1C–F ) . Despite the fact that clonal analysis of the female germ line for both alleles of fandango showed normal oogenesis and egg laying ( Figure 1H ) ( data not shown ) , Fandango protein levels were significantly reduced in the mutant ovaries ( germ-line clones ) ( Figure 1I ) . fandango embryos also failed to initiate germ-band extension after blastoderm cellularization ( data not shown ) . It was previously shown that anterior–posterior ( A–P ) patterning is required for germ-band extension ( Zallen and Wieschaus , 2004 ) . Consistently , fandango embryos showed A–P patterning defects in the early zygotic pair-rule gene even-skipped ( Figure 1M , N ) . The highly conserved NTC/Prp19 and NTC/Prp19-related complexes are essential for pre-mRNA splicing as they facilitate the formation and progression between distinct spliceosome conformations during the splicing reaction ( Chan et al . , 2003; Hogg et al . , 2010 ) . Endogenous Fandango and Prp19 physically interacted in the early embryo ( Figure 3A ) . Moreover , both endogenous Fandango and Prp19 physically interacted with endogenous ISY1 and CDC5L ( Figure 3A ) , confirming that Fandango is a bona fide subunit of Drosophila NTC/Prp19 complexes . Immunoprecipitation of Myc-tagged Fandango and Myc-tagged Prp19 from embryonic protein extracts also identified an identical group of interacting proteins ( Table 1; Supplementary file 1 ) . Whereas Myc-Fandango mostly interacted with the NTC/Prp19-related complex subunits , Myc-Prp19 interacted principally with the NTC/Prp19 complex subunits . This illustrated that , as in humans , distinct but interacting NTC/Prp19 complexes exist in Drosophila , in agreement with the recent suggestion that a remarkable degree of conservation of distinct splicing complexes exists among metazoans ( Herold et al . , 2009 ) . 10 . 7554/eLife . 02181 . 008Figure 3 . Fandango physically interacts with a similar group of splicing proteins during oogenesis and embryogenesis . ( A ) Pull down assay from nuclear-enriched protein extracts using a polyclonal antibody of Prp19 . Endogenous Prp19 interacts physically with Fandango and other subunits of the NTC/Prp19 complexes ( ISY1 and CDC5L ) . Pre-immune serum was used in the control . ( B ) Size-exclusion chromatography of control and fand1 mutant protein extracts from 0–3 hr embryo collections using a Superose 6 10/300 column . After separation , each fraction was analyzed by Western blot . NTC/Prp19 complexes subunits ( Prp19 , Fandango , and ISY1 ) were part of a ∼600–800 kDa complex and also co-purified in a significantly larger complex ( fraction 4 and 5 ) . fand1 mutant protein extracts showed a significant reduction in levels of Fandango and ISY1 subunits and a size reduction of the Prp19-positive ∼600–800 kDa complex . ( C ) Western-blot analysis of total protein extracts from ovaries ( left ) and 0–3 hr embryos ( right ) from control and both fandango alleles , showed a reduction of Fandango and ISY1 protein levels in both tissues . Protein levels of Prp19 and CDC5L were not affected . α-Tubulin was used as loading control . Fandango Western blot is the same as shown in Figure 1I . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 00810 . 7554/eLife . 02181 . 009Table 1 . LC-MS analysis of co-immunoprecipitation assays from ovaries and embryosDOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 009DrosophilaHuman/yeastFandango-mycPrp19-mycCGgeneovariesembryosembryosrep1rep2rep1rep2rep1rep2prp19 complex CG5519prp19PRP19/Prp19+++++++++ CG6905cdc5-likeCDC5L/Cef1+++++++++ CG1796Tango4PLRG1/Prp46++++++ CG4980-BCAS2/Snt309–+--++ CG12135c12 . 1CWC15/cwc15+––-––Prp19 related CG6197FandangoXab2/Syf1++++++++++++++ CG31368–AQR/–++++++++++++++ CG4886cyp33PPIE/–+++++++++ CG9667–ISY1/ISY1++++++ CG8264Bx42SNW1/Prp45++++++ CG14641–RBM22/Cwc2–++++– CG3193CrnCRNKL1/Clf1–+––+– CG13892cyplPPIL1/-–––+–– CG1639l ( 1 ) 10BbBUD31/Bud31––––+–Co-immunoprecipitations were performed using total protein extracts from the different tissues expressing Myc-tagged Fandango or Myc-tagged-Prp19 . Human and yeast homologues and the different sub-complexes are shown as described in ( Herold et al . , 2009 ) . ( − ) , ( + ) , ( ++ ) , ( +++ ) correspond to 0 , 1–9 , 10–19 , and >20 non-repeated peptides respectively . None of the proteins shown were detected in the negative controls ( for detailed LC-MS analysis see Supplementary file 1 ) . The differential requirements of Fandango for pre-mRNA splicing of maternal and early zygotic transcripts potentially suggest distinct interactions between Fandango and other splicing proteins during oogenesis and early embryonic development . Nevertheless , immunoprecipitation of Myc-Fandango specifically expressed in the female germ line during oogenesis and in the early embryo identified a virtually identical group of interacting proteins: mostly subunits of the NTC/Prp19-related complex , and to a lesser extent , subunits of the NTC/Prp19 complex ( Table 1; Supplementary file 1 ) . These results showed that Fandango physically interacts with a similar group of splicing proteins during oogenesis and in the early embryo . To better understand the splicing defects observed in fandango embryos , we investigated if the integrity of NTC/Prp19 complexes was affected in this mutant . Size-exclusion chromatography showed detectable changes in the integrity of NTC/Prp19 complexes in fandango embryos ( Figure 3B ) , with a significant reduction in the levels of ISY1 protein ( Figure 3C ) . ISY1 is a NTC/Prp19-related complex subunit ( Figure 3A ) . The loss of integrity of the ISY1-positive ∼600–800 kDa NTC/Prp19 complex ( Figure 3B ) and concomitant reduction in the stability of some of their subunits , most likely impaired efficient activation of the spliceosome ( Villa and Guthrie , 2005 ) and were likely explanations for the splicing defects observed in fandango embryos . In agreement with the suboptimal spliceosome activation hypothesis , intron retention was the main splicing defect of early zygotic transcripts in fandango embryos ( Figure 2B , C , Figure 2—figure supplement 2B; data not shown ) . Levels of ISY1 were similarly affected in fandango mutants during oogenesis and in the early embryo ( Figure 3C ) , suggesting this decrease did not explain the differential requirements of Fandango for splicing of early zygotic and maternally encoded transcripts . Mutant clonal analysis of a stronger allele of fandango ( nonsense mutation ) , showed a complete loss of the female germ line in adult ovaries ( data not shown ) . This demonstrated that the two isolated alleles of fandango are hypomorphic and suggested that Fandango was required , albeit at lower levels , for splicing of maternal transcripts . We concluded it is unlikely that a differential expression and/or association of core components of the spliceosome could potentially explain the differential requirements for Fandango between oogenesis and the early embryo . The most likely explanation is that Fandango is quantitatively ( but not qualitatively ) differentially required during early embryonic development . Transcriptional elongation can affect co-transcriptional splicing ( de la Mata et al . , 2003; Ip et al . , 2011 ) . It was recently shown that Syf1 , the yeast ortholog of Fandango , is also important for RNApol II transcriptional activity ( Chanarat et al . , 2011; David et al . , 2011 ) , therefore we decided to investigate transcription in fandango embryos . Three intronless early zygotic genes ( nullo , snail , and scute ) and two early zygotic genes with introns ( even-skipped and tailless ) were selected for further analysis by real-time qPCR . During mid/late-syncytial blastoderm ( stage B ) ( Figure 4A , ‘Materials and methods’ ) , no significant differences in transcript abundance were observed between control and fandango ( Figure 4—figure supplement 1A ) , whereas embryos mutant for grapes showed the expected reduction of transcript levels ( Figure 4—figure supplement 1A; Sibon et al . , 1997 ) . 10 . 7554/eLife . 02181 . 010Figure 4 . Early zygotic transcription is not affected during mid/late-syncytial blastoderm in fandango mutants . ( A ) Embryos were divided into three different groups according to developmental stage ( ‘Materials and methods’ ) , stage A: early/mid-syncytial blastoderm embryos , stage B: mid/late-syncytial blastoderm embryos , and stage C: blastoderm cellularization embryos . ( B ) Western blot for RNApol II CTD Ser2 phosphorylation levels . Control and fand1 embryos showed a similar increase in the global levels of RNApol II CTD Ser2 phosphorylation over the course of early embryonic development . α-Tubulin was used as a loading control . ( C ) Quantification of the CTD Ser2 phosphorylation from five independent western blot assays showed no significant difference at any of the embryonic developmental stages analyzed ( Two-way ANOVA p>0 . 05 ns . ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 01010 . 7554/eLife . 02181 . 011Figure 4—figure supplement 1 . Early zygotic transcription is not affected during mid/late-syncytial blastoderm in fandango mutants . ( A ) Real-time qPCR analysis to measure levels of early zygotic transcripts from intronless ( nullo , snail , and scute ) and containing introns genes ( even-skipped and tailless ) during embryonic development . mRNA levels from early zygotic genes were normalized with β-actin mRNA levels . At stage B , when early zygotic genes are transcribed , there was no significant differences in mRNA abundance for any of the genes analyzed in either control or fandango samples ( Two-way ANOVA p>0 . 05 ns . ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 011 During transcriptional elongation , RNApol II is specifically phosphorylated on the Ser2 residue of its carboxy-terminal domain ( CTD ) ( Hsin and Manley , 2012 ) . In agreement with the onset of early zygotic transcription , we observed a significant increase in RNApol II CTD Ser2 phosphorylation as the embryo developed from early/mid-syncytial blastoderm ( stage A ) , into mid/late-syncytial blastoderm ( stage B ) , and blastoderm cellularization ( interphase 14 ) ( stage C ) ( Figure 4A , B ) . Both control and fandango embryos showed a similar increase in global levels of RNApol II CTD Ser2 phosphorylation ( Figure 4B , C ) . As transcriptional regulation during interphase 14 ( stage C ) relies on correct expression of early zygotic genes and degradation of many maternal RNAs ( MZT ) ( Tadros and Lipshitz , 2009 ) , we concluded that transcriptional changes at this stage ( Figure 4—figure supplement 1A ) were most likely a consequence of the widespread defects occurring during mid/late-syncytial blastoderm . Altogether , we concluded that the observed reduction in Fandango levels affects mainly its splicing function . To investigate if the differential requirement of Fandango for splicing of early zygotic and maternally encoded transcripts potentially resulted from distinct transcript sequences , we generated an early zygotic kuk transcript ( kuk-LacZ ) under the control of an UAS/Gal4 inducible promoter , where the open reading frame ( ORF ) was replaced by LacZ ( Figure 5A , see ‘Materials and methods’ for more details ) . As expected , when this construct was expressed zygotically , it was correctly spliced in control but not in fandango embryos ( Figure 5B ) . In contrast , splicing of the kuk-LacZ construct occurred normally in both control and fandango mutants when it was expressed maternally ( Figure 5B ) . Since maternal expression of an early zygotic transcript , in a fandango mutant background , was enough to suppress its splicing defects , we concluded that the differential requirement of Fandango for splicing of early zygotic transcripts was most likely due to the developmental context of gene expression and not a particularity in the early zygotic pre-mRNA sequences . Consistently , we failed to detect differences related to intron size , splice sites consensus , and exon–intron structure between maternal and zygotic transcripts ( Figure 2—figure supplement 2B–D ) . 10 . 7554/eLife . 02181 . 012Figure 5 . Ectopic maternal expression of an early zygotic transcript in the mutant background is sufficient to suppress its splicing defects . ( A ) The kuk-LacZ construct was built using the 5′UTR , the intron and the 3′UTR of the kuk small transcript ( dark gray ) , and replacing the kuk ORF ( black ) by the LacZ coding sequence ( light gray ) . To induce the expression of this construct it was put under the control of the UAS promoter ( green ) to drive the tissue specific expression with GAL4 drivers . Orientation and position of primers used for splicing analysis is indicated ( arrows ) . ( B ) RT-PCR analysis of the kuk-LacZ construct . When it was zygotically expressed , it was correctly spliced in control but not in fand1 embryos ( similarly to the endogenous small kuk transcript ) . Intron retention with e–e primers ( red dots , kuk-st: 596 bp and kuk-LacZ: 869 bp ) and a PCR product with i–e primers ( 751 bp ) were observed in the mutant . When it was maternally expressed , kuk-LacZ construct was correctly spliced both in control and fand1 embryos , being detected just the spliced form of the construct ( green dots , kuk-st: 437 bp and kuk-LacZ: 713 bp ) . In contrast , the endogenous zygotically expressed small kuk transcript ( kuk-st ) is still poorly spliced in fand1 embryos carrying the kuk-LacZ construct . Open circles indicate unspecific PCR products ( confirmed by sequencing ) . Green dots indicate correctly spliced transcripts , whereas red dots indicate unspliced transcripts ( intron retention ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 012 fandango mutants showed a significant reduction in Fandango and ISY1 protein levels ( Figure 3C ) , which most likely impaired efficient activation of the spliceosome ( Villa and Guthrie , 2005 ) . Since mitosis inhibits splicing ( Shin and Manley , 2002 ) , pre-mRNA splicing of early zygotic transcripts needs to be highly efficient for these genes to be correctly expressed . This suggests the existence of a developmental pre-requisite for highly efficient splicing , so that a suboptimal activation of the spliceosome would specifically impair pre-mRNA splicing of early zygotic but not maternal transcripts . Wild-type embryos already showed a detectable amount of intron retention in early zygotic transcripts ( Figure 2B , D , Figure 2—figure supplement 2B ) , which was dramatically exacerbated in fandango embryos ( Figure 2B , D , Figure 2—figure supplement 2B ) . We hypothesized that regardless of transcript size , there was also a constraint on pre-mRNA splicing of early zygotic transcripts in wild-type embryos . We generated a gene where the 5′UTR sequence including the intron of the small zygotic kuk transcript was quadruplicate to test this hypothesis ( Figure 6A , D see ‘Materials and methods’ for more details ) . Quadruplicate introns were linked by in-frame LacZ coding sequences , and the entire construct ( 4x intron kuk-LacZ ) was under the control either of an endogenous early zygotic minimal promoter ( nullo-4x intron kuk-LacZ , ∼2 . 5 Kb ) ( Figure 6A ) or an inducible UAS/Gal4 promoter ( UAS-4x intron kuk-LacZ , ∼2 . 5 Kb ) ( Figure 6D ) . The total size of the encoded pre-mRNAs was comparable to many other endogenous early zygotic genes ( e . g . , kugelkern , runt , krupple ) . As a control , we introduced point mutations in the splice sites of these constructs to generate comparable intronless transcripts ( no intron kuk-LacZ ) ( Figure 6A , D ) . 10 . 7554/eLife . 02181 . 013Figure 6 . A small early zygotic transcript containing four introns is poorly spliced in wild-type embryos . ( A and D ) The 4x intron kuk-LacZ construct was a variant of the kuk-LacZ that contains four copies of kuk small transcript intron ( dark gray ) . Each intron is separated by 201 nucleotides of an in frame Lac-Z sequence ( light gray ) . The no intron kuk-LacZ construct has all splice sites present in the 4x intron kuk-LacZ construct mutated to thymidines . The constructs were fused to a nullo minimal promoter ( blue ) ( A ) , or fused to an inducible UAS promoter ( green ) ( D ) . Orientation and position of primers used for splicing analysis is indicated ( arrows ) . ( B ) RT-PCR analysis showed significant splicing defects ( intron retention ) of the 4x intron kuk-LacZ construct when expressed under the control of an endogenous early zygotic promoter ( nullo promoter ) . The first intron was correctly spliced , being detected mainly the PCR product corresponding to the spliced form ( green dot ) . The remaining introns ( second , third , and fourth ) were completely unspliced ( red dots , intron retention ) . In the intronless ( no intron kuk-LacZ ) construct , under the control of the same nullo promoter were only observed PCR bands whose sizes correspond to unspliced forms ( red dots , intron retention ) . ( C ) Real-time qPCR analysis showed that the 4x intron kuk-LacZ and no intron kuk-LacZ constructs were expressed to the same extent when under the control of the nullo minimal promoter ( t test p>0 . 05 ns . ) . ( E ) RT-PCR analysis of the 4x intron kuk-LacZ construct showed significant splicing defects ( intron retention ) when zygotically expressed in wild-type embryos under the control of an inducible UAS promoter . Although the most 5′-localized introns ( first and second ) were still partially spliced , being observed two PCR bands corresponding to the spliced ( green dots , int1: 191 and int2: 385 bp ) , and unspliced forms ( red dots , int1: 347 and int2: 541 bp ) . The furthest 3′-localized introns ( third and fourth ) were completely unspliced , being only observed one PCR band with the size corresponding to intron retention ( red dots , int3: 463 and int4: 385 bp ) . Maternal expression of the 4x intron kuk-LacZ construct was sufficient to significantly suppress splicing defects in the four introns analyzed ( green dots , spliced forms: int1: 191 , int2: 385 , int3: 307 , int4: 229 bp; red dots , unspliced forms: int1: 347 , int2: 541 , int3: 463 , int4: 385 bp ) . Zygotic and maternal expression of the no intron kuk-LacZ construct only showed PCR bands with sizes corresponding to unspliced forms ( red dots , intron retention ) . ( F ) Real-time qPCR analysis showed that the 4x intron kuk-LacZ and no intron kuk-LacZ constructs were expressed to the same extent both maternally ( Two-way ANOVA p>0 . 05 ns . ) and zygotically ( p>0 . 05 ns . ) in wild-type embryos . All PCR bands shown in these panels were cloned and sequenced to confirm their identity . Green dots indicate correctly spliced transcripts , red dots indicate unspliced transcripts ( intron retention ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02181 . 013 Only the first intron ( int1 ) of the 4x intron kuk-LacZ construct was correctly spliced when it was zygotically expressed in wild-type embryos under the control of an endogenous early zygotic minimal promoter ( Figure 6B ) . Likewise , when the 4x intron kuk-LacZ construct was early zygotically expressed under the control of the inducible promoter UAS/Gal4 there were similar splicing defects ( intron retention ) ( Figure 6E ) . Measurement of in vivo kinetics of mRNA splicing showed that half-lives for splicing reactions are <1 min for the first intron , but 2–8 min for both second and third introns ( Audibert et al . , 2002 ) . Hence , splicing of two or more introns requires more time than transcription and becomes rate limiting . Consistent with the hypothesis of a temporal constraint on pre-mRNA splicing , when the 4x intron kuk-LacZ construct was zygotically expressed , the splicing defects of the firstly transcribed 5′-localized introns ( Int1 and Int2 ) were significantly weaker than those observed in the later transcribed 3′-localized introns ( Int3 and Int4 ) ( Figure 6E ) . Importantly , maternal expression of this construct was sufficient to significantly suppress its splicing defects ( Figure 6E ) . Real-time qPCR analysis showed that these constructs were equivalently zygotically and maternally expressed ( Figure 6C , F ) . This suggested that splicing did not quantitatively impair early zygotic transcription , which was consistent with the observation that the rates of transcriptional elongation proceed independently of splicing ( Brody et al . , 2011 ) . We showed that in wild-type embryos pre-mRNA splicing imposed significant constraints on early zygotic expression , which is a likely explanation why most early zygotic genes are intronless ( De Renzis et al . , 2007 ) . Although a moderate decrease in the length of syncytial blastoderm interphases ( seen in grapes mutant embryos [Sibon et al . , 1997] ) was not sufficient to induce splicing defects in otherwise wild-type embryos ( data not shown ) , we hypothesize that avoidance of pre-mRNA splicing during early zygotic expression is a consequence of the extremely short interphases and frequent mitotic cycles . Similarly to Drosophila , mosquito Aedes aegypti and the zebrafish Danio rerio early zygotic transcripts are frequently intronless when compared with the rest of the transcriptome ( Biedler et al . , 2012; Heyn et al . , 2014 ) . This suggests that highly proliferative tissues need coordination between cell cycle and gene architecture for correct gene expression and avoidance of abnormally processed transcripts . Our results highlight cell cycle constraints during early embryonic development as a force capable of driving changes in gene architecture of multicellular organisms . In unicellular organisms intron paucity correlates with a bias toward the 5′ ends , whereas introns from multicellular genomes are evenly distributed throughout the genes ( Mourier and Jeffares , 2003 ) . This suggests that similar constraints on gene architecture are also likely to exist in yeast and other fast-dividing single cell eukaryotes . The way splicing efficiency might be regulated through changes in constitutive spliceosome factors and how it might influence differential gene expression is a new area of interest . In this study , we present experimental evidence supporting the hypothesis of a requirement for highly efficient pre-mRNA splicing during early embryonic development . Since the NTC/Prp19 complexes are known to be important for efficient spliceosome activation , and our mutant alleles specifically impaired pre-mRNA splicing of early zygotic but not maternally encoded transcripts , we propose that overall requirements for splicing efficiency are likely to vary during development , being the NTC/Prp19 complexes a key modulator of spliceosome activation rates . In agreement with this hypothesis , Prp19 expression varies during neuronal differentiation ( Urano et al . , 2006 ) . In plants it was recently shown that removal of retained introns regulates translation in rapidly developing gametophytes ( Boothby et al . , 2013 ) . In Drosophila , a sub-population of early zygotic transcripts with introns similarly showed some degree of intron retention in wild-type embryos ( Figure 2B , D , Figure 2—figure supplement 2B ) . Our results also suggest that the pre-requisite for highly efficient splicing during early embryonic development is paradoxically also likely to play an important regulatory role in the expression of a subset of early zygotic transcripts , which further supports the possibility that modulation of spliceosome activation per se is important for differential regulation of gene expression during development . Flies were raised using standard techniques . The fandango alleles were isolated in a previously reported maternal screen ( Pimenta-Marques et al . , 2008 ) . Maternal mutant embryos and germ-line mutant clones were generated using the FLP/FRT ovoD system ( Chou and Perrimon , 1992 ) . Germ-line clones of fand1 and fand2 were established by crossing FRT42B fand1/CyO or FRT42B fand2/CyO virgins to hs-FLP; FRT42B ovoD/CyO males and the progeny was heat shocked once at 37°C for 1 hr during second and third larval instar stages . As control we generated germ-line clones with FRT42B by crossing FRT42B/CyO virgins to hs-FLP; FRT42B ovoD/CyO males and followed by the heat shock procedure described before . To generate homozygous mutant clones in ovaries for fand1 ( negative for nuclear GFP label , nGFPminus ) we used FLP/FRT to induce mitotic recombination . Females y , w , hs-FLP; FRT42B nGFP/CyO hs-hid flies were crossed with w; FRT42B , fand1/CyO hs-hid males . Recombination was induced by 1-hr heat shock at 37°C at second and third instar larval stage . Adult ovaries were harvested from 4–5-day-old females and subsequently processed for immunofluorescence . Viability and phenotypes of fandango alleles were complemented by crossing a transgenic fly carrying a genomic fragment construct that contained a wild-type copy of CG6197 ( wt-fandango ) . w; FRT42B , fand1/CyO virgins were crossed to wt-fandango; FRT42B , fand2/CyO males; reciprocal crosses were also performed . Offspring were counted to determine viability . Rescue of maternal phenotypes ( cellularized blastoderm defects and splicing defects in early zygotic transcripts ) was also analyzed in embryos laid by F1 wt-fandango/ +; FRT42B , fand1/FRT42B fand2 females . Germ-line clones of fand1 and fand2 were also rescued ( cellularized blastoderm defects and splicing defects in early zygotic transcripts ) by a copy of wt-fandango in the third chromosome . FRT42B fand1/CyO; wt-fandango or FRT42B fand2/CyO; wt-fandango virgins were crossed with hs-FLP; FRT42B ovoD/CyO males and heat shock performed as described above . To induce maternal and zygotic expression of the UAS-kuk-LacZ construct in control and fandango maternal mutant embryos , we performed the following crosses: Maternal expression in control genetic background: virgin females +/+; Nanos-Gal4 , UAS-kuk-LacZ/TM6B crossed with wild-type males . Zygotic expression in control genetic background: virgin females +/+; actin-Gal4/TM6B crossed with +/+; UAS-kuk-LacZ males . Maternal expression in fandango maternal mutant genetic background: firstly , virgin females FRT42B fand1/CyO; Nanos-Gal4 , UAS-kuk-LacZ/TM6B crossed with hs-FLP; FRT42B ovoD/CyO males , and heat shocked as described above . After eclosion , Cy+ and Tb+ females were selected from the progeny and crossed to wild-type males . Zygotic expression in fandango maternal mutant genetic background: firstly , virgin females FRT42B fand1/CyO; actin-Gal4/TM6B were crossed with hs-FLP; FRT42B ovoD/CyO males , and heat shocked as described . After eclosion , Cy+ and Tb+ virgin females were selected from the progeny and crossed to +/+; UAS-kuk-LacZ males . To induce maternal and zygotic expression of the 4x intron kuk-LacZ and no intron kuk-LacZ constructs we performed following crosses: Maternal expression: firstly , virgin females +/+; actin-Gal4/TM6B crossed with +/+; UAS-4xintron-kuk-LacZ/TM6B or UAS-nointron-kuk-LacZ/TM6B males . After eclosion , females Tb+ ( +/+; actin-Gal4/UAS-4xintron-kuk-LacZ or +/+; actin-Gal4/UAS-nointron-kuk-LacZ ) were selected and crossed with wild-type males . Zygotic expression: virgin females +/+; actin-Gal4/TM6B were crossed with +/+; UAS-4xintron-kuk-LacZ/TM6B or UAS-nointron-kuk-LacZ/TM6B males . To analyze zygotic expression of the 4x intron kuk-LacZ and no intron kuk-LacZ constructs under the control of the minimal promoter of the gene nullo , females carrying the corresponding construct were selected and crossed with wild-type males . To drive embryonic and ovarian expression of Myc-tagged Fandango and Myc-tagged Prp19 proteins , Nanos-Gal4 homozygous virgins were crossed with UAS-Fandango-6xMyc males or UAS-Prp19-6xMyc/TM6B males , respectively . After eclosion females ( in case of Myc-Fandango ) or Tb+ females ( in case of Myc-Prp19 ) were selected , dissected ovaries from 4–5-day-old females , or laid embryos after a cross with wild-type males were used for protein extraction . To identify the gene responsible for lethality in fandango alleles , we performed a complementation analysis using the Bloomington 2R Deficiency kit . Deficiency Df ( 2R ) CX1 ( covering an interval from cytological band 49C1 to 50D2 , Bloomington stock number 442 ) failed to complement zygotic viability of both fandango alleles ( complementation group 7 ) . All additional 22 overlapping deficiencies complemented both fandango alleles . The cytological interval between bands 50B4-B6 ( comprising 6 genes ) was not covered by the 22 deficiencies . We cloned and sequenced these 6 genes from genomic DNA of both control and fandango alleles and identified mutations in gene CG6197 in both fandango alleles . To confirm the identity of our mutants , we digested DNA from genomic clone ( BACR14P04 , Flybase ) with restriction enzymes XbaI and EcoRI to generate a genomic fragment comprising the wild-type gene sequence of CG6197 ( wt-fandango ) . Then we cloned the fragment into pCasper vector and used it to generate transgenic stocks ( Bestgene , Chino Hills , CA , USA ) . A genomic wild-type copy of CG6197 under the control of its endogenous promoter fully complemented all known phenotypes in both fandango alleles . 0–3 hr ( after egg laying ) embryos , both maternally mutant for fandango and control , were fixed and stained using standard procedures ( Pimenta-Marques et al . , 2008 ) . For Neurotactin and Slam immunostaining , the fixation procedure was modified: embryos were added to boiling heat fix solution ( 68 mM NaCl +0 . 1% Triton ) and stirred for 1 min , then cooled by adding an equal volume of cooled fix solution . Immunostaining for oogenesis phenotypic analysis was performed as described in Guilgur et al . ( 2012 ) . Following primary antibodies used were: mouse anti-Neurotactin clone BP106 at 1:133 ( DSHB , Iowa City , Iowa , USA ) ; mouse anti-pTyr at 1:1000 ( 9411; Cell Signaling , Danvers , MA , USA ) , and rabbit anti-slam at 1:1000 ( Ruth Lehman Lab ) . For F-actin staining , a 5-min incubation with phalloidin-Rhodamine at 1:200 dilution ( Sigma , St Louis , MO , USA; stock concentration 1 mg/ml ) was employed at room temperature . For DNA staining , we used SYTOX Green ( Invitrogen , Grand Island , NY , USA ) at 1:5000 dilution with 5 mg/ml RNase A in PBT ( PBS+0 . 1% Tween-20 ) for 30 min at room temperature . Cy3- or Cy5-conjugated secondary antibodies were used at 1:1000 dilution ( Jackson ImmunoResearch , West Grove , PA , USA ) and anti-rabbit Alexa Fluor 488 at 1:1000 dilution ( Molecular Probes , Grand Island , NY , USA ) . The kuk-LacZ construct was synthesized using the 5′UTR and intron of the kuk small transcript ( kuk-RB , Flybase ) . The kuk ORF was replaced by the LacZ coding sequence and was followed by the 3′ UTR of the original transcript ( Figure 5A ) . The kuk-LacZ construct was fused to a UASg promoter ( Gateway system , Invitrogen , Grand Island , NY , USA ) . The 4x intron kuk-LacZ construct was synthesized using 4 repeats of the fragment of 5′UTR and intron of the kuk small transcript , separated by 201 nucleotides of in-frame LacZ sequence . The stop codon is followed by the 3′UTR kuk small transcript sequence and 300 bp of the 3′-located genomic region to promote transcriptional termination ( Figure 6A , D ) . In the case of the no intron kuk-LacZ , all splice sites ( meaning 5′ splice site , branch point , and 3′ splice site ) were mutated to thymidines ( Figure 6A , D ) . To induce expression of these constructs , they were fused to UAS promoter or nullo minimal promoter . The 4x intron kuk-LacZ and no intron kuk-LacZ constructs were cloned into pWALIUM22 . Fandango open reading frame , kuk-LacZ , 4x intron kuk-LacZ , and no intron kuk-LacZ constructs were synthetized ( GenScript , Piscataway , NJ , USA ) . Prp19 open reading frame was cloned into pDONR221 from DGC gold BDGP clone LD09231 . Prp19 and Fandango ORFs were subcloned into a vector containing the UASp promoter and 6x C-terminal Myc-tag ( Gateway , Invitrogen , Grand Island , NY , USA ) . All constructs were then used to generate transgenic flies stocks ( BestGene , Chino Hills , CA , USA ) . Total RNA was extracted from 0–3 hr ( after egg laying ) embryos , unfertilized embryos , and 4-day-old female ovaries mutant for fandango and control ( FRT42B ) following standard procedures ( PureLink RNA Mini Kit , Ambion , Grand Island , NY , USA ) . 1 μg of RNA was then used for reverse transcription with Oligo ( dT ) 12–18 and/or random hexamers primers following the manufacturer’s protocol ( Transcriptor First Strand cDNA Synthesis Kit , ROCHE , Germany ) . Primer combinations used were designed with PrimerSelect ( Lasergene , Madison , WI , USA ) and PCR was performed using GoTaq DNA polymerase ( Promega , Fitchburg , WI , USA ) . Sequences of all primers used are listed in Supplementary file 2 . To measure transcription levels , embryos were staged in three different groups based on the embryonic morphology: stage A ( embryos from cycle 1 to 8 , no pole cells , and no cortical nuclei are observed ) ; stage B ( embryos from cycle 8/9 to 13 , pole cells present , and cortical nuclei are observed ) ; and stage C ( embryos at interphase 14 , blastoderm cellularized ) . Three independent replicas for each stage , containing each 10 manually selected embryos were generated . Three different genetic backgrounds were analyzed ( control ( FRT42B ) , FRT42B fandango , and grapes as positive control ) . To measure fandango mRNA levels , unfertilized eggs were analyzed ( three replicas ) . To measure transcription level of the 4x intron kuk-LacZ and no intron kuk-LacZ constructs , 0–3 hr ( after egg laying ) embryo collections were used to analyze both maternal and zygotic induced expression . Total RNA was extracted from samples and then used for reverse transcription with Oligo ( dT ) 12–18 as described above . Real-time mRNA quantification was performed following the manufacturer’s protocol ( QuantiFast SYBR Green RT-PCR Kit , Qiagen , Germany ) . For analysis of transcription levels of early zygotic genes ( nullo , snail , scute , even-skipped , and tailless ) the Drosophila QuantiTect Primer Assay ( Qiagen , Germany ) was used . For mRNA level measurements of fandango , 4x intron kuk-LacZ and no intron kuk-LacZ constructs primers were designed with Primer3 ( Supplementary file 2 ) . Anti-Fandango and anti-Prp19 rabbit polyclonal antibodies were raised against recombinant proteins corresponding to amino acids 551–750 of Fandango/CG6197-PA , and to amino acids 20–219 of Prp19-PA , respectively ( Metabion international AG , Germany ) . In both cases it was used His-tagged recombinant proteins as antigen and the antibodies were affinity purified . Protein extracts were obtained from 0–3 hr ( after egg laying ) embryos or 4-day-old female ovaries . Embryos were dechorionated with 50% commercial bleach solution and ovaries dissected in PBS , samples then homogenized in NB buffer ( 150 mM NaCl , 50 mM Tris–HCl pH 7 , 5 , 2 mM EDTA , 0 , 1% NP-40 , 1 mM DTT , 10 mM NaF , and EDTA-free protease inhibitor cocktail , Roche , Germany ) , and centrifuge at 20000×g for 3 min . Supernatant was recovered and centrifuged twice . To analyze NTC/Prp19 complex composition ( Table 1 ) , co-immunoprecipitation was done using protein extracts from embryo or ovary tissues expressing Myc-tagged Fandango or Prp19 . Briefly , 1 mg of protein was diluted in 1 ml NB buffer and incubated with 1 µg/ml of mouse c-Myc antiboby ( 9E10 ) ( Santa Cruz Biotechnology , Dallas , Texas , USA ) for 1 hr at 4°C . Subsequently , 0 . 9 mg of Dynabeads Protein G ( Invitrogen , Grand Island , NY , USA ) were added to the immune complex and incubated 1 hr at 4°C . Beads were washed three times with NB buffer and protein elution performed with 50 µl of 100 mM Glycine pH 2 . 5 during 2 min at RT and stopped with 5 µl of 1M Tris–HCl pH 10 . 85 . Eluted proteins were then precipitated in five times the volume of acetone at −20°C and samples analyzed by liquid chromatography coupled to tandem mass spectrometry ( Mass Spectrometry Laboratory , Institute of Biochemistry and Biophysics , Poland ) . To analyze NTC/Prp19 complex composition ( showed in Figure 3A ) , protein co-immunoprecipitation was performed using nuclear protein extracts ( adapted from Kamakaka and Kadonaga , 1994 ) from a collection of 0–3 hr ( after egg laying ) wild-type embryos ( Oregon-R ) . 1 mg of protein extract was incubated with rabbit anti-Prp19 ( 1:1000 dilution ) or the pre-immune ( 1:10 , 000 dilution ) as control , in HNEB2 buffer ( 100 mM NaCl , 2 , 5 mM MgCl2 , 10 mM Tris–HCl pH 7 , 5 , 0 , 5% Triton X-100 , and EDTA-free protease inhibitor cocktail , Roche , Germany ) during 1 hr at 4°C . The procedure was carried out as described in previous co-immunoprecipitation and eluted complexes were boiled in Laemmli sample buffer and analyzed by Western Blot . Size-exclusion chromatography was performed in protein extracts of 0–3 hr ( after egg laying ) embryo collections from FRT42B fand1 mutants or control ( FRT42B ) . Extracts were prepared as described before in NB2 buffer ( 150 mM NaCl , 50 mM Tris–HCl pH 7 , 5 , 2 mM EDTA , 0 , 01% NP-40 , 1 mM DTT , and EDTA-free protease inhibitor cocktail , Roche , Germany ) . Subsequently , 2 mg of protein extract were fractionated using Superose 6 10/300 GL column ( GE Healthcare , United Kingdom ) in NB2 buffer and fractions collected and analyzed by Western blot . To analyze protein amount in ovaries and embryos ( showed in Figures 1I and 3C ) , embryos were dechorionated and ovaries dissected as described above . Samples were homogenized in PBS supplemented with EDTA-free protease inhibitor cocktail ( Roche , Germany ) and centrifuged at 20000×g for 3 min at 4°C . Supernatant was collected and protein concentration determined using the Bradford method ( BioRad , Hercules , CA , USA ) . Samples were immediately boiled in Laemmli sample buffer and 10 µg of protein was run in SDS-PAGE gel and analyzed by immunoblot . Levels of RNApol II CTD Ser2 phosphorylation were analyzed in embryos dechorionated and manually selected at specific developmental stages based on the embryonic morphology ( as described above ) . 15 embryos were selected for each stage and protein sample was obtained by lysing the embryos with a needle in Laemmli sample buffer and heating for 5 min at 100°C . Protein amounts corresponding to ∼7 embryos were running in SDS-PAGE and analyzed by immunoblot . Five independent replicas were analyzed . Antibodies used were: polyclonal rabbit anti-Prp19 at 1:8000 dilution; polyclonal rabbit anti-Fandango at 1:1000 dilution; mouse anti-alpha-Tubulin Dm1A at 1:50 , 000 dilution ( Sigma , St Louis , MO , USA ) ; mouse anti-RNA Polymerase II H5 at 1:500 dilution ( MMS-129R , Covance , Princeton , NJ , USA ) ; rabbit anti-ISY1 at 1:500 dilution ( ab121250; Abcam , United Kingdom ) ; and mouse anti-CDC5L [2136C1a] at 1:200 dilution ( ab51320; Abcam , United Kingdom ) . Peptides mixtures were analyzed by LC-MS-MS/MS ( liquid chromatography coupled to tandem mass spectrometry ) using Nano-Acquity ( Waters , Milford , MA , USA ) LC system and Orbitrap Velos mass spectrometer ( Thermo Electron Corp . , San Jose , CA , USA ) . Prior to analysis , proteins were subjected to standard ‘in-solution digestion’ procedure , during which proteins were reduced with 100 mM DTT ( for 30 min at 56°C ) , alkylated with 0 , 5 M iodoacetamide ( 45 min in darkroom at room temperature ) , and digested overnight with trypsin ( sequencing Grade Modified Trypsin—Promega V5111 ) . The peptide mixture was applied to an RP-18 precolumn ( nanoACQUITY Symmetry C18—Waters 186003514 ) using water containing 0 , 1% TFA as mobile phase , then transferred to nano-HPLC RP-18 column ( nanoACQUITY BEH C18–Waters 186003545 ) using an acetonitrile gradient ( 0%–35% AcN in 180 min ) in the presence of 0 . 05% formic acid with a flow rate of 250 nl/min . The column outlet was directly coupled to the ion source of the spectrometer , operating in the regime of data dependent MS to MS/MS switch . A blank run ensuring no cross contamination from previous samples preceded each analysis . Raw data were processed by Mascot Distiller followed by Mascot Search ( Matrix Science , London , UK , on-site license ) against Flybase database . Search parameters for precursor and product ions mass tolerance were 100 ppm and 0 . 6 Da , respectively , enzyme specificity: trypsin , missed cleavage sites allowed: 0 , fixed modification of cysteine by carbamidomethylation , and variable modification of methionine oxidation . Peptides with Mascot Score exceeding the threshold value corresponding to <5% False Positive Rate , calculated by Mascot procedure , and with the Mascot score above 30 were considered to be positively identified . Human orthologs were determined using DSRC Integrative Ortholog Prediction Tool ( DIOPT ) ( http://www . flyrnai . org/cgi-bin/DRSC_orthologs . pl ) . Only scores above two were considered such as the best matches when there was more than one match per input . Total RNA was isolated from 0–3 hr collections of fandango maternal mutant and control ( FRT42B ) embryos using TRIzol Reagent ( Invitrogen , Grand Island , NY , USA ) , following standard protocol . DNAse I ( Promega , Fitchburg , WI , USA ) treatment was performed during 30 min at 37°C . DNAse was extracted by Phenol-Chloroform extraction; the RNA was precipitated with ethanol , and dissolved it in 25 µl of DEPC water . Bioanalyzer testing was used to analyze quality and concentration of the samples and made up to the volume to 100 µl with water , 50 µl of 7 . 5 M NH4OAc added , 0 . 5 µl of glycogen , and 250 µl of absolute ethanol . cDNA library was generated applying the standard Illumina protocol for RNA-Seq ( polyA RNAs ) and sequenced with an Illumina HiSeq ( Oklahoma Medical Research Foundation , Oklahoma City , OK , USA ) . These generated RNA-Seq data for two biological replicates each of wild-type and fandango mutant ( fand1 ) , consisting of about 150 million illumina paired-end 100 bp reads per sample . Paired-end reads were mapped with tophat ( Trapnell et al . , 2009 ) version 2 . 0 . 3 against the Drosophila melanogaster BDGP5 reference genome , using Flybase gene annotations downloaded from Ensembl e66 as guide . To analyze splicing defects , we first extracted exon–intron boundaries from gene annotations . To avoid potential confounding effects , we removed all boundaries that had overlapping exon sequence ( from different genes or transcripts ) . Subsequent analysis used this set of ‘safe’ exon–intron boundaries . For coverage plots in Figure 2D , we also excluded boundaries where the intron or the exon were less than 50 b long . At each base within 50 bp either side of a splice site we count the number of reads that overlap that base , then divide by the total number of reads within the 100 bp centered around the splice site . To minimize noise , we require that at least 50 reads fall within the −50:50 window around the exon–intron boundary ( reads that only partially overlap the window are also counted ) . To determine the frequency of splicing defects for each boundary , we extracted all reads that overlap the 5′ splice site by at least 10 bp to either side of the boundary . We classified each read as correctly spliced ( if the read is split from the 5′ to the 3′ splice site ) , unspliced ( if the read is not split ) or mis-spliced ( if the read is split but not matching the expected 5′ or/and 3′ sites ) . To reduce noise , we only include an exon–intron boundary if at least 10 fragments overlap that boundary . To determine the exon–intron gene structure ( Figure 2—figure supplement 2D ) , each gene was divided in 1000 equal segments . For each segment of each gene , we checked for the presence ( or absence ) of an exon in that segment . For each segment , we then plotted the frequency of exon presence in all genes . If an exon randomly appears in a given segment , it appears in ∼50% of genes . For example , a set of intronless genes would produce a plot that would be always at 100% . To determine the splice site motif ( Figure 2—figure supplement 2C ) , sequences around exon–intron boundaries were extracted and motifs drawn using Weblogo . Early zygotic and maternal genes were defined using RNA-Seq developmental gene expression data from Flybase ( Graveley et al . , 2011 ) . A gene was defined as an early zygotic gene when its expression at 2–4 hr is at least moderate ( more than 10 expression units , as defined in the Flybase dataset ) and at least 5× greater than its expression at 0–2 hr ( irrespective of the 0–2 hr value ) . Maternal genes are those genes that are not early zygotic and have high expression ( more than 50 expression units ) at 0–2 hr . To avoid potential artifacts , genes that have an extremely high expression ( more than 1000 expression units ) were not considered . Applying this definition we obtained 270 early zygotic genes ( including 43 genes from De Renzis et al . , 2007 ) and 2048 maternal genes . All scripts used for this analysis are available upon request . RNA-Seq data are available in the ArrayExpress database ( www . ebi . ac . uk/arrayexpress ) under accession number E-MTAB-2321 . The procedure has been described in Stein et al . ( 2002 ) . Antisense digoxigenin-labeled RNA probes were synthesized using the DIG RNA labeling Kit ( Roche , Germany ) . eve and nos probes were made from pBluescript plasmids containing the respective cDNAs . Sequences were aligned using ClustalW2 ( http://www . ebi . ac . uk/Tools/msa/clustalw2/ ) and BoxShade 3 . 21 ( http://www . ch . embnet . org/software/BOX_form . html ) for printing and shading of multiple alignment file . Unpaired t test and two-way ANOVA were performed using Prism 5 . 00 for Windows ( GraphPad Software , San Diego , CA , USA ) .	When a fertilized egg develops into an embryo , the expression of many genes must be carefully timed and coordinated . Researchers regularly use a type of fruit fly called Drosophila to study development because it is small , it has a short lifespan , and its whole genome sequence is already known . The development of a Drosophila embryo starts with the nucleus of the fertilized egg , which contains most of the cell’s genetic material , dividing 13 times in quick succession , without the cell itself splitting . These divisions are amongst the fastest known for any animal , and given the fast developmental speed , the embryo must efficiently express all genes it needs to stay alive . Because cell division is known to inhibit gene expression this raises an interesting conundrum about the way fast cell proliferation and gene expression are coordinated . The first step of gene expression involves a length of DNA being transcribed to produce an intermediate molecule called a messenger RNA ( mRNA ) , which is then translated to produce a protein . However , some mRNA molecules contain regions called ‘introns’ that are not translated and must instead be removed via a time-consuming process called ‘splicing’ before the protein is produced . At first a Drosophila embryo uses mRNA molecules that were spliced and packaged inside the egg by the mother , but later it starts to make its own mRNA molecules . The very first mRNA molecules made by the early embryo tend to be short and lack introns . The shortness of these molecules is thought to reflect the fact there is not enough time to produce longer mRNA molecules . Is the same ‘need for speed’ also responsible for the lack of introns in these molecules ? Now , Guilgur et al . have tested this hypothesis by manipulating a gene named fandango , which codes for part of the cellular machinery that removes introns from mRNA molecules , in fruit flies . These mutant fruit flies had less of the Fandango protein than wild-type flies and while they passed through the early stages of development normally , they later developed defects—such as abnormally shaped cells . Guilgur et al . revealed that fandango mutants fail to splice out the introns in the mRNA molecules that are made in the early embryo , whereas similar mRNA molecules from the mother were spliced as normal . Further experiments suggested that wild-type embryos struggled to correctly splice an untypical early gene that had multiple introns . Together the findings of Guilgur et al . suggest that when nuclei ( or cells ) are dividing rapidly , there is a strong selective pressure to splice mRNA molecules quickly in the short time between the divisions . Furthermore , this pressure appears to have shaped the architecture of the earliest genes expressed in the Drosophila embryo , which is why the first mRNA molecules produced by the embryo itself tend not to contain introns .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "chromosomes", "and", "gene", "expression", "developmental", "biology" ]	2014	Requirement for highly efficient pre-mRNA splicing during Drosophila early embryonic development
Many lines of evidence have indicated that both genetic and non-genetic determinants can contribute to intra-tumor heterogeneity and influence cancer outcomes . Among the best described sub-population of cancer cells generated by non-genetic mechanisms are cells characterized by a CD44+/CD24− cell surface marker profile . Here , we report that human CD44+/CD24− cancer cells are genetically highly unstable because of intrinsic defects in their DNA-repair capabilities . In fact , in CD44+/CD24− cells , constitutive activation of the TGF-beta axis was both necessary and sufficient to reduce the expression of genes that are crucial in coordinating DNA damage repair mechanisms . Consequently , we observed that cancer cells that reside in a CD44+/CD24− state are characterized by increased accumulation of DNA copy number alterations , greater genetic diversity and improved adaptability to drug treatment . Together , these data suggest that the transition into a CD44+/CD24− cell state can promote intra-tumor genetic heterogeneity , spur tumor evolution and increase tumor fitness . Differences in the morphology and behavior of cancer cells within tumors were first noted by pathologists in the 1800s . However , advancements in genome-sequencing technologies and the possibility of analyzing the genome at single-cell resolution have affirmed that cancer within a single patient is a heterogeneous mixture of genetically distinct sub-clones that can arise from the accumulation of random mutations during tumor initiation , progression and response to drug treatments ( Navin et al . , 2011; Greaves and Maley , 2012; Burrell et al . , 2013 ) . Tumors have been considered the result of neo-Darwinian ( Zellmer and Zhang , 2014 ) evolution processes within tissues . However , ample evidence has also indicated that non-genetic determinants , including epigenetic modifications ( e . g . chromatin modifications or the activities of microRNAs and other noncoding RNAs ) and the capability of cancer cells to transition between various cell states , contribute to the diversity of cell type and to the cell behaviors observed within tumors ( Almendro et al . , 2013; Kreso and Dick , 2014 ) . One of the best examples of a cell state transition that responds to epigenetic/stochastic events is the differential expression of the cell surface markers CD44 and CD24 ( Al-Hajj et al . , 2003; Polyak and Weinberg , 2009; Yao et al . , 2010; Korkaya et al . , 2011; Brooks et al . , 2015 ) . CD44 and CD24 are cell surface glycoproteins that are reportedly involved in cell-cell and cell-matrix interactions , as well as in the regulation of cancer cell growth , anchorage-independent proliferation and survival ( Alho and Underhill , 1989; Birch et al . , 1991; Sy et al . , 1991; Jain et al . , 1996; Goodison et al . , 1999; Smith et al . , 2006 ) . CD44+/CD24− cell populations are particularly salient in the field of oncology . It has been shown that CD44+/CD24− cancer cells are more adept at forming tumors , are more resistant to chemotherapy and tend to have greater metastatic potential ( Al-Hajj et al . , 2003; Polyak and Weinberg , 2009; Yao et al . , 2010; Korkaya et al . , 2011; Brooks et al . , 2015 ) . CD44+/CD24− cells were initially identified in breast tumors and breast-cancer-derived cell lines and have since been observed in the vast majority of tumors and tumor-derived cell lines , including non-small-cell lung carcinomas ( NSCLCs ) , glioblastomas , neuroblastomas and leiomyosarcomas , as well as pancreatic , colon , prostate and ovarian tumors/cancers . Although CD44+/CD24− cells are the most abundant sub-population of cells in certain tumors , in many cases , they constitute only 3–5% of total tumor cells ( Al-Hajj et al . , 2003; Polyak and Weinberg , 2009; Yao et al . , 2010; Korkaya et al . , 2011; Brooks et al . , 2015 ) . From a molecular standpoint , CD44+/CD24− cells are characterized by mesenchymal-like features including a decreased expression of E-cadherin and an increase in both vimentin and the master regulators of epithelial to mesenchymal transition ( EMT ) : TGF-β , Snail and Zeb2 . These cells also exhibit upregulation of stem-cell markers , such as Oct3/4 , Nanog , Sox2 , IL-6 , C-Myc and BMI-1 , and of anti-apoptotic proteins , such as Mcl1 , Bcl2 and Bcl-XL ( Al-Hajj et al . , 2003; Polyak and Weinberg , 2009; Yao et al . , 2010; Korkaya et al . , 2011; Brooks et al . , 2015 ) . In addition to the aforementioned characteristics , we found that CD44+/CD24− cells have intrinsic defects in their ability to repair DNA lesions . In particular , we observed that these defects could be ascribed to the activation of TGF-β-mediated signaling; we discovered that this signaling pathway was both necessary and sufficient for decreasing the expression of genes such as BLM , BRCA2 , FANCF , NBN , PMS1 , RAD50 , RDM1 , WRN , ATM and ATR , which are essential in homology-directed repair ( HDR ) of DNA double-strand breaks ( DSBs ) . TGF-β was first described in human platelets as a secreted protein that played a potential role in wound-healing responses . Since then , the TGF-betaβ signaling pathway has been the focus of a multitude of studies . Upon activation , TGF-betaβ can bind to the TGF-β receptor II ( TβRII ) , which in turn phosphorylates the TGF-β receptor I ( TβRI ) ( Massagué et al . , 2005 ) . From here , the activated TβRI can trigger a signaling cascade that propagates and amplifies the signal through the phosphorylation of intracellular downstream effectors such as SMAD2 and SMAD3 , AKT and MAPK ( Massagué et al . , 2005 ) . In the case of SMAD2/3 , these effectors can form a complex with the SMAD4 protein that translocates the nucleus where , upon binding to specific DNA sequence motifs and transcriptional regulatory complexes , they can trans-activate target genes ( Massagué et al . , 2005 ) . The activation of the TGF-β signaling axis impacts diverse and often contrasting cellular processes , such as cell proliferation , epithelial to mesenchymal differentiation , migration , apoptosis and ECM remodeling ( Massagué , 2008; Derynck and Miyazono , 2008; Oshimori and Fuchs , 2012 ) . The duality of TGF-β-mediated signaling is best exemplified by the role of TGF-β in tumorigenesis . In fact , depending on the context , TGF-β can either halt or promote tumorigenesis ( Massagué , 2008 ) . Furthermore , TGF-β signaling has recently been shown to mediate resistance to targeted and conventional anticancer agents through the activation of pro-survival pathways ( such as those involving Il-6 , Bcl2 or Mcl-1 ) or of cell metabolism ( Pham et al . , 2007; Franco et al . , 2010; Yao et al . , 2010; Oshimori et al . , 2015 ) . Interestingly , previous observations have already implied a role for TGF-β in DNA repair ( Glick et al . , 1996; Preobrazhenska et al . , 2002; Kirshner et al . , 2006; Wiegman et al . , 2007; Liu et al . , 2014 ) . While some previous studies align with our findings and demonstrate a possible role for TGFβ in reducing the expression and/or activity of certain genes involved in DNA repair , other works have challenged this view by showing that inhibition of TGF-β signaling tends to attenuate DNA damage responses . Owing to these contradictory observations , we utilized a multifaceted approach to investigate whether a decrease in the expression of genes involved in homology directed repair ( HDR ) by TGF-β was sufficient to induce DNA DSBs and an increased accumulation of DNA copy number alterations ( CNAs ) . In fact , because of the decrease in the efficiency of HDR , a switch to other break-repair mechanisms , including non-homologous end joining ( NHEJ ) and microhomology-mediated break-induced replication ( MMBIR ) ( Hastings et al . , 2009; Fitzgerald et al . , 2017 ) , could give rise to an increase in the accumulation of chromosomal translocation and CNAs in CD44+/CD24− cancer cells and cells exposed to TGF-β . Consistent with the observation that TGF-β decreases the expression of BLM , BRCA2 , FANCF , NBN , PMS1 , RAD50 , RDM1 , WRN , ATM and ATR , we found that cells exposed to TGF-β have a decreased capacity to repair DNA DSBs . In addition , we also observed that CD44+/CD24− cells from tumor and tumor-derived cell lines , as well as cells that have been exposed to TGF-β , had an increased clonal genetic diversity . Central to current cancer research is the notion that the acquisition of genetic mutations in cancer cells generate complex populations of cells that are subject to Darwinian evolution ( Cahill et al . , 1999; Almendro et al . , 2013; Burrell et al . , 2013 ) . In keeping with this principle , much like the branching evolution that Darwin described , the progeny of a cancer cell with increased fitness thrives in certain environments and gives rise to a dominant clonal population ( Cahill et al . , 1999; Almendro et al . , 2013; Burrell et al . , 2013 ) . Consistently , we observed that the increased genetic diversity in the cancer cell population that was induced by transient exposure to TGF-β enabled cancer cell populations to better respond to multiple-drug treatments when compared to TGF-β-naïve cells . This observation is particularly interesting because previous works have already shown that TGF-β could induce drug resistance by activating anti-apoptotic pathways ( Yao et al . , 2010 ) . Therefore , the fact that TGF-β could induce drug resistance by accelerating cancer evolution suggests that exposure to TGF-β could possibly provide a survival niche for cancer cells , allowing them to develop more stable genetic traits and increased malignancy . In summary , our findings imply that the tumor microenvironment and gene regulatory networks could generate phenotypic diversity , including cells defined by a CD44+/CD24− state , enabling the development of novel , heritable genotypic profiles that could be selected for . Therefore , our findings could unify the non-genetic ( Lamarckian ) and mutation-driven ( Darwinian ) mechanisms that have been used to explain tumor adaptability and drug resistance . This unified theory surpasses the conventional models of tumorigenesis and somatic evolution in moving beyond simple Darwinian schemes . To gain a better understanding of the biology of CD44+/CD24− cells and to identify genes that are selectively required for their survival , we performed an RNA interference ( RNAi ) -based , forward genetic screen . As an initial paradigm to study CD44+/CD24− vulnerabilities , we used H1650 and H1650-M3 cells . The latter are derived from H1650 cells and harbor both phenotypic characteristics ( i . e . , epithelial to mesenchymal transition , increased metastatic capacity , motility , invasion and resistance to drug treatment ) and molecular characteristics ( i . e . , increased surface expression of CD44 and decreased expression of CD24 , constitutive activation of the TGF-β axis , and autocrine secretion of IL-6 ) of CD44+/CD24− cells ( Figure 1—figure supplement 1A ) ( Yao et al . , 2010 ) . H1650 and H1650-M3 cells were infected with a retroviral shRNA library that was previously constructed in the pSM1 vector ( Paddison et al . , 2004 ) . This library consists of 28 , 000 sequence-verified shRNAs designed to target approximately 9 , 000 genes that have been implicated in tumorigenesis . Each shRNA is linked to a unique 60-nucleotide sequence ( its DNA ‘barcode’ ) that can be used to monitor relative frequencies of individual shRNAs in complex populations over time . Infected cells were grown , after which the representation of individual shRNAs was determined using barcode microarray analysis . DNA was extracted , amplified and hybridized on custom-made microarrays . In the context of our drop-out screen , the decreased abundance of a particular shRNA in the pool indicates that that shRNA is targeting a gene that is essential for proliferation and/or survival . An outline of the experiment is shown in Figure 1—figure supplement 1B . We found that 135 shRNAs reached a p-value of less than 0 . 5 and had reduced representations of more than 1-fold in H1650-M3 cells compared with H1650 cells ( Figure 1A ) . Among the top shRNAs selectively depleted in the CD44+/CD24− H1650-M3 cells , we identified shRNAs targeting IL-6 ( Figure 1A ) . This finding is consistent with previous reports ( both from our group and others ) indicating that the IL-6 axis is required for the survival of cells in a CD44+/CD24− state ( Yao et al . , 2010; Marotta et al . , 2011 ) . Interestingly , we also observed a decreased representation of shRNA-targeted genes in DNA repair/replication pathways such as BRCA1 , ORC5L , RFC3 , POLS , ERCC8 and RPA2 ( Figure 1A ) . 10 . 7554/eLife . 21615 . 003Figure 1 . A genome-wide shRNA screen identifies genes involved in DNA damage repair ( DDR ) that are required for the survival of CD44+/CD24− . ( A ) The graph depicts the relative abundance of barcodes recovered from the screen . Each bar represents fold changes of an shRNA expression vector at T20 ( i . e . , 20 cell passages ) compared with T0 ( time of infection ) in CD44+/CD24− H1650-M3 cells ( upper panel ) and CD44−/CD24+ H1650 cells ( lower panel ) . Dots indicate unique genes , knockdown of which conferred proliferative disadvantage to CD44+/CD24− ( H1650-M3 ) cells . The data are plotted as the means of three biological replicates in ascending order . A FACS profiling of H1650-M3 and H1650 cells , along with a schematic of the shRNA screen , is provided in Figure 1—figure supplement 1 . ( B ) Validation of shRNA screen hits in tumor-derived cell lines characterized by low CD44+/CD24− cell content ( i . e . , MCF7 , A549 and BT474 ) compared to cell lines with high CD44+/CD24− content ( i . e . , NCI-H23 , PC9 , MDA-MB435S and MDA-MB-231 ) . The box plots show the percentage of viable cells 5 days after transfection with the indicated siRNAs relative to the number of control scramble-siRNA transfected cells . Each box is the mean ± SD of data collected from cell lines with either ( CD44+/CD24− ) lo or ( CD44+/CD24− ) hi content , from two independent experiments , each conducted in eight replicates ( p-value <0 . 05 , <0 . 005 , *<0 . 0005 , unpaired t-test ) . FACS profiles for each cell line , relative % of viable cells for each cell line and knockdown efficiency are reported in Figure 1—figure supplement 2 . ( C ) Validation of shRNA screen hits in tumor-derived cell lines FACS-sorted on the basis of the surface expression of CD44 and CD24 . The box plots show the percentage of CD44+/CD24− cells and cells of other immune types upon transfection with the indicated siRNA oligonucleotides relative to control ( scramble ) siRNA . Each box is the mean ± SD of data collected from four cells lines ( A549 , H1650 , PC9 and NCI-H23 ) upon FACS sorting , each from three replicates from two independent experiments . ( p-value <0 . 05 , <0 . 005 , <0 . 0005 , unpaired t-test ) . See Figure 1—figure supplement 3 for more details . ( D ) Schematic of the generation of single cell-derived isogenic cell lines from H1650 cells . See Figure 1—figure supplement 4A for CD44 and CD24 surface marker staining profiles . ( E ) Validation of shRNA screen hits in the FACS-sorted H1650 single cell-derived isogenic clones—Isg-C , Isg-D6 and Isg-E4 . The box plots indicate the percentage of CD44+/CD24− cells and cells of other immune types after transfection with the indicated siRNA oligonucleotides relative to control ( scramble ) siRNA . Each box is the mean ± SD of data collected from three different isogenic cell lines , each from three replicates from two independent experiments ( p-value <0 . 05 , *<0 . 005 , unpaired t-test ) . See Figure 1—figure supplement 4B , C for further details . ( F ) Expression of Pro-caspase three and Cleaved-caspase 3 ( i . e . , cell death marker ) in H1650-M3 ( CD44+/ CD24− ) and H1650 ( CD44−/CD24+ ) cell lines upon knockdown of indicated gene expression . Samples were collected 3 days post-transfection and protein lysates were immune-blotted with the indicated antibodies . Alpha-tubulin is used as the loading control . See Figure 1—figure supplement 5 for quantification . ( G ) Percentage of Cleaved-caspase 3-positive cells , normalized to respective scramble controls ( set at 100% ) , in FACS-sorted CD44+/CD24− and CD44−/CD44+ cells in H1650 cell line . Each bar represents mean ± SD of three replicates from two independent experiments ( p-value <0 . 05 , *<0 . 005 , unpaired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 00310 . 7554/eLife . 21615 . 004Figure 1—figure supplement 1 . Forward genetic screen performed in H1650 and H1650-M3 cell lines . ( A ) FACS analysis of H1650 and H1650-M3 stained with antibodies against cell surface markers CD44 and CD24 . ( B ) A schematic of the RNAi-based , forward genetic screen . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 00410 . 7554/eLife . 21615 . 005Figure 1—figure supplement 2 . Validation of shRNA screen hits in tumor-derived cell lines characterized by low ( H1650 , A549 , MCF7 , and BT-474 ) and high ( PC9 , NCI-H23 , H1650-M3 , MDA-MB-435s and MDA-MB-231 ) content of CD44+/CD24− cells . ( A ) FACS profile of multiple cancer-derived cell lines used in validation of the shRNA screen hits . Cells were stained with antibodies against cell surface markers CD44 and CD24 and analyzed with FACS . ( B ) The charts represent the percentage of viable cells 5 days after transfection with the indicated siRNA oligonucleotides compared to scramble siRNA=transfected control cells . Each bar represents mean ± SD of two independent experiments each conducted in eight- replicates ( n = 16 ) . ( C ) Efficiency of knockdown upon transfection with the indicated siRNA oligonucleotides and compared to scramble siRNA transfected control cells . Cells were collected 3 days post transfection and analyzed for expression of the indicated mRNA by RT-qPCR . Each bar represents mean ± SD of three replicates . p-value <0 . 05 , <0 . 005 , <0 . 0005 , unpaired t-test . Non-significant differences are shown by the symbol ‘#’ or ‘ns’ . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 00510 . 7554/eLife . 21615 . 006Figure 1—figure supplement 3 . Validation of shRNA screen hits in tumor-derived cell lines FACS sorted on the basis of the surface expression of CD44 and CD24 . ( A ) The charts represent the percentage of viable CD44+/CD24− cells and cells of other immune types upon transfection with the indicated siRNA oligonucleotides relative to control ( scramble siRNA ) . Each bar represents mean ± SD of three technical replicates from two independent experiments . 10 , 000 cells were analyzed by FACS for each replicate of each sample . ( B ) Efficiency of knockdown upon transfection with the indicated siRNA oligonucleotides and compared to control cells transfected with scramble siRNA . Cells were collected 3 days post transfection and analyzed for expression of the indicated mRNA . Each bar represents mean ± SD of three replicates . p-value <0 . 05 , <0 . 005 , <0 . 0005 , unpaired t-test . Non-significant differences are shown by the symbol ‘#’ or ‘ns’ . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 00610 . 7554/eLife . 21615 . 007Figure 1—figure supplement 4 . Validation of shRNA screen hits in H1650-derived isogenic clones , FACS sorted on the basis of surface expression of CD44 and CD24 . ( A ) FACS profile of H1650-derived isogenic cell lines stained with CD44 and CD24 antibodies . The numbers in bold represent the percentage of cells that are CD44+/CD24− . ( B ) The charts represent the percentage of viable CD44+/ CD24− cells and cells of other immune types upon transfection with the indicated siRNA oligonucleotides relative to control ( scramble siRNA ) . Each bar represents mean ± SD of three technical replicates from two independent experiments . 10 , 000 cells were analyzed by FACS for each replicate of each sample . ( C ) Efficiency of knockdown upon transfection with the indicated siRNA oligonucleotides and compared to scramble siRNA transfected control cells . Cells were collected 3 days post transfection and analyzed for expression of the indicated mRNA . Each bar represents mean ± SD of three replicates . p-value <0 . 05 , <0 . 005 , <0 . 0005 , unpaired t-test . Non-significant differences are shown by the symbol ‘#’ or ‘ns’ . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 00710 . 7554/eLife . 21615 . 008Figure 1—figure supplement 5 . Quantification of apoptosis-mediated increase in lethality in H1650-M3 and H1650 . The chart depicts quantification of relative amounts of Cleaved-caspase-3 between H1650-M3 and H1650 from the immunoblots indicated in Figure 1F . Levels of intensity of each band were quantified using imajeJ32 software and is represented as a ratio to Pro-Caspase-3 . The data represent mean ± SD from two independent experiments . p-value <0 . 05 , *<0 . 005 , unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 00810 . 7554/eLife . 21615 . 009Figure 1—figure supplement 6 . Validation of shRNA screen hits in H1650 and H1650-M3 cells using the clonogenic assay technique . ( A ) Clonogenic assay performed on H1650 and H1650-M3 to validate siRNA screen hits for BRCA1 , ORC5L , RFC3 , and POLS . The charts represent the percentage of viable cells 15 days after transfection with the indicated siRNA oligonucleotides compared to that of scramble-siRNA-transfected control cells . Each bar represents mean ± SD of data collected from ten fields from three different six-well plates ( n = 30 ) . ( B ) Knockdown efficiency for clonogenic assay upon transfection with the indicated siRNA oligonucleotides . Samples were collected 3 days post-transfection and analyzed by immunoblot analysis with antibodies against Total caspase 3 , alpha-tubulin and Cleaved-caspase 3 . Each bar represents mean ± SD of three replicates . p-value <0 . 05 , <0 . 005 , <0 . 0005 , unpaired t-test . Non-significant differences are shown by the symbol ‘#’ or ‘ns’ . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 009 More specifically , BRCA1 is part of a multi-protein complex that repairs DNA when both strands are broken . Mutations in or downregulation of the expression of this gene is associated with predisposition to cancers ( Ford et al . , 1994; Miki et al . , 1994; Thompson et al . , 1995 ) . ORC5L encodes one of the six subunits that form the origin recognition complex ( ORC ) that is essential for the initiation of DNA replication in eukaryotic cells . In addition , ORC5L has been shown to be involved in other processes such as transcriptional gene silencing and sister chromatid cohesion in Saccharomyces cerevisiae ( Suter et al . , 2004 ) . ERCC8 is part of the nucleotide excision repair ( NER ) pathway , a complex system that eliminates a broad spectrum of structural DNA lesions , including ultraviolet-induced pyrimidine dimers , chemical adducts and DNA cross-links ( Reardon and Sancar , 2005 ) . RPA2 is one of the three components of a protein complex involved in DNA replication , DNA repair and recombination . Interestingly , RPA2 phosphorylation is observed after the exposure of cells to ionizing radiation ( IR ) and other DNA-damaging agents , which suggests that the modified RPA2 protein participates in the regulation of DNA repair and/or DNA replication after DNA damage ( Zou et al . , 2006 ) . When DNA replicative polymerases stall at sites of DNA lesion , translesion synthesis ( TLS ) polymerases such as POLS are recruited . POLS is also reported to be involved in cohesion at the replication fork in S . cerevisiae ( Wang et al . , 2000; Wang and Christman , 2001 ) . Replication factor C3 ( RFC3 ) has been shown to be part of multiple protein complexes that have distinct functions . In conjunction with RFC1 , RFC2 , RFC4 , and RFC5 , it forms a heteropentameric protein complex that is required for the loading of PCNA onto DNA at template-primer junctions and for the polymerase switch between DNA polα and DNA polδ . As part of a complex that contains the RAD17 subunit , it regulates DNA damage checkpoints; in a complex with the CTF18 subunit , it is necessary for sister chromatid cohesion; whereas in a complex with ATAD5 , it aids fork stalling recovery and DNA DSB repair ( Mayer et al . , 2001 ) . The identification of these genes in our drop-out screen was of particular interest as it indicated possible phenotype/cell state dependencies . As a first step to validating these findings , independent siRNAs were used to silence the expression of IL-6 , BRCA1 , ORC5L , RFC3 , POLS , ERCC8 and RPA2 in H1650-M3 and H1650 cells , as well as in seven additional tumor-derived cell lines characterized by low ( A549 , MCF7 and BT474 ) or high ( NCI-H23 , PC9 , MDA-MB-435S and MDA-MB-231 ) content of CD44+/CD24− cells ( Figure 1B and Figure 1—figure supplement 2 ) . Overall , we observed that tumor-derived cell lines that had a high content of CD44+/CD24− cells were more sensitive to the inactivation of these genes . Many studies have shown that stochastic , non-genetic processes can drive the acquisition of phenotypic differences among cancer cells ( Gupta et al . , 2011 ) . This phenomenon is also evident in CD44+/CD24− cells . As shown by Gupta et al . ( 2011 ) , cancer cells grown in a uniform tissue culture in vitro microenvironment , when separated on the basis of the CD44 and CD24 cell surface markers , return to their original equilibrium proportion over a relatively short period of time . Therefore , to determine whether the selective vulnerabilities that we identified in our screen also typified stochastically generated CD44+/CD24− cells , we extended our analysis to four pairs of tumor-derived ( CD44+/CD24− ) lo cell lines that were FACS-sorted on the basis of their surface expression of CD44 and CD24 ( Figure 1C and Figure 1—figure supplement 3 ) . Overall , as illustrated in Figure 1C , we observed that the knockdown of the genes we identified in our original drop-out screen resulted in higher lethality in cells that reside in a CD44+/CD24− state compared to cells of other immune types . Cancer cells harbor many genetic alterations that , while important for tumorigenesis , can render them vulnerable to the loss of function of only one additional gene . To exclude the possibility that the selective vulnerabilities we observed in CD44+/CD24− cells were not due to a concealed genetic mutation but instead to an intrinsic property of the CD44+/CD24− state , we generated multiple single cell-derived cell lines from the H1650 cell line ( Figure 1D ) . Within these isogenic cell lines , we then compared , at early passages , the effect of IL-6 , BRCA1 , ORC5L , RFC3 , POLS , ERCC8 and RPA2 knockdown in FACS-sorted CD44+/CD24− cells and in cells of other immune types . Consistent with our previous data , we observed a decrease in the representation of CD44+/CD24− cells in the H1650 isogenic cell lines ( H1650-Isg-C , -Isg-D6 and -Isg-E4 ) upon silencing of the indicated genes ( Figure 1E and Figure 1—figure supplement 4 ) . The decrease in the number of CD44+/CD24− cells was not due to inter-conversion of cell states or a decrease in cell proliferation; rather , it was the result of increased lethality in the CD44+/CD24− cell populations upon silencing of the indicated genes . In fact , upon inactivation of IL-6 , BRCA1 , ORC5L , RFC3 , POLS , ERCC8 and RPA2 , we observed ( i ) an increase in the expression of the pro-apoptotic marker Cleaved caspase-3 by western blot analysis in H1650-M3 cells compared to parental H1650 cells ( Figure 1F , Figure 1—figure supplement 5 ) ; ( ii ) a higher number of Cleaved caspase-3 positive cells by FACS analysis in H1650 CD44+/CD24− cells compared to CD44−/CD24+ cells ( Figure 1G ) and ( iii ) a decrease in the number of colonies in the CD44+/CD24− cells in a standard clonogenic assay ( Figure 1—figure supplement 6 ) . Mutation analysis of BRCA1 , ORC5L , RFC3 , POLS , ERCC8 and RPA2 indicated that none of these genes were mutated in any of the cell lines utilized in our study ( Supplementary file 1 ) . Moreover , we neither observed any significant differences in the levels of their mRNA expression among the cell lines utilized ( Figure 2A and B and Figure 2—figure supplement 1 ) nor detected significant variations in the efficiency of their knockdown ( Figure 1—figure supplement 2C ) . 10 . 7554/eLife . 21615 . 010Figure 2 . Decreased expression of homology-directed repair ( HDR ) genes in CD44+/CD24− cells results in synthetic lethal interactions . ( A ) The heat map represents a hierarchical cluster analysis of BRCA1 , ORC5L , RFC3 , RPA2 , POLS and ERCC8 mRNA expression in the indicated tumor-derived cell lines and ( B ) in H1650 and A549 cells that were FACS sorted on the basis of their surface expression of CD44 and CD24 . mRNA expression was quantified by SYBR-green-based RT-qPCR . Cell lines with high CD44+/CD24− cell content are indicated in bold . The data represent mean ± SD of three replicates from two independent experiments . See Figure 2—figure supplement 1 for details . ( C ) Hierarchical cluster analysis of the mRNA expression of the indicated HDR genes in multiple tumor-derived cell lines with high ( indicated in bold ) or low content of CD44+/CD24− cells . mRNA expression was quantified by SYBR-green-based RT-qPCR . The data represent the mean ± SD of three replicates from two independent experiments . See Figure 2—figure supplement 2 for details . ( D ) Clustering analysis of mRNA expression of the HDR genes in CD44+/CD24− and CD44−/CD24+ cells FACS-sorted from H1650 , A549 and MCF7 cells . mRNA expression was quantified by SYBR-green-based RT-qPCR . The data represent the mean ± SD of three replicates from two independent experiments . See Figure 2—figure supplement 3A for details . ( E ) Expression of BLM , BRCA2 and NBN genes in FACS-sorted CD44−/CD24+ and CD44+/CD24− cells from four human primary NSCLC tumors . mRNA expression was quantified by SYBR-green-based RT-qPCR . Expression of an indicated mRNA in the CD44+/CD24− cells was calculated relative to its expression in CD44−/CD24+ cells from the respective tumor . Each dot represents mean ± SD of three replicates . See Figure 2—figure supplement 3B , C for details . ( F ) Schematic representation of the functional interactions between genes that we identified in the screen ( green ) and the HDR genes that we found to be downregulated ( orange ) in the H1650-M3 ( and CD44+/CD24− ) cells ( see Supplementary file 2 for details ) . ( G ) The charts depict the percentage of viable cells 5 days after knockdown of the indicated genes relative to a scramble-siRNA control in the A549 cell line . Each bar represents mean ± SD of eight replicates from two independent experiments . See Figure 2—figure supplement 5 for knockdown efficiency ( p-value <0 . 05 , <0 . 005 paired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01010 . 7554/eLife . 21615 . 011Figure 2—figure supplement 1 . mRNA expression analysis of the shRNA screen hits in tumor-derived cell lines and cells that have been FACS sorted on the basis of their surface expression of CD44 and CD24 . ( A ) mRNA expression analysis ( SYBR-green-based RT-qPCR ) of BRCA1 , ORC5L , RFC3 , RPA2 , POLS and ERCC8 in tumor-derived cell lines . The data represent the mean ± SD of three replicates from two independent experiments . ( B ) mRNA expression analysis ( SYBR-green-based RT-qPCR ) of BRCA1 , ORC5L , RFC3 , RPA2 , POLS and ERCC8 in H1650 and A549 that have been FACS sorted on the basis of their surface expression of CD44 and CD24 . Each bar represents the mean ± SD of three replicates from two independent experiments and represents the mRNA expression of the indicated gene normalized to GAPDH expression of each cell line . p-value <0 . 005 , unpaired t-test; ns= non-significant . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01110 . 7554/eLife . 21615 . 012Figure 2—figure supplement 2 . Differential mRNA expression of homology-directed repair ( HDR ) genes in multiple tumor-derived cell lines . The chart represents mRNA expression analysis ( SYBR-green-based RT-qPCR ) of the indicated HDR genes that we observed to be downregulated in the H1650-M3 ( CD44+/CD24− ) cells across the indicated cell lines . The data represent mean ± SD of three replicates from two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01210 . 7554/eLife . 21615 . 013Figure 2—figure supplement 3 . mRNA expression analysis of HDR genes in CD44−/CD24+ and CD44+/CD24− cells FACS sorted from cells lines and patient tumors . ( A ) Comparative expression of the indicated HDR genes in FACS-sorted CD44+/ CD24− cells relative to CD44−/ CD24+ cells from H1650 , A549 and MCF7 . mRNA expression was quantified by RT-qPCR . Each bar is the mean ± SD of three replicates from two different experiments and represents mRNA expression of the indicated gene . p-value <0 . 05 , <0 . 005 , unpaired t-test . ( B ) Schematic of cell sorting from tumors . Tumor-derived single cell suspension was stained with antibodies against CD45 , CD31 , EpCAM , CD44 , and CD24 . CD45-; CD31-; EpCAM+ cells were then FACS sorted according to the immune types CD44+/CD24− and CD44−/CD24+ . ( C ) Expression of BLM , BRCA2 , NBN and B-ACT genes in FACS sorted CD44−/CD24+ and CD44+/CD24− cells from four NSCL tumors . B-ACT was used as a housekeeping control gene . Each bar represents the mean ± SD of three replicates . p-value <0 . 05 , **<0 . 0005 , unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01310 . 7554/eLife . 21615 . 014Figure 2—figure supplement 4 . Differential expression of HDR genes in multiple tumor-derived cell lines at protein level . ( A ) Expression of BRCA2 , WRN and RDM1 in multiple cancer-derived cell lines characterized by different CD44 CD24 immune types . α-Tubulin is used as a loading control . ( B ) The charts depict the quantification of relative amounts of BRCA2 , WRN and RDM1 . Levels of intensity of each band were quantified using imajeJ32 software and are represented as a ratio to α-tubulin . Each bar represents the mean ± SD from two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01410 . 7554/eLife . 21615 . 015Figure 2—figure supplement 5 . Knockdown efficiencies of the indicated siRNAs in A549 cells . Knockdown efficiency upon transfection with the indicated siRNA oligonucleotides and compared to scramble siRNA-transfected control cells . Cells were collected 3 days post transfection and analyzed for expression of the indicated mRNA . Each bar represents mean ± SD of three replicates from two independent experiments and represents mRNA expression of the indicated gene normalized to GAPDH expression of each cell line . p-value <0 . 05 , *<0 . 005 , paired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 015 There is increasing evidence to suggest that mutations or the decreased expression of certain genome-stabilizing genes can weaken the intrinsic robustness of DNA repair mechanisms in tumor cells . Consequently , these cancer cells are rendered dependent on DNA damage/repair components distinct from the originally compromised gene ( Morandell and Yaffe , 2012 ) . Therefore , as a possible mechanism to explain the results of our functional genomic screen , we explored synthetic lethal interactions between gene products that have been reported to participate in DNA repair and thoseidentified in our screen . To this end , we generated gene expression profiles of H1650 and H1650-M3 cells and examined the relative abundance of DNA damage/repair pathway components ( see Supplementary file 2 ) , We did not observe major differences in the expression of genes involved in base excision repair , nucleotide excision repair , mismatch repair , or non-homologous end joining repair ( Supplementary file 2 ) . Yet , genes known to participate in HDR and DNA damage check-point regulation , such as BLM , BRCA2 , FANCF , NBN , PMS1 , RAD50 , RDM1 , WRN , ATM and ATR ( Ciccia and Elledge , 2010 ) , were expressed to a lesser extent in the CD44+/CD24− H1650-M3 cells than in the parental H1650 cells ( Supplementary file 2 ) . We confirmed the observed differences in expression by analyzing these HDR genes by RT-qPCR in the H1650 and H1650-M3 cells ( Figure 2C and Figure 2—figure supplement 2 ) . To determine whether this result was a common feature of cells in a CD44+/CD24− state , we then measured the expression of BLM , BRCA2 , FANCF , NBN , PMS1 , RAD50 , RDM1 , WRN , ATM and ATR ( i ) in the tumor-derived cell lines utilized in our functional genomic screen ( Figure 2C and Figure 2—figure supplement 2 ) , ( ii ) in FACS-sorted H1650 , A549 and MCF7 cells ( Figure 2D and Figure 2—figure supplement 3A ) and ( iii ) in four FACS-sorted primary human NSCLC tumors ( Figure 2E and Figure 2—figure supplement 3C ) . In all cases , we observed that cells lines with ( CD44+/CD24− ) hi content and FACS-sorted CD44+/CD24− cells exhibited decreased expression of these HDR genes . Western blot analysis of BRCA2 , RDM1 and WRN also indicated that the differential mRNA expression levels we observed were reflected by decreased protein expression ( Figure 2—figure supplement 4 ) . This observation was particularly interesting because many of the HDR genes that we detected as being downregulated in H1650-M3 cells have already been described as functionally interacting with the genes that we initially identified in our screen ( Figure 2F ) ( Ciccia and Elledge , 2010 ) . To provide experimental evidence in support of the synthetic lethal interactions between the genes identified in the shRNA screen and the genes that had decreased expression in CD44+/CD24− cells , we partially knocked down expression of BRCA2 and RAD50 in A549 cells and then tested the viability of the cells upon combined silencing with the BRCA1 or ORC5L genes ( essential in CD44+/CD24− cells ) ( Figure 2G and Figure 2—figure supplement 5 ) . We reasoned that decreasing the mRNA levels of BRCA2 and RAD50 would mimic their low expression in CD44+/CD24− cells . As a result , we found that the combined knockdown of BRCA2 and BRCA1 , RAD50 and BRCA1 , or RAD50 and ORC5L resulted in a significant decrease in cell viability , even though the individual inactivation of ORC5L , BRCA1 , BRCA2 or RAD50 had no significant effect on the viability of A549 cells ( Figure 2G ) . Cells in a CD44+/CD24− state are characterized by constitutive activation of the TGF-β axis that is both necessary and sufficient for the maintenance and/or acquisition of many of the features that characterize the CD44+/CD24− cell state ( Figure 3A and B and Figure 3—figure supplement 1 ) ( Mani et al . , 2008 ) . As previous studies have also indicated that TGF-β can modify the activity and expression of certain genes involved in DNA repair ( Glick et al . , 1996; Harris et al . , 1997; Preobrazhenska et al . , 2002; Kirshner et al . , 2006; Wiegman et al . , 2007 ) , we examined whether TGF-β could be responsible for the lower expression of the HDR genes that we observed in CD44+/CD24− cells . 10 . 7554/eLife . 21615 . 016Figure 3 . The TGF-β axis controls the expression of the HDR genes that are downregulated in CD44+/CD24− cells . ( A ) TGF-β signaling is required for CD44+/CD24− cell state transition . The schematic provided here is based on current literature ( Korkaya et al . , 2011; Mani et al . , 2008 ) . ( B ) mRNA expression analysis of well-known TGF-β target genes in CD44+/CD24− cells relative to CD44−/ CD24+ cells FACS-sorted from H1650 and MCF7 cell lines . mRNA expression was quantified by SYBR-green-based RT-qPCR . Each dot represents the mean ± SD of three replicates from two independent experiments . See Figure 3—figure supplement 1 for details . ( C ) FACS analysis of H1650 cells exposed to TGF-β . Cells were treated with TGF-β for the indicated days and stained for the surface expression of CD44 and CD24 and analyzed by FACS . ( D ) mRNA expression analysis of the indicated HDR genes in TGF-β-treated cells relative to vehicle control across multiple tumor-derived cell lines . Cells were treated for 9 hr with TGF-β1 and TGF-β2 ( 1 ng/ml each ) . mRNA expression was quantified by SYBR-green-based RT-qPCR . Each dot represents the mean ± SD of three replicates from two independent experiments . See Figure 3—figure supplement 2B and Figure 3—figure supplement 3 for additional details . ( E ) The inhibition of TGF-β signaling in the CD44+/CD24− H1650-M3 cells results in an increased expression of HDR genes . TGF-β receptor 1 ( TGFBR1 ) kinase activity was blocked by treatment with 20 μM of LY2157299 ( Selleckchem ) for 72–96 hr . Expression of TGF-β signature genes were used as a control for the efficacy of LY2157299 treatment . mRNA expression was quantified through SYBR-green-based RT-qPCR . Each bar represents the mean ± SD of three replicates from three independent experiments ( p-value <0 . 05 , *<0 . 005 paired t-test ) . ( F ) Meta-analysis of human breast tumor dataset ( BRCA ) generated by the TCGA Research Network: http://cancergenome . nih . gov/ . Average Pearson correlation coefficients ( PCCs ) between every pair of genes displayed were calculated and the matrix was generated . Pearson correlation coefficients ranged from −1 to +1 . The matrix indicates an inverse co-regulation of the TGF-β1 and the HDR genes that we found to be downregulated in CD44+/CD24− cells . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01610 . 7554/eLife . 21615 . 017Figure 3—figure supplement 1 . TGF-β signaling is upregulated in FACS-sorted CD44+/CD24− cells . Expression of well-known TGF-β target genes in CD44+/ CD24− cells relative to CD44−/ CD24+ cells sorted from H1650 and MCF7 . mRNA expression was quantified by SYBR-green-based RT-qPCR . Each bar represents the mean ± SD of three replicates from two independent experiments and represents mRNA expression of the indicated gene normalized to GAPDH expression of each cell line ( n = 6 ) . p-value <0 . 05 , *<0 . 005 unpaired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01710 . 7554/eLife . 21615 . 018Figure 3—figure supplement 2 . Active TGF-β signaling control the expression of multiple HDR genes . ( A ) TGF-β signature target genes measured by SYBR-green-based RT-qPCR in H1650 and A549 upon treating the cells with TGF-β for 9 hr . mRNA expression was quantified by RT-qPCR . Each bar represents the mean ± SD of three replicates from three independent experiments and represents mRNA expression of the indicated gene . p-value <0 . 05 , <0 . 005 , <0 . 0005 , paired t-test . ( B ) mRNA expression analysis of indicated HDR genes in TGF-β-treated cells relative to vehicle control in the indicated cell lines . mRNA expression was quantified by SYBR-green-based RT-qPCR . Each bar represents the mean ± SD of three replicates from two independent experiments ( n = 6 ) , normalized to respective GAPDH mRNA expression . Cells were treated for 9 hr with TGF-β ( 1 ng/ml each of TGF-β1 and TGF-β2 ) . p-value <0 . 05 , <0 . 005 <0 . 0005 , paired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01810 . 7554/eLife . 21615 . 019Figure 3—figure supplement 3 . TGF-β signaling controls the expression of multiple HDR genes at the protein level . ( A ) Western blot analysis of the expression of BRCA2 , BLM and RAD50 upon vehicle or TGF-β treatment ( 9 hr , 1 day , 2 day , 3 day or 4 day ) in H1650 cells . Phospho-Smad2 and E-cadherin are used as a marker of active TGF-β signaling and MET , respectively . α-Tubulin is used as a loading control . ( B ) The charts depict the quantification of the relative amounts of BRCA2 , BLM and RAD50 . Levels of intensity of each band were quantified using imajeJ32 software , represented as a ratio of the protein of interest to α-tubulin , and normalized to levels detected in respective TGF-β untreated ( –TGF-β ) samples . The data represent the mean ± SD from two independent experiments . p-value <0 . 005 , **<0 . 001 , unpaired t-test with Welch’s correction . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 01910 . 7554/eLife . 21615 . 020Figure 3—figure supplement 4 . Inhibition of TGF-β signaling in the CD44+/CD24− H1650-M3 cells results in an increased expression of HDR genes . ( A ) TGF-β receptor 1 ( TGFBR1 ) kinase activity was blocked by treatment with 1 uM LY364947 ( Selleckchem ) for 48–72 hr . Expression of TGF-β signature genes were used as a control for the efficacy of LY364947 treatment . mRNA expression was quantified through SYBR-green-based RT-qPCR . Each bar represents the mean ± SD of three replicates from two independent experiments ( p-value <0 . 05 , <0 . 005 paired t-test ) . ( B ) Western blot analysis of H1650 and H1650-M3 cells treated with LY364947 ( 1 uM ) shows that the inhibitor specifically inhibits TGF-β signaling , as phosphorylation of Smad2 and Smad3 are repressed after 48 hr of treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 02010 . 7554/eLife . 21615 . 021Figure 3—figure supplement 5 . Exposure to TGF-β does not alter the cell cycle distribution in cancer cell lines . ( A ) TGF-β treatment does not alter the cell cycle distribution in U2OS cells . The histograms represent U2OS cell cycle distribution without or with TGF-β ( 1 ng/ml of each of TGF-β1 and TGF-β2 , 9 hr ) . 10 , 000 cells were analyzed by FACS for each replicate of each sample . ( B ) Quantification of cell cycle distribution from U2OS and H1650 cells upon 9 hr treatment with TGF-β ( 1 ng/ml of each of TGF-β1 and TGF-β2 ) . Each bar represents the mean ± SD of three replicates from two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 021 By exposing cells to exogenous TGF-β , we confirmed that the TGF-β axis was sufficient not only to induce the acquisition of mesenchymal-like features ( Figure 3—figure supplement 2A ) and the transition into the CD44+/CD24− state ( Figure 3C ) but also to decrease the expression of BLM , BRCA2 , FANCF , NBN , PMS1 , RAD50 , RDM1 , WRN , ATM and ATR in multiple tumor-derived cell lines ( Figure 3D and Figure 3—figure supplement 2B ) . In the case of BRCA2 , BLM and RAD50 , we confirmed that TGF-β also induced a decrease in protein expression as assessed through western blot analyses of cell extracts ( Figure 3—figure supplement 3 ) . Conversely , treatment of the CD44+/CD24− H1650-M3 cells with two selective inhibitors of TGFBRI ( LY2157299 and LY364947 ) augmented the expression of ATR , BLM , BRCA2 , FANCF , NBN , PMS1 and RAD50 and , as expected , decreased the expression of the TGF-β-induced genes SERPINE1 , SNAIL , SKIL and ZEB1 ( Figure 3E and Figure 3—figure supplement 4 ) . Altogether , these data indicate that TGF-β-mediated signaling is sufficient and necessary to decrease the expression of the HDR genes that we observed to be downregulated in CD44+/CD24− cells . Next , to assess the relevance of this observation in human tumors , we mined a human breast tumor dataset ( BRCA ) generated by the TCGA Research Network ( http://cancergenome . nih . gov/ ) ( Figure 3F ) . We found that , consistent with our previous observations , both TGFB1 and the TGF-β signature genes were inversely correlated with the expression of BLM , BRCA2 , FANCF , NBN , PMS1 , RAD50 , RDM1 , WRN , ATM and ATR ( Figure 3F ) . It has been reported that TGF-β can induce cell cycle arrest . Yet , in H1650 and U2OS cells , although TGF-β reduced the expression of DNA repair genes within 9 hr ( Figure 3D ) , it did not induce any significant changes in the distribution of cells in G1/S/G2 within the same timeframe ( Figure 3—figure supplement 5 ) . This observation strongly excludes the possibility that our findings could simply be explained by cell cycle alterations . From a functional standpoint , a decrease in the expression of HDR genes could result in an increased accumulation of DNA DSB ( Kwei et al . , 2010 ) . To test whether this was the case in CD44+/CD24− cells , we compared γ-H2AX foci formation in FACS-sorted CD44+/CD24− and CD44−/CD24+ cells from H1650 . We found that CD44+/CD24− cells had a higher number of γ-H2AX positive foci ( Figure 4A and B ) . By staining the cells with 53BP1 , a scaffold protein for DSB- responsive factors , we next assess if the increase in these foci was a consequence of a defect in DNA repair or DNA shredding ( Panier and Boulton , 2014 ) . In accordance with intrinsic defects in DNA damage/repair , we observed an increased occurrence of 53BP1/γ-H2AX double positive foci in CD44+/CD24− cells ( Figure 4A and B ) . 10 . 7554/eLife . 21615 . 022Figure 4 . CD44+/CD24− cells are characterized by defect in DNA damage repair . ( A ) CD44+/CD24− cells are characterized by increased DNA DSBs compared with CD44−/CD24+ cells . CD44+/CD24− and CD44−/CD24+ cells sorted from H1650 were stained with antibodies against γ-H2AX ( red ) and 53BP1 ( green ) . DAPI ( blue ) was used as a counter-stain . Insets in the left upper corner show a representative nucleus . ( B ) The chart represents quantification of the experiment depicted in ( A ) . Each bar represents the mean ± SD of the percentage of cells with more than three γ-H2AX and 53BP1 double-positive foci per field in CD44−/CD24+ cells and CD44+/CD24− cells , FACS-sorted from the H1650 cell line . Approximately ten fields were counted , for a total of 100 cells ( n = 100 ) ( p-value <0 . 005 , unpaired t-test ) . ( C ) H1650 and A549 cells treated with vehicle ( DMSO ) or TGF-β ( 1 ng/ml of each of TGF-β1 and -β2 , for 4 days ) were stained with antibodies against γ-H2AX ( red ) and 53BP1 ( green ) . DAPI ( blue ) was used as a counter-stain . Insets in the left upper corner show a representative nucleus staining; analysis of an additional cell line is provided in Figure 4—figure supplement 1 . ( D ) The chart represents quantification of the experiment depicted in ( C ) . Each bar represents the mean ± SD of the percentage of cells with more than three γ-H2AX and 53BP1 double-positive foci per field in vehicle or TGF-β treated cells . Approximately ten fields were counted , for a total of 100 cells ( n = 100 ) . ( p-value <0 . 05 , <0 . 005 , paired t-test ) . Doxorubicin treatment ( 10 μM for 24 hr ) was used as a positive control . ( E ) Schematic of Comet assay . ( F ) Comet assay in H1650 , H1650-M3 ( CD44+/CD24− ) and TGF-β-treated H1650 cells ( treatment for 5 days ) , showing increased DNA strand breaks . ( G ) The chart represents quantification of the mean ± SD of tail movement of the samples depicted in ( F ) , conducted in three replicates in two independent experiments . ( p-value <0 . 005 , <0 . 0005 , unpaired t-test ) . ( H ) Schematic of the DR-GFP assay . ( I ) The chart indicates the percentage of GFP-positive cells upon transfection with pCBASce-I ( expressing Sce-I endonuclease ) compared with control cells ( untransfected cells ) . siRNA-mediated knock-down of RAD50 and BRCA2 was used as a homologous recombination ( HR ) efficiency control . Each bar represents mean ± SD of three replicates from two independent experiments ( n = 6 ) ( p-value <0 . 0005 , paired t-test ) . See Figure 4—figure supplement 3 for knockdown efficiencies with the indicated siRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 02210 . 7554/eLife . 21615 . 023Figure 4—figure supplement 1 . TGF-β treatment increases the number of 53BP1/γ-H2AX double-positive foci in MCF7 cells . ( A ) MCF7 cells treated with TGF-β ( 1 ng/ml each of TGF-β1 and TGF-β2 ) or DMSO for 3 days and 4 days were stained with antibodies against γ-H2AX ( red ) and 53BP1 ( green ) . DAPI ( blue ) was used as a counter-stain . Insets in the left upper corner show a representative nucleus . ( B ) The chart represents quantification of the experiment depicted in ( A ) . Each bar represents the mean ± SD of the percentage of cells with more than threeγ-H2AX and 53BP1 double-positive foci per field in vehicle or TGF-β-treated cells . Approximately ten fields were counted , for a total of 100 cells ( n = 100 ) . ( p-value <0 . 005 , paired t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 02310 . 7554/eLife . 21615 . 024Figure 4—figure supplement 2 . TGF-β reduces the efficiency of homologous recombination in a DR-GFP assay in H1650 cells . The chart represents the percentage of GFP-positive cells upon transfection with pDR-GFP or pCBASce-I or both , under vehicle or TGF-β treatment . Each bar represents mean ± SD of three replicates from two independent experiments . ( p-value <0 . 005 , unpaired t-test with desired FDR 1% . ) DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 02410 . 7554/eLife . 21615 . 025Figure 4—figure supplement 3 . Knockdown efficiencies of the indicated siRNAs in U2OS-DR-GFP cells . Efficiency of knockdown upon transfection with the indicated siRNA oligonucleotides , without or with pCBASce-I ( plasmid expressing Sce-I endonuclease ) , and compared to scramble-siRNA-transfected control cells . Cells were collected 3 days post-transfection and analyzed for expression of the indicated mRNA . Each bar represents the mean ± SD of three replicates from two independent experiments . p-value <0 . 005 , *<0 . 0005 , paired t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 025 As an independent approach for evaluating DNA strand breaks , we also performed a comet assay ( Figure 4E ) . The larger mean comet-tail movement in H1650-M3 cells compared with H1650 cells confirmed a higher number of DSBs in CD44+/CD24− cells ( Figure 4F and G ) . As we could show that TGF-β was both necessary and sufficient to regulate HDR genes , we next extended our analysis to A549 , H1650 and MCF7 cells treated with TGF-β . Also in this case , we found that TGF-β treatment was sufficient to increase the number of 53BP1/γ-H2AX double-positive foci and the mean comet tail movement ( Figure 4C , D , F and G and Figure 4—figure supplement 1 ) . In principle , an increased number of DNA breaks could be explained either by a decreased capability of DNA repair or by increased DNA damage . Hence , we used a homologous recombination reporter system ( the DR-GFP reporter system ) to further investigate the possibility that TGF-β exposure could result in defects in DNA repair ( Figure 4H ) ( Pierce et al . , 1999; Nakanishi et al . , 2005; Gunn and Stark , 2012 ) . This system is based on a non-crossover gene-conversion mechanism of two mutated GFP genes: ( i ) the SceGFP that is disrupted by an 18 bp recognition site for the I-SceI endonuclease and ( ii ) the iGFP that is truncated at the 5′ and 3′ ends . Upon transfection with pCBASce-I ( expressing I-SceI ) , the SceGFP is cleaved and , because of an HDR event , a functional GFP gene ( detectable by flow cytometry ) is generated through gene conversion with iGFP . By using this assay , upon transfection with pCBASce-I , we observed a significant reduction in GFP-positive cells among U2OS cells as well as among H1650 cells treated with TGF-β ( Figure 4I and Figure 4—figure supplement 2 ) . As a control , we knocked down expression of BRCA2 and RAD50 , two well-known HDR genes ( Figure 4I and Figure 4—figure supplement 3 ) . Homology-directed repair deficiency is reported to induce a switch to non-homologous break repair mechanisms , such as NHEJ or MMBIR , which can lead to the accumulation of DNA ‘joint points’ where the non-homologous repairs occur ( Daley et al . , 2005; Hastings et al . , 2009 ) . These events have the potential to lead to chromosomal translocation , allelic imbalance and a specific DNA copy number profile named ‘saw-tooth’ ( Kwei et al . , 2010 ) . Whole-genome CNA analysis in FACS-sorted CD44+/CD24− and CD44−/CD24+ cells from three human NSCLC tumors and a tumor-derived cell line showed that the CD44+/CD24− cells possessed a higher content of DNA joint points and the distinctive saw-tooth profile ( Kwei et al . , 2010 ) ( Figure 5A , B , C and F and Figure 5—figure supplement 1A ) . In addition , CNA analysis of a H1650 isogenic cell line ( H1650-Isg-E4 ) also established that this was due neither to concealed genetic mutations ( Figure 5D , E and F and Figure 5—figure supplement 1B ) nor to the expansion of a single clonal population with high DNA DSBs ( Figure 5G and Figure 5—figure supplement 1B ) , but rather to an intrinsic property of CD44+/CD24− cells . 10 . 7554/eLife . 21615 . 026Figure 5 . CD44+/CD24− cells have higher copy number alterations and increased genetic diversity . ( A ) The graph illustrates a representative copy number profile of CD44−/CD24+ cells and CD44+/CD24− cells sorted from an NSCLC patient ( Patient #4 ) . The x-axis corresponds to bins across the genome space from chr1 on the left to the sex chromosomes on the right . The y-axis corresponds to the copy number value at each bin . ( B ) The chart represents the number of DNA joint points in FACS-sorted CD44−/CD24+ cells ( blue circle ) and CD44+/CD24− cells ( orange circle ) from the indicated human NSCLC tumor . Each dot represents the analysis of a single cell . The breakpoint matrix ( utilized to calculate DNA joint points ) , cluster dendogram and heat-map of normalized read counts ( Figure 5—figure supplement 1A ) were generated using Ginkgo , an open-source web platform for interactive analyses of CNA . ( Garvin et al . , 2015 ) ( http://qb . cshl . edu/ginkgo ) . A variable bin size of 175 kb is used . p-value < 0 . 05 , unpaired t-test with Welch’s correction . Error bars indicate standard deviation . ( C ) The chart represents number of DNA joint points in CD44−/CD24+ ( blue circle ) and CD44+/CD24− ( red circle ) cells FACS-sorted from the indicated primary human NSCLC . Each dot represents the analysis of a cell type collected and sequenced in bulk . ( D ) The graph illustrates a representative copy number profile of one CD44−/CD24+ and one CD44+/CD24− FACS-sorted cell from the H1650-derived isogenic cell line H1650-Isg-E4 . ( E ) The chart represents the number of DNA joint points in FACS-sorted CD44−/CD24+ cells ( blue circle ) and CD44+/CD24− cells ( orange squares ) from the H1650-Isg-E4 cell line . Each dot represents the analysis of a single cell . The breakpoint matrix ( utilized to calculate DNA joint points ) is generated along with the cluster dendogram and heat-map of normalized read-counts ( Figure 5—figure supplement 1B ) using Ginkgo . A variable bin size of 175 kb is used . p-value <0 . 05 , unpaired t-test with Welch’s correction . Error bars indicate standard deviation . ( F ) The chart depicts the number of DNA joint points in CD44−/CD24+ ( blue circle ) and CD44+/CD24− ( red square ) cells FACS-sorted from the H1650 ( parental ) and H1650 isogenic Isg-E4 cell lines . Each dot represents the analysis of a cell type collected and sequenced in bulk . ( G ) Cluster dendogram of normalized read-counts across segment breakpoints ( using Euclidian distance and the ward-clustering method ) of CD44−/CD24+ cells ( blue circle ) and CD44+/CD24− cells ( orange squares ) FACS-sorted from the H1650-Isg-E4 cell line . Each dot represents a single cell . The cluster dendogram is generated with Ginkgo . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 02610 . 7554/eLife . 21615 . 027Figure 5—figure supplement 1 . Heat-map of normalized read counts of FACS-sorted cells based on CD44 and CD24 surface markers from patient tumor and from the H1650-derived isogenic cell line H1650-Isg-E4 . Heat-maps of normalized read counts across segment breakpoints ( using Euclidian distance and ward clustering method ) of CD44+/CD24− cells and CD44−/CD24+ cells sorted from ( A ) an NSCL tumor ( patient #4 ) and ( B ) the H1650-derived isogenic cell line ( H1650-isg-E4 ) are displayed . CD44−/ CD24+ cells are marked with a blue circle and CD44+/CD24− cells are marked with an orange circle . The cluster dendogram and heat-map of normalized read counts were generated using Ginkgo , an open-source web platform for interactive analyses of CNA ( Garvin et al . , 2015 ) ( http://qb . cshl . edu/ginkgo ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 027 On the basis of these observations , we concluded that the decreased expression of HDR genes and a reduced DNA repair proficiency could result in a hyper-mutable phenotype and an increased genetic diversity of CD44+/CD24− cells . We showed that TGF-β was both necessary and sufficient to reduce the expression of HDR genes and to increase the number of DNA DSBs in CD44+/CD24− cells . To assess whether TGF-β treatment was also sufficient to increase the number of CNAs and the genetic diversity of cancer cells , we exposed H1650 isogenic cells ( H1650-Isg-D6 ) to TGF-β for six weeks and then performed whole-genome CNA analysis . As genetic alterations are stable , to exclude a possible direct effect of TGF-β-mediated signaling in our analysis of CNA , we performed our studies upon TGF-β removal . Much like our observations in CD44+/CD24− cells , we detected an overall expansion in the number of DNA joint points and an increased genetic heterogeneity of H1650-Isg-D6 cells upon exposure to TGF-β ( Figure 6A , B and C ) . 10 . 7554/eLife . 21615 . 028Figure 6 . Exposure to TGF-β is sufficient to increase CNA and genetic diversity of the cell population . ( A ) The graph illustrates a representative copy-number profile of one TGF-β-naïve cell and one TGF-β-treated ( TGF-b 6w ) cells from the H1650 isogenic/single cell-derived cell line Isg-D6 . H1650-Isg-D6 was treated with vehicle ( DMSO ) or TGF-β ( 1 ng/ml of each of TGF-β1 and -β2 ) for six weeks . Single cell CNA analysis was performed upon TGF-β withdrawal . The x-axis corresponds to bins across the genome space from chr1 on the left to the sex chromosomes on the right . The y-axis corresponds to the copy number value at each bin . ( B ) Heat map of normalized read counts of TGF-β-naïve and TGF-β-treated ( TGF-b 6w ) cells from the H1650-Isg-D6 cell line , across segment breakpoints with a variable bin size of 50 kb ( using Euclidian distance and the ward clustering method ) . Each horizontal line across the y-axis represents an individual cell , whereas the x-axis annotates the CNA across chromosomes from chr1 on the left to the sex chromosomes on the right . ( C ) The chart represents the number of DNA joint points in TGF-β-naïve ( blue circle ) and TGF-b 6w ( purple squares ) cells from the H1650-Isg-D6 cell line . Each dot represents the analysis of a single cell . The breakpoint matrix ( utilized to calculate DNA joint points ) is generated along with the cluster dendogram and heat-map of normalized read-counts ( as shown in ( B ) ) using Ginkgo with a variable bin size of 50 kb ( http://qb . cshl . edu/ginkgo ) . p-value =0 . 0064 , unpaired t-test with Welch’s correction . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 028 Since the conception of evolutionary reasoning , it has become evident that the stability and robustness of ecosystems depend on their diversity . In line with this principle , Luria and Delbrück ( 1943 ) proposed that the genetic diversity of a bacterial population was responsible for the resistance of the population to bacteriophage infection ( Luria and Delbrück , 1943 ) . In keeping with this principle , to evaluate whether the defects in DNA damage/repair and the consequent genetic heterogeneity that we observed in CD44+/CD24− cells were sufficient to increase both the phenotypic diversity and adaptability of a cancer cell population , we artificially induced cells into a CD44+/CD24− state and tested the cells’ capabilities to adapt to different perturbations ( i . e . , exposure to different drugs ) . We took advantage of the fact that when treated with TGF-β , A549 cells transit into a CD44+/CD24− state ( i . e . , differential surface expression of CD44 and CD24 , increased expression of VIM , SNAI1 and IL-6 and decreased expression of both E-cadherin ( CDH1 ) and the HDR genes described previously ) , but upon TGF-β withdrawal , they return to their original epithelial state ( Figure 7A and Band Figure 7—figure supplement 1 ) . We found that transient exposure to TGF-β results in an increased genetic diversity ( Figure 7C ) . The reversibility of the CD44+/CD24− cell state was crucial to test our hypothesis as it enabled us to determine whether an increased drug resistance was due to the effects of TGF-β on genomic instability and clonal diversity , or to the acute and transient activation of drug-resistance signaling pathways . 10 . 7554/eLife . 21615 . 029Figure 7 . The TGF-β-induced CD44+/CD24− cell state increases the adaptability of cell populations . ( A and B ) A549 cells exposed to TGF-β acquire phenotypic and molecular changes characteristic of a CD44+/CD24− cell state . Upon TGF-β withdrawal , the cells return to their original cell state , as indicated by the FACS analysis in ( A ) and the RT-PCR analysis in ( B ) . See Figure 7—figure supplement 1 for quantification of FACS analysis . ( C ) The chart indicates the number of DNA joint points in TGF-β-naïve and TGF-β-treated A549 cells . Each dot represents the analysis of a single cell . The breakpoint matrix ( utilized to calculate DNA joint points ) is generated using Ginkgo ( http://qb . cshl . edu/ginkgo ) . A variable bin size of 175 kb is used . p-value <0 . 05 , unpaired t-test . ( D ) The table depicts the IC50 values of A549 , A549-3W△ and A549-6W△ cells in the context of treatment with the indicated drugs . ( E and F ) TGF-β treatment increased the adaptability of cells . Cells that were transiently exposed to TGF-β for 3 or 6 weeks ( A549-3W△ , A549-6W△ ) were then treated with the indicated drugs upon TGF-β withdrawl . ( E ) The number of colonies ( mean ± SD ) that have survived epitaxol ( 1 . 6 μM ) , tunicamycin ( 3 . 2 μM ) and cisplatin ( 1 mM ) treatment . Notably , the concentration of drugs used in this experiment corresponds to approximately >100X the IC50 . Two independent experiments , each with three replicates , were carried out and approximately 50 fields were counted for each sample . ( p-value <0 . 05 , <0 . 005 , paired t-test . ) ( F ) A549 tunicamycin-resistant clones were grown in regular/drug-free medium for a week and then retested for sensitivity to tunicamycin ( 3 . 2 μM ) or epitaxol ( 1 . 6 μM ) . The plot represents mean ± SD number of colonies surviving 5 days after treatment compared with untreated cells , from two independent experiments each with three replicates ( p-value *<0 . 005 , paired t-test ) . ( G ) Heat map of normalized read counts across segment breakpoints ( using Euclidian distance and the ward-clustering method ) of the indicated cells . Each horizontal line across the y-axis represents an individual cell , whereas the x-axis annotates the CNA across chromosomes from chr1 on the left to the sex chromosomes on the right . A heat map of cisplatin-naïve cells is shown on the left and of cisplatin-resistant cells on the right . ( H ) Cluster dendogram of normalized read counts across segment breakpoints ( using Euclidian distance and the ward-clustering method ) of cisplatin-resistant-A549 , cisplatin-resistant-A549-3W△ and cisplatin-resistant-A549-6w△ . The cluster dendogram and heat-map of normalized read counts were generated using Ginkgo . ( I ) Schematic of proposed model . When cells transit into a CD44+/CD24− state , they acquire mesenchymal-like features and autocrine secretion of TGF-β that leads to the downregulation of HDR genes . This process results in a hyper-mutable phenotype that spurs genetic diversity and intra-tumor clonal heterogeneity . Consequently , following a Darwinian model of cancer evolution , the transition of cancer cells into a CD44+/CD24− state or exposure to TGF-β leads to an increased adaptability to any given perturbation . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 02910 . 7554/eLife . 21615 . 030Figure 7—figure supplement 1 . A549 cells undergo changes in the surface expression of CD44 and CD24 before , during and after exposure to TGF-β . The chart depicts the percentage of of A549 cells in the CD44+/CD24− state and the CD44−/ CD24+ state , before , during and post exposure to TGF-β . A549 cells before , during and post exposure to TGF-β were stained for surface expression of CD44 and CD24 and analyzed by FACS . Each bar represents the mean ± SD of three replicates from three independent experiments . 10 , 000 cells were analyzed for each replicate of each sample . DOI: http://dx . doi . org/10 . 7554/eLife . 21615 . 030 When we treated A549 TGF-β-naïve cells and A549 cells that were transiently exposed to TGF-β for 3 weeks and 6 weeks ( A549-3w△ and A549-6w△ , respectively ) with different concentrations of epitaxol , cisplatin , doxorubicin , erlotinib and tunicamycin , we found that they had similar half maximal inhibitory concentrations ( IC50s ) ( Figure 7D ) This was not the case when cells were treated with very high concentrations of tunicamycin , epitaxol and cisplatin . In this experiment , the number of colonies that could survive the drug treatment was in fact consistently and significantly higher for A549-3w△ and A549-6w△ cells ( Figure 7E ) . In principle , the increased adaptability that we observed in A549-3w△ and A549-6w△ cells upon exposure to high drug concentrations could be explained by the selection of a particular clonal population during TGF-β treatment . If this were the case , cells that were resistant to one drug would be equally resistant to the other drug treatments . But when we tested the sensitivity of tunicamycin-resistant clones to tunicamycin and taxol , we observed that these cells retained resistance to tunicamycin but were still sensitive to taxol . This result indicated that any particular clonal population was not selected but rather presented the possibility that their resistance was due to diverse molecular mechanisms ( Figure 7F ) . Whole genome sequencing of cisplatin-resistant clones further supported this likelihood ( Figure 7G and H ) . Individual malignant cells within a tumor can possess a wide variety of traits ( e . g . , affecting growth , metabolism , motility , morphology , stress responses , etc . ) that could ultimately impact the progression and recurrence of cancer ( Greaves and Maley , 2012; Almendro et al . , 2013; Burrell et al . , 2013; Kreso and Dick , 2014 ) . The acquisition of these distinct features may result from genetic mutations and/or non-genetic determinants , such as the fluctuation among different cell states resulting from epigenetic/stochastic mechanisms or exposure to cues present in the micro-environment . Here , we have shown that cells that reside in a CD44+/CD24− state are characterized by intrinsic defects in their abilities to repair DNA DSBs , leading to both augmented genetic instability and clonal diversity . Our results demonstrate , for the first time , that the oscillation between different cell states is not ‘genetically’ neutral . Instead , much like stress-induced mutagenesis , the oscillation between cell states can both promote the acquisition of new mutations and increase the total phenotypic and genomic diversity of a cancer population ( Meng et al . , 2005; Ponder et al . , 2005; Chan et al . , 2008; Shee et al . , 2011; Gutierrez et al . , 2013; Fitzgerald et al . , 2017 ) . Thus , the CD44+/CD24− state not only can serve as a substrate for but also can spur tumor evolution by promoting continued acquisition of genetic diversity . When we consider that cancer cells within a tumor can die , proliferate , enter dormancy and/or exhaust their long-term clonal growth , then therapy failure and reoccurrence should be attributed not only to the acquisition of new mutations , but also to the surviving cells’ ability to propagate clonally . Consequently , the tumor cells contributing to recurrence must have regenerative potential and behave like cancer stem cells . In light of these considerations , the fact that CD44+/CD24− cells have self-renewal capabilities and have been referred to as ‘cancer stem cells’ is particularly captivating ( Al-Hajj et al . , 2003; Mani et al . , 2008; Korkaya et al . , 2011; Brooks et al . , 2015 ) . Another interesting feature of CD44+/CD24− cells is their general sturdiness and ability to survive exposure to lethal stimuli when compared to cells in other cell states ( Yao et al . , 2010 ) . As noted above , CD44+/CD24− cells express higher levels of genes that are known to protect cells from pro-apoptotic stimuli , including BCL-2 , BCL-XL and MCL-1 ( Keitel et al . , 2014 ) . It is , therefore , tempting to propose that the transition of cancer cells into a CD44+/CD24− state could provide not only a small sub-population of tumor cells that can withstand and survive an initial destructive drug attack , but also the means to accumulate new genetic mutations that will further increase their fitness , ultimately allowing their growth and expansion even in the presence of a drug . Our data indicated that TGF-β-mediated signaling is both necessary and sufficient to reduce the expression of HDR genes and to increase both CNA and the genetic heterogeneity of cancer cell populations ( Figures 6 and 7 ) . TGF-β is not only produced by CD44+/CD24− cells but it is among the many factors that mediate the communication between the cancer cells and their surrounding stroma ( Massagué , 2008 ) . Hence , our findings are particularly exciting because , in principle , they could provide additional support for the role of the tumor microenvironment in shaping tumor fitness . This could be especially relevant in the case of chronic inflammatory conditions and current cancer treatments , as both have been shown to increase TGF-β levels in tissues . Consequently , we can view chemotherapy regimens as double-edged swords . While they can kill cancer cells by inducing inflammation and TGF-β , they can also spur cancer cell evolution and paradoxically contribute to drug resistance . In other words , as noted by Dr Huang in one of his current review articles , our observations may echo Nietzsche’s classic statement: ‘What does not kill me , makes me stronger’ ( Pisco and Huang , 2015 ) . This could be particular relevant when designing cancer treatments because the progression-stimulating effects of a therapy could shape the behavior of the tumor ( Pisco and Huang , 2015 ) . Whereas exposure to TGF-β in malignant cells often elicits pro-tumorigenic activities , in non-transformed cells , it can instead induce cell cycle arrest , senescence and apoptosis through a p53- , p21- and p16-dependent mechanism ( Massagué , 2008 ) . As persistent DSB could elicit these same effects , the downregulation of DNA-repair genes and the consequent induction of DNA damage by TGF-β in non-transformed cells could be part of the tumor-suppressor activities of TGF-β . As often is the case in cancer , a mechanism that has been implemented during evolution to protect tissue homeostasis could become hijacked by cancer cells to promote their survival . In this regard , the observation that the p53 tumor suppressor network is disabled in cancer cells is particularly interesting . TGF-β plays a role not only in the context of tumorigenesis but also during development and in adult organism by regulating tissue-injury responses , adaptive and innate immune responses , tissue homeostasis and so on ( Massagué and Gomis , 2006 ) . The possibility that TGF-β could impair DNA-damage responses not only in the context of cancer cells but also in normal somatic cells is intriguing . This , in fact , could provide a means to activate a tumor suppressor network that is based on DNA-damage responses . Consequently , TGF-β=mediated impaired HDR could be part of an intrinsic cell mechanism that balances cell proliferation in tissues . On the other hand , it could also be possible that when DNA-damage-repair mechanisms are dimmed , the genetic diversity of cells in tissues is increased . Herein , inhibition of HDR by TGF-β could provide a means to increase the heterogeneity of cells within tissues during development . This phenomenon could be important in the context of organs or tissues in which diversity could increase functionality as , for example , in the case of immune cells or neurons ( Rehen et al . , 2001 , 2005 ) . Interestingly in this regard , recent data have indicated that certain somatic cells are not genetically uniform but are characterized by different ploidy ( Duncan et al . , 2010 , 2012 ) . HDR-mediated repair is also utilized by viruses and transposons ( Yant and Kay , 2003 ) . Hence , the downregulation of component of this DNA-repair pathway by TGF-β could be part of innate mechanisms that limit the reactivation of transposable elements and viral integration . Notably , some reports have illuminated the possible role of TGF-β in regulating the expression and/or activities of DNA-repair genes . Similar to our observations , ( Kanamoto et al . , 2002 ) showed TGF-β-dependent downregulation of Rad51 and decreased DNA -epair efficiency in Mv1Lu lung epithelial cells treated with TGF-β . In addition , Liu et al . ( 2014 ) reported that TGF-β regulates the expression and/or activity of BRCA1 , ATM and MSH2 . Conversely , Glick et al . ( 1996 ) reported that mouse keratinocytes deficient in TGF-β show a significant increase in gene amplification in response to the drug PALA . Kirshner et al . ( 2006 ) also showed that TGF-β signaling inhibition attenuates the function of ATM under stress ( ROS stimulation ) , thus impeding the response to DNA damage . Wiegman et al . ( 2007 ) provided evidence that TGF-β can stimulate ATM and p53 phosphorylation in primary irradiated cells in a Smad-independent pathway . Our multi-faceted approaches and data gathering greatly expand upon this existing body of evidence and strongly indicate that TGF-β plays a broader role in decreasing the expression of DNA-repair genes , contributing to CNA accumulation and the clonal diversity of cancer cell populations . As a possible explanation for the differences observed across the studies mentioned here , we noted that those indicating that TGF-β promotes DNA-damage-repair responses were conducted under stress conditions ( i . e . , PALA treatment , irradiation or H2O2 treatment ) and in response to reactive oxygen species ( ROS ) and did not distinguish between the effect of TGF-β on these experimental conditions ( i . e . , on ROS scavengers ) and a direct effect on DNA repair . Our findings are also in accordance with a recent review on the implications of stress-induced mutagenesis ( Fitzgerald et al . , 2017 ) . In summary , our data strongly indicate that the transition into a CD44+/CD24− cell state can promote intra-tumor genetic heterogeneity , spur tumor evolution and increase tumor fitness . These findings have important prognostic and therapeutic implications and speak strongly to the need to target CD44+/CD24− populations in tumors . In this regard , the observation that the decreased expression of HDR genes in CD44+/CD24− cells comes at the expense of increased dependency on other DNA-repair components is particularly notable ( Figures 1 and 2 ) . As some of these genes are potentially druggable , our studies could apprise the development of novel therapeutic options to ameliorate the outcome of cancer based on targeting the processes of cancer evolution , instead of targeting the cancer evolution products . A549 ( RRID: CVCL_0023 ) , H1650 ( RRID: CVCL_1483 ) , HCC4006 ( RRID: CVCL_1269 ) , H23 ( RRID: CVCL_1547 ) , BT-474 ( RRID: CVCL_0179 ) , MDA-MB-231 ( RRID: CVCL_0062 ) , MDA-MB-435S ( RRID: CVCL_0622 ) and NCI-H23 ( RRID: CVCL_1547 ) were obtained from the American Type Culture Collection repository ( ATCC , Manassas , Virginia ) . The MCF-7 ( RRID: CVCL_0031 ) cell line was obtained from the Cold Spring Harbor Laboratory Tissue Culture Facility . The PC9 ( RRID: CVCL_B260 ) cell line was a gift from Dr Jeffrey A . Engelman ( MGH , Charlestown , MA ) . The U2OS-DR-GFP cell line was created by Dr Maria Jasin and was a gift from Dr Agata Smogorzewska ( Rockefeller University , NY ) . The NMuMG ( RRID: CVCL_0075 ) cell line was a gift from Linda Van Aelst ( CSHL , NY ) . The A549 , H1650 , HCC4006 , U2OS , H23 and PC9 cell lines were cultured in RPMI medium supplemented with 10% FBS , glutamine , penicillin and streptomycin . MCF7 , BT-474 , MDA-MB-231 and MDA-MB-435s were cultured in DMEM containing 10% FBS , penicillin , streptomycin and sodium pyruvate . NMuMG was cultured in complete DMEM medium supplemented with 10 µg/ml bovine insulin . These cell lines were monitored for mycoplasma contamination by using the Lonza mycoalert mycoplasma detection kit on a regular basis . All of the cell lines tested negative for Mycoplasma contamination . None of the cell lines used in our studies was mentioned in the list of commonly misidentified cell lines maintained by the International Cell Line Authentication Committee . RNA was isolated from cells using the TRIzol-chloroform method . Cells were lysed with TRIzol reagent ( Invitrogen , Carlsbad , CA ) . After the chloroform extraction , the aqueous phase containing RNA was mixed with a 0 . 7 vol of isopropanol and centrifuged at 13 , 000 rpm for 15 min to precipitate the RNA . Following a wash with 70% ethanol , the RNA pellet was washed with 70% ethanol , air dried , dissolved in distilled water and subjected to DNAseI treatment to remove any contaminating genomic DNA . The quality and concentration of DNA and RNA were assessed using a Nanodrop ND-1000 spectrophotometer ( Nanodrop ) . cDNA was prepared using the ImProm-II cDNA synthesis kit ( Promega , Madison , WI ) . qPCR was carried out using the Power SYBR Green PCR Master MIX ( Applied Biosystems , Carlsbad , CA ) with 7900HT Fast Real-Time PCR system ( Applied Biosystems ) . The primers used to measure mRNA expression at the cDNA levels by RT-qPCR are listed in Supplementary file 3 . Cells were seeded in six-well plates and transfected with siRNA or plasmid or both with Lipofectamine reagent 2000 ( Invitrogen ) for at least 6 hr in antibiotic-free culture media mixed with OptiMEM glutaMax media ( Gibco ) . After transfection , the medium was changed to DMEM or RPMI ( as described above ) supplemented with FBS and antibiotics . At 72 hr post-transfection , the cells were harvested for RNA or protein analysis . To check for apoptosis , the media from the culture vessels were also collected and the dead cells were collected by centrifugation . For survival assays , the cells were trypsinized and counted at least 4 hr after changing the transfection medium . Depending on the cell line , between 2 , 000 and 3 , 000 cells were plated in a 96-well plate and cultured for 96–120 hr . Cells were then washed with PBS once and fixed in 3 . 7% formaldehyde ( Thermo Fisher Scientific , Waltham , MA ) at room temperature for 12 min . Next , cells were stained with Syto-60 Red fluorescent nuclein acid stain ( Thermo Fisher Scientific ) and scanned with ODYSSEY infrared imager ( LI-COR , Lincoln , NE ) at 700 nm . One hundred picomoles ( pmol ) of Stealth RNAi ( Invitrogen ) siRNA against human-IL-6 , -BRCA1 , -ORC5L , -RFC3 , -RPA2 , -NEK9 , -POLS , -ERCCC8 , -SMAD2 , -SMAD3 , -SMAD4 , -RELA , -RFC1 , -ATAD5 , -CTF18 and -RAD17 or 40 pmol of Silencer Select ( Ambion ) siRNA against human-BRCA2 and -RAD50 or 20 pmol of ON-TARGET plus SMART pool ( Dharmacon , Lafayette , Colorado ) against human BLM were used in six-well plates . The siRNA sequences ( sense strands ) are listed below: ORC5HSS181666 . 1 – CGU UUG UCU UAU AUU UCC CUG AUU A ORC5HSS181666 . 2 – UAA UCA GGG AAA UAU AAG ACA AAC G RFC3HSS184273 . 1 – AAG GCU GUA UGA GCU UCU AAC UCA U RFC3HSS184273 . 2 – AUG GAU UAG AAG CUC AUA CAG CCU U RPA2HSS184377 . 1 – CAG AAU UGG GAA UGU UGA GAU UUC A RPA2HSS184377 . 2– UGA AAU CUC AAC AUU CCC AAU UCU G ERCC8HSS174455 . 1 – AGC AGU UUC CUG GUC UCC ACG UUA U ERCC8HSS174455 . 2 – AUA ACG UGG AGA CCA GGA AAC UGC U PAPD7HSS117093 ( POLS ) . 1 – CCU UGG AAU GCU UCU UGU AGA AUU U PAPD7HSS117093 ( POLS ) . 2 – AAA UUC UAC AAG AAG CAU UCC AAG G IL6HSS105337 . 1 – GAG AAA GGA GAC AUG UAA CAA GAG U IL6HSS105337 . 2 – ACU CUU GUU ACA UGU CUC CUU UCU C BRCA1HSS101089 . 1 – GGG CUA UCC UCU CAG AGU GAC AUU U BRCA1HSS101089 . 2 – AAA UGU CAC UCU GAG AGG AUA GCC C BRCA2 . 1 –GGA UUA UAC AUA UUU CGC Att BRCA2 . 2 –CAG UUG AAA UUA AAC GGA Att RAD50 . 1 –GGU AGA CUG UCA UCG UGA Att RAD50 . 2 –GGA AUA GAC UUA GAU CGA Att 1 . 5 × 106 cells were plated in a complete media in 10 cm2 tissue culture dish on day 1 . On day 2 , the medium was changed to –Serum RPMI ( or DMEM ) and the cells were starved overnight . On the morning of day 3 ( 40 hr post seeding ) , cells were treated with rhTGFβ1 and rhTGFβ2 ( R&D systems , Minneapolis , MN ) 1 ng/ml each in complete media , for 9 hr . Following treatment , the cells were harvested for RNA preparation and qRT-PCR , for immunoblotting or for cell cycle analysis . For treatment with LY2157299 ( 20 μM ) and LY364947 ( 1 μM ) ( TGFBR1 kinase inhibitor , Selleckchem , Houston , TX ) , 300 , 000 H1650-M3 cells were plated in a 6 cm2 plate . Inhibitor was added the next day and the mixture was incubated for 3–5 days for LY2157299 and 2–3 days for LY364947 . The cells were lysed with TRIzol and processed for RNA preparation . To determine IC50 values for various drugs ( 17-AAG ( this drug is not mentioned in main text ) , cisplatin , doxorubicin , etoposide , erlotinib , epitaxol and tunicamycin ) , the cells were plated in 96-well plates at 2 , 000 cells/well . The next day , individual drugs were added to the wells at the indicated concentrations and the mixture was incubated for 5 days . The plates were then washed once with PBS , fixed with 3 . 7% formaldehyde and stained with crystal violet . Each stained well was destained in 50–100 µl of 10% acetic acid and the absorbance was read in a spectrophotometer at 590 nm . The cells were trypsinized using TrypLE , incubated at 37°C and re-suspended in PBS containing 1% FBS . Depending on the sample , between 800 , 000–1 , 000 , 000 cells/sample were added to a MACS buffer ( PBS with 0 . 5% BSA and 2 mM EDTA ) . The cells were washed in the MACS buffer twice and re-suspended in 100 µl staining solution ( 100 µl MACS buffer + antibodies ) . Staining was performed for 45 min and the cells were then washed twice with MACS buffer and re-suspended in 400 µl MACS buffer before passing through a 5 ml polystyrene round-bottom tube with cell-strainer cap ( BD-Falcon , Corning , New York ) . Samples were analyzed with the BD LSR II Cell Analyzer . For the cell-sorting procedure , the samples were prepared as above and sorted with the BD FACS Aria ( SORP ) Cell Sorter . Samples were collected and either cultured on an eight-well chamber slide system ( LAB-TEK , Thermo Fisher Scientific ) for 2 days before fixing and staining or subjected to RNA preparation . For the cell-cycle analysis , trypsinized cells were washed in MACS buffer and then fixed in ice-cold 70% ethanol by dropwise addition while vortexing , followed by washing in MACS buffer and staining 1 × 106 cells in PI solution ( PBS with 50 µg/ml propidium Iodide , 0 . 1 mg/ml RNase A , 0 . 05% TritonX-100 ) for 45 min at 37°C , passed through a 5 ml polystyrene round-bottom tube with cell-strainer cap ( B-Falcon ) and run on the BD LSR II Cell Analyzer . Cell cycle arrest was assessed through analysis of the proportion of cells in the G1 , S and G2/M fraction of the cell cycle . Cells were washed with PBS before collection and lysed directly in RIPA buffer containing 0 . 2% SDS for 30 min on ice . Proteins were separated by 6–12% SDS/PAGE , transferred to nitrocellulose membranes ( Bio-Rad , Hercules , CA ) and blotted with antibodies as indicated . To extract BRCA2 , RAD50 , BLM and WRN , cells were lysed in ice-cold NETN-450 buffer ( 450 mM NaCl , 1 mM EDTA , 20 mM Tris pH8 and 0 . 5% Igepal CA-630 ) for 15 min on ice followed by extraction in NETN-0 ( 1 mM EDTA , 20 mM Tris pH8 and 0 . 5% Igepal CA-630 ) for 15 min on ice . CD44+/CD24− cells and CD44−/CD24+ cells from H1650 were FACS-sorted and cultured for 2 days in an eight-well chamber slide system ( LAB-TEK , Thermo Fisher Scientific ) . H1650 , A549 and MCF7 cells were grown on glass coverslips in a 24-well Petri dish and treated with TGF-β or vehicle ( DMSO ) for 3 and 4 days or with 20 µg/ml doxorubicin overnight . Cells were fixed with 3 . 7% formaldehyde and permeabilized in 0 . 1% Triton X-100 in PBS for 10 min . Fixed cells were washed three times in PBS and blocked with 1% BSA in PBS for 1 hr . After washing three times with PBS , the cells were incubated with the primary antibody for overnight at 4°C . Immune complexes were then stained with indicated secondary antibodies ( Invitrogen ) . DAPI was used for nuclear staining . The stained cells were mounted with a Vectashield mounting medium ( Vector Laboratories , Burlingame , CA ) and analyzed using a confocal microscope . Cells embedded in agarose on a microscope slide were lysed with detergent and high salt to form nucleoids containing supercoiled loops of DNA linked to the nuclear matrix . Electrophoresis at high pH results in structures resembling comets that can be observed by fluorescence microscopy . The intensity of the comet tail relative to the head reflects the number of DNA strand breaks . DR-GFP consists of two mutated GFP genes: ( i ) the Sce-GFP that is disrupted by an 18-bp recognition site for the I-SceI endonuclease and ( ii ) the iGFP that is truncated at the 5′ and 3′ ends . Upon transfection with I-SceI , the SceGFP is cleaved and , after an HDR event has occurred , a functional GFP+ gene ( detectable by flow cytometry ) is generated by gene conversion with iGFP . To perform a DR-GFP assay in U2OS cells , U2OS-DR-GFP cells were plated in six wells and treated with vehicle ( DMSO ) or TGF-β ( 1 ng/ml each of TGF-β1 and TGF-β2 ) for a day . Then , cells were transfected with 4 ug of pCBASce-I ( Jasin Lab ) or control plasmid ( pCAG:mRFP1 , data not shown ) for 6 hr , transfection media were removed and cells were grown in media with vehicle ( DMSO ) or TGF-β ( 1 ng/ml each of TGF-β1 and TGF-β2 ) for 3 days . Cells were then harvested and subjected to FACS analysis . To perform a DR-GFP assay in H1650 cells , the cells were plated in triplicate in six wells and treated with vehicle ( DMSO ) or TGF-β ( 1 ng/ml each of TGF-β1 and TGF-β2 ) for a day . Then , cells were transfected with 0 . 5 μg of pDR-GFP ( Addgene ) or 2 μg of pCBASce-I ( Jasin Lab ) or both the plasmids for 6 hr; subsequently , transfection media were removed and cells were grown in media with vehicle ( DMSO ) or TGF-β ( 1 ng/ml each of TGF-β1 and TGF-β2 ) for 3 days . Cells were then harvested and subjected to FACS analysis . 20 , 000 cells were analysed for each sample replicate . H1650 cells were serially diluted in 96 wells such that one well contains one cell . They were then grown for 2 months before the experiments . To perform SNS , cells from culture were stained with antibodies against CD44 and CD24 or nuclei were isolated from cells in culture and stained with 49 , 6-diamidino-2-phenylindole ( DAPI ) . We use FACS to gate a desired population of cells ( for antibody-stained cells , we gated CD44−/CD24+ and CD44+/ CD24− ) or nuclei ( by total DNA content ) and deposited desired cells/nuclei singly into 96-well plates . For sorting cells from tumor , we stained a single-cell suspension derived from a tumor with CD45 , CD31 , EpCAM , CD44 and CD24 antibodies . We gated the CD45−/ CD31−/ EpCAM mid/high population and then gated the desired CD44−/CD24+ and CD44+/ CD24− from the EpCAM mid/high population . After WGA using Sigma GenomePlex , sonication was performed to create free DNA ends without WGA adapters and then we constructed libraries for 76-bp , single-end sequencing using one lane of an Illumina GA2 flow cell per nucleus . For each nucleus , we typically achieve at least 9 million ( mean 59 . 042 million , s . e . m . 60 , 328 , n = 5 , 200 ) uniquely mapping reads using the Bowtie alignment software . These sequences cover approximately 6% ( mean 5 . 95% , s . e . m . 0 . 229 , n = 5 , 200 ) of the genome and are used to count sequence reads in 50 , 000 variable bins . The bin counts are segmented using a KS statistic and then used to calculate integer copy number profiles . Neighbor-joining trees are constructed both from the integer profiles and from the chromosome breakpoint patterns of each cell to infer evolution . The copy-number profiles , dendrograms and heatmaps of single cells were generated using Ginkgo , a visual analytics tool for the analysis of single-cell copy-number variations . Ginkgo automatically corrects for known biases in single-cell data and generates copy number profiles for each cell . To generate copy number profiles , we used Ginkgo with the the following parameters: variable length 500 kb bins simulated using 101 bp reads mapped with bowtie , cell segmented independently with normalized read counts and all bad bins masked . In order to cluster cells to generate dendrograms and heat maps , we using ward linkage and euclidian distance metrics . Quantification of DNA joint points ( referred to as ‘breakpoints’ in the Ginkgo website ) was done by extracting the breakpoint counts for each cell from the ‘Breakpoints’ file under Ginkgo’s ‘Download processed data tab’ . Twenty-five thousand cells were plated in each well of a six-well plate and treated with either 3 . 2 µM tunicamycin ( Sigma-Aldrich , St . Louis , MO ) , 1 . 6 µM epitaxol ( Santa Cruz Biotechnology , Santa Cruz , CA ) or 3 . 2 µM etoposide ( Sigma-Aldrich ) for 3 days . In another assay , 12 , 000 cells were plated in a six-well plate; after the cells formed small colonies with ~10 cells/ colony , they were treated with 3 . 2 µM tunicamycin for 3 days followed by either tunicamycin or epitaxol ( 1 . 6 uM ) treatment for 3 days . The collection of human lung tissue samples and blood for this study was covered by Huntington Hospital/Northwell Health IRB #14–496 ( PI V Singh; approval date 11/14/14 ) . The samples were acquired from patients already undergoing thoracic procedures ( e . g . surgical tumor resection or biopsy ) at Huntington Hospital . All study participants provided informed consent for the use of their lung tissue and blood for research purposes . Participants were informed of study aims , the potential risks and benefits of participation , and that any discoveries facilitated by the analysis of their tissues might be published . The participants were informed that their names would not be associated their samples in any publication or presentation of research findings The following antibodies were used for the flow cytometry analyses: The following antibodies were used for immunofluorescence: The following antibodies were used for immunoblot analysis:	A single tumor can be made up of thousands of cancer cells that look and behave differently from one another . This observed diversity arises from changes in the DNA sequence of particular genes , changes in the activity of genes , and the ability of cells to transition between different states . As a result , individual cells within a tumor may react differently to certain anti-cancer drugs: most of the cells may be sensitive to the treatment and die , whereas some might be resistant and survive . Previous studies on several types of human cancers – including breast , brain and lung cancers – have identified a group of cells called CD44+/CD24- cells that seem to be more aggressive and resistant to therapy than other cancer cells . However , it is currently not known exactly how these CD44+/CD24- cells influence a whole tumor’s resistance to anti-cancer drugs . Certain cancer cells in tumors are exposed to a signal molecule called TGF-β . CD44+/CD24- cells are unusual in that they are able to produce and release this signal molecule themselves . Pal et al . show that in CD44+/CD24- cells from a human lung cancer cell line , TGF-β decreased the activity of genes responsible for accurately fixing breaks in the CD44+/CD24- cells’ DNA . As a result , these cells made more mistakes than other lung cancer cells when repairing damaged DNA and consequently accumulated additional genetic mutations . Furthermore , tumors containing these cells were more likely to survive treatment with chemotherapy . The findings of Pal et al . show that the CD44+/CD24- cells exposed to TGF-β had a survival advantage because they were more genetically diverse and therefore better able to adapt to new drug treatments and other changes in their surroundings . Future experiments may explore how to specifically target and kill the CD44+/CD24- cells from tumors .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "cancer", "biology" ]	2017	TGF-β reduces DNA ds-break repair mechanisms to heighten genetic diversity and adaptability of CD44+/CD24− cancer cells
Animals use spatial differences in environmental light levels for visual navigation; however , how light inputs are translated into coordinated motor outputs remains poorly understood . Here we reconstruct the neuronal connectome of a four-eye visual circuit in the larva of the annelid Platynereis using serial-section transmission electron microscopy . In this 71-neuron circuit , photoreceptors connect via three layers of interneurons to motorneurons , which innervate trunk muscles . By combining eye ablations with behavioral experiments , we show that the circuit compares light on either side of the body and stimulates body bending upon left-right light imbalance during visual phototaxis . We also identified an interneuron motif that enhances sensitivity to different light intensity contrasts . The Platynereis eye circuit has the hallmarks of a visual system , including spatial light detection and contrast modulation , illustrating how image-forming eyes may have evolved via intermediate stages contrasting only a light and a dark field during a simple visual task . Visually guided behavior is widespread in animals ( Ullén et al . , 1997; Garm et al . , 2007; Orger et al . , 2008; Burgess et al . , 2010; Huang et al . , 2013 ) , yet the underlying neuronal circuits and their evolutionary origins remain poorly understood . Spatial vision requires at least two photoreceptors and a neural circuitry capable of making a comparison between the photoreceptor inputs without body movement ( Land and Nilsson , 2002; Nilsson , 2009 ) . The spatial information obtained must then translate to a coordinated motor output . A comprehensive description of the sensory-motor visual circuitry , including all neurons and their synaptic connectivity , is required for a plausible explanation of how visual inputs drive motor output during animal behavior . This can only be achieved using electron microscopic imaging to construct connectomes , comprehensive synaptic-level connectivity maps for large blocks of neural tissue containing behaviorally relevant circuits ( Bock et al . , 2011; Briggman et al . , 2011; Jarrell et al . , 2012; Bumbarger et al . , 2013; Helmstaedter et al . , 2013 ) . However , despite recent advances in the connectomics of visual systems ( Briggman et al . , 2011; Rivera-Alba et al . , 2011; Sprecher et al . , 2011; Takemura et al . , 2011 , 2013 ) , a complete synaptic-level connectivity-map of a visual circuit , including sensory- , inter- , and motorneurons , has not yet been described . Here we reconstruct the neural connectome of the visual eyes in a larva of the marine annelid Platynereis dumerilii using serial-section transmission electron microscopy ( ssTEM ) . Platynereis larvae develop four visual eyes , the ‘adult eyes’ ( henceforth ‘eyes’ ) , that are the precursors of the adult's visual pigment-cup eyes , and are distinct from the more ventrally located ‘eyespots’ ( Jékely et al . , 2008 ) . These eyes consist of only 2–7 photoreceptors , a few shading pigment cells , and a lens , representing the simplest visual eyes described to date ( Rhode , 1992; Arendt et al . , 2002; Randel et al . , 2013 ) . The Platynereis larval visual connectome consists of 71 neurons and 1106 synapses , and was reconstructed from a tissue block containing the larval head and trunk . Using behavioral experiments combined with eye ablations we demonstrate that the eyes mediate spatial vision , whereby light intensities at the left and right eyes are compared to mediate tail bending during phototactic turns . This combination of connectomics and behavioral analysis provides a circuit-level and mechanistic explanation for the regulation of phototactic behavior by spatial vision . The Platynereis visual connectome also provides insights into the origin of visual eyes , and suggests that phototaxis may have been the first visual task in evolution performed by animals . To reconstruct the full chemical synaptic connectivity map of the visual eyes in the Platynereis larva we used ssTEM . Following high-pressure freezing and cryosubstitution , we collected 1690 thin sections ( 40–50 nm ) from a single Epon-embedded 3-day-old larva , encompassing the entire head and part of the first trunk segment . Following contrasting , we imaged the sections at 3 . 7 nm/pixel resolution , in a 140 µm × 140 µm × 80 µm volume ( Figure 1A–C; Video 1 ) . 10 . 7554/eLife . 02730 . 003Figure 1 . Serial-section electron microscopy imaging of the visual eye circuit in a Platynereis larva . ( A ) Scanning electron micrograph of a 72 hr-post-fertilization larva , dorsal view . The boxed volume was sectioned and imaged . ( B ) Anterior view of a 72 hr-post-fertilization larva visualized with differential interference contrast ( DIC ) optics ( grey ) showing the position of the eyes visualized by the reflection of the pigments ( red ) . ( C ) A representative electron micrograph from the series with traced neurons . The boxed area contains the primary optic neuropil . ( D ) Reconstruction of the pigment cup of the left anterior and posterior eyes . Pigment granules from the different pigment cells and photoreceptors are colored differently . ( E ) TEM image of the left primary optic neuropil surrounded by glia ( pink ) , with the photoreceptor projections from the anterior and posterior eye colored differently . ( F ) A virtual cross-section of the primary optic neuropil based on ssTEM shows the anterior-posterior layering of glia , photoreceptor , primary interneuron ( IN1 ) and trans-optic-neuropil interneuron ( INton ) processes , anterior is up . ( G ) TEM image of the secondary optic neuropil , with segmented INton , INsn , INdc and ipsi- and contralateral motorneuron projections . ( H ) TEM image of a neuromuscular synapse from a motorneuron to the ventral longitudinal muscle . Asterisk marks a cluster of synaptic vesicles . Eyeal , anterior-left eye; eyear , anterior-right eye; eyepl , posterior-left eye; eyepr , posterior-right eye; PRC , photoreceptor; IN , interneuron; MN , motorneuron . Scale bars , 50 µm ( A ) , 30 µm ( B and C ) , 5 µm ( D , E , G ) , 1 µm ( F ) , 0 . 5 µm ( H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 00310 . 7554/eLife . 02730 . 004Figure 1—figure supplement 1 . Morphology of photoreceptor cells reconstructed from serial TEM sections . Tracing of the cell skeletons was complemented with partial volume reconstructions . The position of the cell body is marked with an asterisk . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 00410 . 7554/eLife . 02730 . 005Figure 1—figure supplement 2 . Morphology of IN1 , INton , INsn , and INdc cells reconstructed from serial TEM sections . Tracing of the cell skeletons was complemented with partial volume reconstructions . The position of the cell body is marked with an asterisk . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 00510 . 7554/eLife . 02730 . 006Figure 1—figure supplement 3 . Morphology of INint and INvc cells reconstructed from serial TEM sections . Tracing of the cell skeletons was complemented with partial volume reconstructions . The position of the cell body is marked with an asterisk . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 00610 . 7554/eLife . 02730 . 007Figure 1—figure supplement 4 . Morphology of motorneurons reconstructed from serial TEM sections . Tracing of the cell skeletons was complemented with partial volume reconstructions . The position of the cell body is marked with an asterisk . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 00710 . 7554/eLife . 02730 . 008Figure 1—figure supplement 5 . Axon diameter and synapse size in the Platynereis larval connectome . ( A ) Histogram showing axon diameter ( n = 102 ) . ( B ) Histogram showing synapse size defined as the number of consecutive sections in which a synapse was visible ( n = 100 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 00810 . 7554/eLife . 02730 . 009Figure 1—figure supplement 6 . Synapses of photoreceptors and IN1 cells . Examples of PRC to PRC ( A–C ) and PRC to IN1 synapses ( D–G ) . The left and right panels show the same synapse at different resolution ( left: 3 . 7 nm/pixel , right: 0 . 2 nm/pixel ) . Asterisks mark the synapse , which is shown at higher resolution in the right panel . Scale bar , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 00910 . 7554/eLife . 02730 . 010Figure 1—figure supplement 7 . Synapses of IN1 cells . Examples of IN1 to IN1 ( A–D ) and IN1 to INton synapses ( E–H ) . The left and right panels show the same synapse at different resolution ( left: 3 . 7 nm/pixel , right: 0 . 2 nm/pixel ) . Asterisks mark the synapse , which is shown at higher resolution in the right panel . Scale bar , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 01010 . 7554/eLife . 02730 . 011Figure 1—figure supplement 8 . Synapses of INton cells and INsn cells . Examples of INton to INsn ( A–D ) and INsn to MN synapses ( E–H ) . The left and right panels show the same synapse at different resolution ( left: 3 . 7 nm/pixel , right: 0 . 2 nm/pixel ) . Asterisks mark the synapse , which is shown at higher resolution in the right panel . Scale bar , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 01110 . 7554/eLife . 02730 . 012Figure 1—figure supplement 9 . Synapses of motorneurons . Examples of MN to muscle cell ( A–D ) and MN to prototroch cell synapses ( E–H ) . The left and right panels show the same synapse at different resolution ( left: 3 . 7 nm/pixel , right: 0 . 2 nm/pixel ) . Asterisks mark the synapse , which is shown at higher resolution in the right panel . Scale bar , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 01210 . 7554/eLife . 02730 . 013Video 1 . Head and first trunk segment of a 3-day-old Platynereis larva . DIC and reflection imaging of the head and first trunk segment of a 73 hr-post-fertilization Platynereis larva , corresponding to the volume analyzed by ssTEM . The four eyes and the eyespots are visualized based on the reflection of the pigment ( red ) . The frame contains a volume of 151 × 151 × 61 μm . The larva is squeezed dorso-ventrally with the coverslip . Grid spacing is 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 013 We identified the eyes based on the presence of pigment-filled vesicles in the pigment cells , the presence of a lens formed by the apical extensions of the pigment cells , and the presence of the apical microvillar extensions ( rhabdoms ) of the photoreceptors ( Rhode , 1992; Randel et al . , 2013; Figure 1D; Video 2 ) . We then manually traced all neurons lying on synaptic pathways from photoreceptors to locomotor organs ( muscles and ciliary bands ) . We identified 21 photoreceptor cells ( PRC ) , 42 interneurons ( IN ) and 8 motorneurons ( MN ) ( Figure 1—figure supplement 1–4 ) , forming a ‘minimal eye circuit’ from sensors to effectors . We identified 1106 chemical synapses between these cells and their motor targets . We found further sensory neurons other than the photoreceptors providing direct or indirect input to the minimal eye circuit . These upstream circuits will be described in detail elsewhere . In this paper we focus on the analysis of the 71 neurons constituting the minimal eye circuit . 10 . 7554/eLife . 02730 . 014Video 2 . Volume reconstruction of the two left eyecups . The pigment vesicles of the pigment cells and the rhabdoms of the photoreceptors were reconstructed by ssTEM . The pigment of the photoreceptors is on the convex surface of the pigment cup , shown in different colors . The photoreceptor rhabdoms are inside the pigment cup . Scale bar , 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 014 First , we describe the anatomy and synaptic contacts of the eyes and the main neuron types identified as part of the eye circuit . To provide an easily accessible representation of the data we generated a comprehensive 3D anatomical atlas of all neurons , synapses , muscles and ciliated cells using Blender ( http://www . blender . org/ ) ( Randel et al . , 2014; Video 3 ) . Neuron morphologies were reconstructed based on the raw traces ( Figure 1—figure supplement 1–4 ) and represented in the atlas with a simplified morphology ( Figure 2 ) . Users can display groups of neurons , query for pre- and post-synaptic partners of neurons , and display synapses . 10 . 7554/eLife . 02730 . 015Video 3 . Cellular complement of the Platynereis larval visual circuit . 3D reconstruction of the cellular complement of the Platynereis larval visual circuit . The different cell types appear in the following order: photoreceptors , IN1 , INint , INton , INsn , INdc , INvc , motorneurons , synapses , glia , muscles . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 01510 . 7554/eLife . 02730 . 016Figure 2 . Cell complement of the visual circuit . ( A ) Confocal microscopic image of a 3-day-old larva stained with an anti-acetylated tubulin antibody to label neurites and cilia . Anatomical landmarks are labeled . ( B ) Blender visualization of all photoreceptor cells and all synapses of the minimal eye circuit shown in relation to the outline of neuropil , reconstructed by ssTEM . The position of synapses in the visual circuit reveals the primary and secondary optic neuropils . The schematized ciliary band cells are also shown . ( C ) ssTEM reconstruction of photoreceptors , glia cells and synapses . ( D ) ssTEM reconstruction of photoreceptors and primary interneurons ( IN1 ) . ( E ) ssTEM reconstrucion showing the trans-optic-neuropil interneurons ( INton ) connecting the two optic neuropils . ( F ) Confocal microscopic image of a 3-day-old larva stained with phalloidin to label the musculature ( red ) and with DAPI to label nuclei ( cyan ) . ( G ) ssTEM reconstruction of muscles and motorneurons ( MN ) . ( H ) ssTEM reconstruction of the complete cell complement of the minimal visual circuit . Neurons are colored by type . Pigment cups are shown in brown . All images show anterior views . PRC , photoreceptor; IN , interneuron; MN , motorneuron; eyeal , anterior-left eye; eyear , anterior-right eye; eyepl , posterior-left eye; eyepr , posterior-right eye; ON , optic neuropil; cPRC , ciliary photoreceptor; DCC , dorsal branch of the circumesophageal connectives; VCC , ventral branch of the circumesophageal connectives; NSP , neurosecretory plexus; DLM , dorsal longitudinal muscle; VLM , ventral longitudinal muscle . The coloring of cell types is consistent throughout the paper ( PRC , blue; IN1 , yellow; INint , lilac; INton , magenta; INdc , light brown; INvc , cian; INsn , green; MN , red ) . Scale bars , 30 µm . The Blender atlas with the volume rendering of all cells and synapses is available in Randel et al . ( 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 01610 . 7554/eLife . 02730 . 017Figure 2—figure supplement 1 . Muscles and ciliary bands in 3-day old larvae . ( A ) Phalloidin staining of the musculature ( red ) in a 3-day-old larva , ventral view . Arrows point at the most anterior tip of the dorsal longitudinal muscles . Nuclei are labeled with DAPI ( cyan ) . ( B ) SEM micrograph of a 3-day old larva ( ventral view ) showing the prototroch and metatroch ciliary bands . ( C ) ssTEM reconstruction of the prototroch cells . Two MN cells are also shown . One of the prototroch cells extends a long projection towards the MN axons and receives synaptic input there ( arrow ) . ( D and E ) Bilateral divergence of the circuit at the level of the MNs . SN cells connect to both ipsilateral and contralateral MNs at their proximal or distal axon segments respectively . The position of synapses from SN to MN cells are shown . In ( D ) only two MNs are shown for clarity . Scale bars 30 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 017 The neurons of the eye circuit only contain axo-axonal synapses . With the exception of the photoreceptors , all neurons lack dendrites . Axons have a median diameter of 160 nm ( Figure 1—figure supplement 5 ) , allowing reliable tracing of most processes . The morphology of all neurons is very simple , with one primary branch giving rise to several very short secondary branches in the synaptic regions of the optic neuropils ( Figure 1—figure supplement 1–4 ) . These short branches often contain pre- or postsynaptic sites ( Video 4 ) . Only two neurons have a branched main axon ( MNr3 , MNl1 , see ‘Materials and methods’ for nomenclature ) . 10 . 7554/eLife . 02730 . 018Video 4 . Volume rendering of the left primary optic neuropil . The reconstructed volume shows parts of two glia cells ( pink and light blue ) , one IN1 axon ( yellow ) and the photoreceptor axons ( different hues of blue ) forming synapses on the IN1 axon . The position of the individual synapses to the IN1 axon are indicated in different colors for the different photoreceptors . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 018 Synapses were identified as electron dense accumulations of several vesicles close to the presynaptic membrane ( no T-bars as in Drosophila ) . Prominent postsynaptic structures were not detected , in agreement with previous ultrastructural studies in annelids ( Wells et al . , 1972 ) . 69% of the synapses could be identified in at least two and up to five consecutive sections ( Figure 1—figure supplement 5 ) . To further characterize the ultrastructure of synapses we created high-resolution images ( 0 . 2 nm/pixel ) of 60 randomly chosen synapses belonging to the different neuron types ( Figure 1—figure supplement 6–9 ) . In all cases we observed a cluster of vesicles adjacent to the plasmamembrane , but no pre- or post-synaptic specializations . We could not identify gap junctions in the eye circuit . Gap junctions exist in annelids , and we could find several innexin genes , encoding gap junction proteins of invertebrates ( Kandarian et al . , 2012 ) , in the Platynereis transcriptome ( data not shown ) . However , we did not see structures similar to ultrastructurally characterized annelid gap junctions ( Muller and Carbonetto , 1979; Shen et al . , 2002 ) , even in a stack of high-resolution ( 1 . 13 nm/pixel ) images of the primary optic neuropil ( Randel et al . , 2014 ) . If gap junctions are present in the larval stage they may be too small to be distinguished from obliquely cut membranes even at high resolution . The photoreceptors of the four eyes ( 3–7 per eye ) project to a primary optic neuropil area at the center of the larval brain ( Figure 1C , E , Figure 2B ) , consistent with results obtained by the genetic labeling of the photoreceptors ( Backfisch et al . , 2013 ) . All photoreceptors from the same eye project along the same axon bundle , forming four separate eye nerves ( Figure 2A , B ) . The eye nerves from the anterior and posterior eyes from the same side innervate distinct anterior and posterior areas in the primary optic neuropil ( Figure 1E , F ) . Some photoreceptor axons cross the midline and project to the contralateral primary optic neuropil . The primary optic neuropil is surrounded by three giant glia cells . These cells form lamellae that tightly surround the primary and secondary optic neuropils from the dorsal but not the ventral side ( Figure 1F , Figure 2C; Video 4 ) . The main targets of the photoreceptors of the four eyes are four primary interneurons ( IN1 ) , with photoreceptors from one eye forming several ( up to 30 ) synaptic contacts to one primary interneuron with a contralateral cell body ( e . g . , PRCal to IN1pr; Video 4 ) . The four eyes and the four primary interneurons show a crosswise arrangement ( Figure 2D ) . Primary interneurons have mutual synaptic contacts ( see below ) and also form synapses on two other interneuron types ( Figure 2E , H; Video 3; Randel et al . , 2014 ) . One type of interneuron is intrinsic to the optic neuropil ( INint ) with both ipsilateral and contralateral cell bodies and axons crossing the midline . Another target of the primary interneurons are interneurons that project out of the primary optic neuropil into a secondary optic neuropil area ( Figure 1G ) . The major targets of these trans-optic-neuropil interneurons ( INton ) are a group of contralateral interneurons , which we named Schnörkel interneurons ( INsn ) . Schnörkel interneurons project a curved axon to the ipsilateral secondary optic neuropil ( Figure 2H , Figure 2—figure supplement 1 ) . The Schnörkel interneurons are presynaptic to a bilateral group of ventral motorneurons ( MN; Figure 2—figure supplement 1 ) . Two groups of dorsal and ventral interneurons ( INdc , INvc ) also form synapses on the Schnörkel interneurons and motorneurons ( Figure 2H; Video 3 ) . We identified two distinct motorneuron types . The first type sends a contralateral projection to both the ciliary band and the dorsal longitudinal muscle , and branches to send a descending projection to the trunk ( Figure 2G , Figure 2—figure supplement 1 ) . The second type only projects posteriorly along the circumesophageal nerve after crossing the midline . Both motorneuron types form neuromuscular and neurociliary synapses ( Figure 1H , Figure 2—figure supplement 1; Video 3 , Video 5 ) . Due to the lack of further trunk sections we did not trace descending motorneuron axons along their entire length , therefore we cannot exclude the possibility that motorneurons have other synaptic partners in the trunk . Nevertheless , these data represent , to our knowledge , the most complete connectomic reconstruction of a visual circuit to date , from photoreceptors to effector muscles and ciliated cells . 10 . 7554/eLife . 02730 . 019Video 5 . Volume reconstruction of a Schnörkel interneuron and motorneurons . 3D reconstruction of one Schnörkel interneuron and its postsynaptic motorneuron targets . One Schnörkel interneuron is shown with its five postsynaptic motorneurons . The synapses are indicated . The Schnörkel interneuron connects to both ipsilateral and contralateral motorneurons either at a cell body proximal or a cell body distal position along the motorneuron axon . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 019 To gain insights into the nature of synapses in the eye circuit we measured vesicle diameter in the high-resolution images of synapses ( Figure 3—figure supplement 1 ) . The mean diameter of synaptic vesicles in interneurons was significantly larger than in photoreceptors and motorneurons . Photoreceptor synaptic vesicles were not significantly different from neuromuscular synaptic vesicles in motorneurons . However , neurociliary synaptic vesicles in motorneurons were significantly smaller than photoreceptor or neuromuscular synaptic vesicles . These observations indicate the use of different neurotransmitters in the eye circuitry and also suggest that the dual-function muscle- and cilio-motor neurons ( MNr2 , MNl1 ) may have mixed neurotransmitter content . To identify possible neurotransmitters in the eye circuit we performed whole-mount RNA in situ hybridizations with marker genes for various neurotransmitters ( Randel et al . , 2014 ) and compared the expression domains to cell body positions in the TEM series ( Figure 3 ) . In Platynereis larvae the colocalization of distinct gene expression patterns can be determined with near cellular resolution using whole-mount RNA in situ hybridizations and image registration , allowing the molecular profiling of cell types ( Tomer et al . , 2010; Asadulina et al . , 2012 ) . Although expression patterns for several of these genes have already been reported in Platynereis ( Denes et al . , 2007; Jékely et al . , 2008; Tomer et al . , 2010 ) , we acquired new whole-body expression data to analyze patterns in the 3-day-old brain in more detail . We found that vesicular glutamate transporter ( VgluT ) , a marker of glutamatergic neurons , colocalized with r-opsin1 , a marker of eye photoreceptors ( Arendt et al . , 2002; Randel et al . , 2013; Figure 3—figure supplement 2 , Figure 3—figure supplement 3 ) , indicating that the photoreceptors are glutamatergic . In contrast , histidine decarboxylase ( hdc ) , an enzyme catalyzing the synthesis of histamine , the transmitter in arthropod rhabdomeric photoreceptors ( Stuart , 1999 ) , does not localize to the eyes , as shown by image registration and double RNA in situ hybridization ( Figure 3—figure supplement 3 ) . 10 . 7554/eLife . 02730 . 020Figure 3 . Neurotransmitters of the eye circuit . ( A ) Cell body positions of eye circuit neurons relative to the larval axonal scaffold and five large gland cells . Motorneurons are numbered according to the cell identifiers . ( B and C ) Surface representation of the average expression domains of neurotransmitter marker genes relative to the larval axonal scaffold . The following genes are shown: ( B ) histaminergic marker histidine decarboxylase ( hdc; green ) , serotonergic marker tryptophan hydroxylase ( TrpH; red ) , dopaminergic marker tyrosine hydroxylase ( TyrH; yellow ) , adrenergic marker dopamine beta hydroxylase ( dbh; cyan ) , ( C ) glutamatergic marker vesicular glutamate transporter ( VGluT; red ) , cholinergic marker choline acetyltransferase ( ChAT; grey ) , GABAergic marker glutamate decarboxylase ( gad; green ) . The axonal scaffold , based on average acetylated-tubulin signal , is shown in grey . PRC , photoreceptor; cPRC , ciliary photoreceptor; IN , interneuron; MN , motorneuron . In ( B ) dashed ovals mark the position of the eyespots . Black arrows show the ring formed by the circumesophageal connectives . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02010 . 7554/eLife . 02730 . 021Figure 3—figure supplement 1 . Synaptic vesicle diameter for different synapse types . Scatter plot of the diameter of synaptic vesicles of different synapse types . The labels indicate the pre- and post-synaptic neurons . Vesicle diameter was measured from high-resolution ( 0 . 2 nm/pixel ) images . Mean with 95% confidence interval are shown . n >47 vesicles for each synapse type . p-values of an unpaired t test with Welch's correction are indicated relative to synaptic vesicles of the photoreceptors . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02110 . 7554/eLife . 02730 . 022Figure 3—figure supplement 2 . Expression of neurotransmitter markers in the head of Platynereis larva . ( A–H ) Whole mount RNA in situ hybridization in 3-day-old larvae for neurotransmitter marker genes ( red ) , counterstained with acetylated tubulin antibody ( white ) . ( A ) Glutamatergic marker vesicular glutamate transporter ( VGluT ) , ( B and C ) cholinergic markers choline acetyltransferase ( ChAT ) and vesicular acetylcholine transporter ( VAChT ) , ( D ) GABAergic marker glutamate decarboxylase ( gad ) , ( E ) histaminergic marker histidine decarboxylase ( hdc ) , ( F ) serotonergic marker tryptophan hydroxylase ( TrpH ) , ( G ) dopaminergic marker tyrosine hydroxylase ( TyrH ) , and ( H ) adrenergic marker dopamine beta hydroxylase ( dbh ) . ( I ) Ventral view schematic based on EM data of cell body positions of motorneurons ( MN ) and Schnörkel interneurons ( INsn ) relative to the larval axonal scaffold ( as ) and two gland cells ( gc1 , gc5 ) . ( J and K ) Close-up of whole mount RNA in situ hybridization of ( J ) cholinergic marker VAChT and ( K ) GABAergic marker gad . Gland cells ( gc ) are indicated by a white outline . Yellow asterisks mark the ventral domain of gene expression . ( L–O ) Close-up of panels ( E–H ) showing whole mount RNA in situ hybridizations of ( L ) hdc ( M ) TrpH ( N ) TyrH ( O ) dbh . Yellow arrows indicate areas of gene expression in the region of the interneurons . The sensory cilia of the ciliary photoreceptor cells ( cPRC ) are indicated . ( A–H and L–O ) are anterior views , ( I–K ) are ventral views . Scale bar: ( A–H ) 50 µM , ( J–O ) 10 µM . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02210 . 7554/eLife . 02730 . 023Figure 3—figure supplement 3 . Neurotransmitter marker gene expression profiling . The spatial relationships and colocalization of neurotransmitter marker genes were characterized by image registration and double RNA in situ hybridization in 3-day-old Platynereis larvae . ( A–D and F–I ) Average expression patterns of neurotransmitter marker genes and r-opsin1 projected onto a common whole-body nuclear reference template . An average acetylated tubulin signal ( white ) was also projected onto the reference . Colocalization of registered genes in the average gene expression 3D image stacks is indicated by white and highlighted by dashed circles . ( E ) Double whole mount RNA in situ hybridization for hdc ( cyan ) and r-opsin-1 ( red ) counterstained with acetylated tubulin antibody ( white ) . All images are anterior views . Scale bar 50 µM . Average 3D image stacks are available in Randel et al . ( 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 023 In the cell body regions of the interneurons we found expression of several monoaminergic markers in single cells or small cell clusters . These include tryptophan hydroxylase ( TrpH ) , tyrosine hydoxylase ( TyrH ) , hdc , dopa-beta-hydoxylase ( dbh ) , markers for serotonergic , dopaminergic , histaminergic , and noradrenergic neurons , respectively ( Figure 3B , Figure 3—figure supplement 2 , Figure 3—figure supplement 3 ) . This indicates that the interneurons of the eye circuit use several distinct monoamine neurotransmitters . Choline acetyltransferase ( ChAT ) and vesicular acetylcholine transporter ( VAChT ) , two markers of cholinergic neurons , are localized in the cell body region of the ventral motorneurons ( Figure 3C , Figure 3—figure supplement 2 ) . This is consistent with earlier observations showing that the longitudinal trunk muscles , targets of the ventral motorneurons , are influenced by acetylcholine ( Denes et al . , 2007 ) . A marker for GABAergic cells , glutamate decarboxlase ( gad ) , partly colocalizes with the ventral cholinergic domain , in the area of the motorneuron cell bodies ( Figure 3C , Figure 3—figure supplement 2 , Figure 3—figure supplement 3 ) . This suggests that some motorneurons may have mixed GABAergic and cholinergic identity . The difference in mean vesicle size at neuromuscular and neurociliary synapses in motorneurons ( Figure 3—figure supplement 1 ) also supports a mixed neurotransmitter identity . Further cellular resolution mapping will be needed to confirm this and to establish the neurotransmitter identities of individual interneurons and motorneurons . For a detailed analysis of synaptic contacts in the eye circuit , we represented the connectome as a directed graph and as an all-against-all synaptic connectivity matrix ( Figure 4A , Figure 4—figure supplement 1 , Figure 4—source data 1 ) . In the graph , the nodes correspond to neurons and the directed edges to synaptic contacts , weighted by synapse number . We displayed the full cell complement ( 79 nodes; Figure 4A ) and a trimmed graph ( 61 nodes ) where all cells with <3 pre- and post-synaptic sites were omitted and only edges of three or more synapses were shown ( Figure 4—figure supplement 2 ) . Manual or force-field-based clustering was used for graph layout . We also constructed a merged graph where the same neuron types from the left and right body sides were grouped . In this graph the weight of connections is represented as the maximum number of synapses between any individual cell pair of the distinct groups ( Figure 4B , Figure 4—source data 2 ) . 10 . 7554/eLife . 02730 . 024Figure 4 . Network analysis of the visual eye circuitry . ( A ) Full connectomic graph of the visual eye circuit including 71 neurons and 8 effectors ( muscles , ciliary band cells and epithelial cells ) . The edges are directed from presynaptic cell pointing to postsynaptic cells . Edges are weighted by the number of synapses . Inset shows selected network parameters . ( B ) Merged graph representation of the visual circuit . Nodes correspond to neuron classes , edges are weighted by the maximum number of synapses between two neuron types of each class . Nodes are colored following the color scheme used to label cell types . Inset shows the anatomical position of the cell types . PRCal , anterior-left photoreceptors; PRCar , anterior-right photoreceptors; PRCpl , posterior-left photoreceptors; PRCpr , posterior-right photoreceptors; IN , interneuron; MN , motorneuron . Matrix files of the complete and the merged networks are available in Figure 4—source datas 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02410 . 7554/eLife . 02730 . 025Figure 4—source data 1 . All-against-all connectivity matrix of the Platynereis eye circuit . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02510 . 7554/eLife . 02730 . 026Figure 4—source data 2 . Grouped connectivity matrix of the Platynereis eye circuit . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02610 . 7554/eLife . 02730 . 027Figure 4—figure supplement 1 . Synaptic connectivity matrix of the Platynereis larval visual circuit . The synaptic connectivity matrix represents synaptic connections between 71 neurons and 8 effectors . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02710 . 7554/eLife . 02730 . 028Figure 4—figure supplement 2 . Connectivity graphs of the Platynereis larval visual circuit . Nodes correspond to single neurons , edges represent connections , weighted by synapse number . The layout is based on force field clustering . Nodes in the full ( A ) or the trimmed ( B and C ) graphs were colored by eccentricity ( A ) , weighted degree centrality ( B ) or eigenvector centrality ( C ) . In ( A ) all cells and connections of the minimal eye circuit are shown , in ( B and C ) cells connected with <3 synapses and edges <3 synapses were removed . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02810 . 7554/eLife . 02730 . 029Figure 4—figure supplement 3 . Modules in the eye connectivity graph . ( A ) Trimmed connectomic graph of the visual eye circuit including 56 neurons and 5 effectors ( muscles and ciliary band cells ) . Colors indicate the four modules of the network . Edges are colored as their source node . ( B ) Anatomical position of the cells of the four modules . The cells are colored as in ( A ) , according to their module association . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 02910 . 7554/eLife . 02730 . 030Figure 4—figure supplement 4 . Connectivity matrix of the left and right body sides . Synaptic connectivity matrix of cells with a cell body on the left ( A ) and right ( B ) body side . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 03010 . 7554/eLife . 02730 . 031Figure 4—figure supplement 5 . Stereotypy of synapse distribution on IN1 cells . Spatial distribution of synapses onto axons of the posterior right ( A ) , posterior left ( B ) , anterior right ( C ) , and anterior left ( D ) IN1 axons . Synapses of individual PRCs to IN1 axons are shown in different colors . The cell-body proximal synapses from the axons of the respective crosswise IN1 cell are also shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 03110 . 7554/eLife . 02730 . 032Figure 4—figure supplement 6 . Stereotypy of synapse distribution on INton , INsn and MN cells . Spatial distribution of synapses between ( A and B ) IN1 and INton , ( C and D ) INton and INsn and ( E–H ) INsn and MN cells . Asterisks mark the presynaptic neuron . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 032 To analyze information flow we calculated the maximum distance of all nodes to any other node within the directed graph ( eccentricity; Figure 4A ) . The eye circuit is a feed-forward circuit where information flows from the eyes to the muscles and ciliary bands . We also calculated centrality measures to identify strongly connected nodes in the network ( Figure 4—figure supplement 2 ) . With three different centrality measures ( eigenvector centrality , weighted-degree centrality and authority ) the four primary interneurons ( IN1 ) ranked among the top five nodes ( Randel et al . , 2014 ) . These four cells receive several synapses from the photoreceptors , with the highest number of each IN1 received from the crosswise eyes ( Figure 4B , Figure 4—figure supplement 2 ) . The four IN1 cells also have strong mutual connections ( see below ) . A search for highly interconnected nodes or communities ( Blondel et al . , 2008 ) subdivided the eye network into four modules ( Figure 4—figure supplement 3 ) . The four IN1 cells belong to two different modules in the eye circuit ( Figure 4—figure supplement 3 ) , which together contain all neurons with projections intrinsic to the primary optic neuropil . Both modules consist of a crosswise pair of eyes and the associated IN1 cells and trans-optic-neuropil interneurons ( INton ) . The anatomy of the circuitry in the primary optic neuropil thus displays point symmetry rather than bilateral symmetry ( Figure 4—figure supplement 3 ) . The other two modules , the motor modules , contain all neurons intrinsic to the secondary optic neuropil , including Schnörkel interneurons and motorneurons and associated effector organs . A bilateral pair of INton cells links the modules of the primary and secondary optic neuropils , reflecting the anatomy where INton cells receive synapses in the primary and project to the secondary optic neuropil ( Figure 4—figure supplement 3 ) . One of the motor modules provides innervation to the prototroch and metatroch ciliary band cells . The other module contains motorneurons that connect to three out of four bundles of longitudinal muscles ( we could not find neuromuscular synapses on the left dorsal longitudinal muscle ) ( Figure 4A , B , Figure 4—figure supplement 2 ) . The connectome reveals that all four eyes can provide motor input to both body sides . The neuronal pathways from the eyes cross the midline multiple times throughout the circuit . The strongest synaptic connections are formed between photoreceptor—IN1—INton—Schnörkel interneuron , always to the contralateral side in a feed-forward circuit ( Figure 4B ) . However , most Schnörkel interneurons form several synapses on both ipsilateral and contralateral motorneurons in the secondary optic neuropil , representing a bilateral divergence of the connections . Motorneuron axons receive cell-body-proximal synapses from the ipsilateral Schnörkel interneurons , cross the midline , and receive cell-body-distal synapses on the other side of the body from the contralateral Schnörkel interneurons ( Figure 4—figure supplement 6G , H; Video 5 ) . In order to estimate how stereotypic synaptic connectivity is within the circuit we compared the connectivity matrices of neurons with cell bodies on the left or right side of the body . We used the merged graph ( Figure 4B ) since not all neurons had a contralateral pair ( e . g . , we identified three INton cells on the left and two on the right body side ) . We found a strong correlation between the left and right connectivity matrices ( Spearman r = 0 . 67 , p<0 . 0001; Figure 4—figure supplement 4 ) . We also analyzed the stereotypy of the spatial arrangement of pre-and postsynaptic sites among neurons of the same type . The four IN1 cells always receive photoreceptor input in a cluster at the most distal segment of their axon ( Figure 4—figure supplement 5 ) . Synapses onto the IN1 cells from the contralateral photoreceptors are intermingled with presynaptic sites of IN1 cells to INton cells ( Figure 4—figure supplement 6 ) and are not spatially segregated within the cluster . In contrast , IN1 cells receive axo-axonal synaptic input from the crosswise IN1 cell in a cluster positioned more proximal to the cell body ( Figure 4—figure supplement 5 ) . The spatial arrangement of synapses is also similar between the left and right sides for other neuron types ( Figure 4—figure supplement 6 ) . These observations show that both the number and the spatial arrangement of synapses are stereotypic between the left and right body sides . The reconstruction of further individuals will be needed to assess individual-to-individual variation . During development of the Platynereis larva , photoreceptors are continuously added to the eyes at the periphery of the pigment cup ( Rhode , 1992 ) . Individual photoreceptors in an eye are thus at different stages of differentiation within the same individual . We noted that in our reconstructed specimen the photoreceptor rhabdoms have different sizes , and that rhabdom volumes positively correlate with axon length and synapse number ( Figure 5A , B ) . Using rhabdom volume as a proxy to photoreceptor differentiation we then analyzed how photoreceptor connections change during neuron maturation ( Figure 5D ) . 10 . 7554/eLife . 02730 . 033Figure 5 . Maturation of photoreceptor connections . ( A ) Relationship of rhabdom volume to photoreceptor axon length . ( B ) Relationship of rhabdom volume to photoreceptor synapse number . ( C ) Relationship of photoreceptor connectivity-maturation index to rhabdom volume . In A–C the black line shows linear regression with 95% confidence interval ( red dashed lines ) . Pearson r and p-value are shown . ( D ) Connectivity matrix of the photoreceptors . The matrix is ordered from top to bottom by eye and then for each eye by photoreceptor rhabdom size increasing from top to bottom . Eyeal , anterior-left eye; eyear , anterior-right eye; eyepl , posterior-left eye; eyepr , posterior-right eye . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 033 Photoreceptors with short axons connect weakly if at all to the primary interneurons ( IN1 ) , while photoreceptors with longer axons form more synapses on IN1 cells . Photoreceptors with the shortest axons have connections to other interneurons ( e . g . , INintl4 , INintr2 ) that are not observed in photoreceptors with longer axons ( Figure 5D ) . We defined a photoreceptor connectivity-maturation index ( number of crosswise IN1 connections per all connections ) and found that it correlates with rhabdom volume ( Figure 5C ) . These data represent a snapshot during development in one individual , but suggest that during photoreceptor development connections to the IN1 cells get stronger and initial contacts to other interneurons are eliminated . The INton cells also show variation in axon length and connectivity , possibly also reflecting developmental progression . The small number of these interneurons precluded similar analyses . The innervation of the longitudinal trunk muscles and ciliary bands by the eye circuit motorneurons suggested that the eyes could regulate tail bending and ciliary beating during larval swimming . To understand the function of the eye circuit we next analyzed larval photobehavior in detail , in conjunction with eye ablation experiments . We first analyzed swimming trajectories of freely behaving 3- and 4-day-old larvae exposed to light stimuli of alternating directionality . 3- and 4-day-old larvae swim using their cilia while rotating around their anterior–posterior axis . When we used alternating illumination from the two opposite sides of an assay cuvette we observed directional swimming ( phototaxis ) of the larvae . Larvae showed mixed behavior , with some swimming towards and others away from the light source ( Video 6 ) . Such sign-switch in directional swimming responses is common for marine larvae and can be influenced by various environmental stimuli ( Thorson , 1964; Young and Chia , 1982; Marsden , 1990 ) . During reorientations we could observe prominent body bending in 4-day-old larvae and weaker bending in 3-day-old larvae , suggesting that larvae turn by contracting longitudinal muscles while swimming with cilia ( Video 7 and data not shown ) . In agreement with the cholinergic identity of the motorneurons , treatment with mecamylamine , an acetylcholine receptor antagonist , blocked negative phototaxis . Mecamylamine treatment increased swimming speed , probably via influencing cilia ( Jékely et al . , 2008 ) . The effects could be reversed by washout ( Figure 6—figure supplement 1; Randel et al . , 2014 ) . 10 . 7554/eLife . 02730 . 034Video 6 . Mixed positive and negative phototaxis in 3-day-old Platynereis larvae . Larvae were stimulated with alternating directional white light from the left or the right side of the phototaxis cuvette ( shown by white bars on the side ) . Larvae display mixed phototaxis , some negatively phototactic larvae are tracked . Scale bar , 2 mm . Time increment: 0 . 07 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 03410 . 7554/eLife . 02730 . 035Video 7 . A single phototactic turn in 4-day-old larvae . The larvae were exposed to a 180-degree change in the direction of the white stimulus light , eliciting a phototactic turn . The stimulus direction is indicated by the white bars at the side of the video . During the turn the larvae are bending due to the contraction of the longitudinal muscles on one body side . Time increment: 0 . 066 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 035 Next we illuminated the cuvette constantly from both sides but with different left-right light intensities and measured the efficiency of phototaxis for a population of larvae using a phototaxis index . We found that phototactic efficiency increased with increasing contrast between the light intensity at each side of the cuvette , but was independent of total light intensity ( Figure 6—figure supplement 2 ) . To test the contribution of the eyes and the two eyespots ( independent ventral structures that develop in 1-day-old larvae , Figure 1B ) to directional swimming , we laser-ablated the eyes or eyespots and subsequently exposed the larvae to directional illumination . Ablation of the two eyespots required for non-visual phototaxis in 1-day-old larvae ( Jékely et al . , 2008 ) did not abolish phototaxis responses in 3-day-old larvae . In contrast , ablation of all four eyes led to the loss of directional swimming ( Figure 6F ) , demonstrating that only the eyes mediate this behavior . 10 . 7554/eLife . 02730 . 036Figure 6 . Eyes mediate body bending during visual phototaxis . ( A ) Selective eye illumination triggers body bending in an immobilized larva . ( B ) A larva with both left eyes ablated displays body bending upon uniform illumination with white light ( light ) but not with a red filter ( dark ) . An asterisk marks the tip of the tail . ( C and D ) Trajectories of larvae with both left eyes ablated in the dark ( red filter ) ( C ) and upon uniform illumination with white light ( D ) . ( E ) Tail bending upon uniform white light illumination of non-ablated control larvae and larvae lacking the two left eyes . Data are shown as mean ± SEM , n >9 for both condition . p-value of a t test calculated for the last time point is indicated . ( F ) Percentage of phototactic ( red ) and non-phototactic ( black ) larvae among non-ablated control and various eye-ablated larvae . p-values of a chi-square test are indicated relative to non-ablated controls . Only the ‘all eyes’ and ‘two left eyes’ ablated conditions are significantly different from non-ablated control . Number of larvae tested is shown above the columns for each condition . Scale bars , 40 µm ( A and B ) , 1 mm ( C and D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 03610 . 7554/eLife . 02730 . 037Figure 6—figure supplement 1 . Inhibition of phototaxis by a cholinergic antagonist . Phototaxis index ( A ) and swimming speed ( B ) of control 3-day-old larvae and larvae treated with the acetylcholine receptor antagonist mecamylamine ( 50 μm ) . Data are shown as scatter plots with mean ± SEM . The phototaxis index is a population measure , the data are from five replicate experiments , each with >14 larvae . Swimming speed was averaged for single larvae ( n >138 larvae ) . p-values of unpaired t tests relative to control larvae are shown . Source data are available from Randel et al . ( 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 03710 . 7554/eLife . 02730 . 038Figure 6—figure supplement 2 . Efficiency of phototaxis depends on contrast , not absolute light levels . Larval phototaxis was measured in a cuvette illuminated from both sides . The stimulus light was dimmed on one side by adding progressively more 10% or 25% neutral density filters . The experiment was performed with 100% , 50% and 10% starting white light intensity . Data are shown as mean ± SEM . The data points were fitted with a saturation binding-curve . Contrast is defined as ( Imax − Imin ) / ( Imax + Imin ) where Imax and Imin are maximum and minimum intensity . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 038 To directly investigate if the eyes are able to mediate body bending we performed selective eye illumination experiments on larvae held between a slide and a coverslip . Unilateral illumination of the eyes triggered reproducible bending of the body in 3-day-old larvae ( Figure 6A; Video 8 ) . When we used varying stimulus-light intensities for selective eye illumination we obtained graded bending responses with saturation kinetics ( Figure 7D ) . 10 . 7554/eLife . 02730 . 039Video 8 . Selective illumination of eyes triggers body bending . Illumination of the eyes on one body side in the area shown by the green signal triggers body bending on the opposite body side , corresponding to negative phototaxis ( left ) , or the same body side , corresponding to positive phototaxis ( right ) . Scale bar , 50 μm . Time increment: 0 . 43 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 03910 . 7554/eLife . 02730 . 040Figure 7 . An interneuron motif for enhanced contrast detection . ( A ) Strength of reciprocal connections between all neuron pairs in the complete visual circuit . The strength of reciprocal connections was defined as the geometric mean of the number of reciprocal synapses between each neuron pair . The single neuron identifiers are not shown for simplicity . ( B ) The strongest reciprocal motif in the eye circuit is between the crosswise IN1 pairs . The position and polarity of the synapses are indicated . ( C ) Quantification of tail bending in different eye ablated larvae under selective eye illumination with a 488-nm laser . Data are shown as mean ± SEM , n >17 for each condition . Larvae lacking two anterior eyes were compared to non-ablated control larvae at each time point ( p value<0 . 05 , *p value<0 . 01 , unpaired t test ) . ( D ) Signal-response curve of maximum tail bending upon selective eye illumination in immobilized larvae using 488-nm stimuli of different intensities . The data are fitted with a saturation-binding curve . Data are shown as mean ± SEM , n >24 larvae for each condition ( independent from C ) . p-values of unpaired t tests comparing larvae lacking two anterior eyes to the other conditions are shown . Source bending data from ( C and D ) are shown in Figure 7—source data 1 and Figure 7—source data 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 04010 . 7554/eLife . 02730 . 041Figure 7—source data 1 . Source data of tail bending experiments in Figure 7C . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 04110 . 7554/eLife . 02730 . 042Figure 7—source data 2 . Source data of tail bending experiments in Figure 7D . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 042 Unilateral laser ablation of both eyes ( Video 9 ) followed by uniform illumination also triggered strong and persistent body bending ( Figure 6B , E; Video 10 ) . Interestingly , in both experiments , the larvae either bent on or opposite to the illuminated side , with some larvae switching the response during repeated stimulation ( Video 8 and data not shown ) . These results indicate that it is either the contra- or the ipsilateral motor pathway that dominates . Given that the Schnörkel interneurons synapse on both contra- and ipsilateral motorneurons ( Figure 4—figure supplement 6 ) , phototactic sign switching may take place at the level of the motorneurons . 10 . 7554/eLife . 02730 . 043Video 9 . Laser ablation of the eyes . The position of the four eyes and the two eyespots is shown by changing the imaging focus . The right eyes are ablated . The eye pigment is imaged using reflection imaging ( red ) , and is overlaid on the DIC channel . Scale bar , 25 μm . Time increment: 1 . 1 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 04310 . 7554/eLife . 02730 . 044Video 10 . Uniform illumination triggers body bending following unilateral eye ablation . Both left eyes were ablated . The larva is imaged with DIC illumination with a 750 nm long-pass filter . The filter is removed to provide uniform white illumination from the microscope lamp at frame 98 . The pigment of the right eye is visible . Scale bar , 50 μm . Time increment: 0 . 1 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 044 To dissect the interplay of the four eyes during visual phototaxis , we ablated them in different combinations and tested phototactic ability . Larvae with unilateral ablation of both eyes were not phototactic and swam persistently in circles when illuminated ( Figure 6C , D , F ) . The two anterior and two posterior eyes connect to distinct , though partly overlapping , circuits . To test if input to both the anterior and posterior eyes is necessary for directional swimming , we ablated either the anterior or the posterior eye pair . We also performed crosswise eye ablations . All of these ablated larvae showed phototactic responses similar to non-ablated controls ( Figure 6F ) , hence the presence of at least one eye on both body sides is necessary and sufficient for directional turns . To investigate if both dorsal and ventral longitudinal muscles are involved in body bending , we performed calcium-imaging experiments using ubiquitously expressed GCaMP6 ( Chen et al . , 2013 ) . In immobilized larvae we repeatedly stimulated the eyes with 488 nm light on one body side . In agreement with the circuit diagram , upon selective eye stimulation we could observe corresponding calcium signals in both the ventral and dorsal longitudinal muscles on the opposing body side ( Videos 11 and 12 ) . 10 . 7554/eLife . 02730 . 045Video 11 . Calcium-imaging in ventral longitudinal muscles . Calcium-imaging with GCaMP6 in the ventral longitudinal muscle during selective illumination of the right eyes . The larva is ventrally oriented , the eyes are out of focus . The circular ROI shows the illuminated area . The duration of the illuminations is visible in the DIC channel ( brighter overall signal ) . Scale bar , 50 μm . Time increment: 0 . 57 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 04510 . 7554/eLife . 02730 . 046Video 12 . Calcium-imaging in dorsal longitudinal muscles . Calcium-imaging with GCaMP6 in the dorsal longitudinal muscle during selective illumination of the right eyes . The larva is dorsally oriented , the eyes are visible . The circular ROI shows the illuminated area . The duration of the illuminations is visible in the DIC channel ( brighter overall signal ) . Scale bar , 50 μm . Time increment: 0 . 57 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 046 Overall , these experiments demonstrate that the eyes mediate visual turns by comparing simultaneous light inputs at each side of the body . This , together with the shading provided by the pigment cups ( Video 2 ) , allows the detection of the spatial distribution of light , without body movement . The visual contrast then leads to the graded contraction of the longitudinal muscles during turns . The modulation of ciliary beating , not investigated here , may also contribute to phototactic turns , as is known for younger larvae ( Jékely et al . , 2008 ) . This sensorimotor structure is similar to visual phototaxis described in the lamprey ( Ullén et al . , 1997 ) , but is very different from the non-visual phototactic responses in Drosophila larvae ( Kane et al . , 2013 ) . Visual phototaxis is also fundamentally different from the helical phototaxis mediated by the eyespots during early Platynereis larval stages ( 1–2 days ) ( Jékely et al . , 2008 ) . When we analyzed mutual synaptic connections in the eye circuit we found that by far the strongest mutual connections are formed between the crosswise pairs of IN1 cells ( up to 18 synapses one way; Figure 7A , B ) . These mutual , and thus likely inhibitory , connections are formed at the cell-body proximal segment of the IN1 cells' axons , spatially segregated from the more distal sites of photoreceptor input ( Figure 4—figure supplement 5 ) . This reciprocal wiring may provide mutual inhibition and reinforce small differences in activation between the crosswise eyes , potentially representing a network motif for enhancing contrast-detection . Modularity analysis also revealed that the sub-networks of the crosswise eyes are more strongly connected than those of the pairwise eyes ( Figure 4—figure supplement 3 ) . This suggests that two reciprocally connected crosswise eyes are more efficient in visual information processing than a bilateral pair . To test this , we performed various eye ablations and then subjected the larvae to selective eye illumination . We ablated a bilateral pair of eyes , thus eliminating inputs to one side of both crosswise mutual IN1 motifs . We expect that these larvae would have a reduced bending response at the same level of contrast . We also performed crosswise eye ablations , leaving input to one of the crosswise IN1 motifs intact . We then exposed larvae to a 488 nm background illumination and measured their tail bending following unilateral eye illumination with 488 nm stimulus light of varying intensity . Larvae with pairwise ablation of the anterior eyes showed significantly reduced bending relative to non-ablated controls with the average maximum tail displacement reduced to approximately half ( Figure 7C , D ) . In contrast , larvae with crosswise eye ablation showed stronger bending that was not significantly different from non-ablated controls . Although both crosswise and pairwise ablated larvae have only two eyes and are phototactic ( Figure 6F ) , these results show that pairwise ablated larvae have a markedly reduced motor response at the same level of contrast . We conclude that the strong mutual contacts between the crosswise IN1 cells represent a circuit motif that enhances turning responses during phototaxis . Here we described the neural connectome of the visual system in the Platynereis larva . The four eyes mediate visual phototaxis , during which larvae are able to detect and contrast spatial differences in the light field . Although it is not yet possible to directly link all neurons and connections in the Platynereis larva to phototactic behavior , the combined analysis of connectivity and behavior yielded fundamental insights into a visually guided behavior . We concluded that spatial vision is not due to the presence of more ( 3–7 in our larva ) photoreceptors in the eyes , since all photoreceptors from one eye synapse on the same primary interneuron . Instead , spatial vision relies on at least two eyes , pointing in different directions , and the underlying neural circuit that contrasts simultaneous light inputs at the left and right eyes . Furthermore , connectomics and modularity analysis revealed that the four eyes connect to the downstream visual circuitry in a crosswise manner , showing point symmetry . The crosswise-eye-modules show mutual connections at the level of the primary interneurons . In agreement with this , ablation experiments demonstrated that a bilateral pair of eyes elicits a smaller motor response than a crosswise pair , at the same level of light-intensity contrast . The contrast enhancement takes place in the primary optic neuropil circuitry , and the INton may relay a contrasted , one-sided signal to the secondary optic neuropil . In the secondary optic neuropil this one-sided input could lead to the biased activation of the left or right motorneurons . Our circuit reconstructions partly explain the bimodality of the behavior ( either positive or negative phototaxis ) . Such switching behavior requires neuronal connections between the eyes on one body side and the muscles on both body sides . We found that indeed the circuit diverges bilaterally at the level of the Schnörkel interneurons ( INsn ) , when these cells synapse to motorneurons of both the left and right body side . The Schnörkel interneuron synapses form at axon segments either proximal or distal to the motorneuron cell bodies ( Figure 4—figure supplement 6 ) . The spatial organization of synapses may provide an initial bias to the system , favoring bends on one body side . Further modulatory input may influence this bias , inducing a sign switch . There are several sensory neurons , not described here , that feed into the minimal eye circuit at the middle segment of the motorneuron axons ( NR , LABC , and GJ , unpublished ) . Further work will be needed to characterize the possible roles of these neurons in sign switching . The sensory-motor strategy of visual phototaxis is similar to that found in the lamprey ( Ullén et al . , 1997; Figure 8 ) . In this vertebrate , unilateral illumination leads to a lateral turn away from the light during negative phototaxis . However , under some circumstances lampreys display positive phototaxis . Lesion experiments demonstrated the involvement of the pretectum and reticulospinal neurons in phototaxis , the latter forming the descending control system . Surgically severing connections between the pretectum and the ipsilateral reticulospinal neurons ( Figure 8E ) leads to a sign switch in the phototactic response when the ipsilateral side is illuminated . Transection of the ventral rhombencephalic commissure ( Figure 8E ) in turn reduces the turning angle during negative phototaxis . Thus both the Platynereis and the lamprey phototactic circuits are characterized by extensive midline crossing at several levels , bilateral divergence to allow context-dependent sign switching , and bilateral , probably inhibitory , interactions to enhance turning magnitude ( Figure 8D , E ) . 10 . 7554/eLife . 02730 . 047Figure 8 . Comparison of the Platynereis , Drosophila and vertebrate visual circuits . Comparison of the Platynereis visual circuit with the Drosophila and vertebrate visual circuits on the neuronal level ( A–C ) and with the lamprey phototactic circuit on the circuit level ( D and E ) . In ( E ) dashed line ( 1 ) represents a mesencephalic hemisection , severing connections between the pretectum and the ipsilateral reticulospinal neurons , dashed line ( 2 ) represents transection of the ventral rhombencephalic commissure . L , lamina monopolar neuron; Tm , transmedula neuron; Mi , medulla intrinsic neuron; Lcp , lobula complex projection neuron; Lci , lobula complex intrinsic neuron; BPC , bipolar cell; HC , horizontal cell , AC , amacrine cell; RGC , retinal ganglion cell . ( B and C ) after Erclik et al . ( 2009 ) ; ( Sanes and Zipursky , 2010 ) ( E ) after Ullén et al . ( 1997 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02730 . 047 At the cellular level , the Platynereis eye circuit also shows similarity in its multi-layered arrangement to the visual circuits of insects and vertebrates ( Figure 8A–C; Sanes and Zipursky , 2010 ) . In the visual system of Drosophila photoreceptors project to the optic lobe that is organized into distinct ganglia ( the lamina , medulla and lobula complex ) ( Erclik et al . , 2009 ) . Some neurons are intrinsic to one ganglion , others connect two adjacent ganglia ( e . g . , transmedullary neurons ) . Similarly , in the Platynereis circuit the IN1 and INint cells are intrinsic to the primary optic neuropil , and INton cells link the primary and secondary neuropils . With the exception of the photoreceptors ( Arendt et al . , 2002 ) , the evolutionary relationships of the cell types ( ‘cell-type homology’ ) ( Arendt , 2008; Erclik et al . , 2009 ) in the vertebrate , insect , and annelid visual systems are unclear and more comparative work will be needed to assess the evolutionary significance of these similarities . Additionally , our work provides a more general insight about the evolution of higher-resolution image-forming eyes . The current model of eye evolution defines four steps ( Nilsson , 2009 ) , from non-directional photoreception through directional scanning photoreception ( spiral phototaxis ) ( Jékely et al . , 2008; Jékely , 2009 ) and low-resolution spatial vision to high-resolution spatial vision . We now extend this scheme with the concept of the two-pixel visual phototactic eye , likely predating the evolution of low-resolution spatial vision . Simple eyes , similar to the eyes of Platynereis larvae , are widespread in the planktonic larval stages of several bilaterians and may often function in visual phototaxis ( Buchanan , 1986; Blumer , 1994; Lacalli , 2004 ) . More complex image-forming eyes may have repeatedly evolved from such phototactic eyes . Larval eyes sometimes directly develop into the image-forming eyes of the adults ( Cazaux , 1985; Blumer , 1994 ) . In Platynereis , and several other annelids and mollusks , the larval eyes develop into the adult's eyes , which harbor hundreds or thousands of photoreceptors ( Rhode , 1992 ) and mediate low-resolution image-forming vision . In Platynereis , we identified visual phototaxis as the first function during the development of the eyes . As in development , also during evolution the first images seen by animals may have consisted of a dark field at the bottom and a bright field at the top of the ocean . Platynereis larvae were obtained from an in-house breeding culture following an established protocol ( Hauenschild and Fischer , 1969 ) . Larvae were kept at 18°C for development . Behavioral experiments were performed at room temperature . 72 hr post fertilization Platynereis larvae , reared at 18°C , were fixed using a high-pressure freezer ( HPM 010; BAL-TEC , Balzers , Liechtenstein ) and transferred to liquid nitrogen . Frozen samples were cryosubstituted with 2% osmium tetroxide in acetone and 0 . 5% uranyl acetate in a cryosubstitution unit ( EM AFS-2; Leica Microsystems GmbH , Wetzlar , Germany ) over a regime of gradually rising temperatures . Samples were embedded in Epon . 40–50-nm serial sections were cut starting from the anterior end of a larva ( Platynereis HT-9-3 ) on a Reichert Jung Ultracut E microtome . The sections were collected on single-slotted copper grids ( NOTSCH-NUM 2 × 1 mm , Science Service , Munich , Germany ) with Formvar support film , contrasted with uranyl acetate and Reynold's lead citrate , and carbon coated to stabilize the film . Image acquisition of serial sections was performed at a pixel size of 3 . 72 nm on a FEI TECNAI Spirit transmission electron microscope ( FEI , Hillsboro , Oregon ) equipped with an UltraScan 4000 4 × 4k digital camera using the image acquisition software Digital Micrograph ( Gatan , Pleasanton , CA ) . Stitching and alignment were accomplished using TrakEM2 ( Cardona et al . , 2010 ) . The first 1245 layers were cut at 50 nm and aligned using an elastic alignment algorithm ( Saalfeld et al . , 2012 ) . We later found that the larva was slightly tilted , and we did not reach the muscles and the ciliary band on the right side of the animal . We therefore sectioned a further 445 sections at 40 nm , and aligned these using the rigid alignment algorithm in TrakEM2 . All structures were segmented manually as area-lists or area-trees by an expert tracer ( NR ) . Traces were exported into 3Dviewer , Imaris and Blender . Given the simple anatomy of the neurons and the large average diameter of axons tracing was in many cases unambiguous . Neurons with several short branches or with uncertain continuation were traced again by another expert tracer ( LABC ) . The minimal eye circuit of the eyes contains only postsynaptic neurons downstream of photoreceptors . Neurons presynaptic to the minimal eye circuit at any level ( with possible modulatory functions ) are not included here . Five sensory cells , which are both pre- and post-synaptic to the minimal eye circuit were also excluded . 12 postsynaptic targets of the MNs in the ventral nerve cord and five between the secondary optic neuropils as well as 32 postsynaptic targets of interneurons were also excluded because of low connectivity and the occurrence of only individual synapses outside the optic neuropils . 48 neurite fragments could not be traced completely . Most of these fragments are short ( <5 µm ) with 1–5 synapses . They occur in regions where neurons have small short branches to receive or provide synapses . These fragments likely belong to identified neurons . Nine of the non-traceable fragments were longer ( up to 44 µm ) . We classified neurons according to their morphology , position , projection pattern and connectivity . Photoreceptors were identified based on the presence of a rhabdom located in the eye pigment cup . Their dominant targets are the four primary interneuron ( IN1 ) cells . IN1 cells are named based on their position ( anterior-left etc . ) . IN1 cells are defined as a separate group based on their unique connectivity patterns , including strong input from photoreceptors and strong reciprocal contacts between the crosswise cells . The second group of interneurons is classified as INints , with projections intrinsic to the primary optic neuropil and weak connectivity to photoreceptors and IN1 cells . A third group is formed by the INton cells , distinguished by their unique projection pattern from the primary to the secondary optic neuropil . Schnörkel interneurons ( INsn ) form a distinct class with ventral cell bodies and curved axons ( ‘Schnörkel’ is German for ‘curlicue’ ) projecting to the ipsilateral secondary optic neuropil and synapsing on the motorneurons . The dorsal interneurons ( INdc ) have dorsally located cell bodies and axons that project to the secondary optic neuropil . The ventral interneurons ( INvc ) have ventrally located cell bodies and axons that project to the secondary optic neuropil . Motorneurons were recognized based on their innervation of dorsal and ventral longitudinal muscles as well as multi-ciliated cells ( prototroch or metatroch ) . Motorneurons form two bilateral clusters . For some motorneurons we could not find motor synapses , however , we could classify these cells as motorneurons based on their cell body position and posteriorly projecting axons . Neurons of all types are labeled based on body side ( left [l] or right [r] ) and numbered ( e . g . , INtonl1 ) . Network analyses were done in Gephi 0 . 8 . 2 . Modules were detected with an algorithm described in Blondel et al . ( 2008 ) with randomization on , using edge weights and a resolution of 1 . 2 . Force-field-based clustering was performed using the Force Atlas 2 Plugin . Neurons and their connections reconstructed in TrakEM2 were exported as 3D objects in a wavefront ( . obj ) format and imported into Blender . Cell bodies were approximated with ellipsoids and axons and dendrites with smooth curves . The approximations were performed either manually or automatically using Python scripting embedded in Blender . Muscles were kept in the original form , pigment cups and ciliated cells were approximated with simplified 3D shapes based on TrakEM2 reconstructions . The model was extended with the connectivity and synapse position information obtained from TrakEM2 . The virtual atlas ( Randel et al . , 2014 ) can be opened with Blender ( http://www . blender . org/ ) . Users new to Blender may consider watching a tutorial ( http://www . blendtuts . com/2010/06/blender-25-interface . html ) . After opening the model file , two panels are visible in the Tool shelf: Hide/Show groups and Highlight subnetworks ( Press ‘T’ if Tool shelf did not appear ) . If the panels do not appear , they can be loaded with the button Reload Trusted in the main menu panel ( on the top ) . If menus still do not appear , the Panel_show_group . py and Panel_synapses . py python scripts need to be run by selecting the Scripting view in the main menu panel instead of Default . In the panel Hide/Show groups the cell types can be selected and hidden or displayed by pressing the Apply button . The panel Highlight subnetworks enables querying pre- and post-synaptic cells by selecting a cell body of interest ( right click ) and pressing the Postsynaptic or Presynaptic buttons . The complete up- or down-stream network of a cell can also be displayed by selecting a cell body and pressing the Pre . subnetwork or Post . subnetwork buttons . The Show connectors button visualizes the connectors between two cells after selecting the presynaptic cell first and then the postsynaptic cell . The Show out connectors button will reveal all outgoing synapses for a selected cell . Laser ablation was performed on an Olympus FV1000 confocal microscope equipped with a SIM scanner ( Olympus Corporation , Tokyo , Japan ) . Larvae were immobilized between a slide and a coverslip in NSW containing 100 mM MgCl2 . Larvae were imaged with an UPLSAPO 60X NA:1 . 20 water immersion objective using a 635-nm laser at 2–5% and transmission imaging with DIC optics . A 351-nm pulsed laser ( Teem Photonics™ , Grenoble , France ) at 8–15% power was used , coupled via air to the SIM scanner for controlled ablation in a region of interest . 12% corresponds to a beam power of 168 µwatts as measured with a microscope slide power sensor ( S170C; Thorlabs , Newton , NJ ) . During eye ablations we also imaged the eye pigment by reflection imaging of the 635-nm light using a PMT . Ablated larvae were placed into NSW for recovery ( 1–6 hr ) before behavioral experiments . Selective eye illumination was performed on an Olympus FV1000 confocal microscope equipped with a SIM scanner . Larvae were held by trapping them between a slide and a coverslip in natural seawater using several layers of tape as spacer . Larvae were imaged with an UPLSAPO 40X NA:0 . 90 air objective using a 635 nm laser at 1% and transmission imaging with DIC optics . The scanning speed was 2 μm/msec and we recorded 256 × 256 pixel images with a time increment of 0 . 43 s . For measuring the stimulus–response curve we used background illumination with the 488 nm laser at 10% power corresponding to 17 . 8 μwatts in the main scanner . For stimulations we used the 488 nm laser at 1–65% power via the SIM scanner , corresponding to a beam intensities of 0 . 53–20 . 9 μwatts . A circular region of interest of 25-pixel diameter ( area of 490 pixels ) covering the eyes was used for controlled illumination . The pixel dwell time of the beam was 10 μsec and the eyes were stimulated during 150 frames with the SIM scanner corresponding to a total exposure time of 735 msec during a 9 . 8 s trial . For the bending Video 8 the stimulus laser power was at 10% . The laser power was measured with a microscope slide power sensor ( S170C; Thorlabs ) . To record the spatial and temporal extent of the stimulus the reflected 488-nm light was imaged with a PMT . To prevent the detection of the 488-nm stimulus light by the transmitted light detector , we placed a red long-pass filter in front of the detector . Whole-mount in situ hybridization and image registration were performed as previously described ( Asadulina et al . , 2012 ) . For probe generation we used expressed sequence tag clones or PCR-amplified fragments of already published Platynereis genes . The GeneBank accessions are: TyrH—JZ446954 , TrpH–JZ446141 , VGlut—JZ395359 , hdc—JZ396646 , VAChT—JZ402823 , ChAT—JZ402100 , gad—GU169427 , dbh—KJ855061 . For registering gene expression patterns of 3-day-old larvae we generated a reference template ( Randel et al . , 2014 ) where we removed the bias present in the earlier reference , due to the use of a single larva as a starting template ( Asadulina et al . , 2012 ) . In brief , we identified a median-size larva and registered it to every single image stack used for template generation ( 40 stacks ) , computed the average transformation and applied it to the median-size larva . The corrected larva was used as the starting point for the iteratively template generation ( Asadulina et al . , 2012 ) . We registered minimum four individual larval whole-body scans for each gene to create an average expression pattern . Larval phototaxis was assayed in a horizontal 6 × 15 mm rectangular glass cuvette with 3 mm high walls . The cuvette was illuminated uniformly through a light diffuser with white light from a 150-watt halogen cold light source ( Schott KL 2500 LCD , Schott AG , Mainz , Germany ) . Larval behavior was recorded at 21 frames per second on a Zeiss Stemi 2000-CS stereomicroscope equipped with an AxioCam MRm camera ( Carl Zeiss AG , Jena , Germany ) . Phototaxis efficiency was measured using a custom ImageJ macro that calculated larval density at the left and right side of the cuvette before and after the light stimulus ( Randel et al . , 2014 ) . Fertilized eggs were injected with capped and polyA-tailed GCaMP6 ( Chen et al . , 2013 ) RNA generated from a vector ( pUC57-T7-RPP2-GCaMP6 ) containing the GCaMP6 ORF fused to a 169 base pair 5′ UTR from the Platynereis 60S acidic ribosomal protein P2 . 3-day-old larvae were held as described above . Imaging was performed on an Olympus FV1000 confocal microscope equipped with a SIM scanner . Larvae were imaged with an UPLSAPO 60X NA:1 . 20 water immersion objective using a 635-nm laser at 10% and transmission imaging with DIC optics . The GCaMP6 signal was imaged with a 488-nm laser at 2% intensity . The eyes were stimulated with the 488-nm laser in a region of interest using the SIM scanner . GCaMP6 signals were recorded simultaneously .	Many animals show automatic responses to light , from moths , which are attracted to light sources , to cockroaches , which are repelled by them . This phenomenon , known as phototaxis , is thought to help animals navigate through their environment . It is an evolutionarily ancient behavior , as revealed by its widespread presence in the animal kingdom . One animal with a simple visual system for phototactic behavior is the marine worm Platynereis dumerilii . Platynereis is a segmented worm ( annelid ) with four eyes on the top of its head , two on the right and two on the left . Exposure to light triggers the contraction of muscles that run along the length of the body , causing the worm to bend and thus change the direction it is swimming in . Now , using a combination of high-resolution microscopy and behavioral experiments in larvae , Randel et al . have mapped the neural circuits underlying the worm's phototactic behavior . A 3-day-old Platynereis larva was sectioned to produce almost 1700 slices , each less than 50 nanometers thick , which were then viewed under a transmission electron microscope . By tracing individual neurons from one slice to the next , it was possible to reconstruct the entire visual system and all of its connections . This ‘visual connectome’ consisted of 71 neurons—21 light-sensitive cells , 42 interneurons , and 8 muscle-controlling motorneurons—organized into a circuit with 1106 connections . Shining light onto living larvae triggered phototaxis , with some larvae consistently swimming towards the light and others away from it . Using a laser to destroy all four eyes abolished this behavior , as did the removal of both eyes on either side of the head . By contrast , removing one eye from each side had no effect . This was because these larvae were still able to simultaneously compare the amounts of light reaching the left and right sides of their body , and to use any difference in these levels as a directional cue to guide swimming . By revealing the circuitry underlying phototaxis in a marine worm , Randel et al . have provided clues to the mechanisms that support this behavior in other species . The data could also provide insights into the processes that contributed to the evolution of more complex visual systems .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Neuronal connectome of a sensory-motor circuit for visual navigation
Myofibrils are huge cytoskeletal assemblies embedded in the cytosol of muscle cells . They consist of arrays of sarcomeres , the smallest contractile unit of muscles . Within a muscle type , myofibril diameter is highly invariant and contributes to its physiological properties , yet little is known about the underlying mechanisms setting myofibril diameter . Here we show that the PDZ and LIM domain protein Zasp , a structural component of Z-discs , mediates Z-disc and thereby myofibril growth through protein oligomerization . Oligomerization is induced by an interaction of its ZM domain with LIM domains . Oligomerization is terminated upon upregulation of shorter Zasp isoforms which lack LIM domains at later developmental stages . The balance between these two isoforms , which we call growing and blocking isoforms sets the stereotyped diameter of myofibrils . If blocking isoforms dominate , myofibrils become smaller . If growing isoforms dominate , myofibrils and Z-discs enlarge , eventually resulting in large pathological aggregates that disrupt muscle function . Inside cells , proteins are assembled into complex functional units . The correct assembly of these units is crucial for their function ( Marsh and Teichmann , 2015 ) . Myofibrils are highly organized assemblies of cytoskeletal proteins forming an array of sarcomeres that are embedded in the cytosol of myotubes and mediate contractility ( Huxley and Niedergerke , 1954b; Huxley and Hanson , 1954a; Huxley , 2004; Lemke and Schnorrer , 2017 ) . Sarcomeres are composed of actin-containing thin filaments and myosin-containing thick filaments arranged into antiparallel cables . Thin filaments are anchored to a large multiprotein complex called the Z-disc ( Luther , 2009 ) , and thick filaments are anchored to another large multiprotein complex called the M-line ( Agarkova and Perriard , 2005 ) . Anchoring of myofibrils to the exoskeleton provides the mechanical tension that aligns sarcomeres into myofibrils and coordinates their development ( Weitkunat et al . , 2017; Weitkunat et al . , 2014 ) . Once aligned , sarcomeres grow to their final size ( Lemke and Schnorrer , 2017 ) . Electron and confocal microscopy studies showed that sarcomeres form initially from small structures called Z-bodies that grow eventually into mature Z-discs to which thin filaments are anchored ( Loison et al . , 2018; Orfanos et al . , 2015; Reedy and Beall , 1993; Shafiq , 1963; Sparrow and Schöck , 2009 ) . The size of the Z-disc therefore sets the diameter of the myofibril ( Agarkova and Perriard , 2005; Luther , 2009 ) . While mechanisms have been proposed that set the length of sarcomeres ( Fernandes and Schöck , 2014; Gokhin and Fowler , 2013; Tskhovrebova and Trinick , 2017 ) , Z-disc growth and growth termination is poorly understood . A hallmark of genetically caused myopathies is the appearance of large aggregates composed mainly of Z-disc proteins ( Kley et al . , 2016; Maerkens et al . , 2016 ) . Interestingly , many myopathy-associated mutations encode Z-disc proteins . Mutations in any of the four α-actinin genes or in Zasp and other Alp/Enigma family genes in humans cause myopathies characterized by the presence of large aggregates ( Murphy and Young , 2015; Selcen and Engel , 2005 ) . The aggregation phenotype is conserved among animals: fruit flies with mutations in myopathy-related genes also develop Z-disc aggregates ( González-Morales et al . , 2017 ) . α-Actinin and Z-disc Alternatively Spliced Protein ( Zasp ) are conserved proteins that coordinate Z-disc formation ( Faulkner et al . , 1999; Katzemich et al . , 2013; Murphy and Young , 2015 ) . α-Actinin forms a rod-shaped antiparallel homodimer at the Z-disc , where it crosslinks and serves as an attachment point for actin filaments ( Djinović-Carugo et al . , 1999; Luther , 2009; Murphy and Young , 2015; Ribeiro et al . , 2014; Rusu et al . , 2017; Takahashi and Hattori , 1989 ) . Zasp and other members of the Alp/Enigma family of proteins are scaffolding proteins with an α-actinin-binding PDZ domain ( InterPro: IPR001478 ) , an uncharacterized Zasp Motif ( ZM; InterPro: IPR031847 and IPR006643 ) domain , and one to four protein-protein interaction LIM domains ( InterPro: IPR001781 ) ( Finn et al . , 2017; Klaavuniemi et al . , 2004; Liao et al . , 2016 ) . Zasp and α-actinin proteins are present at the earliest stages of Z-disc formation , and are required for Z-disc assembly ( Dabiri et al . , 1997; Katzemich et al . , 2013 ) . Vertebrates have seven Alp/Enigma genes , each encoding several isoforms . The Drosophila genome has three Zasp genes , Zasp52 , Zasp66 , and Zasp67 , which encode 21 , 12 , and 4 isoforms , respectively ( Gramates et al . , 2017 ) . Zasp66 and Zasp67 are duplications of Zasp52 and resemble the smallest isoforms of Zasp52 ( González-Morales et al . , 2019 ) . The number of isoform variants and genes adds an additional layer of complexity and regulation to sarcomere formation . We used Drosophila indirect flight muscles and asked if the mechanism that controls Z-disc size relates to the pathological aggregation behavior known for Z-disc-related myopathies . We show that accumulation of multivalent Zasp growing isoforms ( with multiple LIM domains ) causes Z-disc growth , whereas upregulation of monovalent blocking isoforms at later developmental time points terminates Z-disc growth . An imbalance of growing and blocking Zasp isoforms results either in aggregate formation , enlarged Z-disc size or reduced Z-disc size . We propose that this mechanism has wide implications for diseases caused by aggregate formation . Dominant Zasp mutations cause aggregate formation in human myopathies ( Selcen and Engel , 2005 ) . To examine if this holds true in Drosophila , we overexpressed a full-length Zasp52-PR transgene consisting of a PDZ , a ZM , and 4 LIM domains in the indirect flight muscles ( IFM ) of adult flies , which causes formation of large aggregates ( Figure 1A , B , I , Figure 1—figure supplement 1 ) . To determine which domain is responsible for aggregation , we deleted them individually: the absence of LIM domains ( Zasp52-PP ) and ZM domain ( Zasp52-PR∆ZM ) abolished aggregate formation , whereas removal of the PDZ domain only slightly reduced the number of aggregates ( Zasp52-PR∆PDZ , Figure 1C–E , I ) . The PDZ domain is required for binding α-actinin , a crucial crosslinker of actin filaments at the Z-disc ( Katzemich et al . , 2013; Liao et al . , 2016 ) . Reducing α-actinin levels by RNAi in Zasp52-PR overexpression reduces aggregate formation to a similar level ( Figure 1—figure supplement 2 ) , indicating that α-actinin increases aggregate number , but is not required for their initial formation . Expression of the paralogous genes Zasp66 and Zasp67 ( González-Morales et al . , 2019 ) , which lack LIM domains like Zasp52-PP , also does not result in aggregate formation ( Figure 1F , G and I ) . Finally , we tested if multiple LIM domains are required for aggregate formation by overexpressing Zasp52-PK consisting of PDZ , ZM and 1 LIM domain . No aggregates form ( Figure 1H and I ) indicating that the ZM domain and multiple LIM domains are required for aggregate formation . To test if aggregates formed as a result of Z-disc overgrowth , we tested the overexpression of Zasp52-PR using UH3-Gal4 , a driver line that has the same temporal and spatial expression pattern as Act88-Gal4 but is expressed at a much lower level . To better capture the size variation , we measured the Z-disc height that corresponds to the disc diameter and categorized them into specific size categories . In this condition , Z-discs are bigger than the control , but aggregates are not present ( Figure 1J–L ) , indicating that Z-disc growth and pathological aggregation correlate with the amount of Zasp52 protein . Next , we employed the yeast two-hybrid system ( Y2H ) to determine if Zasp forms oligomers by interacting with itself . Zasp52-PK and Zasp52-PE , another full-length isoform , can interact with each other , whereas controls do not interact ( Figure 2A ) . Full-length isoforms of Zasp66 and Zasp67 ( Zasp66-PK and Zasp67-PD ) can also interact with Zasp52-PK and Zasp52-PE ( Figure 2A ) . These data suggest that Zasp proteins can homo- and heterodimerize . We confirmed these results by co-immunoprecipitation of Flag-tagged Zasp52 with GFP-tagged Zasp52 from thorax muscle extracts ( Figure 2—figure supplement 1A ) . Furthermore , bacterially purified His-Zasp52-PK-Flag dimerized in vitro in a chemical crosslinking assay ( Figure 2—figure supplement 1B ) . To identify the domains involved in homo- and heterodimerization , we tested the interaction of Zasp52-PK , Zasp66 and Zasp67 with each individual domain of Zasp52 . Zasp52-PK interacts with the ZM domain and LIM domains , whereas Zasp66 and Zasp67 only interact with LIM domains ( Figure 2B ) . This suggests that dimerization is mediated by a ZM-LIM domain interaction , and because Zasp66 and Zasp67 lack LIM domains , they cannot interact with the ZM domain ( Figure 2C ) . To confirm this hypothesis , we next tested the interaction of LIM2A , LIM2B , LIM3 and Zasp52-PK with a series of Zasp66 deletion constructs . As soon as the ZM domain of Zasp66 is deleted , the interaction with LIM domains is abolished ( Figure 2D ) . Finally , we tested Zasp66-PH , a ZM-only isoform of Zasp66 for interaction with the individual domains of Zasp52 . Zasp66-PH can only interact with LIM domains of Zasp52 ( Figure 2E ) . Thus , LIM-ZM binding mediates homo- and heterodimerization between Zasp proteins . Next , we investigated if LIM-ZM interaction observed in a heterologous system also occurs in vivo at the Z-disc . We hypothesized that an endogenous GFP-tagged Zasp66 will be recruited to the Z-discs by Zasp52 . We analyzed two Zasp52 mutants that differentially affect sarcomere structure: Zasp52MI02988 disrupts N-terminal isoforms and only partially affects the last three LIM domains , whereas Zasp52MI00979 introduces a stop codon before the last three LIM domains ( Figure 3—figure supplement 1A ) ( Liao et al . , 2016 ) . Zasp66-GFP fluorescence is mildly reduced in Zasp52MI02988 mutants ( Figure 3A , B and D ) . In contrast , Zasp66-GFP fluorescence is strongly reduced in Zasp52MI00979 mutants ( Figure 3C and D ) , suggesting that Zasp52 recruits other Zasp proteins through its LIM domains , consistent with the Y2H results . α-Actinin localization at the Z-disc was not decreased in any of the Zasp mutants ( Figure 3—figure supplement 1B , C ) . To determine if LIM and ZM domains directly interact in flies at the Z-disc , we employed bimolecular fluorescence complementation ( BiFC ) assays ( Figure 3—figure supplement 1D ) ( Ciruela , 2008; Gohl et al . , 2010 ) . We first confirmed that the ZM-only isoform Zasp66-PH fused to a Venus tag localized to Z-discs ( Figure 3—figure supplement 1E ) . We then generated transgenic flies expressing Zasp52-PK and Zasp66-PH tagged with either the C-terminus or the N-terminus of yellow fluorescent protein ( NYFP or CYFP ) and quantified the fluorescence of reconstituted YFP at the Z-disc . Zasp52-PK-NYFP and Zasp52-PK-CYFP show specific fluorescence at the Z-disc , whereas controls do not ( Figure 3E–G ) . More importantly , the ZM-only isoform Zasp66-PH-NYFP interacts specifically with Zasp52-PK-CYFP at the Z-disc ( Figure 3H ) . Thus , ZM-LIM interaction of Zasp proteins occurs in vivo at the Z-disc . LIM domains are well-known protein-protein interaction domains ( Kadrmas and Beckerle , 2004 ) . In contrast , functional information on the ZM domain is scarce . It is an uncharacterized short motif of 26 amino acids found only in Alp/Enigma proteins ( Klaavuniemi et al . , 2004; Letunic et al . , 2015 ) . Alignment of Drosophila Zasp-encoding genes to the ZM consensus sequence shows weak conservation ( Figure 4A ) . To determine the phylogenetic distribution of the ZM domain , we compiled the presence of ZM domains in all branches of metazoans using data from the PFAM database ( Bateman et al . , 2004; El-Gebali et al . , 2019 ) . In contrast to the universal presence of LIM domains in eukaryotes , the ZM domain is only present in a subset of animal genomes and absent from plants , fungi and unicellular metazoans ( Figure 4B ) . Muscle striation evolved independently in cnidarians and bilaterians , with only the latter showing a canonical Z-disc structure ( Steinmetz et al . , 2012 ) . Intriguingly , the ZM domain is restricted to bilateral animals ( Figure 4B ) . Thus , our correlative data suggest that canonical Z-discs and ZM domains arose together during evolution . Zasp isoforms can be divided into isoforms with multiple LIM domains ( e . g . Zasp52-PR ) and isoforms with just one or no LIM domain ( e . g . Zasp52-PK , Zasp52-PP , Zasp66 , Zasp67 ) . We named the former growing isoforms and the latter blocking isoforms . We then considered what mechanism might coordinate Z-disc growth: we hypothesize that multiple LIM domain-containing Zasp proteins recruit other Zasp proteins by interacting with their ZM domains . If the recruited proteins are also multivalent ( contain two or more LIM domains ) , more Zasp proteins will be recruited leading to Z-disc growth . However , if recruited proteins lack LIM domains , they block further recruitment of Zasp proteins , and Z-disc growth terminates ( Figure 5A ) . To regulate Z-disc growth , growing isoforms should be overrepresented at earlier stages of Z-disc formation , and blocking isoforms at later stages , to stop Z-disc growth . We used an RNAseq dataset from developing IFM that covers the whole Z-disc formation process , from 16 hr after puparium formation ( APF ) to newly eclosed flies ( Spletter et al . , 2018 ) . We observed earlier expression of the growing isoforms compared to the blocking isoforms ( Figure 5B ) . For example , multivalent Zasp52 growing isoforms are already strongly expressed at 24 hr APF , whereas some blocking isoforms without any LIM domains are expressed only after 60 hr APF ( Figure 5B ) . Z-discs grow at the periphery of the disc , starting as a small Z-body ( Orfanos et al . , 2015; Shwartz et al . , 2016 ) . Our model therefore predicts growing isoforms to be enriched at the centre and blocking isoforms to be enriched at the periphery of the final-sized Z-disc . To analyze the distribution of Zasp proteins within Z-discs , we made cross sections of myofibrils and evaluated the localization pattern of blocking isoforms ( using Zasp66-GFP and Zasp67-GFP ) , and growing isoforms ( using Zasp52-GFP; Zasp52ZCL423 ) at the level of the Z-disc . To label the entire disc , we used actin staining as counterstain ( Figure 5C ) . To compensate for the low signal-to-noise ratio , we used a smoothening algorithm that takes advantage of the geometrical properties of discs ( Figure 5—figure supplement 1 ) . Actin is evenly distributed in the Z-disc ( Figure 5D , E ) . The growing isoforms are present throughout the disc , but most concentrated at the centre of the Z-disc ( Figure 5D , E ) . The blocking isoforms Zasp66 and Zasp67 are partially excluded from the centre of the disc and form a ring-like pattern ( Figure 5D , E ) . Zasp66 localizes more at the periphery compared to Zasp67 , which correlates with its later expression peak compared to Zasp67 ( middle panels of Figure 5B ) . We then asked whether decreasing the levels of growing isoforms would result in smaller Z-discs . We measured individual Z-discs in control and Zasp52 mutants ( Figure 6A–C ) . Interestingly , only the Zasp52MI00979 mutant , which deletes multiple LIM domains , had smaller Z-discs compared to the control ( Figure 6B , C ) . In Zasp52MI00979 , the smaller size categories were significantly elevated , whereas Zasp52MI02988 was comparable to the control ( Figure 6C ) . We next determined if the opposite approach , increasing the levels of blocking isoforms , also results in smaller Z-discs . Overexpressing the Zasp52-PP blocking isoform led to smaller Z-discs than the control ( Figure 6D ) . This effect requires the ZM domain , because overexpression of Zasp52-Stop143 containing only a functional extended PDZ domain , causes a milder phenotype than Zasp52-PP ( Figure 6D , Figure 1—figure supplement 1 ) . Like Zasp52-PP , overexpression of Zasp66 or Zasp67 resulted in smaller Z-discs ( Figure 6E ) . We also confirmed these phenotypes by measuring the relative fluorescence of Zasp52-mCherry at the Z-disc . Upon overexpression of Zasp52-PP , but not of Zasp52-Stop143 , Zasp52-mCherry recruitment to the Z-disc is reduced ( Figure 6F , Figure 6—figure supplement 1A ) . Likewise , overexpression of Zasp66 and Zasp67 reduces Zasp52-mCherry recruitment to the Z-disc ( Figure 6G , Figure 6—figure supplement 1A ) . In contrast , α-actinin levels were unchanged in Zasp overexpression conditions ( Figure 6—figure supplement 1B , C ) . These data indicate that increasing the ZM/LIM ratio in IFM results in smaller Z-discs . Finally , we wondered if depleting the blocking isoforms encoded by Zasp66 and Zasp67 can generate bigger Z-discs or aggregates using CRISPR Zasp66 and Zasp67 null mutants ( González-Morales et al . , 2019 ) . We observed rare aggregates in Zasp66 and Zasp67 single mutants ( Figure 7A , B , D ) . As these two proteins likely have some redundant roles , we also analyzed the Zasp66 Zasp67 double mutant , where we observed frequent aggregates ( Figure 7C , D ) . Additionally , the single mutants had normal-sized Z-discs , whereas the double mutant had enlarged Z-discs ( Figure 7E ) . Thus , blocking isoforms are required to prevent Z-disc overgrowth . As many diseases are believed to be caused by the formation of aggregates ( Baba et al . , 1998; Selcen , 2008 ) , we asked if we can suppress aggregate formation in our model of overexpressed GFP-Zasp52-PR ( Figure 1B ) . Co-overexpression of both growing and the Zasp52-PP blocking isoform suppressed aggregate formation compared to a LacZ control ( Figure 7F–I ) . The mutated blocking isoform Zasp52-Stop143 without the ZM domain did not suppress aggregation ( Figure 7H , I ) . Co-overexpression of Zasp66 or Zasp67 also suppressed aggregation ( Figure 7I ) . Our findings indicate that Z-disc formation and growth is driven by multivalent oligomerization of Zasp proteins with multiple LIM domains , and eventually terminated at the proper Z-disc size by the upregulation of blocking isoforms without LIM domains ( see model in Figure 5A ) . The process of Z body and Z-disc formation is reminiscent of membraneless organelles with compositions distinct from the surrounding cytosol , which form through a mechanism of phase separation ( e . g . Cajal bodies or P bodies ) ( Boeynaems et al . , 2018; Weber , 2017 ) . Both have sharp boundaries between themselves and the cytoplasm , they form and organize as discrete puncta in the cytosol , and multivalent protein domains are often involved in their formation ( Li et al . , 2012; Weber , 2017 ) . Sarcomere size is stereotyped in a given muscle type but distinct among different muscles ( Schönbauer et al . , 2011 ) . How can our model explain differences in sarcomere sizes ? The sarcomere grows , while the Zasp growing isoforms dominate the Zasp isoform pool . Different sarcomere sizes can be achieved in two ways . First , the sarcomere growth period – the window of time in which Zasp growing isoforms dominate , might vary among muscle types . Second , the speed at which new Zasp molecules are recruited to the Z-disc , might be different among muscle types , while the growth period remains constant . Given the diversity of muscle types and therefore sarcomere sizes that exist , it is likely that a combination of these two strategies occur simultaneously . Finally , apart from the ZM-LIM mechanism described here , additional redundant mechanisms to control Z-disc growth might exist , as evidenced by the observation that Zasp52-PR overexpression makes big Z-discs and aggregates , while the mutant removing Zasp52 LIM domains reduces Z-disc size to a comparatively small degree . Redundant mechanisms might operate through other LIM domain proteins , or the coordination of Z-disc and M-line growth , all of which may provide important buffering functions to ensure proper myofibril size , which is crucial for fully functional muscles . The ZM domain is a conserved domain without a clearly defined function . On its own , the ZM domain from two mouse Zasp proteins localizes to the Z-disc ( Klaavuniemi et al . , 2004; Klaavuniemi and Ylänne , 2006 ) . Our data suggest that Z-disc localization is a conserved feature of ZM domains from vertebrates to insects . ZM-containing proteins are tethered to the Z-disc by the physical interaction with the LIM domains of other Zasp proteins . In sum , the LIM domain serves as a recruitment signal for Zasp proteins and potentially other unidentified ZM-containing proteins to join the Z-disc . In addition , given the appearance of the ZM domain in bilateral animals with canonical Z-discs , we postulate that a conserved mechanism involving LIM-ZM binding underlies Z-disc growth and growth termination . In vertebrates , the Zasp proteins are very diverse and are better known as Alp/Enigma family: ZASP/Cypher/Oracle/LDB3/PDLIM6 , ENH/PDLIM5 , PDLIM7/ENIGMA/LMP-1 , CLP36/PDLIM1/Elfin/hCLIM1 , PDLIM2/Mystique/SLIM , ALP/PDLIM3 , and RIL/PDLIM4 . The ZM/DUF4749 motif occurs in ZASP , CLP36 , PDLIM2 , ALP and RIL ( Cheng et al . , 2010; D'Cruz et al . , 2016; Faulkner et al . , 1999; Vallenius et al . , 2004; Zheng et al . , 2010; Zhou et al . , 2001 ) . The LIM domain occurs in all Zasp proteins , either as one domain in Alp family members or as three domains in Enigma family members ( Zheng et al . , 2010 ) . We identified two Zasp genes that encode only blocking isoforms in fruit flies: Zasp66 and Zasp67 , and one gene , Zasp52 , that encodes blocking and growing isoforms . Although Zasp66 and Zasp67 genes are insect-specific ( González-Morales et al . , 2019 ) , vertebrate Alp/Enigma genes also express isoforms without LIM domains that could fulfill a blocking isoform function ( Cheng et al . , 2011; Zheng et al . , 2010 ) . In addition , because Zasp52-PK , which only contains one LIM domain , behaves as a blocking isoform , the Alp members with only one LIM domain might also behave as blocking isoforms . The function of the growing isoforms of Zasp requires multiple functional LIM domains . As the Enigma family members contain three C-terminal LIM domains , they are the ideal candidates to fulfill the growing role in vertebrates . Three Enigma proteins exist in vertebrates: PDLIM7/Enigma/LMP-1 , ENH/PDLIM5 and ZASP/Cypher/Oracle/LDB3/PDLIM6 . Functional redundancy between them at the Z-disc is likely common and demonstrated in one case ( Mu et al . , 2015 ) . In Cypher knockout mice sarcomere assembly occurs normally during development , followed by immediate sarcomere failure after postnatal onset of contractility ( Zhou et al . , 2001 ) . ENH mutants exhibit cardiac dilation and abnormal Z-disc structure in the heart ( Cheng et al . , 2010 ) . Intriguingly , in both Cypher and ENH single mutants , as well as Cypher ENH double mutants , sarcomeres look considerably smaller in diameter in electron microscopy images ( Cheng et al . , 2010; Mu et al . , 2015; Zhou et al . , 2001 ) . Thus , a similar role for Enigma proteins in setting sarcomere diameter in vertebrates appears likely . Is the mechanism that controls Z-disc size related to the protein aggregation defects in human myopathies ? Our Z-disc oligomerization hypothesis agrees well with the observation that many myopathies present aggregates , and several human ZASP mutations have been linked to aggregate-forming myopathies ( Murphy and Young , 2015; Selcen and Engel , 2005 ) . Many ZASP mutations linked to disease lie within the ZM domain or one of the LIM domains ( Selcen and Engel , 2005; Theis et al . , 2006; Vatta et al . , 2003 ) . Protein aggregation in myopathy patients might be a consequence of an imbalance in the mechanism that controls sarcomere size , favoring the growing over the blocking isoforms . If this were the case , our data points to a potential therapeutic avenue: blocking the growing isoforms with short peptides containing a ZM domain . In conclusion , we propose that a conserved mechanism involving LIM-ZM binding underlies Z-disc growth and therefore myofibril diameter . We used Drosophila melanogaster as a model organism for most experiments . Fly stocks and crosses were raised at 25°C . A comprehensive list of all strains used and generated can be found in the Key Resources Table . We used Saccharomyces cerevisiae for the yeast two-hybrid assays . Escherichia coli BL-21 strain was used to express recombinant proteins . Transgenic flies were generated by site-directed integration using the PhiC31 integrase method into either M[3xP3-RFP . attP]ZH-58A- or M[3xP3-RFP . attP]ZH-86Fb-bearing flies to ensure comparable expression levels ( Bischof et al . , 2007; Markstein et al . , 2008 ) . Unless stated otherwise , Act88F-Gal4 was used to drive strong transgene expression in the IFM ( Bryantsev et al . , 2012 ) . Partial or complete coding sequences for Zasp52 , Zasp66 and Zasp67 were cloned into Y2H vectors either by PCR-ligase cloning or through Gateway cloning . Constructs and cloning details are listed in the Key Resources Table . All constructs were verified by sequencing . Selected constructs were transformed into the Matchmaker Y2H Gold strain using the lithium acetate method ( Gietz and Schiestl , 2007 ) . Double transformant colonies were selected and amplified in media lacking leucine and tryptophan ( -Leu/-Trp ) . Serial dilutions of the selected double transformants were grown in plates lacking leucine , tryptophan , histidine and adenine ( -Ade/-His/- Leu/-Trp ) to test for protein-protein interactions . The experiments were done at least three times . Recombinant 6xHis-Zasp52-PK-FLAG was expressed in Escherichia coli BL21 bacteria . Then , 6xHis-Zasp52-PK-FLAG was purified from the protein extracts using Ni-NTA agarose beads ( Qiagen ) for 3 hr at 4°C . The purified protein was then dialyzed overnight at 4°C . Finally , purified 6xHis-Zasp52-PK-FLAG was either incubated with ethylene glycol bis-sulfosuccinimidyl succinate ( EGS ) or alone . Then , the protein samples were analyzed by denaturing SDS-PAGE followed by western blotting with anti-Flag antibody ( 1:5000 ) . The IFM were dissected as previously described ( González-Morales et al . , 2017; Xiao et al . , 2017 ) . All images were acquired with comparable parameters . In agreement with previous studies , at least 10 thoraces were dissected for each condition . Samples were allocated into experimental groups according to their genotype . A big number of flies were collected for each experimental group and a subsample was randomly selected for dissection . Zasp52-PK , Zasp52-PE and Zasp66-PH Gateway pENTRY clones were obtained from the DGRC ( GEO02280 , GEO12859 and GEO14752 ) . The BIFC pDEST vectors pUAS-RfB-HA-CYFP-attB and pUAS-RfB-myc-NYFP-attB vectors were a kind gift from Sven Bogdan ( Gohl et al . , 2010 ) . Constructs were cloned using Gateway technology , and the final vectors were verified by sequencing . The Key Resources Table contains a list of transgenic flies used in this study . The Act88F-Gal4 driver line was used to express CYFP- and NYFP-tagged proteins in the IFM . Then , the YFP fluorescence of individual Z-discs was quantified using the ImageJ plot profile tool ( Schindelin et al . , 2012 ) . At least 20 samples were used for each condition . The data was normalized to the basal noise levels and plotted in R software . To estimate the size of individual Z-discs and the fluorescence intensity levels of Z-disc proteins we first used a segmentation method that allows individual Z-discs to be measured ( Xiao et al . , 2017 ) . Briefly , images from fluorescent Z-disc proteins were passed through a threshold filter , and then individual Z-discs were separated and measured using the analyze particles tool . Both tools can be found in any recent ImageJ/Fiji distribution ( Schindelin et al . , 2012 ) . To estimate the Z-disc size , we measured the Z-disc height , that corresponds to the disc diameter . Then , the data was analyzed and plotted in R software . To estimate the relative fluorescence intensities of Z-disc proteins , we first obtained the mean intensity raw values for each Z-disc . Then , we normalized the values to a control genotype . The resulting data was then analyzed in R software . Rare segmentation mistakes result from Z-discs that are in close proximity . We could filter these mistakes out by size exclusion in the control images , but not in mutant conditions . We decided to leave the segmentation mistakes for consistency . If anything , this method underestimates the size differences between control and mutants . Batch Macro processing code for measuring individual Z-disc side-views in ImageJsetAutoThreshold ( "Default dark" ) ; run ( "Analyze Particles . . . " , "size = 0 . 2 Infinity show = Outlines display exclude" ) ; To estimate the density of aggregates , present in a given IFM sample , we used confocal images of GFP- or mCherry-tagged Zasp . Each 36 × 36 µm image was divided into 256 2 × 2 µm images . An automatic threshold filter based on the original image was then used on the small images , and the area above the given threshold was measured . Then , we used R software to count the number of images with aggregates . Batch Macro processing code for aggregation estimates in ImageJrun ( "Montage to Stack . . . " , "images_per_row = 16 images_per_column = 16 border = 0" ) ; macro "Measure Stack" { saveSettings; setOption ( "Stack position" , true ) ; for ( n = 1; n <= nSlices; n++ ) { setSlice ( n ) ; setAutoThreshold ( "Default dark stack" ) ; //run ( "Threshold . . . " ) ; run ( "Create Selection" ) ;; run ( "Measure" ) ; } restoreSettings; } Thoraces of Zasp52-GFP ( Zasp52ZCL423 ) , Zasp66-GFP ( Zasp66ZCL0663 ) , and Zasp67-GFP ( Zasp67 fTRG ) flies were fixed in paraformaldehyde and embedded into low-melting agarose blocks . Then , we sectioned the thoraces into 100-µm-thick sections using a standard vibratome . The sections were stained with TRITC-phalloidin and imaged using a LEICA SP8 confocal microscope and a 63x/1 . 4 oil objective . Images were then processed to isolate individual properly oriented Z-discs . To avoid artifacts from misoriented Z-discs , we selected only those in which the GFP signal coming from the Z-disc would cover the whole sarcomere area stained with phalloidin . To remove the background noise from cross-section Z-disc images , we took advantage of the geometrical properties of the Z-disc . We drew a selection circle around individual Z-discs and rotated the image by 2 degrees 180 times . Then , we calculated the average from all the rotated images . We compared the Z-disc size categorical data using Fisher's exact test for count data , p-values above 0 . 001 were noted as not significant . We used Welch’s two-sample t-test to compare the relative fluorescence intensities between groups . We used Welch’s two-sample t-test to compare the aggregation frequency estimates . All species that contain at least one protein with either the ZM ( PF15936 ) or the LIM ( PF00412 ) domain were downloaded from the PFAM web ( Bateman et al . , 2004; El-Gebali et al . , 2019 ) . A general phylogenetic tree was plotted based on the NCBI taxonomy data ( Federhen , 2012 ) using the ape:plot . phyllo tool ( Paradis , 2012 ) . We uploaded all individual RNA-seq SRA reads from the project: PRJNA419412 ( GEO: GSE107247 ) ( Spletter et al . , 2018 ) . Each SRA read corresponds to a different time point in IFM development . We extracted the single-end reads using the fastq-dump tool . We then used Salmon v0 . 8 . 2 ( Patro et al . , 2017 ) to quantify the expression of all transcripts as Relative Transcripts Per Million ( TPM ) values . We used all the cDNAs from the D . melanogaster r6 reference transcriptome in fasta format ( Adams et al . , 2000 ) . Then , the individual transcripts corresponding to the three Zasp genes were grouped according to their encoded protein domain architectures . Finally , we calculated the mean TPM values between similar isoforms and plotted the results over developmental time using R software . The accession number for all reads used are: SRR1665023 , SRR1665024 , SRR1665025 , SRR1665026 , SRR1665027 , SRR1665028 , SRR6314253 , SRR6314254 , SRR6314255 , SRR6314256 , SRR6314257 , SRR6314258 , SRR6314259 , SRR6314260 , SRR6314261 , SRR6314262 , SRR6314263 , SRR6314273 , SRR6314274 , SRR6314275 , SRR6314276 , SRR6314277 , SRR6314278 .	Muscles are made up of many long muscle fibers , each containing thousands of cylindrical segments called sarcomeres . When animals move , proteins in the sarcomere move past each other , shortening the muscles . Inside each muscle , all sarcomeres have the same length and diameter . The protein titin controls the length of each sarcomere , but it was unknown what controls the diameter . At the end of each sarcomere is a structure called the Z-disc that is composed of many muscle proteins . Mutations in Z-disc proteins are often involved in diseases called myopathies , where muscle structure breaks down . As the size of the Z-disc determines sarcomere diameter , improper regulation of sarcomere diameter could contribute to myopathies . One Z-disc protein called Zasp is a candidate for controlling diameter and can have many different forms in the same cells . Zasp has a similar role in most animals including humans , mice and flies . González-Morales et al . investigated Zasp in the muscles of the fruit fly , Drosophila melanogaster . Gene editing was used to vary the amounts of different forms of Zasp inside the muscles . The results revealed two types of Zasp , those that make sarcomeres wider , and those that limit growth . Reducing the second type of Zasp resulted in bigger Z-discs and in muscle aggregates similar to the ones seen in patients with certain myopathies . This study reveals a mechanism for coordinating the development of muscle . It also reveals the likely cause of certain myopathies and suggests a possible target for future treatment through regulation of Zasp proteins .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "cell", "biology" ]	2019	Myofibril diameter is set by a finely tuned mechanism of protein oligomerization in Drosophila
Posttranslational modifications ( PTMs ) play a crucial role in a wide range of biological processes . Lysine crotonylation ( Kcr ) is a newly discovered histone PTM that is enriched at active gene promoters and potential enhancers in mammalian cell genomes . However , the cellular enzymes that regulate the addition and removal of Kcr are unknown , which has hindered further investigation of its cellular functions . Here we used a chemical proteomics approach to comprehensively profile ‘eraser’ enzymes that recognize a lysine-4 crotonylated histone H3 ( H3K4Cr ) mark . We found that Sirt1 , Sirt2 , and Sirt3 can catalyze the hydrolysis of lysine crotonylated histone peptides and proteins . More importantly , Sirt3 functions as a decrotonylase to regulate histone Kcr dynamics and gene transcription in living cells . This discovery not only opens opportunities for examining the physiological significance of histone Kcr , but also helps to unravel the unknown cellular mechanisms controlled by Sirt3 , that have previously been considered solely as a deacetylase . Histone posttranslational modifications ( PTMs ) play a crucial role in regulating a wide range of biological processes , such as gene transcription , DNA replication , and chromosome segregation ( Kouzarides , 2007 ) . Increasing evidence has indicated that PTMs of histones can serve as a heritable ‘code’ ( so-called ‘histone code’ ) , which provides epigenetic information that a mother cell can pass to its daughters ( Jenuwein and Allis , 2001 ) . Histone code is ‘written’ or ‘erased’ by enzymes that add or remove the modifications of histones ( Goldberg et al . , 2007; Kouzarides , 2007 ) . Meanwhile , ‘readers’ of histone code recognize specific histone modifications and ‘translate’ the code by executing distinct cellular programs necessary to establish diverse cell phenotypes , while the genetic code ( DNA ) is unaltered ( Seet et al . , 2006; Taverna et al . , 2007 ) . Lysine acetylation ( Kac ) was among the first covalent modification of histones to be described ( Allfrey and Mirsky , 1964; Allfrey et al . , 1964 ) . Since its identification , histone Kac has been correlated with gene expression . However , the mechanistic insights into the regulation and functions of histone Kac remained challenging and elusive , until the identification and characterization of the enzymes responsible for the addition and removal of this PTM , which are now known as histone acetyltransferases ( Roth et al . , 2001 ) and deacetylases ( Sauve et al . , 2006; Yang and Seto , 2008b; Haberland et al . , 2009 ) , respectively . Extensive studies have now revealed that Kac plays an important role in controlling chromatin structure and gene transcription ( Grunstein , 1997; Yang and Seto , 2008a ) . By neutralizing positively charged lysine residues , acetylation alters the coulumbic interactions between basic histones and the negatively charged DNA , and thereby influences the structure of chromatin compaction ( Ura et al . , 1997; Shogren-Knaak et al . , 2006 ) . In addition , acetylation may serve as a docking site for ‘reader’ proteins ( e . g . , bromodomain containing proteins ) , which are recruited onto chromatin to carry out downstream cellular processes , such as gene transcription ( Dhalluin et al . , 1999; Marmorstein and Berger , 2001; Zeng et al . , 2010 ) . Lysine crotonylation ( Kcr ) is a newly discovered histone PTM that is specifically enriched at active gene promoters and potential enhancers in mammalian cell genomes ( Tan et al . , 2011 ) . In postmeiotic male germ cells , Kcr specifically marks testis specific X-linked genes , suggesting it is likely that it is an important histone mark for male germ cell differentiation . However , further mechanistic and functional studies of histone Kcr have been limited by a lack of knowledge of the enzymes that catalyze the addition or removal of Kcr in cells . In a systematic screening of the activities of the 11 human zinc-dependent lysine deacetylases ( i . e . , HDAC1–HDAC11 ) against a series of C-terminal lysine acylated peptides , Olsen et al . found that HDAC3 in complex with nuclear receptor corepressor 1 ( HDAC3–NCoR1 ) had detectable decrotonylase activity towards a model peptide substrate in a fluorometric assay ( Madsen and Olsen , 2012 ) . Recently , using a radioactive thin layer chromatography based assay , Denu et al . demonstrated that Sirt1 and Sirt2 can catalyze the removal of a crotonyl group from a histone H3K9Cr peptide ( Feldman et al . , 2013 ) . However , this discovery was based on a single peptide substrate . Due to lack of further characterization of these identified enzymes , their mechanisms of catalysis and the molecular bases of substrate recognition remain unclear . More importantly , since both discoveries relied on peptide based in vitro screening assays , there is still an essential need to identify endogenous histone decrotonylases . To fill this knowledge gap , a method to profile ‘eraser’ enzymes that recognize Kcr is needed . A Cross-Linking Assisted and Stable isotope labeling of amino acids in cell culture ( SILAC ) based Protein Identification ( CLASPI ) approach has recently been reported to identify histone PTM ‘readers’ ( Li et al . , 2012; Li and Kapoor , 2010 ) . However , this approach has not previously been explored to identify histone PTM ‘erasers’ , which are likely involved in weak and transient interactions . Here we present the application of an optimized CLASPI approach to comprehensively profile ‘eraser’ enzymes that recognize histone Kcr marks . We identified human Sirt1 , Sirt2 , and Sirt3 as decrotonylases in vitro and examined the molecular basis for how the enzymes recognize Kcr using X-ray crystallography . Furthermore , we demonstrated that Sirt3 can function as an ‘eraser’ enzyme to regulate histone crotonylation dynamics in living cells . We first focused on a crotonylation mark discovered on histone H3K4 ( Tan et al . , 2011 ) . We designed a peptide probe ( probe 1 , Figure 1A ) to convert non-covalent protein–protein interactions mediated by this Kcr into irreversible covalent linkages through photo-cross-linking . The probe is based on the unstructured N-terminal region of histone H3 , with lysine-4 crotonylated , a photo-cross-linker ( benzophenone ) appended to alanine-7 , and a bio-orthogonal handle ( alkyne ) at the peptide C terminus to enable selective isolation of captured binding partners . To identify proteins that bind H3K4Cr with high selectivity and high affinity , we performed two types of CLASPI experiments with cell lysates derived from HeLa S3 cells grown in medium containing either ‘heavy’ ( 13C , 15N-substitued arginine and lysine ) or ‘light’ ( natural isotope abundance forms ) amino acids ( Figure 1B ) . 10 . 7554/eLife . 02999 . 003Figure 1 . Cross-Linking Assisted and SILAC based Protein Identification ( CLASPI ) strategy . ( A ) Chemical structures of probe 1 and probe C . ( B ) Schematic diagram illustrating the CLASPI strategy to profile proteins that bind H3K4Cr with high selectivity and affinity in whole-cell proteomes . LC-MS , liquid chromatography–mass spectrometry . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 003 In a ‘selectivity filter’ CLASPI experiment , the ‘heavy’ and ‘light’ cell lysates were photo-cross-linked with probe 1 and an unmodified H3 control probe ( probe C , Figure 1A ) , respectively , and pooled for the subsequent steps . The captured proteins were then conjugated to biotin using click chemistry , followed by affinity purification , gel electrophoresis , and in-gel trypsin digestion . The digested peptide mixtures were separated by high performance liquid chromatography ( HPLC ) and analyzed with a LTQ-Orbitrap mass spectrometer . Using this method , proteins that show a high SILAC ratio of heavy/light ( H/L ) are likely H3K4Cr selective binders . To further distinguish the high affinity interactions , we performed an ‘affinity filter’ CLASPI experiment , in which both lysates were photo-cross-linked with probe 1 but the ‘light’ sample also contained H3K4Cr peptide as a competitor ( 30 μM ) ( Figure 1B ) . We expected that the addition of the competitor peptide in the ‘light’ lysate would effectively inhibit 1-induced cross-linking of H3K4Cr binders that have high affinity ( Kd < 30 μM ) towards the H3K4Cr peptide , and should thereby produce a high SILAC ratio of H/L for these proteins . Together , we consider a protein as a selective and tight binder of H3K4Cr when it shows high SILAC ratios of H/L in both ‘selectivity filter’ and ‘affinity filter’ experiments ( Figure 2—source data 1 ) . A two-dimensional plot with logarithmic ( Log2 ) SILAC ratios of H/L of the identified proteins in the ‘selectivity filter’ and ‘affinity filter’ experiments , along the x axis and y axis , respectively , is shown in Figure 2A . As expected , the majority of identified proteins did not show significant differences between the signal intensities of their ‘heavy’ and ‘light’ forms ( i . e . , H/L close to 1:1 ) , suggesting they are not likely to be H3K4Cr binding proteins . In contrast , three nicotinamide adenine dinucleotide ( NAD ) -dependent deacetylases ( Imai et al . , 2000; Landry et al . , 2000; Sauve et al . , 2006 ) , Sirt1 , Sirt2 , and Sirt3 , were enriched by more than 10-fold by the K4 crotonylated probe ( 1 ) in the ‘selectivity filter’ experiment ( Figure 2A , B and Figure 2—figure supplement 1 ) , indicating that they preferentially bind to this histone Kcr mark . However , among these three selective H3K4Cr binders , only Sirt3 showed the highest SILAC ratio of H/L and thereby appeared as an outlier outside of the background in the ‘affinity filter’ experiment ( Figure 2A , B and Figure 2—figure supplement 1 ) . This result indicates that Sirt3 is likely a selective and relatively tight binding partner of H3K4Cr . 10 . 7554/eLife . 02999 . 004Figure 2 . Identification of Sirt3 as a selective and tight binding partner of lysine-4 crotonylated histone H3 . ( A ) A two-dimensional plot showing the Log2 values of the stable isotope labeling of amino acids in cell culture ( SILAC ) ratios ( heavy/light ( H/L ) ) of each identified protein for the ‘selectivity filter’ ( x axis ) and ‘affinity filter’ ( y axis ) experiments . ( B ) Representative mass spectrometry ( MS ) spectra of a peptide , 225LYTQNIDGLER235 , from Sirt3 identified in both the ‘selectivity filter’ and ‘affinity filter’ experiments . The ‘light’ and ‘heavy’ peptide isotopes are indicated by blue and red dots , respectively . ( C ) Recombinant Sirt3 was selectively labeled in vitro by crotonylated probe 1 ( 2 μM ) and the labeling by probe 1 was inhibited by a H3K4Cr peptide ( 30 μM ) . ( D ) Determination of IC50 for inhibition of probe 1 induced labeling of Sirt3 by H3K4Cr peptide ( n=3 , mean±s . e . ) . ( E ) Isothermal titration calorimetry measurement for the binding affinity of Sirt3 towards the H3K4Cr peptide . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 00410 . 7554/eLife . 02999 . 005Figure 2—source data 1 . Proteins quantified in the ‘selectivity filter’ and ‘affinity filter’ Cross Linking Assisted and SILAC based Protein Identification ( CLASPI ) experiments . Proteins are sorted according to their ratio in the ‘selectivity filter’ experiment . Known contaminants such as keratin are not included . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 00510 . 7554/eLife . 02999 . 006Figure 2—figure supplement 1 . Representative mass spectrometry spectra for peptides from Sirt1 and Sirt2 . Representative mass spectrometry spectra for peptides from Sirt1 ( A ) and Sirt2 ( B ) , identified in both ‘selectivity filter’ and ‘affinity filter’ Cross Linking Assisted and SILAC based Protein Identification ( CLASPI ) experiments . The ‘light’ and ‘heavy’ peptide isotopes are indicated by blue and red dots , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 00610 . 7554/eLife . 02999 . 007Figure 2—figure supplement 2 . ITC measurement for the binding affinity of Sirt1-3 toward H3K4Cr peptide . ( A–C ) Isothermal titration calorimetry measurement for the binding affinities of Sirt1 ( A ) , Sirt2 ( B ) , and Sirt3 ( C ) towards H3K4Cr . ( D ) A summary of the dissociation constants ( Kd ) , enthalpy changes ( ΔH ) , and andentropy changes ( ΔS ) of Sirt1 , Sirt2 , and Sirt3 binding to H3K4Cr . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 007 We next examined whether Sirt3 can directly and selectively bind to this crotonylated histone peptide in vitro . As shown in Figure 2C , the recombinant Sirt3 was captured by probe 1 but not by probe C , and the cross-linking was competed by the H3K4Cr peptide with an IC50=32 . 3 μM ( Figure 2D ) , verifying a direct and selective interaction between Sirt3 and the K4 crotonylated H3 peptide . Indeed , the direct measurement of binding affinity using isothermal titration calorimetry showed that Sirt3 bound to the H3K4Cr peptide with Kd=25 . 1 μM ( Figure 2E ) . Consistent with our ‘affinity filter’ CLASPI analysis , Sirt1 and 2 showed lower affinities towards the H3K4Cr peptide ( Figure 2—figure supplement 2 ) , indicating that they are selective but relatively weak binders towards this histone Kcr mark . To study the molecular basis for the recognition of H3K4Cr by Sirt3 , we determined the crystal structure of human Sirt3 in complex with an H3K4Cr peptide to 2 . 95 Å resolution ( PDB 4V1C ) . The asymmetric unit consists of six molecules , each containing one Sirt3–H3K4Cr complex . The two globular domains of Sirt3 composed of an NAD binding Rossmann fold and a zinc binding motif are similar to other sirtuins ( Figure 3A ) ( Avalos et al . , 2004; Du et al . , 2011; Yuan and Marmorstein , 2012; Jiang et al . , 2013 ) . Residues 2RTKQTAR8 of the H3K4Cr peptide were clearly identified based on electron density . The way that the substrate is bound is similar to the published complex structure of Sirt3 , with a lysine acetylated AceCS2 peptide ( PDB 3GLR ) ( Figure 3—figure supplement 1 ) ( Jin et al . , 2009 ) . The crotonyl lysine is located in a binding pocket formed by hydrophobic residues Phe180 , Ile230 , His248 , Ile291 , and Phe294 of Sirt3 ( Figure 3B ) . Residue His248 , a catalytic residue for the deacetylation activity of Sirt3 , interacts with the crotonyl amide oxygen via hydrogen bonding in the structure ( Figure 3B ) . Strikingly , the phenyl ring of residue Phe180 aligns parallel to the planar crotonyl group and has a short distance of 3 . 6 Å to its conjugated carbon–carbon double bond ( C=C ) ( Figure 3C , D ) , indicating a robust π-π stacking interaction between the two functional groups . Interestingly , a primary sequence alignment of all sirtuins revealed that the phenylalanine residue ( Phe180 ) of Sirt3 is conserved in Sirt1 and Sirt2 , but not in other sirtuins ( Figure 3—figure supplement 2 ) , which may explain why Sirt4–Sirt7 were not identified in our CLASPI experiments . This π-π interaction therefore underlies the mechanism for the recognition of crotonyl lysine by Sirt1 , Sirt2 , and Sirt3 . 10 . 7554/eLife . 02999 . 008Figure 3 . Structural basis for how Sirt3 recognizes lysine crotonylation . ( A ) Overall structure of the complex of Sirt3 ( gray ) with H3K4Cr peptide ( green ) . ( B ) The binding pocket formed by hydrophobic residues ( orange ) that accommodate the crotonyl lysine . A side view ( C ) and top view ( D ) of a π-π stacking interaction between residue Phe180 of Sirt3 and the crotonyl group . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 00810 . 7554/eLife . 02999 . 009Figure 3—figure supplement 1 . Detailed structural analysis for Sirt3 in complex with H3K4Cr peptide . ( A ) Structural alignment between the Sirt3–AceCS2 complex colored in cyan ( PDB 3GLR ) and the Sirt3–H3K4Cr complex colored in orange . ( B ) Interactions between Sirt3 and H3K4Cr peptide . The residues of Sirt3 are colored in yellow and H3K4Cr peptide is colored in green . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 00910 . 7554/eLife . 02999 . 010Figure 3—figure supplement 2 . Sequence alignment of human Sirt1–Sirt7 . The included sequences are Sirt1-197-542 ( full length , 747 amino acids ( a . a . ) ) , Sirt2-54-370 ( 389 a . a . ) , Sirt3-107-391 ( 399 a . a . ) , Sirt4-31-312 ( 314 a . a . ) , Sirt5-20-305 ( 310 a . a . ) , Sirt6-20-305 ( 355 a . a . ) , and Sirt7-71-347 ( 400 a . a . ) . The residues of Sirt3 that interact with crotonyl lysine in the crystal structure of the Sirt3–H3K4Cr complex are indicated ( ▼ ) above the residues and labeled with the Sirt3 sequence number . Highlighted in green is the Phe180 of Sirt3 , which is involved in recognition of the crotonyl lysine via a π−π stacking interaction , and is conserved in Sirt1 and Sirt2 , but not in the other sirtuins . The alignment was done by T-coffee ( Notredame C , Higgins DG , and Heringa , J . 2000 , Journal of Molecular Biology 302 , 205–17 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 010 Inspired by the fact that Sirt3 binds crotonyl lysine at its catalytic pocket that is known for hydrolysis of acetyl lysine , we next tested whether Sirt3 has decrotonylation activity . Liquid chromatography–mass spectrometry ( LC-MS ) was used to monitor hydrolysis of the H3K4Cr peptide by Sirt3 . As expected , Sirt3 efficiently catalyzed the hydrolysis of the crotonyl peptide only in the presence of NAD ( Figure 4A ) , suggesting an NAD-dependent decrotonylation mechanism . The steady state kinetic analysis revealed that the kcat , Km , and kcat/Km for Sirt3 catalyzed decrotonylation of H3K4Cr were 0 . 010 s−1 , 12 . 6 μM , and 783 s−1 M−1 , respectively ( Figure 4—figure supplement 1 ) . In addition , we detected O-crotonyl-adenosine 5ʹ-diphosphoribose ( O-Cr-ADPR ) as a product of this hydrolysis reaction ( Figure 4—figure supplement 2 ) . A mutation of the catalytic residue ( H248Y ) that is crucial for the deacetylation activity of the enzyme also completely abolished its decrotonylation activity ( Figure 4C ) . These data indicate that Sirt3 hydrolyzes crotonyl lysine with the same mechanism as it hydrolyzes acetyl lysine ( Figure 4—figure supplement 3 ) ( Tanner et al . , 2000; Tanny and Moazed , 2001 ) . In addition to H3K4Cr , we also examined the activity of Sirt3 to hydrolyze a collection of crotonyl histone peptides ( Tan et al . , 2011 ) . As shown in Figure 4D–G , Sirt3 manifested varied decrotonylation activities towards these peptides and this substrate selectivity can be partially explained by the binding affinities of Sirt3 to these peptides ( Figure 4—figure supplement 4 ) . The observation that Sirt3 binds a crotonylated peptide by recognizing both the modification site and its surrounding residues was also supported by the extensive hydrophobic and hydrogen bonding interactions between Sirt3 and the peptide side chains in the Sirt3–H3K4Cr complex structure ( Figure 3—figure supplement 1B ) . 10 . 7554/eLife . 02999 . 011Figure 4 . Sirt3 catalyzes the hydrolysis of crotonyl lysine in vitro . ( A–C ) The hydrolysis of the crotonylated peptides by Sirt3 was analyzed by liquid chromatography–mass spectrometry . The hydrolysis of H3K4Cr was observed with Sirt3 in the presence ( A ) , but not absence of nicotinamide adenine dinucleotide ( NAD ) ( B ) , or with the mutated Sirt3 , H248Y ( C ) . ( D–G ) Sirt3 showed varied decrotonylation activities towards H2BK5Cr ( D ) , H3K9Cr ( E ) , H3K27Cr ( F ) , and H4K8Cr ( G ) peptides . Black traces show total ion intensity for all ion species with m/z from 300 to 2000 ( i . e . , total ion counts , TIC ) ; pink traces show ion intensity ( 5× magnified ) for the masses of decrotonylated ( unmodified ) peptides; and blue traces show ion intensity ( 5× magnified ) for the masses of crotonylated peptides . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01110 . 7554/eLife . 02999 . 012Figure 4—figure supplement 1 . Michaelis–Menten plots showing the kinetics of Sirt3 and mutant Sirt3 ( F180L ) decrotonylation on H3K4Cr . The enzyme concentrations and reaction times were used as indicated in ‘Materials and methods’ . For kinetic parameters , values are reported as mean ± s . e . ( n=3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01210 . 7554/eLife . 02999 . 013Figure 4—figure supplement 2 . Detection of O-crotonyl-adenosine 5ʹ-diphosphoribose ( O-Cr-ADPR ) by liquid chromatography–mass spectrometry ( LC-MS ) . The reaction mixture of Sirt3 catalyzed nicotinamide adenine dinucleotide ( NAD ) dependent decrotonylation of H3K4Cr peptide was analyzed by LC-MS . Partial chromatogram of the reaction mixture with UV ( 260 nm ) ( A ) and mass detector ( B ) . In ( B ) , the black trace shows total ion intensity for all ion species with m/z from 300 to 2000 ( i . e . , total ion counts , TIC ) ; blue trace shows ion intensity ( 5× magnified ) for the mass of NAD ( m/z = 664 ) ; and pink trace shows ion intensity ( 5× magnified ) for the mass of O-Cr-ADPR ( m/z = 628 ) . ( C ) ESI-MS spectra of NAD and O-Cr-ADPR . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01310 . 7554/eLife . 02999 . 014Figure 4—figure supplement 3 . Proposed mechanism of Sirt3 catalyzed nicotinamide adenine dinucleotide dependent decrotonylation . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01410 . 7554/eLife . 02999 . 015Figure 4—figure supplement 4 . ITC measurement for the binding affinity of Sirt3 toward crotonylated histone peptides . ( A ) Isothermal titration calorimetry measurement for the binding affinities of Sirt3 towards H3K4Cr , H3K9Cr , H3K27Cr , and H4K8Cr . ( B ) A summary of dissociation constants ( Kd ) , enthalpy changes ( ΔH ) , and entropy changes ( ΔS ) of Sirt3 for the crotonylated peptides . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01510 . 7554/eLife . 02999 . 016Figure 4—figure supplement 5 . Decrotonylation activity of sirtuins in vitro . The hydrolysis of the H3K4Cr peptide by sirtuins was analyzed by liquid chromatography–mass spectrometry . ( A–C ) Sirt1 , Sirt2 , and Sirt3 showed significant decrotonylation activities towards H3K4Cr . ( D , E ) Sirt5 and Sirt6 showed little decrotonylation activities towards H3K4Cr . Black traces show total ion intensity for all ion species with m/z from 300 to 2000 ( i . e . , total ion counts , TIC ) ; pink traces show ion intensity ( 5× magnified ) for masses of decrotonylated ( unmodified ) peptides; and blue traces show ion intensity ( 5× magnified ) for masses of crotonylated peptides . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01610 . 7554/eLife . 02999 . 017Figure 4—figure supplement 6 . Michaelis–Menten plots showing the kinetics of Sirt3 and mutant Sirt3 ( F180L ) deacetylation on H3K4Ac . The enzyme concentrations and reaction times were used as indicated in ‘Materials and methods’ . For kinetic parameters , values are reported as mean ± s . e . ( n=3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 017 We next investigated whether other members of the sirtuin family could also function as decrotonylases . Consistent with the work of Denu and coworkers , Sirt1 and Sirt2 also catalyzed the hydrolysis of the H3K4Cr peptide , although they were relatively weaker binders towards this substrate ( Figure 2—figure supplement 2 ) . In contrast , for Sirt5 or Sirt6 , little hydrolysis of the crotonyl peptide was observed ( Figure 4—figure supplement 5 ) . These results agree well with the observation that the phenylalanine residue , which is involved in recognition of crotonyl lysine via π−π stacking interaction ( Figure 3C , D ) , is only conserved in Sirt1–Sirt3 ( Figure 3—figure supplement 2 ) . To further examine the importance of this conserved phenylalanine to the decrotonylase activity of the enzyme , we mutated Phe180 of Sirt3 to a leucine residue ( F180L ) , which lacks an aromatic ring as a π donor but retains a similar hydrophobicity . We then carried out kinetic studies on this F180L mutant Sirt3 . The steady state kinetic data showed that the catalytic efficiency of Sirt3 F180L mutant ( kcat/Km=21 s−1 M−1 ) for the hydrolysis of the H3K4Cr peptide was about 40-fold lower than that of wild-type Sirt3 ( Figure 4—figure supplement 1 ) , indicating a critical role of the phenylalanine mediated π−π interaction in the decrotonylation activity of the enzyme . Interestingly , the F180L mutation caused only about a two-fold decrease in the deacetylation activity of the enzyme ( Figure 4—figure supplement 6 ) . This result rules out the possibility that the observed significant decrease in the decrotonylase activity of the enzyme is caused by a potential disruption of the NAD binding pocket in the mutated Sirt3 . To test whether Sirt1–Sirt3 decrotonylate proteins , we incubated whole-cell proteins that were resolved in a sodium dodecyl sulfate–polyacrylamide gel electrophoresis ( SDS-PAGE ) gel and transferred onto a poly ( vinylidene fluoride ) ( PVDF ) membrane with the enzymes in the presence of NAD . A pan antibody against Kcr was used to assess protein crotonylation levels . While the incubations with Sirt1–Sirt3 had little influence on lysine crotonylation in most of the protein bands , substantial reductions in Kcr levels were observed in two bands with a molecular mass of approximately 15 kDa ( Figure 5A ) . Considering Sirt1–Sirt3 can decrotonylate histone peptides in vitro , we speculated that these 15 kDa proteins with reduced Kcr levels could be histones . We therefore examined the decrotonylation activity of Sirt1–Sirt3 using purified core histone proteins as substrates . Indeed , Sirt1–Sirt3 not only reduced global Kcr levels of all core histones , they also showed robust decrotonylation activity towards two known histone Kcr sites , H3K4Cr and H3K27Cr ( Figure 5B and Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 02999 . 018Figure 5 . Sirt3 regulates histone lysine crotonylation and gene expression . ( A ) Western blot analyses showing Sirt1–3 catalyzed decrotonylation of whole-cell lysates on the membrane . The arrows indicate the protein bands with reduction of crotonylation levels . Ponceau S staining was used as the loading control . ( B ) Western blot analyses showing the decrotonylation of purified histones catalyzed by Sirt1–3 in the reaction buffer . Ponceau S staining was used as the loading control . ( C ) Western blot analyses showing that Sirt3 knockdown caused the accumulations of H3K4Cr without effect on H3K4Ac , H3K4Me3 , or H3K27Cr levels . ( D ) Western blot analyses showing that Sirt3 knockdown caused global histone crotonylation ( as indicated by the arrows ) . H3 and actin were used as loading controls . ( E ) Chromatin immunoprecipitation–quantitative PCR analyses showing the changes in the histone Kcr levels of the indicated chromatin loci on Sirt3 knockdown . Quantitative PCR signal was normalized by a non-Sirt3 bound region of Gapdh . ( F ) Real time PCR analyses showing the changes in mRNA level of the indicated genes on Sirt3 knockdown . Quantitative PCR signal was normalized by Gapdh . Error bars indicated ± s . e . from four ( E ) or three ( F ) independent biological replicates . The p values are based on the Student's t test . p<0 . 05 , p <0 . 01 , **p <0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01810 . 7554/eLife . 02999 . 019Figure 5—figure supplement 1 . Western blot analyses showing Sirt1-3 catalyzed decrotonylation of extracted histones on membrane . Ponceau S staining was used as the loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 01910 . 7554/eLife . 02999 . 020Figure 5—figure supplement 2 . Analysis of the decrotonylation activities of Sirt1 and Sirt2 in cells . ( A , B ) Western blot analyses showing the influence of Sirt1 ( A ) or Sirt2 ( B ) knockdown on H3K4Cr and H3K27Cr levels in HeLa cells . ( C , D ) Western blot analyses showing that Sirt1 ( C ) or Sirt2 ( B ) knockdown did not cause an increase in global histone crotonylation levels ( as indicated by the arrows ) . H3 and actin were used as loading controls . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 02010 . 7554/eLife . 02999 . 021Figure 5—figure supplement 3 . Subcellular locolization of Sirt3 . ( A ) Analysis of Sirt3 distribution in HeLa cells by fluorescence microscopy using anti-Sirt3 N-term antibody . Blue channel: DAPI . Red channel: Sirt3 . ( B ) Western blot analyses showing the nuclear localization of endogenous Sirt3 using both anti-Sirt3 C-term and N-term antibodies . Western blot analyses of fibrillarin/histone H3 and HSP60 were used to show the purity of nuclear ( NE ) and mitochondrial extraction ( Mito ) , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02999 . 021 We next examined whether Sirt1–Sirt3 regulate histone lysine crotonylation in cells . Although Sirt1 and Sirt2 can decrotonylate histone peptides and proteins in vitro , their knockdowns by siRNA did not cause an appreciable increase in crotonylation levels for both the global histone and the two tested Kcr ( i . e . , H3K4Cr and H3K27Cr ) sites ( Figure 5—figure supplement 2 ) . In contrast , Sirt3 knockdown caused accumulation of global histone crotonylation and the H3K4Cr mark , while the histone H3K4Ac and H3K4Me3 levels were unaltered ( Figure 5C , D ) , suggesting that Sirt3 selectively targets histone crotonylation . Interestingly , the crotonylation level on H3K27 was not influenced by the knockdown of Sirt3 , which may be explained by the observation that Sirt3 showed weaker activity towards the H3K27Cr peptide in vitro ( Figure 4—figure supplement 3 ) . It should be noted that Sirt3 was found to localize predominantly to mitochondria and was mainly involved in metabolic regulations through controlling protein acetylation dynamics . However , recent evidence has suggested that Sirt3 can also be present in the nucleus in its full length form ( Scher et al . , 2007; Iwahara et al . , 2012 ) . Indeed , using an antibody that targets the N-terminal region of Sirt3 , we detected endogenous full length Sirt3 in the nucleus of HeLa cells by both immunofluorescence and western blotting analyses ( Figure 5—figure supplement 3 ) . Taken together , these data suggest that endogenous Sirt3 can function as an ‘eraser’ enzyme to regulate histone crotonylation dynamics in cells . Finally , we sought to determine the potential biological consequence of histone decrotonylation mediated by Sirt3 . It has been reported that Sirt3 can bind to chromatin and cause repression of the neighboring genes in U2OS cells ( Iwahara et al . , 2012 ) . We therefore hypothesized that Sirt3 could regulate gene transcription via controlling local histone Kcr levels . To test this hypothesis , we focused on seven candidate genes , Baz2a , Brip1 , Corin , Ptk2 , Tshz3 , Wapal , and Zfat , whose transcription start sites are close to the Sirt3 enriched region . Chromatin precipitation ( ChIP ) coupled with quantitative PCR ( qPCR ) was performed in U2OS cells with the pan anti-Kcr antibody to measure Kcr levels near the transcription start sites of the candidate genes . As shown in Figure 5E , Sirt3 knockdown by siRNA resulted in significant increases in Kcr levels of five of the seven genes analyzed , indicating that Sirt3 may directly regulate crotonylation dynamics at the genomic loci where it binds . Interestingly , the mRNA levels of the three candidate genes with increased Kcr levels , Ptk2 , Tshz3 , and Wapal , were also significantly upregulated on Sirt3 knockdown ( Figure 5F ) . Given that histone Kcr is enriched at active gene promoters and potential enhancers ( Tan et al . , 2011 ) , this positive correlation between the gene transcription level and the nearby histone Kcr level on Sirt3 knockdown suggests that Sirt3 might relieve a repressive effect on these target genes through ‘erasing’ histone Kcr ‘marks’ . We have established a robust chemical proteomics approach to comprehensively profile histone decrotonylases . There have been important advances in our ability to detect PTMs . However , we currently lack reliable methods to identify , without bias , enzymes that regulate the addition and removal of PTMs , as interactions between PTMs and their regulating enzymes can be weak and transient , thereby limiting the applicability of conventional biochemical ‘pull-down’ methods . Our CLASPI approach overcame this difficulty by applying photo-cross-linking chemistry to convert weak and transient enzyme–PTM interactions into irreversible covalent linkages , and enabled a systematic profiling of the ‘erasers’ of protein PTMs . The present study has also largely broadened the scope of CLASPI technology from finding PTM ‘readers’ ( Li et al . , 2012; Li and Kapoor , 2010 ) , which are usually involved in relatively more stable protein–protein interactions , to identifying dynamic and transient interactions between PTMs and their ‘erasers’ . We anticipate that this approach can be used to comprehensively profile ‘erasers’ of other PTMs , such as arginine demethylases . Siruins were initially recognized as NAD-dependent deacetylases ( Imai et al . , 2000; Landry et al . , 2000; Sauve et al . , 2006 ) . However , emerging evidence revealed that some sirtuins that displayed weak deacetylation activity had substrate specificity towards other acyl groups attached to lysine residues . For examples , Lin et al . recently demonstrated that Sirt5 can preferentially hydrolyze malonyl and succinyl lysine ( Du et al . , 2011; Peng et al . , 2011 ) , and Sirt6 can remove long chain fatty acyl groups ( e . g . , myristoyl group ) from lysine residues ( Jiang et al . , 2013 ) . In this study , we demonstrated that the three human sirtuins , Sirt1–Sirt3 , catalyzed the hydrolysis of crotonyl lysine . This newly discovered decrotonylase activity broadens the landscape of PTMs that are targeted by sirtuins , and it also provides new impetus to investigate the cellular mechanisms and functions of Sirt1–Sirt3 , which to date have been considered solely as deacetylases . This finding is also partially in agreement with the work of Denu and coworkers , in which only Sirt1 and Sirt2 exhibited decrotonylase activity , whereas Sirt3 was totally inactive , towards a histone H3K9Cr peptide in their radioactive [32P]-NAD thin layer chromatography assay . In contrast , Sirt3 displayed robust decrotonylase activity against a variety of crotonylated histone peptides , including an H3K9Cr peptide in this study ( Figure 4 ) . Given the fact that the activity of Sirt3 can be peptide sequence-dependent ( Figure 4 ) , this discrepancy may be caused by the different H3K9Cr peptide substrates used in Denu's and this study , which consisted of amino acid residues 5–13 and 1–15 of histone H3 , respectively . We have demonstrated that endogenous Sirt3 functions as an ‘eraser’ to regulate histone crotonylation in cells . This finding opens new opportunities to investigate the cellular mechanisms and functions of histone crotonylation . In contrast , while the knockdowns of Sirt1 and Sirt2 did not cause accumulation of histone global or H3K4 crotonylation ( Figure 5—figure supplement 2 ) , we cannot rule out the possibility that these two sirtuins could target other histone crotonylation sites . Future studies are therefore needed to systematically profile mammalian crotonylome and analyze the lysine crotonylation sites that are targeted by Sirt1 , Sirt2 , and Sirt3 , by comparing the corresponding wild-type and genetic knockout cells or tissues in conjunction with quantitative proteomics approaches . The seven human sirtuins have distinct subcellular localizations . Sirt1 , Sirt6 , and Sirt7 are in the nucleus , Sirt3–Sirt5 localize to the mitochondria , and Sirt2 is primarily found in the cytoplasm ( Houtkooper et al . , 2012 ) . However , Sirt3 , in its full length form , has recently been found in the nucleus , and nuclear Sirt3 can associate with chromatin and result in repression of nearby genes ( Scher et al . , 2007; Iwahara et al . , 2012 ) . Based on the focused analysis at several Sirt3 target gene loci , the current study suggests a potential correlation of the transcriptional upregulation and the increase in local histone Kcr levels on Sirt3 knockdown . It also generates a hypothesis that Sirt3 could lead to silencing through ‘erasing’ Kcr at target genes . To test this hypothesis and examine the correlation between Sirt3 catalyzed histone deacrotonylation and gene expression genome-wide requires comprehensive profiling of global histone Kcr and gene expression regulated by Sirt3 using ChIP coupled to high throughput sequencing , in combination with RNA sequencing in future studies . In addition , the same type of PTM at different modification sites of histones may have distinct effects on gene expression . For example , trimethylation at histone H3 Lys-4 ( H3K4Me3 ) ‘marks’ genes that are being actively transcribed , whereas the same modification at H3 Lys-27 ( H3K27Me3 ) ‘marks’ transcriptionally silent chromatin ( Martin and Zhang , 2005 ) . By analogy , it is possible that crotonylation at specific lysine sites of histones could also play different roles in the regulation of gene expression . This possibility may account for the fact that the transcription of the two genes ( i . e . , Brip1 and Zfat ) with elevated Kcr levels was not influenced in our study . The study of the effects of site-specific histone Kcr ‘marks’ ( e . g . , H3K4Cr ) targeted by Sirt3 on the regulation of gene expression is an important next step . Unless otherwise noted , all chemical reagents were purchased from Sigma–Aldrich ( St . Louis , MO ) . Dulbecco's Modified Eagle Medium ( DMEM ) was purchased from Life Technologies . Ethylene diamine tetraacetic acid ( EDTA ) free protease inhibitor was purchased from Roche Applied Science ( Germany ) . Pre-stained protein ladder was purchased from Bio-Rad ( Hercules , CA ) . Pre-cast polyacrylamide gels ( 4–12% NuPAGE Bis-Tris gels ) were purchased from Life Technologies . Mass spectrometry grade trypsin was purchased from Promega ( Madison , WI ) . High capacity streptavidin beads were purchased from ThermoFisher Scientific ( Waltham , MA ) . Antibodies were purchased from Santa Cruz Biotechnologies ( Santa Cruz , CA ) ( anti-Sirt1 , anti-Sirt2 , and anti-γ-actin antibodies ) , Cell Signaling Technology ( Danvers , MA ) ( anti-Sirt3 , anti-HSP60 , anti-fibrillarin , and anti-histone H3 antibodies ) , Abcam ( United Kingdom ) ( anti-H3K4Ac and anti-H3K4Me3 antibodies ) , or PTM BioLabs ( Chicago , IL ) ( anti-H3K4Cr , anti-H3K27Cr , and pan anti-crotonyllysine antibodies ) . Anti-Sirt3 N-term antibody was a generous gift from Dr Danny Reinberg ( New York University , New York , United States ) . In-gel fluorescence scanning was performed using a Typhoon 9410 variable mode imager ( excitation 532 nm , emission 580 nm ) . Isothermal titration calorimetry measurements were performed on a MicroCal iTC200 titration calorimeter ( Malvern Instruments , United Kingdom ) . Peptides were purified on a preparative HPLC system with Waters ( Milford , MA ) 2535 Quaternary Gradient Module , Waters 515 HPLC pump , Waters SFO System Fluidics Organizer , and Waters 2767 Sample Manager . Enzymatic reactions were monitored by an LC-MS system with Waters 1525 Binary HPLC Pump , Waters 2998 Photodiode Array Detector , and Waters 3100 Mass Detector . Detection of O-Cr-ADPR was carried out by Agilent ( Santa Clara , CA ) 1260 Infinity HPLC system connected to a Thermo Fisher Scientific LCQ DecaXP MS detector . All peptides were synthesized on Rink-Amide MBHA resin following a standard Fmoc based solid phase peptide synthesis protocol . Removal of protecting groups and cleavage of peptides from the resin were done by incubating the resin with a cleavage cocktail containing 95% trifluoroacetic acid ( TFA ) , 2 . 5% triisopropylsilane , 1 . 5% water , and 1% thioanisole for 2 hr . Peptides were purified by preparative HPLC with an XBridge Prep OBD C18 column ( 30 mm×250 mm , 10 μm; Waters ) . Mobile phases used were water with 0 . 1% TFA ( buffer A ) and 90% acetonitrile ( ACN ) in water with 0 . 1% TFA ( buffer B ) . Peptides containing photo-cross-linker ( benzophenone ) were eluted with gradient 15–40% buffer B in 40 min; all other peptides were eluted with gradient 5–35% buffer B in 40 min . The elution rate was 15 mL/min . The purity and identity of the peptides were verified by LC-MS . HeLa S3 , HEK293T , and HeLa cells were cultured in DMEM supplemented with 10% fetal bovine serum ( FBS ) , 100 U/mL penicillin , and 100 μg/mL streptomycin . Cells were maintained in a humidified 37 °C incubator with 5% CO2 . HeLa S3 cells were grown in suspension at 37°C in a humidified atmosphere with 5% CO2 in DMEM medium ( –Arg , –Lys; Life Technologies ) containing 10% dialyzed fetal bovine serum ( Life Technologies ) , penicillin–streptomycin , and supplemented with 22 mg/L 13C615N4-L-arginine ( Cambridge Isotope Laboratories , Tewksbury , MA ) and 50 mg/L 13C615N2-L-lysine ( Cambridge Isotope ) or the corresponding non-labeled amino acids ( Peptide International , Louisville , KY ) . Harvested cell pellets were washed with ice cold phosphate buffered saline ( PBS ) and frozen in liquid N2 . The cell powder grinded with a Ball Mill ( Retch MM301 ) was stored at −80 °C until use . To prepare whole-cell lysates , the frozen cell powder was first resuspended in a hypotonic buffer ( 10 mM HEPES , pH 7 . 5 , 2 mM MgCl2 , 0 . 1% Tween-20 , 20% glycerol , 2 mM phenylmethylsulfonyl fluoride ( PMSF ) , and Roche Complete EDTA free protease inhibitors ) and incubated for 10 min at 4 °C . The suspension was centrifuged at 16 , 000×g for 15 min at 4 °C and the supernatant was kept for use later . The pellet was resuspended in a high salt buffer ( 50 mM HEPES , pH 7 . 5 , 420 mM NaCl , 2 mM MgCl2 , 0 . 1% Tween-20 , 20% glycerol , 2 mM PMSF , and Roche Complete EDTA free protease inhibitors ) and incubated for 30 min at 4 °C . The suspension was centrifuged at 16 , 000×g for 15 min at 4 °C , and the supernatant was combined with the soluble fraction in hypotonic buffer to give the whole-cell lysates . In a ‘selectivity filter’ experiment , probe 1 and probe C were incubated with heavy and light SILAC whole-cell lysates , respectively , in the binding buffer ( 50 mM HEPES , pH 7 . 5 , 168 mM NaCl , 2 mM MgCl2 , 0 . 1% Tween-20 , 20% glycerol , 2 mM PMSF , and Roche Complete EDTA free protease inhibitor cocktail ) for 15 min at 4 °C . The samples were then irradiated at 365 nm using a Spectroline ML-3500S UV lamp for 15 min on ice . In ‘an affinity filter experiment’ , the heavy and light SILAC lysates were reacted with probe 1 in the absence and presence , respectively , of H3K4Cr ( 1–15 ) peptide ( 30 μM ) as a competitor . After photo-cross-linking , the heavy and light lysates were pooled . Briefly , to the prepared samples , 100 μM of rhodamine azide for in-gel fluorescence scanning or cleavable biotin-azide for streptavidin enrichment were added , followed by 1 mM tris ( 2-carboxyethyl ) phosphine and 100 μM tris[ ( 1-benzyl-1H-1 , 2 , 3-triazol-4-yl ) methyl]amine , and the reactions were initiated by the addition of 1 mM CuSO4 . The reactions were incubated for 1 . 5 hr at room temperature . After the click chemistry with cleavable biotin-azide , the reaction was quenched by adding 4 volumes of ice cold acetone to precipitate the proteins . After washing with ice cold methanol twice , the air dried protein pellet was dissolved in PBS with 4% SDS , 20 mM EDTA , and 10% glycerol by vortexting and heating . The solution was then diluted with PBS to give a final concentration of SDS of 0 . 5% . High capacity streptavidin agarose beads ( Thermo Fisher Scientific ) were added to bind the biotinylated proteins with rotating for 1 . 5 hr at room temperature . To remove non-specific binding , the beads were washed with PBS with 0 . 2% SDS , 6 M urea in PBS with 0 . 1% SDS , and 250 mM NH4HCO3 with 0 . 05% SDS . The enriched proteins were then eluted by incubating with 25 mM Na2S2O4 , 250 mM NH4HCO3 , and 0 . 05% SDS for 1 hr . The eluted proteins were dried down with SpeedVac . The dried proteins were resuspended in 30 μL of lithium dodecyl sulfate sample loading buffer ( Life Technologies ) with 50 mM dithiothreitol ( DTT ) , heated at 75 °C for 8 min , and then reacted with iodoacetamide in the dark for 30 min to alkylate all of the reduced cysteines . Proteins were then separated on a Bis-Tris gel , followed by fixation in a 50% methanol/7% acetic acid solution . The gel was stained by GelCode Blue stain ( Pierce ) . The diced 1 mm ( Goldberg et al . , 2007 ) cubes of gels were then destained by incubating with 50 mM ammonium bicarbonate/50% acetonitrile for 1 hr . The destained gel cubes were dehydrated in acetonitrile for 10 min and rehydrated in 25 mM NH4HCO3 with trypsin for protein digestion at 37 °C overnight . The resulting peptides were enriched with StageTips . The peptides eluted from the StageTips were dried down by SpeedVac and then resuspended in 0 . 5% acetic acid for analysis by LC-MS/MS . Mass spectrometry was performed on an LTQ-Orbitrap Velos mass spectrometer ( Thermo Fisher Scientific ) . First , peptide samples in 0 . 1% formic acid were pressure loaded onto a self-packed PicoTip column ( New Objective , Woburn , MA ) ( 360 μm od , 75 μm id , 15 μm tip ) , packed with 7–10 cm of reverse phase C18 material ( ODS-A C18 5-μm beads from YMC America , Allentown , PA ) , rinsed for 5 min with 0 . 1% formic acid , and subsequently eluted with a linear gradient from 2% to 35% B for 150 min ( A=0 . 1% formic acid , B=0 . 1% formic acid in ACN , flow rate ∼200 nL/min ) into the mass spectrometer . The instrument was operated in a data-dependent mode , cycling through a full scan ( 300–2000 m/z , single μscan ) followed by 10 CID MS/MS scans on the 10 most abundant ions from the immediate preceding full scan . Cations were isolated with a 2 Da mass window and set on a dynamic exclusion list for 60 s after they were first selected for MS/MS . The raw data were processed and analyzed using MaxQuant ( version 1 . 2 . 2 . 5 ) . A human fasta file ( ipi . HUMAN . v . 3 . 68 . fasta ) was used as the protein sequence searching database . Default parameters were adapted for protein identification and quantification . In particular , parent peak MS tolerance was 6 ppm , MS/MS tolerance was 0 . 5 Da , minimum peptide length was 6 amino acids , and maximum number of missed cleavages was 2 . The proteins quantified were supported by at least two quantification events . Both the ‘selectivity filter’ and ‘affinity filter’ experiments were repeated twice , and only the proteins that were identified and quantified in all experiments were reported . The click chemistry reactions were quenched by adding 1 volume of 2×sample buffer . The proteins were heated at 85 °C for 8 min , and resolved by SDS-PAGE . The labeled proteins were visualized by scanning the gel on a Typhoon 9410 variable mode imager ( excitation 532 nm , emission 580 nm ) . Plasmids of Sirt1 ( 193–747 ) , Sirt2 ( 36–356 ) , Sirt5 ( 34–302 ) , and Sirt6 ( 1–314 ) for Escherichia coli expression were generated as previously described ( Finnin et al . , 2001; Du et al . , 2011; Hubbard et al . , 2013; Jiang et al . , 2013 ) . Plasmids of Sirt3 ( 102–399 ) cloned in pTrcHis 2C vector for E . coli expression and full length Sirt3 ( wide-type and mutant H248Y ) cloned into pcDNA3 . 1 vector for mammalian cell expression were generous gifts from Dr Eric Verdin ( University of California , San Francisco ) . Sirt3 mutant F180L was generated by site directed mutagenesis . All of the proteins were expressed in E . coli Rosetta cells . To induce expression of target proteins , isopropyl β-D-1-thiogalactopyranoside was added to a final concentration of 0 . 2 mM when OD600 reached 0 . 6 , and the culture was grown at 15 °C ( Sirt3 at 25 °C ) for 16–18 hr . Cells were harvested and resuspended in lysis buffer A ( 50 mM Tris–HCl , pH 7 . 5 , 500 mM NaCl , 1 mM PMSF , and Roche EDTA free protease inhibitors , for Sirt1 , Sirt2 , and Sirt6 ) or buffer B ( 50 mM Tris–HCl , pH 7 . 5 , 150 mM NaCl , 1 mM PMSF , and Roche EDTA free protease inhibitors , for Sirt3 and Sirt5 ) . Following sonication and centrifugation , the supernatant was loaded onto a nickel column pre-equilibrated with lysis buffer . The column was washed with 5 column volumes of wash buffer ( lysis buffer with 30 mM imidazole ) and then the target proteins were eluted with elution buffer ( lysis buffer with 250 mM imidazole ) . After purification , Sirt2 was digested by UPL1 at 4 °C overnight and purified by a Highload 26/60 Superdex75 gel filtration column ( GE Healthcare Life Sciences , United Kingdom ) . Sirt6 was purified by SP column and Superdex75 gel filtration column . Others were loaded onto a Superdex75 gel filtration or Highload 26/60 Superdex200 ( for Sirt1 ) column . After concentration , the target proteins were frozen and stored at −80 °C . Experiments were performed at 25 °C on a MicroCal iTC200 titration calorimeter ( Malvern Instruments ) . The reaction cell containing 200 μL of 100–200 μM proteins was titrated with 17 injections ( firstly 0 . 5 μL , and all subsequent injections 2 μL of 1 . 5–2 . 5 mM peptides ) . The binding isotherm was fit with Origin 7 . 0 software package ( OriginLab , Northampton , MA ) that uses a single set of independent sites to determine the thermodynamic binding constants and stoichiometry . Sirt3/H3K4Cr mixtures were prepared at a 1:20 protein/peptide molar ratio and incubated for 60 min on ice . Crystals of Sirt3 ( 102–399 ) complexed with H3K4Cr ( 1–10 ) peptide were obtained by the hanging drop vapor diffusion method at 291 K using commercial screens from Hampton Research ( Aliso Viejo , CA ) . Each drop , consisting of 1 μL of 10 mg/mL protein complex solution ( 20 mM Tris–HCl , pH 7 . 4 , 100 mM NaCl , and 5 mM DTT ) and 1 μL of reservoir solution , was equilibrated against 400 μL of reservoir solution . The qualified crystals of Sirt3 grew with a cube profile within 1 week with a reservoir containing 12% PEG4K , 0 . 1 M sodium malonate , pH 6 . 5 , and 5% isopropanol . The mixture of 25% glycerol with the reservoir solution above was used as the cryogenic liquor . The X-ray diffraction data were collected at 100 K in a liquid nitrogen gas stream using the Shanghai Synchrotron Radiation Facility beamline 17U ( λ = 0 . 9791 Å ) . A total of 120 frames were collected with a 1° oscillation and the data were indexed and integrated using the program HKL2000 ( Otwinowski and Minor , 1997 ) . The complex structure of Sirt3 with H3K4Cr peptide was solved by molecular replacement using the program Molrep from the CCP4 Suit ( Collaborative Computational Project , Number 4 , 1994 ) , with the published Sirt3 structure ( PDB: 3GLR ) ( Jin et al . , 2009 ) as the search model . Refinement and model building were performed with REFMAC5 and COOT from CCP4 . The X-ray diffraction data collection and structure refinement statistics are shown in Supplementary file 1 . The enzymatic activities of human sirtuins were measured by detecting the removal of the crotonyl group from peptides ( Du et al . , 2011 ) . Sirtuin protein ( 5 μM ) was incubated with 500 μM of corresponding crotonylated peptides and 1 mM of NAD in a reaction buffer containing 20 mM Tris–HCl buffer ( pH 7 . 5 ) and 1 mM DTT at 37 °C for 2 hr . The reactions were stopped by adding one-third reaction volume of 20% TFA and immediately frozen in liquid N2 . For Sirt3 , samples without NAD or without enzyme were treated under the same conditions as the controls . Samples were then analyzed by LC-MS with a Vydac 218TP C18 column ( 4 . 6 mm×250 mm , 5 μm; Grace Davison , Columbia , MD ) . Mobile phases used were 0 . 05% TFA in water ( buffer A ) and 0 . 05% TFA in ACN ( buffer B ) . The flow rate for LC was 0 . 6 mL/min . The peptide mixtures were eluted by buffer A for 10 min and then 0–30% buffer B over 10 min . MS started to record at 10 min for each injection . Enzyme was incubated with different concentrations of corresponding peptides bearing two tryptophans at the C terminus ( 20 , 40 , 60 , 80 , 100 , 200 , and 500 μM ) and 1 . 0 mM NAD in 20 mM Tris–HCl buffer ( pH 7 . 5 ) containing 1 mM DTT in 25 μL reaction at 37°C for a certain period of time within the initial linear range . The enzyme concentration and reaction time used were: Sirt3–H3K4Ac: 1 μM enzyme , 5 min; Sirt3–H3K4Cr: 1 μM enzyme , 20 min; Sirt3 ( F180L ) –H3K4Ac: 0 . 8 μM enzyme , 5 min; and Sirt3 ( F180L ) –H3K4Cr: 5 μM enzyme , 20 min . The reactions were stopped by adding one-third reaction volume of 20% TFA and immediately frozen in liquid N2 . Samples were then analyzed by HPLC with a Vydac 218TP C18 column ( 4 . 6 mm×250 mm , 5 μm; Grace Davison ) . Mobile phases used were water with 0 . 1% TFA ( buffer A ) and 90% ACN in water with 0 . 1% TFA ( buffer B ) . The wavelength for UV detection was 280 nm . The analysis gradient for deacetylation samples was 16% buffer B for 20 min with a flow rate at 1 . 5 mL/min . The analysis gradient for decrotonylation samples was 15–35% buffer B in 12 min with a flow rate at 1 . 0 mL/min . Sirt3 ( 5 μM ) was incubated with 500 μM of H3K4Cr ( 1–15 ) peptide and 1 mM of NAD in a reaction buffer containing 20 mM Tris–HCl buffer ( pH 7 . 5 ) and 1 mM DTT at 37°C for 2 hr . The reactions were stopped by immediately freezing in liquid N2 . Sample was then analyzed by LC-MS with a VisionHT C18 column ( 2 . 1 mm×150 mm , 3 μm; Grace Davison ) on an Agilent 1260 Infinity HPLC system , followed by Thermo Fisher Scientific LCQ DecaXP MS Detector . Mobile phases used were 0 . 02% TFA in water ( buffer A ) and 90% ACN in water with 0 . 02% TFA ( buffer B ) . The flow rate for LC was 0 . 2 mL/min . The sample was eluted by buffer A for 10 min and then 0–10% buffer B over 10 min . The wavelength for UV detection was 260 nm . MS started to record at 10 min . Sirt1 siRNA 15 nM ( Santa Cruz Biotechnologies ) , Sirt2 siRNA 30 nM ( Thermo Fisher Scientific ) , or Sirt3 siRNA 30 nM ( Thermo Fisher Scientific ) was transfected into a HeLa cell line with DharmaFECT 1 Transfection Reagent ( Thermo Fisher Scientific ) , according to the manufacturer's instructions . Corresponding concentrations of control siRNA were used as negative controls . Following transfection , cells were then maintained in a humidified 37 °C incubator with 5% CO2 for another 48 hr ( for Sirt1 and Sirt2 ) or 72 hr ( for Sirt3 ) . An acid extraction method was used to isolate histones from HeLa S3 cells ( Shechter et al . , 2007 ) . Briefly , the harvested HeLa S3 cell pellet was resuspended with lysis buffer ( 10 mM Tris–HCl pH 8 . 0 , 1 mM KCl , 1 . 5 mM MgCl2 , 1 mM DTT 2 mM PMSF , and Roche Complete EDTA free protease inhibitors ) and incubated at 4 °C by rotating for 1 hr . The intact nuclei were pelleted by centrifuging at 10 , 000×g for 10 min at 4 °C . To extract histones , 0 . 4 N H2SO4 was added to resuspend the nuclei , followed by rotating at 4°C overnight . After centrifuging to remove the nuclei debris , histones were precipitated by adding 100% trichloroacetic acid drop by drop ( trichloroacetic acid final concentration 33% ) . The precipitated histones were pelleted at 16 , 000×g for 10 min at 4 °C and washed with ice cold acetone twice . The air dried protein pellet was dissolved with ddH2O and stored at −80 °C for later use . HeLa S3 whole-cell lysate ( 20 μg ) or 5 μg of extracted histones were resolved by SDS-PAGE gel and transferred to PVDF membranes . The membranes were incubated with or without 0 . 1 μM of Sirt3 in reaction buffer ( 25 mM Tris–HCl , 130 mM NaCl , 3 mM KCl , 1 mM MgCl2 , and 1 mM DTT , pH 7 . 5 ) containing 1 mM NAD at 37 °C for 2 hr . Extracted histones ( 4 μg ) were incubated with or without 1 μM or 5 μM of Sirt3 in reaction buffer ( 25 mM Tris–HCl , 130 mM NaCl , 3 mM KCl , 1 mM MgCl2 , and 1 mM DTT , pH 7 . 5 ) containing 1 mM NAD at 37 °C overnight . HeLa cells grown on coverslips were fixed with 3 . 7% polyformaldehyde in PBS , permeabilized with 0 . 1% Triton X-100 in PBS , and blocked for 30 min at room temperature using 5% bovine serum albumin ( dissolved with PBS containing 0 . 1% Triton X-100 ) . Cells were incubated with primary antibody overnight at 4°C and washed trice with PBST ( 0 . 1% Tween-20 in PBS ) prior to secondary antibody ( containing DAPI for nucleus staining ) incubation at room temperature for 1 hr . Washed cells were then subjected to a Zeiss LSM 510 laser scanning confocal microscope . In brief , HeLa cells were harvested by centrifugation and washed with PBS twice; all subsequent steps were performed at 4 °C . Cells were then suspended in 5 cell pellet volumes of buffer A ( 10 mM HEPES , pH 7 . 9 at 4 °C , 1 . 5 mM MgCl2 , 10 mM KCl , and 0 . 5 mM DTT ) followed by incubation for 10 min . After centrifugation , cells were resuspended in 2 cell pellet volumes of buffer A and lysed by Dounce homogenizer ( B type pestle ) with homogenate checked by microscopy . The cell lysis was layered over 30% sucrose in buffer A and then centrifuged for 15 min at 800×g . The resulting pellet was recovered from the sucrose phase , washed by buffer A twice , and then extracted by buffer C ( 20 mM HEPES , pH 7 . 9 , 25% ( vol/vol ) glycerol , 0 . 42 M NaCl , 1 . 5 mM MgCl2 , 0 . 2 mM EDTA , 0 . 5 mM PMSF , and 0 . 5 mM DTT ) for 30 min at 4 °C . After centrifugation at 12 , 000×g for 30 min , the supernatant was termed the nuclear fraction . The resulting supernatant was centrifuged twice at 800×g to complete the pellet nuclei and intact cell . The supernatant was then centrifuged at 7 , 000×g to pellet the mitochondria followed by washing twice with buffer A . The mitochondria were then lysed by TXIP-1 buffer ( 1% Triton X-100 ( vol/vol ) , 150 mM NaCl , 0 . 5 mM EDTA , and 50 mM Tris–HCl , pH 7 . 4 ) . Protein concentration was determined by BCA assay . Proteins separated by SDS-PAGE were transferred onto a PVDF membrane which was then blocked ( 5% non-fat dried milk and 0 . 1% Tween-20 in PBS ) for 1 hr at room temperature . The membrane was incubated with primary antibody diluted in PBST with 2% bovine serum albumin , followed by washing with PBST for 5 min trice , incubated with goat anti-rabbit-horseradish peroxidase conjugated secondary antibody ( 1:20000; Santa Cruz Biotechnologies ) , or rabbit anti-mouse- horseradish peroxidase conjugated secondary antibody ( 1:5000; Santa Cruz Biotechnologies ) diluted in PBST for 1 hr at room temperature , and then visualized with western blotting detection reagents ( Thermo Fisher Scientific ) . Total RNA was isolated using TRIzol Reagent ( Life Technologies ) . RNA was reverse transcribed into cDNA by M-MLV Reverse Transcriptase ( Life Technologies ) using oligo ( dT ) primers . qPCR was performed using Power SYBR Green PCR Master Mix ( Life Technologies ) on an ABI StepOnePlus system following the manual’s instructions . All primers used are listed in Supplementary file 2 . Cells were cross linked by 1% formaldehyde for 10 min and quenched by 0 . 125 M glycine for 5 min at room temperature . Cells were then lysed by ChIP lysis buffer ( 5 mM PIPES pH 8 . 0 , 85 mM KCl , and 1% IGEPAL CA-630 ) and homogenized using a glass Dounce homogenizer ( type B pestle ) . The nuclear fraction was precipitated and lysed in nuclei lysis buffer ( 50 mM Tris–HCl , pH 8 . 0 , 10 mM EDTA , and 1% SDS ) for 30 min at 4 °C . The nuclear lysis was sonicated to a chromatin ranging from 600 bp to 800 bp . Immunoprecipitation was done in immunoprecipitation dilution buffer ( 50 mM Tris–HCl , pH 7 . 4 , 150 mM NaCl , 1% IGEPAL CA-630 , 0 . 25% deoxycholic acid , and 1 mM EDTA ) using Dynabeads coupled with Protein G ( Life Technologies ) . Chromatin ( 5 μg ) and 8 μg of pan anti-crotonyllysine antibody were used for each ChIP reaction . Chromatin complex was eluted from beads by ChIP elution buffer ( 50 mM NaHCO3 and 1% SDS ) and added to 5 M NaCl to a final concentration of 0 . 54 M . To reverse cross links of protein/DNA complex to free DNA , samples were incubated at 65 °C for 2 hr followed by 95 °C for 15 min . After incubation with RNase ( Thermo Fisher Scientific ) for 20 min at 37 °C , DNA was recovered and used for qPCR , as described above . All primers used are listed in Supplementary file 2 .	Most of the DNA in a cell is wound around histone proteins to form a compacted structure called chromatin . Enzymes can modify the histones by adding small chemical tags on to them , and these histone modifications can cause the chromatin to either become more tightly packed or more open . Opening up the chromatin makes the DNA more accessible to the cellular machinery involved in gene expression . Thus , cells can regulate which genes they express , and by how much , by modifying the histone proteins . Like all other proteins , histones are made of smaller molecules called amino acids . Specific amino acids within histone proteins can be modified in a number of different ways , with different effects . For instance , adding a chemical tag called an acetyl group onto an amino acid in a histone weakens the interaction between the histone and the DNA , which opens up the chromatin and increases gene expression . Another way that histones can be modified is by the addition of crotonyl groups . These chemical tags have not been examined much because the enzymes that add or remove them remain to be identified . However , it was recently suggested that enzymes called sirtuins—which are known to remove acetyl groups from histones—might also remove the crotonyl groups . Finding histone-modifying enzymes is challenging because the interactions between these enzymes and the histones are both weak and brief . Bao , Wang , Li , Li et al . have now overcome this challenge by developing a method to firmly link any protein that interacts with a crotonylated histone to the histone . Three out of the seven sirtuin enzymes found in humans were revealed to bind to crotonylated histones . All three of these enzymes—called Sirt1 , Sirt2 and Sirt3—could remove crotonyl groups from histones in a test-tube , and Sirt3 could also do the same in living cells . Further biochemical experiments suggested that the mechanism used by these enzymes to remove crotonyl groups is the same as the mechanism they use to remove acetyl groups . Bao , Wang , Li , Li et al . then uncovered the three-dimensional structure of the Sirt3 enzyme bound to a crotonylated histone , and revealed that the enzyme recognizes the crotonyl group on the histone via a unique interaction between the crotonyl group and a specific amino acid in the binding pocket of Sirt3 . This amino acid is also found in Sirt1 and Sirt2 , but not in other sirtuins; this interaction can thus explain why decrotonylation activity was only detected for these three enzymes . Moreover , the levels of crotonylated histones and gene expression were higher in cells that lacked Sirt3 , but not in those lacking Sirt1 or Sirt2 . By identifying Sirt3 as the main decrotonylation enzyme in living cells , the role of histone crotonylation can now be investigated in greater detail .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology" ]	2014	Identification of ‘erasers’ for lysine crotonylated histone marks using a chemical proteomics approach
The mammalian heartbeat is thought to begin just prior to the linear heart tube stage of development . How the initial contractions are established and the downstream consequences of the earliest contractile function on cardiac differentiation and morphogenesis have not been described . Using high-resolution live imaging of mouse embryos , we observed randomly distributed spontaneous asynchronous Ca2+-oscillations ( SACOs ) in the forming cardiac crescent ( stage E7 . 75 ) prior to overt beating . Nascent contraction initiated at around E8 . 0 and was associated with sarcomeric assembly and rapid Ca2+ transients , underpinned by sequential expression of the Na+-Ca2+ exchanger ( NCX1 ) and L-type Ca2+ channel ( LTCC ) . Pharmacological inhibition of NCX1 and LTCC revealed rapid development of Ca2+ handling in the early heart and an essential early role for NCX1 in establishing SACOs through to the initiation of beating . NCX1 blockade impacted on CaMKII signalling to down-regulate cardiac gene expression , leading to impaired differentiation and failed crescent maturation . The heart is the first organ to form and function during mammalian embryonic development . In the mouse , mesoderm originating from the primitive streak forms a bilateral pool of progenitor cells that at E7 . 5 give rise to the cardiac crescent ( CC ) . The CC subsequently expands and migrates to the midline whereupon , between E8 . 25 and E8 . 5 , the two sides of the CC fuse and form the linear heart tube ( LHT ) ( reviewed in Buckingham et al . , 2005 ) . The first cardiac contractions have been described during the transition from CC to LHT . Studies of heart development in model organisms have historically focused on the origin and spatial-temporal allocation of cardiac progenitors and cardiovascular lineage determination ( Buckingham et al . , 2005; Saga et al . , 1996; Cai et al . , 2003; Meilhac et al . , 2004; Wu et al . , 2006; Moretti et al . , 2006; Evans et al . , 2010; Devine et al . , 2014 ) . Whilst insight into the identification and regulation of cardiac cell types is important for improved understanding of congenital heart disease ( Bruneau , 2008 ) , an anatomical and cellular bias has overlooked a role for the onset of cardiac function . Early descriptions of initial cardiac contractions ( Navaratnam et al . , 1986 ) , including suggested pacemaker activity on either side of the embryonic midline ( Goss , 1952 ) , as well as optical mapping of spontaneous action potentials performed in both chicken ( 7 somite stage ) and rat ( 3-somite stage ) ( Fujii et al . , 1981; Hirota et al . , 1985 ) , have been informative , but lack resolution . Subsequent studies in mouse ( 3-somite stage ) could only infer early electrical activity based on irregular fluctuations in basal Ca2+ ( Nishii and Shibata , 2006 ) . Given the forming heart contracts from an early stage , this raises the important question of when and how contractile activity of cardiomyocytes is first initiated during development and to what extent this influences the progression of differentiation and subsequent cardiac morphogenesis . This is especially important as the forces exerted by cardiac contractions have been shown in several models to be required for proper heart development ( Granados-Riveron and Brook , 2012 ) , and to modulate gene expression ( Miyasaka et al . , 2011 ) at later developmental stages . In mature cardiomyocytes , coordinated electrical excitation is coupled to physical contraction in a process termed excitation contraction coupling ( ECC ) ( Bers , 2002 ) . ECC relies on changes in the intracellular concentration of the second messenger Ca2+ via release from the sarcoplasmic reticulum ( SR ) in a process termed Ca2+ induced Ca2+ release ( CICR ) ( Fabiato and Fabiato , 1979 ) . Increases in the concentration of intracellular Ca2+ result in cardiomyocyte contraction due to Ca2+ binding to troponin and myofilament activation . ECC involves a number of specific proteins including L-type Ca2+ channels ( LTCC , sarcolemmal Ca2+ influx ) , ryanodine receptors ( RyR2 , SR Ca2+ release ) , the sarco ( endo ) plasmic reticulum Ca2+ ATPase ( SERCA , SR Ca2+ uptake ) and the Na+/Ca2+ exchanger ( NCX , Sarcolemmal Ca2+ efflux ) . Targeted disruption of genes encoding ECC proteins in mice has shown that contractile activity of immature cardiomyocytes does not require ECC . Embryonic cardiomyocytes have a less developed SR and T-tubule system as well as an increased requirement for sarcolemmal Ca2+ flux ( Conway et al . , 2002; Seki et al . , 2003 ) and whilst they express a variety of ion channels and exchangers present in the adult heart ( Seisenberger et al . , 2000; Cribbs et al . , 2001; Linask et al . , 2001 ) , the expression and activity of these proteins is distinct from that in mature cardiomyocytes ( Liang et al . , 2010 ) . Using isolated cells as well as genetically manipulated animals , two contrasting mechanisms have been proposed for how Ca2+ transients are generated in the developing heart from approximately E8 . 5 onwards . Early studies suggested that myocyte contraction is triggered by sarcolemmal Ca2+ influx through voltage activated Ca2+ channels with little or no contribution from the SR ( Nakanishi et al . , 1988; Takeshima et al . , 1998 ) . In contrast , more recently it has been shown that at ~E8 . 5–9 , Ca2+ transients originate from the SR , via RyR together with InsP3 channels , to trigger electrical activity as well as contraction ( Viatchenko-Karpinski et al . , 1999; Méry et al . , 2005; Sasse et al . , 2007; Rapila et al . , 2008 ) . Whilst these studies characterised SR function at ~E8 . 5–9 , they did not investigate how Ca2+ transients are regulated at the earliest stages of cardiac crescent development when contraction is initiated , and relied on experiments performed using isolated cells cultured for between 12 to 70 hr ( Sasse et al . , 2007; Rapila et al . , 2008 ) . Thus there is a lack of cellular resolution in vivo and no current mechanistic insight into the onset of Ca2+ handling and its impact on differentiation and cardiogenesis . We report here , for the first time , high-resolution live imaging of Ca2+ transients during the earliest manifestation of murine heart development well before any indication of spontaneous cardiac contractions . We employed the use of multiple pharmacological inhibitors to address the contribution of the NCX1 and LTCC Ca2+ channels during this process and reveal an essential early role for NCX1-dependent Ca2+ handling on downstream cardiac differentiation and morphogenesis . It is commonly stated that initiation of contraction begins with the formation of the LHT ( Bruneau , 2008 ) , and whilst cardiac contractions have been reported just prior to the ‘linear heart tube’ stage ( Navaratnam et al . , 1986; Nishii and Shibata , 2006; Linask et al . , 2001; Kumai et al . , 2000; Porter and Rivkees , 2001 ) , a precise study on the initiation of cardiac function has not been conducted . A difficulty with these reports is the use of ‘embryonic day’ or ‘somite number’ to stage the developing heart . Somite number is variable in its correlation to the overall embryonic stage ( Kaufman and Navaratnam , 1981 ) , can depend on genetic background ( Méry et al . , 2005; Porter and Rivkees , 2001 ) and importantly , is not a sufficiently fine-grained proxy for the developmental stages of the heart . This can lead to ambiguities , as a ‘3-somite’ embryo may range from the cardiac crescent to early LHT stages . We , therefore , created a staging system specific to the early heart , from early crescent to LHT ( Supplementary file 1a ) , similar to studies at later stages when a more precise morphological characterization is necessary ( Biben and Harvey , 1997 ) . On this basis , we defined four stages ( 0 , 1 , 2 and 3 ) of cardiac crescent development prior to the LHT stage , based on clear morphological differences . Stage 0 hearts represented the first discernible crescent structure situated beneath the developing head folds , being the widest ( 360–390 µm along the medio-lateral axis ) and thinnest ( 70–80 µm along the rostro-caudal axis ) of the crescent stages ( Supplementary file 1a ) . Whilst stage 1 was morphologically similar to stage 0 , the cardiac crescent had become narrower ( 300–370 µm ) and thicker ( 75–95 µm ) . By stage 2 , folding of the cardiac crescent is evident based on the formation of a trough at the embryonic midline and two lobes on either side . As the embryo transitions to stage 3 this trough becomes less obvious with a rostral-caudal elongation of the heart as the LHT begins to form . Transition from stage 3 to the LHT was defined by the complete fusion of the two lobes and loss of the central trough ( Figure 1; Supplementary file 1a ) . 10 . 7554/eLife . 17113 . 003Figure 1 . Sarcomeric assembly occurs in the forming cardiac crescent during heart development . Maximum intensity projections of alternating myomesin ( Myom ) and sarcomeric alpha-actinin ( α-Actinin ) immunostaining from cardiac crescent formation to the linear heart tube stage ( LHT; A–E; A , 11 stacks; B , 36 stacks; C , 35 stacks; D , 31 stacks; E , 36 stacks ) . Analysis by qRT-PCR revealed a significant increase in the expression of Myom1 , Actn2 , Tnnt2 ( encoding Myomesin , sarcomeric alpha-actinin and cardiac troponin t ) , in isolated cardiac crescents between stage 0 and stage 1 ( F ) . cc , cardiac crescent ( lateral plate mesoderm ) ; hf , head folds ( neural ectoderm ) . Scale bars: A–E , 100 μm , A’–E’ , 10 μm . Statistics: ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; *p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 00310 . 7554/eLife . 17113 . 004Figure 1—figure supplement 1 . Sarcomeric assembly occurs in the forming cardiac crescent during heart development . Maximum intensity projections of cardiac troponin ( cTnT ) immunostaining revealed progressive differentiation and sarcomeric assembly during stages of cardiac crescent through to linear heart formation ( LHT; A–E; A , 30 stacks; B , 31 stacks; C , 40 stacks; D , 26 stacks; E , 21 stacks ) . qRT-PCR of Myom1 , Actn2 and Tnnt2 expression during stages of cardiac crescent through to linear tube formation ( F , n = 5 per stage ) . cc , cardiac crescent; hf , head folds . Scale bars: A , top panel 100 μm , bottom panel 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 004 To molecularly characterize these stages , we performed immunostaining for three proteins of the cardiac contractile machinery: sarcomeric α-actinin ( α-Act ) , a protein of the Z-line; Myomesin ( Myom ) , a protein of the M-line and cardiac Troponin T ( cTnT ) , the Tropomyosin binding subunit of the troponin complex ( Figure 1; Figure 1—figure supplement 1 ) . At stage 0 , cTnT was the most evident contractile protein within the early cardiac crescent , albeit without sarcomeric banding ( Figure 1—figure supplement 1A ) . Both α-Actinin and Myomesin appeared in small clusters of cells at stage 0 ( Figure 1A ) . Sarcomeric banding of these proteins , indicative of contractile capability , only manifested later in discrete regions within stage 1 and 2 crescents ( Figure 1B , C; Figure 1—figure supplement 1B , C ) . This was surprising given that the crescent stages of early heart development are thought to correspond to non-differentiated cardiac mesoderm ( reviewed in Harvey , 2002 ) , without prior reports of contractile machinery or functionality . By stage 3 , sarcomere assembly and myofibrilar banding became more uniform , coincident with coalescence of the paired crescent primordia to the embryonic midline . This increased through to the linear heat tube stage , consistent with progressive cardiomyocyte maturation ( Figure 1D , E; Figure 1—figure supplement 1D , E ) . To further characterize stage development in relation to sarcomere assembly , qRT-PCR was performed on isolated cardiac crescents , to assess the corresponding gene expression of Myom1 ( encoding Myom ) , Actn2 ( encoding α-actinin ) and Tnnt2 ( encoding cTnT ) ( Figure 1F ) . Myom1 , Actn2 and Tnnt2 expression significantly increased between stages 0 and 1 and continued to increase until formation of the LHT ( Supplementary file 1b ) . Coincident with sarcomere formation at stage 1 ( Figure 1B; Figure 1—figure supplement 1B ) , we observed the onset of beating in discrete foci in the lateral regions of cardiac crescents of cultured mouse embryos , by differential interference contrast ( DIC ) imaging ( Video 1 ) , significantly earlier than previously described ( Navaratnam et al . , 1986; Nishii and Shibata , 2006 ) . These foci generally contracted at the same rate , indicative of either synchronization or a shared intrinsic beat rate for nascent cardiomyocytes ( Figure 2—figure supplement 1 ) . The earliest cardiac contractions had a frequency of approximately 30 beats per minute ( BPM ) , which increased significantly by stage 3 ( after approximately 5 hr ) to around 60 BPM ( Figure 2A ) . Contractile activity requires cytosolic Ca2+ flux , therefore , to investigate the earliest manifestations of Ca2+ handling within the cardiac crescent , we loaded embryos with the fluorescent Ca2+ indicator Cal-520 followed by live imaging using confocal microscopy . From stage 1 crescents until late stage 3 , lateral propagation of Ca2+ transients was observed across the embryonic midline ( Figure 2B ) , even through regions of non-contractile tissue ( Video 2 ) . As the cardiac crescent starts to fuse to form the LHT , the transients switch from a lateral to a more caudal-rostral propagation ( data not shown ) . At stage 0 spontaneous asynchronous Ca2+ oscillations ( SACOs ) were observed within the forming crescent , in the absence of contractile activity and sarcomeric banding of cTnT ( Figure 2C; Video 3 ) . SACOs were observed in all stage 0 cardiac crescents imaged ( n = 35 ) , propagated within individual cells ( Figure 2D’’ ) and displayed a range of Ca2+ dynamics with variable frequencies and durations ( Figure 2D’ , Video 4 ) . Compared to Ca2+ transients at later stages ( Figure 2A ) SACOs were significantly slower , with fluorescence reaching peak intensity between 0 . 79 and 11 . 9 s and decreasing with a similar slow rate of efflux ( Figure 2—figure supplement 1C ) . During a ~20 s recording period we observed only 10 . 3 ± 0 . 7 individual SACOs per embryo ( n = 35 ) occurring in different sites . Consecutive SACOs in the same site were rarely observed within the 20 s imaging window and , therefore , we conclude that SACOs in individual cells occur at a frequency < 3 bpm . The appearance of Ca2+ transients in the embryonic heart prior to beating ( Figure 2A , B ) is consistent with the idea that Ca2+ signalling within early cardiac progenitors may be important to promote sufficient differentiation for subsequent contractile function ( Mesaeli et al . , 1999; Li et al . , 2002 ) . 10 . 7554/eLife . 17113 . 005Video 1 . Representative movie of beating regions in the cardiac crescent at Stage 1 . DIC imaging of a stage 1 embryo highlighting beating regions in the lateral regions ( dotted circle ) of the developing cardiac crescent . Acquisition was performed at 10 frames per second ( fps ) with a 20x objective and movie played at 8 fps . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 00510 . 7554/eLife . 17113 . 006Figure 2 . Initiation of contraction begins within the forming cardiac crescent and is preceded by spontaneous asynchronous Ca2+ oscillations during heart development . Quantitative analysis from the onset of cardiac contraction at stage 1 of crescent formation to formation of the LHT ( see Figure 1 and Supplementary file 1a for morphological staging . Stage 1 , n = 12; stage 2 , n = 8; stage 3 , n = 10; LHT , n = 7 ) , revealed a significant increase in heart rate from stages 2 to 3 ( A ) . Ca2+ signal following Cal520 loading of stage 1 embryos revealed lateral propagations of transients across the crescent that correlated with the onset of beating at stages significantly earlier than previously described ( B ) . Ca2+ signal following Cal-520 loading of stage 0 embryos revealed spontaneous asynchronous Ca2+ oscillations in individual cells prior to beating , highlighted by white arrows ( C ) , temporal maximum intensity projection of Cal-520 fluorescence ( MaxFluo ) over a period of 30 s . Higher resolution imaging of stage 0 single cell Ca2+ oscillations represented as a temporal maximum intensity projection over a period of 100 s ( D ) revealed variation in SACO transient size and frequency ( D’ ) and could be observed slowly propagating throughout cells ( D’’ ) . Scale bars: B , C , 100 μm , D , 20 μm . Statistics: ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 00610 . 7554/eLife . 17113 . 007Figure 2—figure supplement 1 . Discrete foci either side of the embryonic midline beat at the same rate within the stage 1 cardiac crescent . DIC image of a stage 1 embryo with foci of beating highlighted on the left ( red arrowhead ) and right ( black arrowhead ) of the embryonic midline and cardiac crescent outline highlighted by white dashed line ( A ) . Discrete foci on the left and right of lateral region of the crescent beat at the equivalent frequency ( B ) , indicative of synchronization across the midline or intrinsic pacing . hf , head folds; em , embryonic midline . ( C ) Single cell Ca2+ oscillations were observed in all stage 0 cardiac crescents ( n = 35 ) and varied in time to peak ( TTP ) and time to ½ maximal ( TT1/2M ) fluorescent intensity ( n = 145 cells ) , revealing variable Ca2+ wave duration as well as differences in speed of Ca2+ influx and efflux . All scale bars 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 00710 . 7554/eLife . 17113 . 008Figure 2—figure supplement 2 . Principal component analysis of temporal gene expression profiles to cluster embryonic stage with stage of ESC-derived cardiomyocyte differentiation . Principal component analysis ( PCA ) comparing embryonic stages E7 . 5-E8 . 5 ( n = 5 per stage ) and days 0–14 ( n = 5 per stage ) of ESC induced cardiomyocyte differentiation ( embryoid bodies; EBs; A ) . Groups were calculated by hierarchical clustering: E7 . 5 embryos clustered weakly with day 4/5 EBs , E8 . 0 embryos clustered with day 6/7 EBs and E8 . 5 embryos clustered with day 14 EBs as an outlier ( B ) . Temporal gene expression profiles were assessed on whole embryos by qRT-PCR: significantly increased expression of Myh6 ( B ) , Tnnt2 ( C ) and Mef2c ( D ) was evident from E7-5-E8 . 5 ( n = 5 per stage ) , coincident with increased cardiomyocyte differentiation . In contrast Mesp1 was significantly down-regulated from E7 . 5 following early cardiac specification ( E ) . Comparative gene expression across the time course of ESC differentiation ( n = 5 per stage ) , revealed equivalent trends of increased Myh6 ( F ) , Tnnt2 ( G ) and Mef2c ( H ) from days 4 through 14 , consistent with cardiomyocyte differentiation and embryonic stages E7 . 5-E8 . 5 . Mesp1 was similarly downregulated from day 4 of differentiation ( I; as mapped against E7 . 5 from the PCA A ) and this followed Brachyury expression indicating ( lateral plate ) mesoderm formation ( J ) and loss of pluripotency , as demarcated by a down-regulation of Pou5f1 ( encoding Oct-4 ) from day 4 ( K ) . Ca2+ handling genes were expressed during the time course of ESC differentiation , with Slc8a1 encoding NCX1 ( L ) expressed from day 4 and significantly increased from day six , Cacna1c encoding LTCC ( M ) and Ryr2 ( encoding the Ryanodine Receptor; N ) significantly increased from day 7 . All error bars are mean ± S . E . M . Statistics: one-way ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 00810 . 7554/eLife . 17113 . 009Figure 2—figure supplement 3 . Cardiomyocyte formation , onset of beating and Ca2+ transients are evident by day 7 of ESC-derived cardiomyocyte differentiation . Immunocytochemistry revealed the expression of sarcomeric α-actinin in Nkx2 . 5 positive regions , indicating formation of ESC-derived cardiomyocytes ( A ) ; higher magnification images of α-actinin filaments revealed the formation of sarcomeres ( B ) . Beating was first observed in day 7 ( D7 ) ESC-derived cardiomyocytes . Rate of beating was assessed using bright field movies and calculated at day 7 ( n = 176 EBs ) and day 14 ( n = 180 EBs ) , revealing a significant increase between the two time points ( C ) . Comparative beat rates were observed in ESC derived cardiomyocytes and stage 1 embryos ( n = 12 ) . Cal520 Ca2+ imaging of day 7 ESC-derived cardiomyocytes revealed large fast propagating Ca2+ waves ( D ) as well as slow Ca2+ oscillations in isolated single cells ( E ) . Representative Cal-520 traces for the time series shown in D ( F ) and E ( G ) . Scale bars: A , D , 100 μm; B , 25 μm; E , 10 μm . All error bars are mean ± standard deviation . Statistics: ANOVA and Tukey test for multiple comparisons ( p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 00910 . 7554/eLife . 17113 . 010Video 2 . Representative movie of a Ca2+ transient at Stage 1 . Confocal time-lapse of a stage 1 embryo loaded with Cal-520 . Cal-520 emission ( rainbow ) was captured simultaneously with DIC imaging ( gray ) . Embryo the same as that shown in Figure 2D . Acquisition was performed at 10 fps with a 40x water immersion objective . Background fluorescence was removed by subtracting the signal at a resting phase . Movie played at 8 fps to better show the propagation of the Ca2+ transient . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 01010 . 7554/eLife . 17113 . 011Video 3 . Representative movie of SACOs at Stage 0 . Confocal time-lapse of a stage 0 embryo loaded with Cal-520 . Cal-520 emission ( rainbow ) was captured simultaneously with DIC imaging ( gray ) . Acquisition was performed at 10 fps with a 20x objective . Background fluorescence was removed by subtracting the signal at a resting phase . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 01110 . 7554/eLife . 17113 . 012Video 4 . Representative high-resolution movie of SACOs at Stage 0 . Confocal time-lapse of a stage 0 embryo loaded with Cal-520 . Cal-520 emission ( rainbow ) was captured simultaneously with DIC imaging ( grey ) . Acquisition was performed at 10 fps with a 20x objective . Movie playback is at 5x original speed . Background fluorescence was removed by subtracting the signal at a resting phase . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 012 We next used an embryonic stem cell ( ESC ) -derived model of cardiomyocyte development to complement the embryo studies and provide further insight into the stages of cardiac contraction coincident with cardiomyocyte specification and differentiation . While ESC-derived cardiogenesis is aligned with stages of mesoderm induction , pre-cardiac mesoderm , cardiomyocyte specification and differentiation based on temporal patterns of gene expression ( reviewed in Kattman and Keller , 2007; Willems et al . , 2009; Van Vliet et al . , 2012 ) , this has not been rigorously mapped onto embryonic stages of heart development . We , therefore , performed a principal component analysis ( PCA ) and hierarchical clustering of 12 cardiac related gene expression profiles of whole embryos across embryonic stages E7 . 5 to E8 . 5 , compared with ESC-derived embryoid bodies ( EBs ) from days 0 to 7 inclusive and day 14 of differentiation . Whole embryos were used to reflect the myriad of cell types present in the ESC-cardiomyocyte differentiation assay . Hierarchical gene clusters were evident with the onset of beating at E7 . 5 and day 4 and 5 of EB formation , coincident with cardiac progenitor gene expression ( Figure 2—figure supplement 2A ) . Expression profiles by days 6 and 7 correlated with E8 . 0 ( stage 0 to stage 2 ) when beating was well established in both ESC-derived EBs and the embryonic heart , whereas day 14 was equivalent to the more mature E8 . 5 ( Figure 2—figure supplement 2A ) . Key cardiac genes Myh6 , Tnnt2 and Mef2c revealed comparable trends of increased expression over time of differentiation in both EBs and embryos ( Figure 2—figure supplement 2 ) . The cardiac specification program , characterized by expression of Mesp1 , was evident by day 4 in EBs ( Figure 2—figure supplement 2I ) following Brachyury expression , an indicator of mesoderm formation , and loss of pluripotency markers such as Pou5f1 ( encoding Oct-4; Figure 2—figure supplement 2J–K ) . Mesp1 up-regulation in EBs was consistent with expression at E7 . 5 in the embryo ( Figure 2—figure supplement 2E ) and preceded beating at day 6 and 7 . These later stages revealed an up-regulation of Ca2+-handling genes such as Slc8a1 , Cacna1c and Ryr2 ( Figure 2—figure supplement 2L–N ) . Sarcomere assembly in ESC-derived cardiomyocytes accompanied the onset of beating at day 7 ( Figure 2—figure supplement 3A , B ) and , interestingly , the rate of beating was comparable at the outset ( day 7 of differentiation ) with that observed in the developing heart at stage 1 ( Figure 2—figure supplement 3C ) , consistent with an intrinsic rate for early cardiomyocytes contractions . Furthermore , Ca2+ transients were observed in day 7 EBs as both large propagating waves ( Figure 2—figure supplement 3D , F; Video 5 ) and within small regions of cells prior to beating , similar to that observed in the stage 0 embryonic heart ( Figure 2—figure supplement 3E , G; Video 6 ) . 10 . 7554/eLife . 17113 . 013Video 5 . Representative movie of a propagating Ca2+ transient at day eight of ESC cardiomyocyte differentiation . Confocal time-lapse of a day eight EB loaded with Cal-520 . Cal-520 emission ( rainbow ) was captured simultaneously with DIC imaging ( grey ) . Acquisition was performed at 10 fps with a 20x Objective . Background fluorescence was removed by subtracting the signal at a resting phase . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 01310 . 7554/eLife . 17113 . 014Video 6 . Representative movie of SACOs at day six of ESC cardiomyocyte differentiation . Confocal time-lapse of a day six EB loaded with Cal-520 . Cal-520 emission ( rainbow ) was captured simultaneously with DIC imaging ( grey ) . Acquisition was performed at 10 fps with a 20x objective . Background fluorescence was removed by subtracting the signal at a resting phase . Scale bar: 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 014 ECC components , and specifically NCX1 , have not previously been implicated in the initiation event of cardiac contraction , nor investigated at the earliest stages of heart development coincident with the onset of beating . We , therefore , assessed ECC gene expression on embryonic hearts at the different stages defined herein; focusing specifically on Slc8a1 ( encoding NCX1 ) , Cacna1c and Cacna1d ( encoding the LTCC subunits , CaV1 . 2 and CaV1 . 3 respectively; Figure 3A ) , Atp2a2 ( encoding SERCA2 ) , Iptr2 ( encoding InsP3 type 2 channels ) and Ryr2 ( encoding RyR2 ) that collectively are key components of SR Ca2+ regulation ( Figure 3—figure supplement 1A–D; Supplementary file 1b ) . Between E7 . 75 ( as defined by the presence of clear head-folds but not a cardiac crescent ) and stage 0 , Slc8a1 significantly increased 21 fold ( p-value<0 . 001 ) , whilst Cacna1c revealed a modest but non-significant increase of 2 . 5 fold ( p-value>0 . 05 ) and Cacna1d a significant 4-fold increase ( p-value <0 . 01 ) . From stage 1 onwards Slc8a1 , Cacna1c and Cacna1d expression continued to significantly increase until stage 3 ( Figure 3A , B; Figure 3—figure supplement 1A , Slc8a1 , 234-fold; Cacna1c , 38-fold; Cacna1d , 41-fold ) . Between stage 3 and the LHT , expression of Slc8a1 and Cacna1c was maintained whilst Cacna1d significantly decreased ( Supplementary file 1b; p-value<0 . 05 ) . Previous studies have suggested a non-ECC dependent role for the SR during development , implicating both InsP3 and RyR channels ( Méry et al . , 2005; Sasse et al . , 2007; Rapila et al . , 2008 ) . While Ryr2 significantly increased 30 fold between E7 . 75 and stage 0 ( p-value<0 . 001 ) , expression of Atp2a2 only increased 1 . 7 fold ( p-value<0 . 05 ) and Iptr2 did not significantly change ( p-value>0 . 05; Figure 3—figure supplement 1C , D ) . 10 . 7554/eLife . 17113 . 015Figure 3 . The ECC components NCX1 and LTCC are expressed within the early embryonic heart and ESC-derived cardiomyocytes . Analyses by qRT-PCR revealed a significant increase in the expression of Slc8a1 ( encoding NCX1 ) in the heart from E7 . 75 to stage 0 ( A , n = 5 per stage ) and from day 2 of differentiation of ESC-derived cardiomyocytes ( B , n = 5 per stage ) . In contrast expression of Cacna1c ( encoding the LTCC subunit CaV1 . 2 ) , increased at a later stages from stage 0 to stage 1 and from day 4 of ESC-derived cardiomyocyte differentiation ( C , n = 5 per stage ) . Confocal imaging section of a stage 0 embryo following immunostaining for cTnT ( red ) and NCX1 ( green ) , indicated membrane localization of NCX1 within the forming crescent ( D; white arrows in lower panel ) , whereas CaV1 . 2 ( green ) was absent from cTnT+ ( red ) regions at the same stage ( E ) . A maximum intensity projection of day 7 ESC-derived cardiomyocytes revealed complete overlap of staining for cTnT ( red ) and NCX1 ( green; F , 33 stacks ) , whereas CaV1 . 2 ( green ) overlapped in part with cTnT ( red ) but there were also extensive cTnT+/CaV1 . 2- regions ( dotted box ) emphasizing the later requirement for LTCC ( G , 22 stacks ) . Confocal imaging section of a stage 2 embryo following immunostaining for both NCX ( H ) and CaV1 . 2 ( I ) revealed the expression of both proteins at later stages of heart development . cc , cardiac crescent; hf , head folds . Scale bars: D , F 50 μm , H , I 100 μm . All error bars are mean ± S . E . M; Statistics: one-way ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; *p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 01510 . 7554/eLife . 17113 . 016Figure 3—figure supplement 1 . ECC component expression increases during cardiac crescent formation . Analyses by qRT-PCR revealed a significant increase in the expression of the Cacna1d ( encoding the LTCC subunit Cav1 . 3 ) from E7 . 75 to the stage 0 cardiac crescent ( A ) , suggesting both L-type Ca2+ channel subunits Cav1 . 2 and Cav1 . 3 maybe required for cardiac function from stage 1 onwards . Expression of Atp2a2 ( encoding the SR Ca2+ pump SERCA2a; B ) and Ryr2 ( C ) significantly increased between E7 . 75 and stage 0 , whilst Itpr2 ( encoding the IP3 type 2 receptor ) was unchanged over the complete timecourse from E7 . 75 to the LHT ( D ) . All error bars are mean ± S . E . M . Statistics: A–D , one-way ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 01610 . 7554/eLife . 17113 . 017Figure 3—figure supplement 2 . Expression of Slc8a1 and Cacna1c does not increase in the head folds between E7 . 75 and the LHT stages of development . Analysis by qRT-PCR showed that expression of Slc8a1 ( encoding NCX1 was maintained throughout later stages of embryonic development with no significant change at birth ( newborn pups , P0 ) or into adulthood ( A ) , in contrast to Cacna1c ( encoding the LTCC subunit Cav1 . 2; B ) and Ryr2 ( encoding the Ryanodine receptor; C ) which increased significantly at P0 . In the head folds , qRT-PCR revealed a significant increase in the expression of the neural ectoderm marker Sox1 from E7 . 75 to stage 0 ( A ) , however , in contrast expression of both Slc8a1 ( B ) and Cacna1c ( C ) were unchanged over the entire time course . All error bars are mean ± S . E . M , n = 3 , A–C; n = 5 , D–F . Statistics: one-way ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 017 We performed a similar analysis on EB derived cardiomyocytes . Slc8a1 expression increased significantly in EBs prior to Cacna1c ( day 2 versus day 4; Figure 3B , C ) and to a much greater extent with the onset of beating ( 110 fold versus 38 fold; Figure 3B , C ) , suggesting NCX1 might play a more immediate role in the onset of beating . This was supported by the lack of any further increase in Slc8a1 expression from E9 . 5 through to birth ( P0 ) and adulthood ( Figure 3—figure supplement 2 ) , whereas Cacna1c expression fluctuated across later developmental stages and significantly increased post-natally ( Figure 3—figure supplement 1A , B ) with increased maturation . Since the sarcolemmal channel genes revealed the most significant increases over the developmental timecourse , we proceeded to investigate spatiotemporal protein expression of NCX1 and CaV1 . 2 in embryos ( Figure 3D , E ) , and ESC-derived cardiomyocytes ( Figure 3F , G ) . Whilst NCX1 was clearly detectable within the cardiac crescent at stage 0 ( Figure 3D ) , CaV1 . 2 was absent ( Figure 3E ) . Differences in the expression of NCX1 and LTCC prior to beating were not maintained at later stages , after established contraction ( stage 2 onwards ) , when both channels were expressed ( Figure 3H , I ) . In EBs at day 7 of culture , both NCX1 and CaV1 . 2 were expressed , but whilst NCX1 was consistently co-expressed with cTnT ( Figure 3F ) , there were cTnT+ foci that were negative for CaV1 . 2 ( Figure 3G ) . Collectively , this expression data suggests that NCX1 precedes Cav1 . 2 within the developing cardiac crescent at stage 0 , being present prior to and at the onset of both SACOs and beating , whereas Cav1 . 2 became upregulated later from stage 1 onwards . Of note , while NCX1 and CaV1 . 2 signal was detected in other regions of the embryo , notably in the head folds , in contrast to the heart it was not membrane localised in these regions , indicative of non-functional protein . Furthermore qRT-PCR data for isolated head folds did not reveal any significant increases in Slc8a1 or Cacna1c expression , from E7 . 75 through to LHT stages ( p-value>0 . 05; Figure 3—figure supplement 2E , F; Supplementary file 1c ) even though the head folds were maturing , as shown by morphological changes and expression of the neural ectoderm marker Sox1 ( E7 . 75 versus LHT , p-value<0 . 001; Supplementary file 1c ) To functionally assess the roles of the sarcolemmal and SR channels in establishing and maintaining heartbeat , we employed pharmacological blockade on embryos maintained ex-vivo and ESC-derived cardiomyocyte cultures . We inhibited NCX1 using the specific inhibitors CB-DMB ( Secondo et al . , 2009 ) or KB-R7943 ( Kimura et al . , 1999 ) , the LTCC using nifedipine ( McDonald et al . , 1994 ) and Ryanodine together with 2-APB to simultaneously block RyR and InsP3 , similar to that described previously ( Sasse et al . , 2007; Rapila et al . , 2008 ) . Treated embryos were imaged for contractile activity by DIC imaging and Ca2+ transients were recorded in parallel with confocal fluorescence imaging of Cal-520 . Acute treatment ( for a maximum of 30 min ) with NCX inhibitors ( CB-DMB and KB-R7943 ) or LTCC inhibitor ( Nifedipine ) , affected the embryos in a stage dependent-manner ( Figure 4 ) . Both NCX1 and the LTCC were required for Ca2+ transients during stages 1 and 2 ( Figure 4A ) , and their inhibition led to an initial confinement of Ca2+ transients to one side of the crescent ( within approximately 5 min of treatment , which persisted through to 15 min ) followed shortly afterwards by complete loss of Ca2+ signal ( Figure 4C , D ) . At stage 3 and later , only the LTCC was required for Ca2+ transient generation ( Figure 4B ) . Whilst NCX1 can function in both forward ( Ca2+ efflux from the cell ) and reverse ( Ca2+ influx into the cell ) modes , our data suggested that NCX1 in the early cardiac crescent was functioning in reverse mode , as KB-R7943 , that is reported to specifically inhibit reverse mode NCX1 function ( Hoyt et al . , 1998; Iwamoto , 2004 ) , recapitulated the results with CB-DMB and acted as a control for off-target effects of the latter . These data suggests that reverse mode NCX function may contribute to Ca2+ transient generation at the earliest stages of cardiac contraction , potentially via inward Ca2+ flux . The inhibitor experiments were repeated on day 7 and day 14 EBs . The relative effects of CB-DMB , KB-R7943 and nifedipine were consistent with those observed in stages 1 and 2 of the embryonic heart ( Figure 4E–G ) . Treatment with Ryanodine + 2-APB only affected more mature embryos that had already undergone both cardiac looping and embryonic turning ( Figure 4—figure supplement 1A ) and did not prevent Ca2+ transients within cardiac crescents at stage 3 ( Figure 4—figure supplement 1B ) . This suggest that a functional SR is only required once cardiac looping has been fully initiated , more than 18 hr later than the first observable SACOs within the cardiac crescent . 10 . 7554/eLife . 17113 . 018Figure 4 . Both NCX1 and LTCC are required for Ca2+ transients associated with beating cardiomyocytes during cardiac crescent development . Inhibition of Ca2+ transients upon treatment of stage 1-LHT embryos with either NCX1 inhibitors CB-DMB , KB-R7943 or the LTCC inhibitor nifedipine , relative to DMSO control after 5 , 15 and 30 min of drug application ( A , B ) . Inhibition of NCX1 with either CB-DMB ( 20 μM ) or KB-R7943 ( 30 μM ) affected only stage 1 and 2 embryos ( A ) , whereas inhibition of LTCC with nifedipine ( 10 μM ) effected both stages 1 and 2 and the later stage 3/LHT ( B; stage1/2: DMSO , n = 5; CB-DMB , n = 9; KB-R7943 , n = 6; nifedipine , n = 6; stage3/LHT: DMSO , n = 9; CB-DMB , n = 19; KB-R7943 , n = 8; nifedipine , n = 8 ) . Time series of Ca2+ transients on stage1-2 embryos at different time points of either CB-DMB or nifedipine treatment , revealed a confinement to the right side of the embryo prior to complete block ( C , D ) . ESC-derived cardiomyocytes at different days of differentiation were treated with the same channel blockers: inhibition of NCX1 with CB-DMB ( 10 μM ) significantly reduced contractions only in day 7 cardiomyocytes , whereas KB-R7943 ( 30 μM ) affected cardiomyocytes at both day 7 and 14 ( E ) . Inhibition of LTCC with nifedipine ( 10 μM ) significantly reduced contractions in both day 7 and 14 cardiomyocytes and to a much greater extent than KB-R7943 at the later stage ( F; day 7: DMSO , n = 38; CB-DMB , n = 36; KB-R7943 , n = 7; nifedipine , n = 10; day 14: DMSO , n = 25; CB-DMB , n = 36; KB-R7943 , n = 15; nifedipine , n = 15 ) . Time series of Ca2+ transients on day 7 ESC-derived cardiomyocytes at different time points of either CB-DMB or nifedipine treatment , revealed a confinement prior to complete block ( F , G ) , equivalent to that observed in the treated embryos ( C , D ) . Treatment of stage 0 embryos prior to the onset of beating with CB-DMB ( 20 μM ) and KB-R7943 ( 30 μM ) resulted in inhibition of slow asynchronous Ca2+ transients after 15 min application relative to baseline ( H , I; DMSO , n = 10; CB-DMB , n = 8; KB-R7943 , n = 9 ) whereas treatment with nifedipine ( 10 μM ) had no discernible effect on the slow transients ( H , J; nifedipine n = 8 ) , supporting the earlier role for NCX1 in initiating Ca2+ handling and beating . All scale bars 100 μm . Statistics: Freeman-Halton extension of Fisher exact probability test for embryos for embryos; Chi-square test with Bonferroni correction for ESCs ( p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 01810 . 7554/eLife . 17113 . 019Figure 4—figure supplement 1 . Contribution of the sarcoplasmic reticulum ( SR ) to Ca2+ transients does not occur until looping stages of heart development . The functional role of SR derived Ca2+ was assessed using pharmacological inhibition of RyR using Ryanodine and IP3Rs with 2-APB . Inhibition of RyR and IP3 receptors prevented beating in the looping hearts of turned embryos relative to DMSO ( Ryanodine + 2-APB , n = 20; DMSO , n = 20; A ) and significantly after formation of the LHT . In contrast , application of ryanodine and 2-APB did not inhibit contraction of looping hearts in non-turned embryos ( Ryanodine + 2-APB , n = 24; DMSO , n = 13 ) or block Ca2+ transients in stage 3 cardiac crescents or LHT ( Ryanodine + 2-APB , n = 13; DMSO , n = 9 ) , as shown in representative Ca2+ traces ( B , scale bar 5 s ) . All error bars are mean ± S . E . M . Statistics: A , Freeman-Halton extension of Fisher exact probability test ( p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 019 Taking into account the earlier upregulation of NCX1 relative to Cav1 . 2 at stage 0 ( Figure 3A ) , we tested the possibility that NCX1 might be pivotal in initiating the SACOs observed in the early cardiac crescent ( Figure 2C ) . Treatment of stage 0 embryos with CB-DMB and KB-R7943 resulted in a 90% decrease in the number of cells with SACOs after drug application when compared to baseline . This was significantly greater than following application of nifedipine to inhibit LTCC , or the DMSO control ( CB-DMB , 90% inhibition + 6 . 6 , mean + SEM , n = 10 embryos , p-value <0 . 01; KB-R7943 , 91% inhibition + 7 . 0 , n = 9 embryos , p-value <0 . 001; nifedipine , 39% inhibition + 11 . 4; n = 8 embryos , p-value=0 . 44; DMSO , 29% inhibition + 7 . 6 , n = 10; Figure 4H–J; Supplementary file 1d ) . These data suggest that NCX1 is required for establishing the earliest pre-contractile SACOs . Chronic treatment of Eomes-GFP ESC cultures ( ESCs containing a knock-in of a GFP reporter into the endogenous Eomes locus ( Arnold et al . , 2009 ) ; marking nascent mesoderm ( Ciruna and Rossant , 1999 ) , with CB-DMB or Nifedipine , from days 0 to 14 of differentiation , resulted in a reduction in the percentage of beating EBs at day 14 , down to 22% and 52% ( Percentage beating EBs; DMSO , 76 . 53% ( n = 196 ) ; CB-DMB , 22 . 37% , p-value<0 . 001 ( n = 152 ) ; Nifedipine , 52 . 11% , p-value<0 . 001 [n = 71] ) in the presence of CB-DMB and nifedipine respectively , compared to controls ( Figure 5A ) , whilst not effecting cell number ( Figure 5—figure supplement 1A–C ) . The reduction in beating EBs was supported by a down-regulation of key cardiac genes Myh6 ( mean+ SEM: 0 . 57 ± 0 . 058; p-value<0 . 05; n = 6; Figure 5B ) and Tnnt2 ( 0 . 66 ± 0 . 095; p-value<0 . 05; n = 6; Figure 5B ) specifically following treatment with CB-DMB relative to control DMSO treatment . The decrease in cardiac gene expression associated with CB-DMB treatment could be observed from day 4 of Eomes-GFP ESC differentiation , at the initiation of cardiomyocyte differentiation ( Figure 5—figure supplement 1D–N ) . In an Nkx2 . 5-EGFP ESC line ( ESCs containing a vector expressing eGFP under the control of a murine Nkx2 . 5 promoter and regulatory region; marking cardiac progenitor cells [Wu et al . , 2006] ) this was associated with loss of GFP+ cardiac progenitors by day 14 ( Figure 5—figure supplement 2A–E ) and the significant down-regulation of key cardiac markers following treatment with CB-DMB relative to control DMSO treatment ( Figure 5—figure supplement 2F ) . Importantly not all cardiac genes were effected by chronic exposure to the channel inhibitors including Slc8a1 , Cacna1c and Camk2d downstream of Ca2+ signalling ( Figure 5—figure supplement 3A ) , suggesting that the inhibitors did not simply have a global negative effect on gene expression or cardiomyocyte survival . 10 . 7554/eLife . 17113 . 020Figure 5 . Influx of Ca2+ and CaMKII signalling are required for early and late cardiac gene expression and crescent formation . Following chronic exposure of embryoid bodies for 14 days to CB-DMB ( 1 μM ) there was a significant decrease in the incidence of beating from 80% to 22% as compared to a reduction to 52% following nifedipine treatment ( 10 μM ) ( A; DMSO , n = 196; CB-DMB , n = 152; nifedipine , n = 71 ) . Prolonged exposure resulted in a significant decreases in mature cardiomyocyte genes , Mef2c , Myh6 , Myh7 and Tnnt2 ( B ) gene expression following treatment with CB-DMB ( n = 6 ) but not with nifedipine ( n = 6 ) . EBs cultured for 14 days in different concentrations of extracellular Ca2+ ( 1 . 8 mM is the normal culture medium concentration ) revealed significantly decreased incidence of beating following culture with reduced Ca2+ when assessed in media containing 1 . 8 mM Ca2+ ( C; 1 . 8 mM , n = 74; 1 . 0 mM , n = 61; 0 . 1 mM n = 80 ) . Cultured ESC derived-cardiomyocytes exposed to NCX1 inhibitors effected downstream Ca2+ signalling via alterations in the levels of phosphorylated CaMKII ( pCaMKII; D ) . pCamKII to total CamKII ratio was decreased in the presence of 1 μM CB-DMB as compared to 10 μM nifedipine and DMSO ( D’; n = 3 ) . E7 . 5 embryos were dissected and cultured for 12 hr in media containing either DMSO , nifedipine ( 10 μM ) or CB-DMB ( 3 μM ) and stained for cTnT and Nkx2 . 5 ( E; maximum intensity projections , 30 stacks each ) . Embryos developed normally in culture , as indicated by head fold formation , coalescence of the cardiac crescent and addition of somites ( not shown ) . Embryos cultured in CB-DMB were delayed in terms of cardiac crescent formation and show a weaker cTnT signal compared to either DMSO alone or nifedipine-treated ( E; number of affected embryos: DMSO – 1/7; CB-DMB – 7/8; Nifedipine – 1/6 ) . Cultured E7 . 5 embryos in the presence of either CB-DMB or nifedipine for 12 hr , revealed that CB-DMB significantly down-regulated the expression of both early Nkx2 . 5 and Mef2c and late Tnnt2 ( F , G ) cardiac genes , coincident with impaired cardiac crescent formation , whereas nifedipine-treatment did not appear to have any effect on cardiac gene expression ( H ) . All error bars are mean ± S . E . M . Statistics: B , D , G and H: one-way ANOVA and Tukey test for multiple comparisons; A , C: Chi-square test with a Bonferroni correction for multiple comparisons ( p<0 . 05; p<0 . 01; p<0 . 001 ) . CC , Cardiac crescent; HF , Head folds; EM , Embryonic midline . All scale bars 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 02010 . 7554/eLife . 17113 . 021Figure 5—figure supplement 1 . Inhibition of NCX1 or Cav1 . 2 did not overtly effect embryoid body formation or cell outgrowth during Eomes-GFP ESC differentiation . Analysis of Eomes-GFP ESC differentiation using qRT-PCR at multiple timepoints revealed the CB-DMB inhibition blocks the expression of key cardiac genes during the formation of cardiomyocytes ( Myh6 ( A ) , Mef2c ( B ) , Nkx2 . 5 ( C ) ; n = 3 ) . Treatment of Eomes-GFP ESCs with 1 μm CB-DMB or 10 μm nifedipine did not prevent EB formation or cell outgrowth . Bright field images confirmed that cultures were grossly unaffected by CB-DMB or nifedipine at all stages of differentiation ( D–L ) . The number of cells per EB were also unaffected at both day 4 ( DMSO , 20744 cells/EB ( n = 9 ) ; CB-DMB , 19 , 184 cells/EB ( n = 7 ) ; nifedipine , 26 , 359 cells/EB ( n = 3 ) ; L ) and day 7 ( n , DMSO , 46 , 800 cells/EB ( n = 10 ) ; CB-DMB , 44 , 922 cells/EB ( n = 8 ) ; nifedipine , 50761 cells/EB ( n = 3 ) ; M ) . Scale bars: 500 μm . All error bars are mean ± S . E . M . Statistics: A–C , two-way repeated measures ANOVA and Tukey test for multiple comparisons; M–N , one-way ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 02110 . 7554/eLife . 17113 . 022Figure 5—figure supplement 2 . NCX1 blockade from the outset of ESC differentiation reduces the incidence of Nkx2 . 5+ cardiac progenitors . Control Nkx2 . 5-EGFP EBs ( DMSO-treated ) reveal extensive GFP+ cardiac differentiation after 14 days in culture ( A , B ) as compared to a loss of GFP+ cells when treated with CB-DMB ( day 0–14 ) ( C , D ) . Fluorescence ( A , C ) and bright field ( B , D ) images shown for each treatment group; bright field confirmed the cultures were grossly unaffected by CB-DMB treatment . To quantify levels of GFP expression , the percentage area of Nkx2 . 5-EGFP regions was calculated from epifluorescent images using Image J and revealed a significant decrease in the percentage area of GFP+ cardiac differentiation ( n = 9; E ) . This was also observed in gene expression as assessed by qRT-PCR , with significant decreases in cardiomyocyte genes , Nkx2 . 5 , Mef2c , Tnnt2 and Myh6 ( F ) , following treatment with CB-DMB ( n = 5 ) . All error bars are mean ± S . E . M . Statistics: one-way ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; p<0 . 001 ) . Scale bar: A , 500 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 02210 . 7554/eLife . 17113 . 023Figure 5—figure supplement 3 . Late administration of NCX1 and LTCC inhibitors did not affect gene expression in ESC-derived cardiomyocytes whereas NCX1 inhibition reduced Slc8a1 in E7 . 5 embryos ex vivo . Expression of Slc8a1 , Cacna1c or Camk2d ( A ) , as determined by qRT-PCR on day 14 ESC-derived cardiomyocytes were unaffected by culture in the presence of either DMSO , CB-DMB or nifedipine ( n = 6 ) . Activation of CaMKII as determined by the ratio of pCaMKII/total CaMKII levels was unaltered in ESC-derived cardiomyocytes after four days in culture in the presence of either DMSO , CB-DMB ( 1 μM ) or nifedipine ( 10 μM ) ( B , n = 3 ) . In E7 . 25 embryos cultured for 12 hr in the presence of either DMSO , CB-DMB or nifedipine , there was a significant decrease in Slc8a1 in embryos cultured in CB-DMB ( C , n = 3 ) consistent with loss of crescent cells ( see Figure 5F ) , but no change with treatment of nifedipine ( D , n = 4 ) consistent with the lack of effect on crescent formation ( see Figure 5F ) . Cacna1c remained unchanged in the presence of either inhibitor ( C , D ) , suggesting non-cardiac ( neural fold ) expression in CB-DMB treated embryos ( see Figure 3 ) . All error bars are mean ± S . E . M . Statistics: one-way ANOVA and Tukey test for multiple comparisons ( p<0 . 05; p<0 . 01; *p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17113 . 023 Given the proposed role for inward Ca2+ via NCX1 , we cultured ESCs in media containing reduced concentrations of Ca2+ ( 0 . 1 mM and 1 . 0 mM ) relative to the normal level of 1 . 8 mM ( Figure 5C ) . At the end of the differentiation protocol ( 14 days ) we accessed the percentage of beating EBs after returning to media with normal levels of Ca2+ for 2 hr . Culture in 0 . 1 mM Ca2+ resulted in an inhibitory effect on EB beating to the same extent ( percentage beating EBs; 1 . 8 mM , 78 . 38% , n = 74; 0 . 1 mM , 28 . 75% , n = 80; p-value<0 . 001 ) as with CB-DMB treatment ( Figure 5C ) and culture in 1 . 0 mM resulted in an equivalent inhibitory effect to that following treatment with nifedipine ( Figure 5C; 1 . 0 mM , p-value<0 . 01; n = 61 ) . This suggested that influx of external Ca2+ is required for cardiomyocyte maturation/beating and supports a potential role for NCX1 working in reverse mode to bring Ca2+ into early cardiac progenitors . We next investigated whether Ca2+ handling prior to , and concurrent with , the earliest contractile function impacted on known Ca2+-signalling pathways to activate foetal gene expression , as has been reported during adult pathological hypertrophy ( Molkentin et al . , 1998 ) . Calmodulin-dependent kinase II ( CaMKII ) is a key component of Ca2+ and calcineurin signalling which directly impacts on downstream hypertrophic gene expression by induction of Mef2c ( Wu et al . , 2006; Molkentin et al . , 1998; Passier et al . , 2000; Zhang , 2007 ) . Following CB-DMB-induced NCX1 inhibition at day 7 of EB differentiation , activated phospho-CaMKII ( pCaMKII ) levels were significantly reduced compared to either DMSO control or nifedipine treatment ( mean ± SEM: DMSO , 1 . 18 ± 0 . 13; CB-DMB , 0 . 69 ± 0 . 15; nifedipine , 1 . 13 ± 0 . 06 . DMSO vs . CB-DMB , p-value<0 . 01; CB-DMB vs . nifedipine , p-value<0 . 05; n = 3; Figure 5D , D' ) . NCX1 inhibition at day 4 , a stage prior to cardiac progenitor specification , had no effect on pCaMKII levels ( Figure 5—figure supplement 3B , B' ) . NCX1 inhibition reduced Ca2+ influx and decreased activation of CaMKII accompanied by significantly reduced expression of Mef2c and Myh7 relative to DMSO control ( mean ± SEM: Mef2c , 0 . 71 ± 0 . 13; Myh7 , 0 . 55 ± 0 . 18; p-value<0 . 05; n = 6 ) which , whilst previously associated with pathological cardiac hypertrophy , contribute here towards ( physiological ) cardiomyocyte differentiation ( Figure 5B ) . In contrast , nifedipine treatment resulted in decreased Myh7 , whereas Mef2c remained unaffected at early stages ( Figure 5B ) . To determine the effect of NCX1 blockade on the subsequent development of the heart , we cultured embryos isolated at E7 . 25 ( the onset of head fold formation and pre-cardiac crescent stage ) in the presence of DMSO , CB-DMB or Nifedipine for 12 hr ( Figure 5E ) . Whilst the embryos remained viable during the culture period , initiation of heart development was impaired in those treated with CB-DMB compared to DMSO controls ( Figure 5E ) . This was in contrast to treatment with Nifedipine alone , in which the crescent developed normally and progressed to an equivalent stage as DMSO treated embryos ( Figure 5E ) . More specifically embryos cultured in the presence of CB-DMB had reduced expression of cTnT compared to embryos cultured in other conditions . However , most cells were still positive for Nkx2 . 5 , suggesting delayed or impaired differentiation likely accounted for the failure of crescent formation under conditions of NCX1 blockade . The relative effects of NCX1 and LTCC inhibition on developmental progression and crescent formation were supported by corresponding gene expression data from cultured embryos ( Figure 5F , G ) . While Tnnt2 was down-regulated in both CB-DMB and Nifedipine treated embryos , Nkx2 . 5 was only down-regulated in CB-DMB treated embryos ( Figure 5F , G ) . Equally Slc8a1 , was down-regulated exclusively in the presence of CB-DMB ( Figure 5—figure supplement 3C ) , whereas Cacna1c was unaltered in the presence of either inhibitor ( Figure 5—figure supplement 3C , D ) . These data collectively support a role for NCX in cardiomyocyte differentiation and crescent formation . Previous studies have attempted to investigate how cardiac function develops within the early embryo . Whilst these studies are informative they rely on the dissociation and culture of embryonic cardiomyocytes to facilitate physiological measurements resulting in the loss of critical spatial and temporal information regarding Ca2+-handling and downstream changes in gene expression and morphology . Using a staging method based on morphological landmarks ( stages 0–3 ) , we characterized in detail the in vivo progression of physiological activity during early heart development . The stage series defined correlates with gradual expression of several cardiac related genes and sarcomere assembly . We observed spontaneous asynchronous Ca2+ oscillations ( SACOs ) at stage 0 in the developing cardiac mesoderm , before any detectable cardiac contractions . These transients appeared sporadically in individual cells within the forming cardiac crescent and did not appear to be synchronized . At stage 1 , periodic Ca2+ transients began to be propagated laterally through the cardiac crescent , and traversed the midline of the crescent where there were no visible contractions . This was a rapid and dynamic process occurring through stages 0–3 that involved sarcolemmal ion channel and exchanger function . NCX1 was exclusively required for SACOs at stage 0 , whereas both NCX1 and LTCC appeared to play equally important roles at stage 1 and 2 . NCX1 was no longer required from stage 3 onwards when the LTCC channels maintained Ca2+ transients . We observed no contribution of the SR in regulating Ca2+ transients within the cardiac crescent , as assessed by simultaneous blockade of RyR2 and InsP3 . However , consistent with previous studies ( Sasse et al . , 2007; Rapila et al . , 2008 ) , our data did show that the SR becomes functional at later stages during cardiac looping ( ~E8 . 5 ) . This may represent a period in which SR Ca2+ filling is required before a threshold for Ca2+ release from the SR can occur , as has been proposed in the adult setting ( Stokke et al . , 2011 ) . Overall our data reveals that during the time frame from the earliest cardiac morphogenesis ( crescent maturation ) to looping of the heart ( approximately 18 hr ) there is a rapid transition in the mechanism by which the intracellular concentration of Ca2+ is elevated , based on distinct channel and exchanger function . The observation of spontaneous asynchronous calcium oscillations ( SACOs ) within the forming cardiac crescent was a surprising finding that , to the best of our knowledge , has not previously been reported in any type of excitable cell . The specific role of SACOs is currently unknown and we hypothesise that they are required in cells that need to optimally activate Ca2+-dependent signalling ( via the CAMKII pathway ) in order to up-regulate genes necessary for further differentiation and morphogenesis . This would explain why SACOs are present much earlier than complete sarcomere assembly . An alternative hypothesis is that SACOs are a by-product of cells that are already committed to specific cardiac lineages and , therefore , arise with the expression of specific channels required for future function . This could explain the variation in duration and frequencies of SACOs observed within the same embryo . At this point , quite how SACOs become synchronised transients is unknown . We speculate that release of a Ca2+-dependent signal from a 'pacemaker' cell may entrain neighbours to have synchronised transient periodicity . To this end we observed highly variable Ca2+ periodicity but also regions containing cells of similar periodicity ( Video 2 ) . Since blockade of NCX1 prior to the formation of the cardiac crescent , and chronic treatment in ESCs , leads to impaired cardiac differentiation it is also possible that SACOs may be present in mesoderm cells earlier than reported here , which we were not able to image due to the limitations of the current experimental set-up . In the embryonic heart , Ca2+ handling is assigned to sequential roles for the NCX and LTCC channels , whereby NCX has been assumed to compensate , at least in part , for the rudimentary ( non-functional ) SR in the developing mouse heart ( Conway et al . , 2002 ) . In zebrafish embryos it has recently been demonstrated that Ca2+ handling and not contraction per se is essential for regulating cardiomyocyte development ( Andersen et al . , 2015 ) . Previous studies characterised functional expression of NCX from E8 . 5-E9 . 5 ( post-LHT formation ) in mouse ( Linask et al . , 2001; Reppel et al . , 2007 ) coincident with cardiac looping and significantly later than the onset of the first heat beat described herein . NCX1 knock-out mice have been generated by multiple groups , with conflicting results in regards to the phenotype , extent of mutant heart development and the stage at which embryonic lethality occurs ( Wakimoto et al . , 2000; Koushik et al . , 2001; Reuter et al . , 2002; Cho et al . , 2003 ) . This suggests that loss of NCX1 may be compensated for at the level of early cardiomyocyte specification , differentiation and contraction . In mammals there are three different Na+-Ca2+ exchangers ( NCX1 , NCX2 and NCX3 ) , and it has been previously demonstrated that these three exchangers share similar physiological properties ( Linck et al . , 1998; Lytton , 2007 ) . Furthermore , NCX1 has several splicing variants with exon 1 being mutually exclusive to exon 2 ( Lytton , 2007; Quednau et al . , 1997 ) . It is , therefore , possible that either a different NCX or an alternative splice variant will compensate for loss of the NCX1 variant targeted in the previous studies . Indeed , in all of the previous NCX1 generic loss-of-function studies , mutant mice were created by targeting exon 2 , supporting the possibility that an alternately spliced variant may be able to compensate . Furthermore , precedent for genetic ablation being compensated for over the course of development exists , whereby patterning defects were muted over time ( Bloomekatz et al . , 2012 ) . In contrast , our use of pharmacological channel blockers CB-DMB and KB-R7943 , resulted in an acute stage-specific-loss of NCX1 function which attributed an essential role for NCX1 in generating SACOs , and potentially acting as a trigger to establish the onset of beating in the cardiac crescent and subsequent cardiomyocyte differentiation and morphogenesis . These findings are consistent with the situation in tremblor ( tre ) zebrafish which have a mutation in NCX1 leading to absence of normal calcium transients and cardiac fibrillation ( Langenbacher et al . , 2005 ) as well as earlier findings in rodents demonstrating a requirement for elevated cytoplasmic calcium to drive cardiac myofibrillogenesis in developing cardiomyocytes ( Webb and Miller , 2003 ) . Parallel analyses on ESC-derived cardiomyocytes revealed that lowering the concentration of Ca2+ in the media had a similar effect on reducing the number of beating EBs as treatment with CB-DMB . This both reinforced the relative importance of NCX1 and also provided the first indication that the exchanger may be working in reverse mode , facilitating inward Ca2+ as necessary for cardiomyocyte differentiation . Furthermore , our use of the inhibitor KB-R7943 , previously reported to specifically inhibit reverse mode NCX1 activity ( Brustovetsky et al . , 2011 ) , replicated observations made with CB-DMB , such that at stage 1 and 2 both LTCC and NCX1 are required for sufficient Ca2+ influx . Whilst inhibition of NCX1 clearly blocked SACOs in stage 0 cardiac crescents , the mechanism by which NCX1 alone could lead to periodic oscillations in Ca2+ is still unclear , and may require oscillations in other ions , such as Na+ , and/or contribution of other sarcolemmal proteins , such as the Plasma membrane Ca2+ ATPase ( PMCA ) , to regulate Ca2+ efflux while NCX1 is working in reverse mode . Due to the slow nature of SACOs , oscillations in energy production along with adenosine triphosphate levels could also be involved in SACO generation , especially with reduced SR function . Whilst SR inhibition did not prevent stage 3 Ca2+ transients , we have not tested the involvement of SR function at stage 0 and , therefore , cannot fully exclude that periodic Ca2+ releases from the SR are involved in the generation of SACOs . The latter has been reported in cultured E8 . 5–9 embryonic cardiomyocytes as a pace making mechanism ( Sasse et al . , 2007; Rapila et al . , 2008 ) at later stages than were the focus in this study . To further understand the mechanism of NCX1 in SACO generation will require in vivo electrophysiological analysis to more accurately determine whether the inhibitory effect of NCX1 blockade with CB-DMB and KB-R7943 is due to effects on reverse mode NCX1 function ( inhibition of translemmal Ca2+ influx ) or forward mode NCX1 function ( inhibition of Ca2+ efflux leading to Ca2+ overload ) , imbalance of Ca2+ homeostasis or off-target effects . Although the existence of the NCX1 acting in reverse mode is contentious , and the precise function of KB-R7943 is still debated in the field , it is difficult to otherwise explain how inhibition of NCX1 can block SACOs . That said , regardless of mode of action , the observation that at stage 0 SACOs were abolished with both CB-DMB and KB-R7943 treatment , but persisted when treated with nifedipine , suggests that NCX1 , and not the LTCC , plays a major role in Ca2+ transient generation within the early cardiac crescent . As cardiomyocytes matured , transition to LTCC became the predominant mechanism for inward Ca2+ entry , as demonstrated by nifedipine-induced inhibition of beating at later stages in both the embryo and ESC-derived cardiomyocytes . NCX1 expression was further maintained at high levels during more advanced stages ( stage 0-LHT ) , suggesting a later role in ensuring Ca2+ removal via its forward mode of action . Of note , the early versus late roles for mammalian NCX1 and LTCC , respectively , are further supported by studies on tre zebrafish , whereby sarcomeric assembly defects in developing cardiomyocytes following loss of NCX1 function are not recapitulated by mutations in LTCC ( Ebert et al . , 2005 ) . Increased expression of NCX1 has also been linked with pathological hypertrophy and heart failure ( reviewed in Sipido et al . , 2002 ) , whereby elevated NCX1 is thought to compensate for defective ECC and depressed function of SERCA but also produce arrhythmogenic-delayed after-depolarisations ( Gómez et al . , 1997; Schultz et al . , 2004; Venetucci et al . , 2007 ) . Transgenic mice over-expressing NCX1 within the myocardium exhibited a proportional decrease in contractile function and increased incidence of heart failure , suggesting a decompensatory mechanism with regards to Ca2+ handling ( Roos et al . , 2007 ) . Delta isoforms of CaMKII predominate in the heart and are involved in multiple signalling cascades to regulate gene expression , as well as cardiomyocyte physiology including Ca2+ and Na+ homeostasis ( Wagner et al . , 2006; Aiba et al . , 2010 ) . In this study , CaMKII activation was impaired following NCX1 inhibition in embryonic stem cells , and , moreover , canonical hypertrophy genes , including Mef2c and Myh7 were down-regulated in response to impaired NCX function during development . Whilst changes in gene expression did differ between embryos and ESCs in response to treatment with different drugs , this likely reflects inherent differences between the in vitro versus in vivo models , as well as the timing and length of drug exposure . ESCs were exposed at a relatively earlier stage of cardiac lineage specification and for longer periods , which was not feasible in the live embryo cultures , resulting in a stronger down-regulation of the genes investigated . Overall though these findings suggest analogous roles for NCX1 in early heart development and adult hypertrophy , with regards Ca2+-handling and foetal cardiac gene induction to promote physiological or pathological myocyte growth , respectively . There is recent precedent for a role for Ca2+ in the establishment of other embryonic lineages; most notably Ca2+ signals are involved in the earliest steps of neurogenesis , including neural induction and the differentiation of neural progenitors into neurons ( Leclerc et al . , 2012 ) . Here we show that pharmacological inhibition of NCX1 and dysregulation of Ca2+ handling from the outset had an adverse effect on early cardiomyocyte differentiation and led to impaired cardiogenesis in the embryo ( Figure 5 ) . Thus , an early induction of Ca2+-handling preceding beating within cardiac muscle is pivotal for subsequent terminal differentiation and normal heart development . All animal experiments were carried out according to UK Home Office project license PPL 30/3155 and 30/2887 compliant with the UK animals ( Scientific Procedures ) Act 1986 and approved by the local Biological Services Ethical Review Process . To obtain wild-type embryos C57BL/6 males ( in house ) were crossed with CD1 females ( Charles River , England ) . All mice were maintained in a 12-hr light-dark cycle . Noon of the day finding a vaginal plug was designated E0 . 5 . In order to dissect the embryos the pregnant females were culled by cervical dislocation in accordance with the schedule one of the Animal Scientific Procedures Act . Embryos of the appropriate stage were dissected in M2 medium ( Sigma-Aldrich , England ) . Progressive crescent stages were defined based on morphological criteria: the length ( medio-lateral axis ) and the maximum height ( rostral-caudal axis ) of the cardiac crescent was measured for each embryo . Embryos were considered to be at the LHT stage once both sides of the cardiac crescent were completely folded and fused . The ratio between width and maximum height measurements ( µm ) was used to categorize the embryos from stage 0 through to stage 3 ( Supplementary file 1a ) . Live imaging of embryos , including Ca2+ imaging was performed as previously described , with some adaptations ( Chen et al . , 2014 ) . Briefly , freshly dissected embryos were imaged in a mix of 50% phenol red-free CMRL ( PAN-Biotech , Germany ) supplemented with 10 mM L/glutamine ( Sigma-Aldrich ) and 50% Knockout Serum Replacement ( Life Technologies , England ) . Initial characterisation of cardiac contractions was performed using Differential Interference Contrast ( DIC ) imaging on a Spinning Disk Confocal microscope at 37°C and an atmosphere of 5% CO2 + Air . Images were acquired at 10 frames per second ( fps ) for up to 20 s . For all experiments involving Ca2+ imaging , embryos were loaded with 8 μM of Cal-520 by incubating the embryos in 50% CMRL + 50% Knockout Serum Replacement with the dye for 15 min at 37°C and an atmosphere of 5% CO2 + Air . The embryos were then transferred to fresh media in a MatTek dish ( MatTek Corporation , Framingham , MA ) without the dye and imaged . To image Ca2+ transients in embryoid bodies ( EBs ) , ESCs were cultured in hanging drops in MatTek dishes for four days under differentiation conditions . At day 7 , the cells were loaded with 8 μM Cal-520 for 30 min in media without serum at room temperature and then transferred into media with serum and cultured in the presence of the dye for 10 min at 37°C and an atmosphere of 5% CO2 + Air . All Ca2+ imaging was performed with a Zeiss 710 LSM fitted with an environmental chamber to maintain the embryos at 37°C at 5% CO2 . Embryos were imaged with a 20× air objective ( 0 . 6 NA ) with a single optical section every 97 ms ( ~10 frames per second ) . Images were captured at 256 × 256 pixel dimensions , with a 2× line step and no averaging to increase the scan speed . For acute inhibition of Ca2+ transients in embryos ( stage 1 to LHT ) and EBs , Ca2+ transients were first imaged at baseline , drug containing media was then added and imaging was performed at 5 , 15 and 30 min post drug treatment . Inhibition was defined as the complete cessation as well as confinement of Ca2+ transients to small regions of the cardiac crescent . For inhibitor experiments involving stage 0 embryos , imaging was carried out at baseline , 5 and 15 min post treatment . For all the experiments involving calcium imaging of embryo the overlying yolk sac endoderm had to be dissected out in order to allow proper drug and dye penetration , as well as better visualisation of the cardiac mesoderm . For experiments involving embryos Nifedipine ( Sigma-aldrich ) was used at a final concentration of 10 μM , CB-DMB ( Sigma-aldrich ) at 20 μM , KB-R7943 ( Sigma-aldrich ) at 30 μM , Ryanodine ( Tocris Bioscience , England ) at 100 μM and 2-APB at 200 μM , all diluted in DMSO ( Sigma-aldrich ) . We found that most drugs had to be used at a higher concentration than previously used in studies involving isolated cells , presumably due to penetration difficulties inherent with using whole embryos . Control experiments for Nifedipine , CB-DMB and KB-R7943 were performed with 0 . 002% DMSO , and for experiments involving dual inhibition with Ryanodine + 2-APB with 0 . 6% DMSO . All experiments involving embryos were repeated with at least three different litters ( 6–10 embryos ) . Both embryo and ESC experiments were performed on at least three independent days . Embryos were cultured in the presence of drug ( CB-DMB or nifedipine ) for 12 hr , from E7 . 5 to E8 . 0 using a rolling culture system . Embryos were cultured in a mix of 50% CMRL and 50% Knockout Serum Replacement at 37°C and with an atmosphere of 5% CO2+ Air . Only embryos at E7 . 5 , prior to cardiac crescent and head fold formation were cultured . For these experiments embryos were cultured in the presence of either in 10 μM Nifedipine , 3 μM CB-DMB or 0 . 0003% DMSO . Experiments for immunostaining were performed on three independent days , while experiments for qRT-PCR were performed on four different days for Nifedipine treatments and five days for CB-DMB treatments . Cardiomyocyte differentiation from ESCs was carried out as previously described using an Nkx2 . 5-GFP ( Moretti et al . , 2006 ) and Eomes-GFP ( Arnold et al . , 2009 ) cell line . Briefly , ESCs were maintained in an undifferentiated state by culturing with KO-DMEM ( Gibco , England ) supplemented with glutamax ( 2 mM; Gibco ) , embryomax FBS ( 15%; Millipore ) , nonessential amino acids ( 0 . 1 mM; Invitrogen , England ) , penicillin ( 60U/mL; Gibco ) , streptomycin ( 60 μg/mL; Gibco ) , β-mercaptoethanol ( 0 . 1 mM; Sigma-aldrich ) and Leukaemia inhibitory factor ( 1000 U/ml; Millipore , England ) . Cardiomyocyte differentiation was induced using the hanging drop culture method ( Kattman and Keller , 2007 ) . Approximately 500 ESCs were plated in 20 μl drops of differentiation media ( 4 . 5 g glucose/DMEM , embryomax FBS ( 20% ) , nonessential amino acids ( 0 . 1 mM ) , penicillin ( 60 U/ml ) , streptomycin ( 60 μg/ml ) , β-mercaptoethanol [0 . 1 mM] ) on the lids of petri dishes and cultured as hanging drops throughout the first four days of differentiation , allowing embryoid bodies ( EBs ) to form . At day 4 , the EBs were transferred onto 0 . 1% gelatin coated plates for a further 10 days of culture before being collected at day 14 of differentiation . 1 μM CB-DMB and 10 μM Nifedipine were added to the differentiation media at day 0 . For control experiments , DMSO was added at the same concentration as drug containing media . Once the EBs had been plated at day 4 , drug-containing media was changed every two days up until day 14 . To assess the percentage of EBs which were beating , drug containing media was removed 2 hr prior to assessment and replaced with fresh drug-free culture media . Culturing of ESCs in different Ca2+ concentrations was achieved using Ca2+ free DMEM ( Gibco ) with Ca2+ being added back to the desired concentration ( 0 . 1 mM , 1 . 0 mM , 1 . 8 mM ) . For immunostaining , ESCs were cultured on 0 . 1% gelatin coated glass coverslips before being fixed with 3 . 7% PFA for 30 min on ice . For RNA isolation tissue samples were dissociated for 2 min using 0 . 25% trypsin-EDTA at 37°C prior to snap freezing . Dissected embryos were fixed for 1 hr at room temperature with 4% PFA in PBS . The embryos were then washed 3x in PBT-0 . 1% ( PBS with 0 . 1% Triton X-100 ) for 15 min , permeabilised in PBT-0 . 25% for 40 min and washed again 3x in PBT-0 . 1% . The embryos were transferred to blocking solution ( 5% donkey serum , 1%BSA in PBT-0 . 1% ) overnight ( o/n ) at 4°C . Primary antibodies ( Supplementary file 1e ) were then added to the solution and incubated o/n at 4°C . The embryos were washed 3x in PBT-0 . 1% and incubated o/n 4°C in PBT-0 . 1% with the secondary antibodies ( Supplementary file 1e ) , then subsequently washed 3x PBT-0 . 1% for 15 min and mounted in Vectashield mounting media with DAPI for at least 24 hr at 4°C . After fixation ESC samples were rinsed with PBS before being permeabilised with PBT-0 . 1% for 10 min , followed by blocking with 10% goat serum , 1% BSA in PBT-0 . 1% for 1 hr . Incubation with primary antibodies , diluted in blocking buffer , was carried out o/n at 4°C . After incubation with primary antibodies samples were washed for 3x for 10 min with PBT-0 . 1% and then incubated with secondary antibodies ( Supplementary file 1e ) diluted in blocking buffer for 1 hr at room temperature . After incubation with secondary antibodies samples were washed 5x for 5 min in PBS . Samples were mounted using Vectashield mounting media with DAPI for at least 24 hr prior to imaging with either a 40x oil ( 1 . 36 NA ) or water ( 1 . 2 NA ) objective . Images were captured at a 512 × 512 pixel dimension and tiled 2x2 with a Z-step of 1 . 5 µm . For embryos each staining was repeated for at least three litters ( 6–10 embryos per litter ) . Both experiments involving embryos and ESCs were performed on at least three independent days and with two different secondary antibodies . RNA extraction of whole embryos , embryonic hearts , head folds and ESC samples was performed using an RNeasy Micro Kit ( Qiagen , England ) according to manufacturer’s instructions: briefly , homogenisation was carried out with a 21G needle and the extract run through an on-column DNase I treatment . For P0 and adult heart samples , RNA extraction was performed using Trizol ( Invitrogen , England ) , according to the manufacturer’s instructions , additional DNase I ( Promega , England ) treatment on Trizol-extracted RNAs was carried out to eliminate genomic DNA contamination . In order to collect enough RNA for isolated cardiac crescent and head fold development , each biological replicate was composed of 10 individual cardiac crescents or head folds . For both types of extraction , RNA pellets were dissolved in RNase-free water and the RNA quality and quantity determined by Nanodrop readings at 260 , 280 and 230 nm wavelengths . cDNA was generated from 1 μg of RNA using random primers and SuperScript III polymerase ( Invitrogen ) . The expression of mRNAs for genes of interest ( Supplementary file 1f ) , together with endogenous controls ( treated embryos and differentiated ES cells , HPRT , GAPDH , 18 s; in vivo timecourse , GAPDH & HPRT; ES cell differentiation , GAPDH & 18 s ( Murphy and Polak , 2002 ) were measured in triplicate for each sample by quantitative real-time PCR using SYBR Green ( Applied Biosystems , England ) . Each reaction contained: 8 ng cDNA , 0 . 5 μl of each primer , 6 . 5 μl water and 12 . 5 μl 2 x SYBR Green , made up with H2O to a final volume of 22 μl . Primers ( Sigma-aldrich ) were either designed using Primer-BLAST ( National Center for Biotechnology Information , National Institutes of Health ) or obtained from PrimerBank ( http://pga . mgh . harvard . edu/primerbank; Primer 3 ) or previous publications ( primer sequences in Supplementary file 1f ) . Primers were designed to span exon-intron boundaries , have annealing temperatures around 60°C and generate amplicons between 50–200 bp . The reaction mixture and samples were loaded into either a MicroAmp Optical 96-Well Reaction Plates or MicroAmp Fast Optical 96-Well Reaction Plates and sealed with Optical Adhesive Films ( Life Technologies ) . Quantification was performed on a ViiA 7 Detection System ( Applied Biosystems ) using a PCR programme of 95°C 15 min followed by 40 cycles of ( 95°C 15 s melting phase and 60°C 1 min annealing and extension ) . Amplification of a single amplicon was confirmed by obtaining dissociation curve ( melt curve ) profiles as well as using gel electrophoresis to separate the reaction product . Cycle threshold ( Ct ) values were generated using either Viia7 software ( Applied Biosystems ) . Relative gene expression levels were obtained using the ΔΔCt method , in which expression of each gene of interest was normalised to endogenous controls ( Schmittgen and Livak , 2008 ) ( Murphy and Polak , 2002 ) , and presented as fold change over a reference sample . For time course data fold-change was calculated in relation to the earliest stages ( E7 . 75 in vivo , D0 ESC models ) , whilst for drug treated experiments fold change was calculated relative to control ( DMSO ) samples . Non-template controls were performed by replacing cDNA with water , to test for non-specific amplification . Protein was extracted on ice using direct lysis of cells with NP-40 extraction buffer ( 150 mM NaCl , 1 . 0% NP-40 , 50 mM Tris ( pH8 . 0 ) ) and 1x protease and phosphatase inhibitor cocktail ( Roche , England ) . Lysates were span at 10 , 000 rpm for 20 min at 4°C and supernatant collected . An aliquot was taken for protein quantification using a DC protein assay ( Bio-rad , England ) . Supernatants were prepared for SDS-PAGE with the addition of 4x Laemmli sample buffer and boiling at 95°C for 5 min . Western blotting was performed using standard SDS-PAGE methods using HRP-conjugated secondary antibodies and enhanced chemiluminescence detection ( GE Healthcare , England ) . All data involving beat rate and qPCR gene levels was compared using one-way ANOVA followed by a Tukey test for multiple comparisons . In cases where the raw data failed to map to a normal distribution with consistent variance , we applied Taylor’s law to choose the best transformation for the data . All data to be analysed passed the Shapiro-Wilk normality test and Bartlett test for homogeneous variances . To compare the number of affected embryos upon acute treatments , a Freeman-Halton extension of Fisher exact probability test was applied due to a smaller number of samples . To compare the effect of different treatments on the number of beating EBs , due to a large number of samples , a Chi Square test with a Bonferroni correction for multiple comparisons was performed instead . Principal Component Analysis ( PCA ) was carried out to directly compare temporal gene expression data from EBs with that derived from embryos; each biological replicate for embryonic stage was composed of 5–6 embryos and for EBs was composed of 10–80 EBs , depending on the day of differentiation . The data from each sample was normalized by assessing the ratio with the maximum value of all samples . A log transformation was applied to the normalized data and the principal components were calculated using R environment . The 3D representation of the PCA was constructed by plotting the 3 first components given that these explained more that 95% of the observed variance . A hierarchical clustering analysis was performed using Ward’s minimum variance method as a more precise clustering algorithm to divide the samples in different groups ( Ward , 1963 ) . A variant of absolute image filter was used to visualize and plot measurements of cardiac contractions in the developing cardiac crescent as described elsewhere ( Chen et al . , 2014 ) . Briefly , pixel displacement , indicative of contractions , was visualized and represented by increased grey levels within the crescent . Change in pixel intensity was assessed in a selected region , to reveal the contraction dynamics . Background Ca2+ signal was subtracted from all frames of a given time-lapse using ImageJ . To obtain the profiles for Ca2+ transients , regions of interest were plotted using the ratio between observed fluorescent and minimum fluorescent ( F/F0 ) after background subtraction .	The heart is the first organ to form and to begin working in an embryo during pregnancy . It must begin pumping early to supply oxygen and nutrients to the developing embryo . Coordinated contractions of specialised muscle cells in the heart , called cardiomyocytes , generate the force needed to pump blood . The flow of calcium ions into and out of the cardiomyocytes triggers these heartbeats . In addition to triggering heart contractions , calcium ions also act as a messenger that drives changes in which genes are active in the cardiomyocytes and how these cells behave . Scientists commonly think of the first heartbeat as occurring after a tube-like structure forms in the embryo that will eventually develop into the heart . However , it is not yet clear how the first heartbeat starts or how the initial heartbeats affect further heart development . Tyser , Miranda et al . now show that the first heartbeat actually occurs much earlier in embryonic development than widely appreciated . In the experiments , videos of live mouse embryos showed that prior to the first heartbeat the flow of calcium ions between different cardiomyocytes is not synchronised . However , as the heart grows these calcium flows become coordinated leading to the first heartbeat . The heartbeats also become faster as the heart grows . Using drugs to block the movement of calcium ions , Tyser , Miranda et al . also show that a protein called NCX1 is required to trigger the calcium flows prior to the first heartbeat . Moreover , the experiments revealed that these early heartbeats help drive the growth of cardiomyocytes and shape the developing heart . Together , the experiments show that the first heartbeats are essential for normal heart development . Future studies are needed to determine what controls the speed of the first heartbeats , and what organises the calcium flows that trigger the first heartbeat . Such studies may help scientists better understand birth defects of the heart , and may suggest strategies to rebuild hearts that have been damaged by a heart attack or other injury .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2016	Calcium handling precedes cardiac differentiation to initiate the first heartbeat
AMPA-type glutamate receptors ( AMPARs ) mediate fast excitatory neurotransmission and are selectively recruited during activity-dependent plasticity to increase synaptic strength . A prerequisite for faithful signal transmission is the positioning and clustering of AMPARs at postsynaptic sites . The mechanisms underlying this positioning have largely been ascribed to the receptor cytoplasmic C-termini and to AMPAR-associated auxiliary subunits , both interacting with the postsynaptic scaffold . Here , using mouse organotypic hippocampal slices , we show that the extracellular AMPAR N-terminal domain ( NTD ) , which projects midway into the synaptic cleft , plays a fundamental role in this process . This highly sequence-diverse domain mediates synaptic anchoring in a subunit-selective manner . Receptors lacking the NTD exhibit increased mobility in synapses , depress synaptic transmission and are unable to sustain long-term potentiation ( LTP ) . Thus , synaptic transmission and the expression of LTP are dependent upon an AMPAR anchoring mechanism that is driven by the NTD . AMPA receptors ( AMPARs ) are embedded at postsynaptic sites , aligned with the presynaptic glutamate release machinery for optimal signaling ( Lisman et al . , 2007 ) . Their activation drives propagation of presynaptic impulses through depolarization of the postsynaptic membrane ( Traynelis et al . , 2010 ) . As AMPARs have low apparent glutamate affinity , and rapidly diffuse in the plane of the membrane , they require trapping at synaptic sites in order to effectively contribute to signal transmission ( Choquet and Triller , 2013; Heine et al . , 2008 ) . Synapse strengthening , as occurs during learning , results from the recruitment of additional AMPARs and their enrichment at synapses ( Chater and Goda , 2014; Huganir and Nicoll , 2013; Kessels and Malinow , 2009 ) . Hence , the mechanisms underlying AMPAR positioning are fundamental to synaptic transmission and plasticity . Signaling properties and synaptic delivery depend on AMPAR composition . AMPARs are tetramers , assembled from the core GluA1-GluA4 ( pore-forming ) subunits , which associate with a varying set of auxiliary subunits , such as transmembrane AMPAR regulatory proteins ( TARPs ) ( Jackson and Nicoll , 2011 ) . Each core subunit consists of four domains – a short cytosolic C-terminus ( CTD ) , the transmembrane ion channel domain ( TMD ) , and two extracellular domains: the ligand-binding domain ( LBD ) and the distal N-terminal domain ( NTD ) ( Figure 1A ) . 10 . 7554/eLife . 23024 . 003Figure 1 . NTD deleted GluA2 is robustly expressed on the cell surface . ( A ) AMPA receptor schematic detailing the four-domain structure ( NTD - N-terminal domain; LBD - Ligand-binding domain; TMD - Transmembrane domain; CTD - C-terminal domain ) . ( B ) Single-cell electroporated CA1 pyramidal neurons in an organotypic slice culture . Scale bar = 50 μm . ( C1 ) I/V curves of glutamate-evoked AMPAR currents recorded from outside-out patches of untransfected , GluA2Q and GluA2Q ΔNTD-expressing cells . ( C2 ) AMPAR currents from transfected neurons show strong inward-rectification on GluA2 construct expression ( Rectification index ( RI ) : untrans . : 0 . 62 ± 0 . 03 ( n = 5 ) ; GluA2Q: 0 . 13 ± 0 . 02 ( n = 8 ) ; GluA2Q ΔNTD: 0 . 15 ± 0 . 02 ( n = 13 ) ; One-way ANOVA , p<0 . 0001 ) . Significance ( * ) indicates difference to untransfected cells . ( D1 ) Ratio of response amplitude to kainic acid and glutamate , indicative of auxiliary subunit association , from somatic patches is unchanged on receptor overexpression ( KA/Glu: untrans . : 0 . 48 ± 0 . 03 ( n = 5 ) ; GluA2Q: 0 . 41 ± 0 . 03 ( n = 8 ) ; GluA2Q ΔNTD: 0 . 42 ± 0 . 05 ( n = 9 ) ; One-way ANOVA , p=0 . 51 ) . Example traces showing glutamate ( Glu ) and kainic acid ( KA ) application are shown left . Scale bar = 50 ms and 100 pA . ( D2 ) Amplitudes of surface patch AMPAR glutamate responses are apparently elevated on GluA2Q or GluA2Q ΔNTD overexpression ( untrans . : 398 ± 50 pA; GluA2Q: 886 ± 153 pA; GluA2Q ΔNTD: 763 ± 145 pA ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 00310 . 7554/eLife . 23024 . 004Figure 1—figure supplement 1 . NTD deletion construct screening . ( A ) GluA1 and GluA2 NTD-LBD linker sequences . ( B1 ) Fluorescence histogram of two GluA2 NTD-deletion constructs as measured using flow cytometry . Amino acid number indicates first residue of deletion constructs . Background fluorescence of untransfected ( untrans . ) cells is shown for reference . ( B2 ) Surface expression of NTD deletion constructs of GluA2 normalized to best expressing construct . Data represents an average of two repetitions . L378 construct was used henceforth . ( B3 ) Normalized surface expression of GluA1 NTD deletion constructs ( two preparations averaged ) . A374 construct was used henceforth . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 004 The sequence-diverse C-termini mediate subtype-selective AMPAR trafficking , and their role in recruitment of specific subunits during synaptic plasticity has been extensively studied ( Derkach et al . , 2007; Kessels and Malinow , 2009; Newpher and Ehlers , 2008; Shepherd and Huganir , 2007 ) . The CTD interacts with postsynaptic scaffolding proteins ( Anggono and Huganir , 2012; Shepherd and Huganir , 2007 ) , but deletion of this region is not a prerequisite for receptor clustering ( Bats et al . , 2007; MacGillavry et al . , 2013 ) , and how critical these interactions are in plasticity is not fully understood ( Boehm et al . , 2006; Granger et al . , 2013; Kim et al . , 2005 ) . Currently , the best-described anchoring mechanism is mediated by TARP γ−2 , which interacts via its C-terminus with the scaffolding protein PSD-95 , and limits diffusion of synaptic AMPARs ( Opazo et al . , 2012; Schnell et al . , 2002 ) . Like the CTD , the NTD is highly sequence-diverse between the four AMPAR subunits , offering great capacity for subunit-specific control . This domain projects into the crowded environment of the synaptic cleft , providing a large , structurally dynamic docking platform ( García-Nafría et al . , 2016a ) . For example , neuronal pentraxins ( NPs ) interact with the AMPAR NTD and mediate clustering at interneuron synapses , but the underlying mechanism remains to be clarified ( Chang et al . , 2010; O'Brien et al . , 1999; Sia et al . , 2007 ) . Here we show that synaptic delivery of GluA1 and GluA2 , prominent AMPAR subunits in CA1 pyramidal neurons ( Lu et al . , 2009 ) , is dependent on their NTDs in a subunit-specific manner . Although receptors lacking the NTD accumulate at the extra-synaptic surface , they cannot effectively contribute to synaptic transmission and are unable to sustain LTP . We initially focused on the NTD of GluA2 , a subunit most commonly incorporated into AMPARs ( Isaac et al . , 2007 ) . To compare GluA2 wild-type ( WT ) to a mutant lacking the NTD , we first tested the surface trafficking capacity of GluA2 constructs bearing an NTD deletion , a modification that does not impair AMPAR function ( Pasternack et al . , 2002 ) . Deletion of amino acids 1–377 of the mature polypeptide resulted in optimal surface expression in HEK293T cells ( Figure 1—figure supplement 1A–B2 ) and was used throughout this study ( designated GluA2Q ΔNTD ) . When expressed in CA1 pyramidal neurons , this GluA2Q ΔNTD construct produced inwardly rectifying responses in somatic outside-out patches , matching the responses from neurons expressing full-length GluA2Q ( Figure 1C ) . As the functional properties of neuronal AMPARs and their expression at synapses are modulated by auxiliary subunits , most prominently by TARPs , we determined whether exogenous GluA2 was still TARP-associated . A signature of TARP action is an increased efficacy of the partial agonist kainic acid ( KA ) ( Tomita et al . , 2005; Turetsky et al . , 2005 ) , and the ratio of kainate and glutamate response amplitudes ( KA/Glu ratio ) is a measure of AMPAR/TARP stoichiometry ( Shi et al . , 2009 ) . The KA/Glu ratio suggests full TARP occupancy for both GluA2Q and GluA2Q ΔNTD ( Figure 1D1 ) , implying that TARPs are not limiting in our system and NTD deletion does not affect TARP association . Amplitudes of glutamate-evoked currents from GluA2Q and GluA2Q ΔNTD-expressing somatic patches were similar , further confirming that expression levels are comparable , and were approximately double that of untransfected neurons ( Figure 1D2 ) , which can be explained by the greater single channel conductance of ( unedited ) GluA2Q homomers than native receptors ( see below; Swanson et al . , 1997 ) . To assay synaptic responses , we stimulated Schaffer collateral fibers and simultaneously recorded whole-cell AMPAR responses from pairs of transfected and untransfected neurons . As observed for somatic receptors , both GluA2Q and GluA2Q ΔNTD-expressing cells produced strongly rectifying responses relative to untransfected neurons ( Figure 2A ) , demonstrating that GluA2Q homomers lacking the NTD reach synapses . However , while EPSC amplitudes were elevated in neurons expressing GluA2Q relative to paired untransfected cells ( 166 ± 13% ) , EPSCs were significantly reduced in GluA2Q ΔNTD neurons ( 58 ± 5%; Figure 2B1–2 versus Figure 2C1–2 ) . This effect was specific to the synaptic AMPAR component , as NMDAR EPSCs were unchanged in both conditions ( Figure 2B3 and C3 ) . Therefore , GluA2Q ΔNTD receptors reach synapses but interfere with synaptic transmission . 10 . 7554/eLife . 23024 . 005Figure 2 . Expression of NTD deleted GluA2 causes large reduction in synaptic currents . ( A ) Synaptic RI measured from pairs of untransfected and transfected cells indicate both homomeric GluA2Q and GluA2Q ΔNTD are inserted into synapses . ( A1 ) Untrans . , 0 . 60 ± 0 . 19; GluA2Q , 0 . 17 ± 0 . 10; n = 13 pairs; paired t-test , p=0 . 001 . ( A2 ) Untrans . , 0 . 60 ± 0 . 25; GluA2Q ΔNTD , 0 . 22 ± 0 . 11; n = 16 pairs; paired t-test , p<0 . 0001 . Sample traces and construct schematics are shown on the left . Scale bar = 10 ms , 50 pA ( grey and GluA2Q ) or 20 pA ( GluA2Q ΔNTD ) . ( B–C ) Scatter plots and bar charts of EPSC amplitudes from pairs of cells , showing single pairs ( open circles ) and mean values ± SEM ( filled circles ) . Sample traces inset . Scale bar = 10 ms , 30 pA . ( B1 ) GluA2Q-expressing cells have increased AMPAR EPSCs relative to untransfected cells ( untrans . : 26 . 8 ± 3 . 1 pA; GluA2Q: 38 . 8 ± 3 . 0 pA; n = 22 pairs; paired t-test , p<0 . 0001 ) . ( B2 ) Bar chart of AMPAR EPSCs from B1 . ( B3 ) NMDAR-mediated EPSCs remain unchanged ( untrans . : 27 . 2 ± 2 . 7 pA; GluA2Q: 24 . 0 ± 2 . 2 pA; n = 20; paired t-test , p=0 . 107 ) . ( C1 ) GluA2Q ΔNTD-expressing cells show AMPAR EPSC amplitude depression ( untrans . : 44 . 3 ± 4 . 3 pA; GluA2Q ΔNTD: 23 . 0 ± 2 . 5 pA; n = 22; paired t-test , p<0 . 0001 ) . ( C2 ) Bar chart of AMPAR EPSCs from C1 . ( C3 ) NMDAR EPSCs show no amplitude change ( untrans . : 41 . 8 ± 9 . 0 pA; GluA2Q ΔNTD: 35 . 8 ± 6 . 1 pA; n = 14; paired t-test , p=0 . 243 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 00510 . 7554/eLife . 23024 . 006Figure 2—figure supplement 1 . Measurement of paired-pulse ratio and spine density . ( A ) Average paired-pulse ratio of neighboring untransfected and transfected cells ( untrans . : 1 . 98 ± 0 . 11 ( n = 13 ) ; GluA2Q: 1 . 95 ± 0 . 10 ( n = 6 ) ; GluA2Q ΔNTD: 2 . 04 ± 0 . 06 ( n = 7 ) ; One-way ANOVA: p=0 . 876 ) , with example traces from pairs of cells shown on the left . Scale bars = 20 ms and 30 pA . ( B ) Bar chart of dendritic spine density on untransfected and transfected cells ( untrans . : 12 . 0 ± 0 . 7 ( n = 8 cells ) ; GluA2Q: 11 . 8 ± 0 . 6 ( n = 8 ) ; GluA2Q ΔNTD: 12 . 2 ± 0 . 7 ( n = 8 ) ; One-way ANOVA: p=0 . 916 ) . Sample images are shown on the left . Scale bar = 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 006 As a change in AMPAR EPSC amplitude could occur for a variety of reasons , we sought to identify the mechanism for this effect . Paired-pulse ratios were unchanged in cells expressing either GluA2 construct and were comparable to untransfected cells ( Figure 2—figure supplement 1A ) , suggesting a postsynaptic locus for the effect . As the GluA2 NTD has been implicated in spine formation ( Passafaro et al . , 2003; Saglietti et al . , 2007 ) , we assayed spine density , which was unchanged between the three conditions ( GluA2Q , GluA2Q ΔNTD and untransfected; Figure 2—figure supplement 1B ) . Since NMDAR EPSCs were also unaffected in these neurons , a change in the number of synapses cannot explain this effect . To characterize the postsynaptic response in greater detail we recorded AMPAR miniature EPSCs ( mEPSCs ) . In line with the changes in evoked transmission , spontaneous transmission was dramatically impaired on NTD deletion . mEPSC amplitudes of GluA2Q ΔNTD cells were significantly reduced relative to GluA2Q-expressing cells ( Figure 3A ) , while decay kinetics was unaffected by NTD deletion ( Figure 3C ) . We also noted a highly significant decrease in mEPSC frequency ( Figure 3B ) between GluA2Q and GluA2Q ΔNTD-expressing cells . While an increase in mEPSC amplitude on GluA2Q expression ( as per evoked EPSCs ) was not observable , an increase in mEPSC frequency was apparent , which could be caused by an increase in mEPSC amplitude , resulting in small events emerging from below the detection limit ( see Materials and methods ) . 10 . 7554/eLife . 23024 . 007Figure 3 . NTD deleted GluA2 is detrimental to spontaneous transmission . ( A1 ) Example traces of mEPSCs recorded from untransfected , GluA2Q and GluA2Q ΔNTD-expressing cells . Scale bar = 0 . 5 s , 5 pA . ( A2 ) Bar chart of mEPSC amplitude with event detection limit indicated ( dotted line ) ( untransfected: 17 . 5 ± 0 . 8 pA ( n = 25 cells ) ; GluA2Q: 18 . 9 ± 0 . 8 pA ( n = 23 ) ; GluA2Q ΔNTD: 13 . 2 ± 0 . 5 pA ( n = 16 ) ; One-way ANOVA , p<0 . 0001 ) . ( A3 ) Cumulative frequency distribution of mEPSC amplitude data from A2 . ( B ) Bar chart of mEPSC frequency ( untransfected: 0 . 52 ± 0 . 06 Hz; GluA2Q: 0 . 70 ± 0 . 08 Hz; GluA2Q ΔNTD: 0 . 32 ± 0 . 04 Hz; One-way ANOVA , p=0 . 002 ) . ( C ) Example traces of scaled mEPSCs from untransfected ( grey ) and GluA2 construct-expressing cells . Scale bar = 3 ms . Bar chart shows cell averaged mEPSC decay times ( untrans . : 11 . 14 ± 0 . 27 ms ( n = 25 ) ; GluA2Q: 8 . 86 ± 0 . 23 ms ( n = 23 ) ; GluA2Q ΔNTD: 8 . 08 ± 0 . 26 ms ( n = 16 ) ; One-way ANOVA , p<0 . 0001 ) . ( D1 ) Amplitude vs . variance plot for non-stationary fluctuation analysis ( NSFA ) of scaled mEPSCs with parabolic fits from representative cells . ( D2 ) Single-channel conductance of synaptic AMPARs show comparable conductance between GluA2Q and GluA2Q ΔNTD-expressing cells ( untransfected: 11 . 2 ± 0 . 82 pS ( n = 7 ) ; GluA2Q: 25 . 4 ± 1 . 92 pS ( n = 8 ) ; GluA2Q ΔNTD: 22 . 8 ± 2 . 70 pS ( n = 7 ) ; One-way ANOVA , p=0 . 0002 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 007 To determine whether the depression of EPSC amplitudes could be explained by reduced single-channel conductance , caused by deleting the NTD , we conducted non-stationary fluctuation analysis ( NSFA ) from mEPSCs ( Figure 3D ) . The synaptic AMPAR single-channel conductance was significantly increased in both groups of transfected cells , as expected from the expression of Q/R-unedited receptors ( Swanson et al . , 1997 ) . However , there was no difference between GluA2Q and GluA2Q ΔNTD that could explain the substantial drop in synaptic AMPAR current amplitudes in GluA2Q ΔNTD-expressing neurons ( Figure 2C ) . Based on this accumulated evidence , a change in the number of receptors at the synapse most feasibly explains the observed effect on EPSC amplitude , whereby significantly less GluA2Q ΔNTD receptors are present at the synapse than GluA2Q . Since AMPARs have a relatively low affinity for L-glutamate ( e . g . Jonas , 2000 ) , the NTD may stabilize and cluster the receptor in proximity to presynaptic release sites to enable optimal receptor activation . To test whether the NTD stabilizes AMPARs at synapses , we assayed receptor mobility using fluorescence recovery after photobleaching ( FRAP ) in cultures of dissociated hippocampal neurons . GluA2Q and GluA2Q ΔNTD were tagged at their N-termini with Super Ecliptic pHluorin ( SEP ) , a pH-sensitive GFP variant . SEP is quenched at low-pH ( as found in endosomal transport vesicles ) facilitating visualization of surface-expressed receptors , which are exposed to neutral pH ( Figure 4A ) ( Ashby et al . , 2006; Makino and Malinow , 2009 ) . 10 . 7554/eLife . 23024 . 008Figure 4 . The GluA2 NTD controls synaptic immobilisation . ( A ) Example images of FRAP on dendritic spine from cells expressing SEP-tagged AMPAR constructs , where t = 0 indicates time and square indicates location of photobleaching . Red channel: cytosolic mCherry; green: SEP fluorescence . Scale bar = 1 μm B1 , SEP fluorescence over time in bleached regions of dendrite or spine , normalized to pre-bleaching fluorescence . Orange vertical line indicates onset of photobleaching ( time constant of fit τ: spine GluA2 = 197 . 3 , spine GluA2 ΔNTD = 65 . 4 , dendritic GluA2 = 98 . 1 , dendritic GluA2 ΔNTD = 83 . 1 ) . ( B2 ) , Fluorescence at 600s averaged by cell shows greater recovery for GluA2Q ΔNTD than full-length GluA2Q ( spine GluA2Q: 0 . 63 ± 0 . 03 ( n = 22 cells ) ; spine GluA2Q ΔNTD: 0 . 93 ± 0 . 05 ( n = 19 ) ; dendrite GluA2Q: 0 . 80 ± 0 . 04 ( n = 6 ) ; dendritic GluA2Q ΔNTD: 0 . 92 ± 0 . 04 ( n = 5 ) ; One-way ANOVA , p<0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 008 In dendritic spines , SEP-GluA2 recovered from bleaching with a time constant of τrec=197 s , and recovery was incomplete after 10 min with an immobile fraction of ~40% ( at t = 600 s; Figure 4B ) . These values are in line with previous studies ( Kerr and Blanpied , 2012; Makino and Malinow , 2009; Zhang et al . , 2013 ) . In sharp contrast , recovery of SEP-GluA2 ΔNTD was rapid ( τrec= 65 s ) and the immobile fraction at 600 s was reduced to just 7% . Thus , NTD-deleted receptors are poorly confined in spines when compared to full-length GluA2Q . Interestingly , this difference was specific to spine fluorescence: in extra-synaptic ( dendritic ) regions , SEP-GluA2 exhibited rapid ( τrec = 98 s ) and almost complete recovery ( immobile fraction = 18% ) , which was not significantly different to that of SEP-GluA2 ΔNTD ( Figure 4B ) . Moreover , diffusion in dendrites resembled the behavior of spine localized SEP-GluA2 ΔNTD . These data support the hypothesis that the NTD plays a role in specifically stabilizing AMPARs at postsynaptic sites . We next extended our experiments to GluA1 , which differs from GluA2 in primary sequence chiefly in the NTD and the CTD . Each subunit exhibits different trafficking properties that have been exclusively ascribed to their CTD ( Shepherd and Huganir , 2007; Shi et al . , 2001 ) . Similar to GluA2 , NTD deletion ( Figure 1—figure supplement 1B3 ) does not affect GluA1 trafficking to the cell surface , and the RI of the NTD-deleted receptors was comparable to GluA1 in somatic patches of CA1 pyramidal cells ( RI GluA1: 0 . 23 ± 0 . 03 , GluA1 ΔNTD: 0 . 19 ± 0 . 02 ) ( Figure 5A ) . Moreover , as was the case for GluA2 , the KA/Glu ratio of neurons expressing either GluA1 construct was similar to untransfected cells ( Figure 5B1 ) and current amplitudes were approximately doubled on expression of GluA1 or GluA1 ΔNTD ( Figure 5B2 ) . Again , TARPs do not appear to be limiting and expression levels of exogenous subunits were comparable . 10 . 7554/eLife . 23024 . 009Figure 5 . The NTD is essential for synaptic anchoring of GluA1 . ( A1 ) I/V relationships of glutamate-evoked AMPAR-mediated current from outside-out patches of untransfected , GluA1 and GluA1 ΔNTD expressing cells . ( A2 ) Average RI of surface currents from above neurons ( untrans . : 0 . 61 ± 0 . 05 ( n = 5 ) ; GluA1: 0 . 23 ± 0 . 03 ( n = 5 ) ; GluA1 ΔNTD: 0 . 19 ± 0 . 02 ( n = 7 ) ; One-way ANOVA , p<0 . 0001 ) . ( B1 ) KA/Glu ratio from somatic patches is unchanged on GluA1 construct overexpression ( KA/Glu: untrans . : 0 . 47 ± 0 . 02 ( n = 5 ) ; GluA1: 0 . 43 ± 0 . 03 ( n = 9 ) ; GluA1 ΔNTD: 0 . 39 ± 0 . 02 ( n = 9 ) ; One-way ANOVA , p=0 . 16 ) . Example traces showing glutamate ( Glu ) and kainic acid ( KA ) application are shown left . Scale bar = 50 ms and 300 pA . ( B2 ) AMPAR surface patch amplitudes are similarly elevated on GluA1 or GluA1 ΔNTD overexpression ( untrans . : 421 ± 90 pA; GluA1: 849 ± 154 pA; GluA1 ΔNTD: 878 ± 173 pA ) . ( C1 ) Synaptic RI from pairs of untransfected and GluA1-expressing cells ( untransfected: 0 . 57 ± 0 . 04; GluA1: 0 . 33 ± 0 . 03; n = 20; paired t-test , p=0 . 0001 ) , with example traces and construct schematic shown on the left . Scale bars for panels C and D = 10 ms and 15 pA C2 Scatter plot of AMPAR EPSC amplitudes from pairs of untransfected and GluA1-expressing cells ( untransfected: 36 . 4 ± 3 . 5 pA; GluA1: 29 . 4 ± 3 . 0 pA; n = 34; paired t-test , p=0 . 013 ) . ( D1 ) Synaptic RI from pairs of untransfected and GluA1 ΔNTD-expressing cells ( untransfected: 0 . 51 ± 0 . 04; GluA1 ΔNTD: 0 . 48 ± 0 . 03; n = 18; paired t-test , p=0 . 072 ) , with example traces shown on the left . ( D2 ) Scatter plot of AMPAR EPSC amplitudes from pairs of untransfected and GluA1 ΔNTD-expressing cells ( untransfected: 32 . 6 ± 2 . 1 pA; GluA1 ΔNTD: 26 . 9 ± 2 . 7 pA; n = 21; paired t-test , p=0 . 014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 009 Contrasting with previous studies ( Hayashi et al . , 2000; Shi et al . , 2001 ) , we find that the RI of GluA1 expressing cells is consistently lower than untransfected cells , ( RI - Untrans . : 0 . 57 ± 0 . 04 , GluA1: 0 . 33 ± 0 . 03 ) , although not to the same extent as GluA2 ( Figure 5C1 versus Figure 2A1 ) . Strikingly , in contrast to GluA2 ΔNTD , GluA1 ΔNTD expression was unable to change the rectification index , which remained comparable to untransfected cells ( Figure 5D1 ) , suggesting that synaptic anchoring of GluA1 was completely dependent upon its NTD . Unlike GluA2Q , EPSC amplitudes were not elevated upon GluA1 expression but were slightly decreased relative to untransfected neurons; an effect that was evident for both GluA1 constructs ( Figure 5C2 and D2 ) . These data demonstrate an essential role for the GluA1 NTD in synaptic incorporation . To corroborate these observations , we also examined NTD-dependent anchoring of GluA1 and GluA2 in an AMPAR null background , using organotypic slices from conditional GluA1-3 knockout mice ( Gria1lox/lox; Gria2lox/lox; Gria3lox/lox , denoted Gria1-3fl ) ( Lu et al . , 2009 ) . Both overexpression and knockout/rescue approaches provide complementary information on receptor function . While knockout/rescue permits unequivocal quantification and interpretation of receptor contributions without interference from endogenous receptors , overexpression prevents any compensatory effects of receptor removal . Additionally , competition with endogenous subunits for synaptic slots facilitates identification of subtle deficits caused by receptor mutation that would be of lesser consequence using a knockout approach . Gria1-3 genes were excised by viral injection of Cre-recombinase into P0 mouse pups ( AAV-Cre-GFP ) and AMPAR null neurons were rescued by single-cell electroporation of AMPAR constructs into organotypic slices 12 days later ( Figure 6—figure supplement 1A ) . Successful GluA1-3 deletion was confirmed , as CA1 neurons expressing Cre-GFP alone showed almost complete loss of AMPAR responses , as assayed using the ratio of AMPAR and NMDAR EPSC amplitudes ( AMPAR/NMDAR ) ( Figure 6—figure supplement 1B ) . GluA2Q transfection rescued the AMPAR EPSC to levels comparable with uninfected ( Cre-negative ) neurons of each paired recording ( Figure 6A1 ) , whereas transfection of GluA2Q ΔNTD did not ( Figure 6A2 ) . Normalization to the untransfected cell of each pair revealed that GluA2Q ΔNTD rescue was less than half that of GluA2Q ( Relative AMPAR/NMDAR - GluA2Q: 0 . 99 ± 0 . 10 , GluA2Q ΔNTD: 0 . 44 ± 0 . 05; Figure 6A3 ) . This difference between GluA2Q and GluA2Q ΔNTD responses closely matches our previous observations ( Figure 2B , C ) . 10 . 7554/eLife . 23024 . 010Figure 6 . NTD dependent interactions enhance synaptic AMPAR anchoring in a knockout background . Paired recordings from Gria1-3fl neurons infected with AAV-Cre and rescued with AMPAR subunits , or uninfected and untransfected ( uninf . ) . Example traces show current responses at −60 mV ( AMPAR ) and +40 mV ( NMDAR ) holding potentials . Scale bars = 30 ms and 50 pA . ( A1 ) Rescue with GluA2Q restores the ratio of AMPAR to NMDAR currents to levels of uninfected neurons ( AMPAR/NMDAR , ( n = 11 pairs ) : uninf . : 1 . 64 ± 0 . 16; GluA2Q: 1 . 55 ± 0 . 17; paired t-test , p=0 . 567 ) . ( A2 ) Rescue with GluA2Q ΔNTD cannot fully restore AMPAR currents relative to NMDAR ( AMPAR/NMDAR , ( n = 11 pairs ) : uninf . : 2 . 28 ± 0 . 26; GluA2Q ΔNTD 0 . 97 ± 0 . 15; paired t-test , p=0 . 0001 ) . ( A3 ) Normalization of synaptic currents to uninfected cells reveals that GluA2Q NTD deletion reduces synaptic AMPAR rescue ( Relative AMPAR/NMDAR ratio: GluA2Q: 0 . 99 ± 0 . 10; GluA2Q ΔNTD: 0 . 44 ± 0 . 05; unpaired t-test , p=0 . 0001 ) . ( B1 ) Rescue of synaptic currents by GluA1 transfection ( AMPAR/NMDAR , ( n = 13 pairs ) : uninf . : 1 . 72 ± 0 . 15; GluA1: 0 . 75 ± 0 . 06; paired t-test , p<0 . 0001 ) . ( B2 ) Rescue of synaptic currents with GluA1 ΔNTD ( AMPAR/NMDAR , ( n = 9 pairs ) : uninf . : 2 . 48 ± 0 . 23; GluA1 ΔNTD: 0 . 83 ± 0 . 19; paired t-test , p<0 . 0001 ) . ( B3 ) GluA1 rescues synaptic currents to a greater extent than GluA1 ΔNTD ( Relative AMPAR/NMDAR ratio: GluA1: 0 . 46 ± 0 . 04; GluA1 ΔNTD: 0 . 28 ± 0 . 03; unpaired t-test , p=0 . 007 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 01010 . 7554/eLife . 23024 . 011Figure 6—figure supplement 1 . Overview and characterization of conditional AMPAR knockout in Gria1-3fl organotypic slices . ( A ) Protocol schematic for AMPAR knockout and rescue in Gria1-3fl neurons , with timings shown as mouse postnatal age ( P ) and days in vitro ( DIV ) . ( B ) Paired recordings from AAV-Cre and uninfected ( uninf . ) neurons from Gria1-3fl slices shows almost complete loss of AMPAR EPSCs ( normalized to NMDAR current amplitude ) ( AMPAR/NMDAR , ( n = 7 pairs ) : uninf . : 2 . 68 ± 0 . 48; Cre: 0 . 16 ± 0 . 06; paired t-test , p=0 . 022 ) as seen by scatter ( left ) and bar charts ( right ) . Example traces show currents at −60 mV and +40 mV holding potentials for each condition . Scale bar = 30 ms and 50 pA . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 011 Contrasting with GluA2Q , a complete rescue could not be achieved with GluA1 , and rescue was further impaired when expressing the GluA1 ΔNTD mutant ( Figure 6B1–3 ) , as seen with the overexpression data ( Figure 5C and D ) . These results underscore the AMPAR’s dependence on its NTD for synaptic anchoring . To examine the subunit selectivity of the NTD further , we swapped this domain between GluA1 and GluA2 and expressed the resulting swap mutants in WT neurons . Both mutants readily trafficked to the cell surface and altered the RI in somatic patches similar to the WT subunits ( Figure 7—figure supplement 1A ) . Transplanting the GluA2 NTD onto GluA1 ( GluA1 +A2NTD ) enhanced synaptic inward rectification ( RI 0 . 22 ± 0 . 02 , Figure 7A1 ) and increased response amplitudes relative to GluA1 WT ( Figure 7A2 and A3 ) , further demonstrating that the GluA2 NTD is able to promote AMPAR incorporation into synapses . In line with this , in neurons expressing GluA2Q +A1NTD ( i . e . the GluA2 NTD swapped for that of GluA1 ) response amplitudes were reduced relative to WT GluA2Q , approaching values obtained with GluA1 ( Figure 7B2 and B3 ) . Therefore , the NTD of GluA2 appears to confer a unique ‘synapto-sticky’ phenotype that efficiently drives the receptor into synapses . 10 . 7554/eLife . 23024 . 012Figure 7 . NTD-dependent synaptic anchoring is subunit-specific . Synaptic EPSC properties of neurons expressing chimeric AMPAR constructs formed by exchanging NTD sequences ( see construct schematics ) . ( A1 ) Synaptic RI from pairs of untransfected cells and cells expressing GluA1 +A2NTD ( untransfected: 0 . 56 ± 0 . 04; GluA1 +A2NTD: 0 . 22 ± 0 . 02; n = 23; paired t-test , p<0 . 0001 ) , with example traces and construct schematic shown on the left . GluA1 and GluA2Q RI values are indicated for reference . Scale bars = 10 ms and 20 pA . ( A2 ) Scatter plot of AMPAR EPSCs from pairs of untransfected and transfected cells expressing GluA1 +A2NTD ( untrans . : 31 . 6 ± 2 . 7 pA; GluA1 +A2NTD: 34 . 0 ± 3 . 3 pA; n = 24; paired t-test , p=0 . 521 ) . ( A3 ) Bar chart of AMPAR EPSC amplitudes of transfected cells normalized to untransfected cells of paired recordings ( GluA1: 0 . 88 ± 0 . 07 ( n = 34 ) ; GluA1 +A2NTD: 1 . 25 ± 0 . 15 ( n = 24 ) ; unpaired t-test , p=0 . 017 ) . Value for GluA2Q is indicated by red line for reference . ( B1 ) Synaptic RI from pairs of untransfected and GluA2Q +A1NTD-expressing cells ( untransfected: 0 . 55 ± 0 . 03; GluA2Q +A1NTD: 0 . 17 ± 0 . 03; n = 19; paired t-test , p<0 . 0001 ) , with example traces shown on the left . ( B2 ) Scatter plot of AMPAR EPSCs from pairs of untransfected and GluA2Q +A1NTD ( untrans . : 32 . 7 ± 4 . 1 pA; GluA2Q +A1NTD: 31 . 6 ± 2 . 6 pA; n = 22; paired t-test , p=0 . 795 ) . ( B3 ) Bar chart of AMPAR EPSCs amplitudes normalized to untransfected cell of paired recording ( GluA2Q: 1 . 66 ± 0 . 13 ( n = 22 ) ; GluA2Q +A1NTD: 1 . 19 ± 0 . 13 ( n = 22 ) ; unpaired t-test , p=0 . 013 ) . Value for GluA1 is indicated by blue line for reference . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 01210 . 7554/eLife . 23024 . 013Figure 7—figure supplement 1 . Investigation of GluA2Q ΔNTD + A1CTD . ( A ) Average RI of surface currents for chimeric construct expression ( untrans . : 0 . 60 ± 0 . 08 ( n = 5 ) ; GluA1 +A2NTD: 0 . 23 ± 0 . 07 ( n = 6 ) ; GluA2Q +A1NTD: 0 . 16 ± 0 . 08 ( n = 5 ) ; GluA2Q ΔNTD + A1CTD; 0 . 24 ± 0 . 06 ( n = 5 ) ; One-way ANOVA , p<0 . 0001 ) . ( B ) Synaptic EPSC properties of cells expressing chimeric construct formed by exchange of GluA2Q ΔNTD’s C-terminal sequence with that of GluA1 ( GluA2Q ΔNTD + A1CTD; see construct schematic ) . ( B1 ) RI of synaptic AMPAR-mediated EPSCs from pairs of untransfected and transfected cells . GluA1 ΔNTD and GluA2Q ΔNTD RIs are shown as lines for reference ( untrans . : 0 . 57 ± 0 . 03; GluA2Q ΔNTD + A1CTD: 0 . 42 ± 0 . 02; n = 20; paired t-test , p=0 . 0002 ) . ( B2 ) Scatter plot of AMPAR EPSCs from above cells showing no change in amplitude ( untrans . : 47 . 2 ± 4 . 3 pA; GluA2Q ΔNTD + A1CTD: 39 . 5 ± 3 . 9 pA; n = 21; paired t-test , p=0 . 068 ) . ( B3 ) AMPAR EPSCs normalized to untransfected cell of pair for GluA2Q ΔNTD and GluA2Q ΔNTD + A1CTD expressing cells showing loss of AMPAR EPSC depression by GluA2Q ΔNTD on CTD exchange ( GluA2Q ΔNTD: 0 . 57 ± 0 . 05 ( n = 22 ) ; GluA2Q ΔNTD + A1CTD: 0 . 88 ± 0 . 08 ( n = 21 ) ; unpaired t-test , p=0 . 002 ) . GluA1 ΔNTD value is shown as a line for reference . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 013 We note that the rectification index of GluA2Q +A1NTD expressing cells was comparable to GluA2Q WT ( Figure 7B1 ) . This is unsurprising , as GluA2 can traffic to synapses without any NTD ( see GluA2Q ΔNTD; Figure 2A2 ) . NTD-independent GluA2 trafficking , causing the strong rectification and synaptic depression of GluA2Q ΔNTD-expressing cells is likely to be caused in part by its C-terminal tail . Indeed , replacing the CTD of GluA2Q ΔNTD with that of GluA1 ( GluA2Q-ΔNTD +A1CTD ) greatly reduced inward rectification and alleviated synaptic depression of postsynaptic responses ( Figure 7—figure supplement 1B ) . LTP expression requires recruitment of additional AMPARs to synapses ( Huganir and Nicoll , 2013; Kessels and Malinow , 2009 ) . This has been long associated with the GluA1 subunit ( Hayashi et al . , 2000; Zamanillo et al . , 1999 ) , and explained by activity-dependent trafficking requiring the GluA1 CTD ( Shi et al . , 2001 ) . However , this model has recently been challenged ( Granger et al . , 2013 , see also Kim et al . , 2005 ) . Our experiments show that the GluA1 NTD is essential for synaptic anchoring under basal conditions ( Figure 5 ) . To assess whether the NTD also plays a role in synaptic plasticity , we utilized two potentiation protocols to compare GluA1 and GluA1 ΔNTD: ( i ) expression of constitutively active CaMKII ( tCaMKII ) ( Hayashi et al . , 2000 ) , a kinase essential for LTP ( Hell , 2014 ) , and ( ii ) electrical stimulation using a pairing protocol . Neurons transfected with tCaMKII gave rise to significantly enhanced EPSCs ( Figure 8—figure supplement 1A1 ) , in line with earlier work ( Hayashi et al . , 2000 ) . Expression of tCaMKII together with GluA1 similarly potentiated responses ( Figure 8A1 ) and showed inward rectification ( Figure 8A2 cf . Figure 8—figure supplement 1A2 ) . RI does not appear to differ from expression of GluA1 alone ( Figure 5C1 ) , indicating that the potentiation is mediated by both recombinant GluA1 homomers , and native AMPARs ( such as GluA1/2 heteromers ) . However , synapses expressing tCaMKII with GluA1 ΔNTD failed to potentiate ( Figure 8B1 ) . Given that the EPSCs showed rectification ( Figure 8B2 ) , the potentiating stimulus appears to drive these receptors into the synapse , as described previously ( Hayashi et al . , 2000 ) , yet without their NTD they are unable to maintain a potentiated state . 10 . 7554/eLife . 23024 . 014Figure 8 . LTP is impaired without the NTD of GluA1 . ( A1 ) Co-expression of GluA1 and tCaMKII increases AMPAR EPSC amplitude ( untransfected: 32 . 1 ± 2 . 0 pA; GluA1+tCaMKII: 44 . 9 ± 4 . 4 pA; n = 42; paired t-test , p=0 . 006 ) . Scale bar = 10 ms and 20 pA . ( A2 ) RI from above cells ( untransfected: 0 . 47 ± 0 . 02; GluA1 + tCaMKII: 0 . 34 ± 0 . 03; n = 42; paired t-test , p=0 . 0003 ) . ( B1 ) Potentiation mediated by tCaMKII is impaired in GluA1 ΔNTD expressing cells ( untransfected: 39 . 3 ± 3 . 3 pA; GluA1 ΔNTD + tCaMKII: 31 . 6 ± 3 . 7 pA; n = 32; paired t-test , p=0 . 117 ) . Scale bar = 10 ms and 20 pA . ( B2 ) RI from above cells ( untransfected: 0 . 49 ± 0 . 04; GluA1 ΔNTD + tCaMKII: 0 . 38 ± 0 . 02; n = 33; paired t-test , p=0 . 019 ) . ( C1 ) AMPAR EPSC amplitudes from cells expressing GluA1 or GluA1 ΔNTD over time , averaged in one-minute bins . At time = 0 LTP was induced using a pairing protocol ( 2 Hz , 100 s at −10 mV holding potential ) . EPSC amplitudes are normalized to pre-induction amplitude . LTP is maintained past 35 min in cells expressing GluA1 , but not GluA1 ΔNTD . Normalized amplitude at 45 mins: GluA1: 2 . 35 ± 0 . 49 ( n = 16 ) ; GluA1 ΔNTD: 1 . 20 ± 0 . 17 ( n = 14 ) . ( C2 ) Representative AMPAR EPSC traces from cells expressing GluA1 or GluA1 ΔNTD . Traces show EPSCs before induction ( baseline ) and 15 and 40 min after induction . Scale bar = 10 ms and 20 pA . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 01410 . 7554/eLife . 23024 . 015Figure 8—figure supplement 1 . Synaptic potentiation using tCaMKII and an electrical pairing protocol . ( A1 ) Scatter plot of AMPAR EPSCs from tCaMKII-EGFP and untransfected cells showing successful potentiation ( untrans . : 33 . 9 ± 2 . 9 pA , tCaMKII: 46 . 4 ± 2 . 8 pA; n = 28; paired t-test , p=0 . 0018 ) . ( A2 ) RI is unchanged on tCaMKII expression ( untrans . : 0 . 51 ± 0 . 03 , tCaMKII: 0 . 48 ± 0 . 02; n = 28; paired t-test , p=0 . 369 ) . ( B1 ) Average AMPAR EPSC amplitudes over time normalized to pre-LTP amplitudes ( with LTP induced at t = 0 ) from untransfected cells shows potentiation at 45 mins post induction ( 1 . 87 ± 0 . 34 , n = 9 ) . ( B2 ) AMPAR EPSCs from a subset of GluA1 and GluA1 ΔNTD-expressing cells ( see Figure 8C1 ) . Control pathways from both conditions remained comparable and stable over recording duration . Only test pathway of GluA1 remained potentiated at 45 mins ( GluA1: n = 7; GluA1 ΔNTD: n = 10 ) . ( C1 ) Construct schematic and RI of GFP-GluA2Q expression ( untrans . : 0 . 477; GFP-GluA2Q: 0 . 298; n = 9; paired t-test , p=0 . 264 ) . ( C2 ) Scatter plot of AMPAR EPSCs from GFP-GluA2Q-expressing and untransfected cells showing no change in amplitude ( untrans . : 43 . 3 pA; GFP-GluA2Q: 46 . 4 pA; n = 16; paired t-test , p=0 . 389 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 01510 . 7554/eLife . 23024 . 016Figure 8—figure supplement 2 . The sequence diversity of the AMPAR NTD mediates subunit-selective synaptic anchoring . ( A ) AMPAR structural model colored by sequence conservation across AMPAR paralogs and orthologs highlighting the potential of the NTD and CTD for subunit specific interactions ( Ashkenazy et al . , 2016 ) , while the channel core ( LBD and TMD ) is highly conserved . ( B ) A hypothetical model of synaptic AMPAR anchoring , whereby CTD interactors ( brown ) facilitate postsynaptic localization , however stabilization of the receptor to engage in functional transmission requires N-terminal interactions , with factors yet to be identified ( grey ) ( Post = postsynapse , Pre = presynapse ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23024 . 016 To investigate this observation further , we examined LTP in neurons expressing GluA1 either with or without its NTD . Using a pairing protocol , LTP could be reliably induced in both untransfected cells ( Figure 8—figure supplement 1B1 ) and cells expressing GluA1 ( Figure 8C ) , with enhanced transmission maintained for at least 45 min after induction . However , although neurons in which the extrasynaptic pool of receptors contained GluA1 ΔNTD showed a transient potentiation , EPSC amplitudes had returned to baseline levels 30–35 min after induction ( Figure 8C ) . In a subset of recordings , a second stimulation pathway was included , in which no potentiation was induced ( Figure 8—figure supplement 1B2 ) . This pathway showed similar amplitudes and no LTP induction in either condition , confirming the stability of recordings , yet the effect on LTP seen in test pathways was clearly exhibited . Thus , taken together with tCaMKII expression data , NTD-dependent interactions are essential for AMPA receptor anchoring , which is a prerequisite for synaptic potentiation . AMPAR insertion into synapses has emerged as a central mechanism underlying the expression of LTP ( Durand et al . , 1996; Isaac et al . , 1995; Liao et al . , 1995 ) . Regulation of AMPAR trafficking to ( and from ) synapses involves lateral diffusion ( Choquet and Triller , 2013 ) and vesicular trafficking ( Newpher and Ehlers , 2008 ) . These events have been mostly ascribed to the receptor CTD: 50–80 residue long cytosolic extensions that vary in sequence , are selectively phosphorylated , and interact with scaffolding and actin-binding proteins in a subunit-selective manner ( Anggono and Huganir , 2012; Shepherd and Huganir , 2007 ) . Recent work suggests that the tails only play a modulatory role in LTP , raising the possibility that other segments of the receptor are essential for synaptic targeting ( Granger et al . , 2013 ) . Here we demonstrate that the N-terminal domain is a central player in this process , and participates in AMPAR anchoring at synapses in a subunit-selective manner . AMPARs rapidly diffuse in the plane of the membrane and are trapped at postsynaptic sites upon LTP ( Choquet and Triller , 2013; Opazo et al . , 2012 ) . Receptor trapping and clustering may occur selectively opposite presynaptic release sites , to ensure optimal receptor activation on neurotransmitter release ( Lisman et al . , 2007; Raghavachari and Lisman , 2004; Tang et al . , 2016 ) . The NTD , which projects about mid-way into the synaptic cleft , is ideally suited to engage interaction partners . Structural data ( Dürr et al . , 2014; Herguedas et al . , 2016; Meyerson et al . , 2014; Nakagawa et al . , 2005; Sukumaran et al . , 2011 ) and receptor simulations ( Dutta et al . , 2015; Krieger et al . , 2015 ) have shown that this domain layer is highly dynamic and could therefore ‘sample’ the local environment for binding partners ( García-Nafría et al . , 2016a ) . Some NTD-interactors have been identified: secreted pentraxins ( O'Brien et al . , 1999; Sia et al . , 2007 ) , the adhesion molecule N-cadherin ( Saglietti et al . , 2007 ) , and AMPAR auxiliary subunits ( Cais et al . , 2014 ) , but whether any of these molecules participate in selective synaptic anchoring remains to be established . The NTD is highly sequence diverse ( Figure 8—figure supplement 2A ) with only 56% sequence identity between subunits in rodents . Our data are most compatible with an NTD-anchoring mechanism that is subunit-selective , mediated by this diversity , and supports some of the previous work on subunit-specific AMPAR trafficking ( Malinow et al . , 2000; Shi et al . , 2001 ) . In line with these studies , we show that GluA2 integrates into synapses more efficiently than GluA1 , an observation that is supported by previous AMPAR knockout data . In a conditional GluA2/3 knockout , GluA1 homomers are unable to maintain full synaptic transmission , despite providing a complete extrasynaptic pool of receptors ( Lu et al . , 2009 ) , a deficit that is alleviated with GluA2 also present . As we show that replacement of the GluA1 NTD with that of GluA2 facilitates more robust incorporation into the synapse , N-terminal domain interactions appear to effectuate this subunit-specific anchoring . We also demonstrate that the GluA2 CTD facilitates synapse targeting , but alone , it is unable to stably position the receptor in the absence of the NTD . The depression in basal transmission seen in GluA2Q ΔNTD expressing cells presumably occurs through sequestration of critical , subunit-specific interactors from native receptors , as this can be alleviated through CTD exchange with GluA1 ( Figure 7—figure supplement 1B ) . Interestingly , the GluA1 CTD is insufficient to deliver GluA1 to synapses and this subunit strictly depends on its NTD for targeting and anchoring . In potentiating the synapse the GluA1 CTD has been described as both essential ( Hayashi et al . , 2000; Shi et al . , 2001 ) and dispensable ( Granger et al . , 2013 ) . While we observe that tCaMKII expression appears to drive synaptic incorporation of GluA1 ΔNTD , without the AMPAR NTD , potentiation does not occur . Interestingly , potentiation cannot be achieved by native receptors in these cells , indicating a possible sequestration of important CTD interactors by GluA1 ΔNTD in a similar manner to GluA2 under basal conditions , which is reminiscent of the conclusions of Shi et al . ( 2001 ) . It is clear from somatic patch recordings that exogenously expressed receptors comprise the vast majority of the extrasynaptic pool . When LTP is induced in GluA1 ΔNTD-expressing cells , rapid short-term potentiation is observed , yet this potentiation cannot be maintained . The transient short-term potentiation most likely requires NTD-independent AMPAR trafficking mechanisms , such as TARP phosphorylation ( Opazo et al . , 2010 ) , highlighting the fine interplay of interactions that dictates AMPAR delivery to synapses . However , the synaptic rearrangements required for LTP appear critically dependent on the AMPAR NTD . We propose that a key requirement for LTP expression is stable receptor anchorage via the NTD , when the synapse is rearranged to a potentiated state . Based on these data , we hypothesize that CTD interactions are important in accruing receptors at postsynaptic sites , but NTD interactions are key for positioning or stabilizing the AMPAR for effective transmission ( Figure 8—figure supplement 2B ) . The GluA2 NTD’s affinity for the synaptic sites may allow critical control of synaptic signaling . As GluA2 renders AMPARs calcium-impermeable , this interaction has the potential to bias for GluA2-containing receptors , preventing excitotoxicity and controlling the potential signaling of calcium-permeable AMPARs ( Cull-Candy et al . , 2006 ) . Our results shed light on some controversies in the literature . Whereas GFP-GluA1 is unable to traffic to the synapse without a potentiating stimulus ( Hayashi et al . , 2000; Shi et al . , 2001 ) , constitutive synaptic trafficking of GluA1 has been described , with the discrepancy being attributed to the N-terminal GFP tag ( Granger et al . , 2013 , but see Nabavi et al . , 2014 ) . This can be explained by the essential requirement for the GluA1 NTD that we describe ( Figure 5 ) . Similarly , we report an increase in AMPAR EPSC amplitude upon expression of GluA2Q ( Figure 2B ) , likely mediated by the increased channel conductance , which was not seen using N-terminally GFP-tagged GluA2Q ( Shi et al . , 2001 ) . The authors reported changes in rectification ( and hence synaptic delivery of GluA2-GFP ) but their construct did not give rise to elevated AMPAR amplitudes . We tested whether the presence of the tag could explain the discrepancy with our data . Indeed , inserting GFP upstream of the GluA2 NTD reduced current amplitudes to levels of untransfected control neurons ( Figure 8—figure supplement 1C ) , supporting the hypothesis that the GFP tag interferes with GluA2 synaptic anchoring . This effect has implications for interpretation of our FRAP data , but given the clear functional difference between GFP-GluA2 and GluA2 ΔNTD , it appears that the GFP tag impairs but does not abolish GluA2 anchorage . Recent studies have highlighted a fundamental role for glutamate receptor NTDs in synapse operation and architecture ( Elegheert et al . , 2016; Matsuda et al . , 2016 ) . In each case , presynaptic cell adhesion molecules have been identified as critical interaction partners . Given the recently emerging role for synaptic adhesion molecules in LTP ( Aoto et al . , 2013; Shipman and Nicoll , 2012; Soler-Llavina et al . , 2013 ) and structured alignment of the postsynapse with presynaptic release machinery ( Tang et al . , 2016 ) , transsynaptic interactions are likely to play a key role in controlling AMPAR signaling . The N-terminal domain now emerges as a prime candidate to mediate these effects . Rat sequence AMPAR subunits GluA1 and GluA2 ( flip and R/G edited ) were expressed from the pRK5 vector . All cloning procedures were performed using IVA cloning ( García-Nafría et al . , 2016b ) . GluA2 mutation R586Q was used for all experiments ( GluA2Q ) . GluA1 and GluA2 ΔNTD were created by simultaneous deletion of the NTD coding region ( GluA1 residues 1–373 , GluA2 1–377 ) and replacement with a c-myc epitope sequence immediately after the signal sequence . The tCaMKII-EGFP construct was a gift from José Esteban , and has been described previously ( Shi et al . , 2001 ) . GluA1/2 NTD swap constructs were created by exchange of NTD and NTD-LBD linker sequences ( GluA1 residues 1–390 , GluA2 1–394 ) GluA2 ΔNTD +A1CTD tail swap construct was created by exchange of the entire C-terminal sequence of GluA2 ( 813–862 ) with that of GluA1 ( 809–889 ) . Super-ecliptic pHluorin ( SEP ) or GFP tagged GluA2 was produced by insertion of fluorescent protein coding region between the third and fourth residues of the mature GluA2 protein , preceded and followed by an ‘Ala-Ser’ dipeptide linker . SEP sequence was a gift from Jonathan Hanley . All procedures were carried out under PPL 70/8135 in accordance with UK Home Office regulations . Experiments conducted in the UK are licensed under the UK Animals ( Scientific Procedures ) Act of 1986 following local ethical approval . Organotypic slice cultures were prepared as described previously ( Stoppini et al . , 1991 ) . Briefly , hippocampi from P6-8 C57/Bl6 mice were isolated in high-sucrose Gey’s balanced salt solution containing ( in mM ) : 175 Sucrose , 50 NaCl , 2 . 5 KCl , 0 . 85 NaH2PO4 , 0 . 66 KH2PO4 , 2 . 7 NaHCO3 , 0 . 28 MgSO4 , 2 MgCl2 , 0 . 5 CaCl2 and 25 glucose at pH 7 . 3 . Hippocampi were cut into 300 μm thick slices using a McIlwain tissue chopper and cultured on Millicell cell culture inserts ( Millipore Ltd ) in equilibrated slice culture medium ( 37°C/5% CO2 ) . Culture medium contained 78 . 5% Minimum Essential Medium ( MEM ) , 15% heat-inactivated horse serum , 2% B27 supplement , 2 . 5% 1 M HEPES , 1 . 5% 0 . 2 M GlutaMax supplement , 0 . 5% 0 . 05 M ascorbic acid , with additional 1 mM CaCl2 and 1 mM MgSO4 ( all from Thermo Fisher Scientific; Waltham , MA ) . Medium was refreshed every 3–4 days . Cultures were transfected at 4–7 days in vitro ( DIV ) by single-cell electroporation ( SCE ) and recordings were performed 4–6 days after transfection . Mice with floxed loci at Gria1 , 2 and 3 genes [Gria1lox/lox ( RRID:IMSR_JAX:019012 ) , Gria2lox/lox ( RRID:IMSR_EM:09212 ) , Gria3lox/lox ( RRID:IMSR_EM:09215 ) ] were a gift from Rolf Sprengel ( MPI - Heidelberg ) and were interbred to produce mice homozygous for all floxed alleles ( Gria1lox/lox; Gria2lox/lox; Gria3lox/lox , denoted Gria1-3fl ) . 0 . 5 μl of AAV9-hSyn-Cre-GFP ( Penn Vector Core , USA ) ( titre - 2 × 1012 GC/ml ) was injected into each hippocampus of Gria1-3fl mice at postnatal day 0–1 ( P0/1 ) using a borosilicate glass micropipette and a 5 µL syringe ( Model 75 , Hamilton Company; Reno , NV ) . Pups were anaesthetized with 4 % Isoflurane in an anesthetic induction chamber for 3–4 min and subsequently transferred to a stereotactic rig where they were subjected to intracerebral injection , with anesthetic maintained throughout the procedure . Following recovery , pups were returned to their home cage and were used at P6-8 for the preparation of organotypic slices . Organotypic slices were transfected using an adapted version of the single-cell electroporation method described in ( Rathenberg et al . , 2003 ) . DNA plasmids were diluted to 33 ng/µL with potassium-based intracellular solution and the mixture was back-filled into borosilicate microelectrode pipettes . Slices were submerged in HEPES-based artificial cerebrospinal fluid ( aCSF ) containing ( in mM ) : 140 NaCl , 3 . 5 KCl , 1 MgCl2 , 2 . 5 CaCl2 , 10 HEPES , 10 Glucose , 1 sodium pyruvate , 2 NaHCO3 , at pH 7 . 3 . Plasmids were introduced into individual cells by the application of a short burst of current pulses ( 60 pulses at 200 Hz ) while in cell-attached mode . To visualize transfected cells , pN1-EGFP ( Clontech; Mountain View , CA ) was routinely mixed with AMPAR-expressing plasmids at a base pair ratio of 1:7 . In the CaMKII experiments , the ratio between tCaMKII-EGFP and AMPAR-expressing plasmids was 1:1 . All procedures were carried out in accordance with UK Home Office regulations . Briefly , E18 Sprague Dawley rats were sacrificed , embryonic hippocampi were isolated in HEPES-buffered Hank’s balanced saline solution ( Thermo Fisher Scientific ) and hippocampal cells were dissociated using 0 . 25 % trypsin ( Thermo Fisher Scientific ) . Cells were cultured on glass coverslips ( Hecht Assistent; Germany ) coated with poly-L-lysine ( Sigma-Aldrich; UK ) and maintained in equilibrated culture medium ( 37°C/5% CO2 ) containing Neurobasal Medium , B27 supplement ( 0080085SA ) and GlutaMax ( all from Thermo Fisher Scientific ) . Cultures were transfected using Lipofectamine 2000 ( Thermo Fisher Scientific ) at 14–16 days in vitro and used 3–6 days after transfection . Transfected hippocampal slice cultures were submerged in aCSF containing ( in mM ) : 125 NaCl , 2 . 5 KCl , 1 . 25 NaH2PO4 , 25 NaHCO3 , 10 glucose , 1 sodium pyruvate , 4 CaCl2 , 4 MgCl2 and 0 . 001 SR-95531 at pH 7 . 3 and saturated with 95% O2/5% CO2 . 100 μM D-APV was used to isolate AMPAR currents for mEPSC and rectification index recordings . With the exception of mEPSC recordings , 2 μM 2-chloroadenosine was added to aCSF to dampen epileptiform activity . 1 μM tetrodotoxin was included in aCSF for miniature EPSC ( mEPSC ) recordings . All drugs were purchased from Tocris Bioscience . 3–6 MΩ borosilicate pipettes were filled with intracellular solution containing ( in mM ) : 135 CH3SO3H , 135 CsOH , 4 NaCl , 2 MgCl2 , 10 HEPES , 4 Na2-ATP , 0 . 4 Na-GTP , 0 . 15 spermine , 0 . 6 EGTA , 0 . 1 CaCl2 , at pH 7 . 25 . Paired recordings involved simultaneous recording from a neighboring pair of GFP positive and negative cells . EPSCs were evoked by simulation of Schaffer collaterals in the stratum radiatum of CA1 using a monopolar glass electrode , filled with aCSF . Recordings were collected using a Multiclamp 700B amplifier ( Axon Instruments ) . Recordings during which the series resistance varied by more than 20% or exceeded 20 MΩ were discarded . mEPSC detection was conducted using a template-based search in Clampfit ( Molecular Devices ) . Cumulative frequency plot was produced using equal numbers of events from all cells within each condition to prevent misrepresentation . Regarding interpretation of mEPSC data , it is of note that changes in mEPSC amplitude and frequency require careful interpretation due to the event detection limit . A postsynaptic increase in event amplitude will cause previously sub-threshold events to be detected , and therefore , while the average event amplitude will not change , this would instead be represented as an increase in mEPSC frequency . Rectification index was calculated by recording AMPAR currents from cells held at −60 , 0 and +40 mV , using the following equation:RI= − ( I+40−I0 ) ( I−60−I0 ) AMPAR/NMDAR EPSCs were compared by recording synaptic currents at −60 and +40 mV . AMPAR current amplitudes were quantified as the peak current at −60 mV . NMDAR amplitudes are measured at +40 mV , 100 ms after response initiation . Paired-pulse ratio was calculated from two AMPAR currents , stimulated at an interval of 50 ms . For LTP recordings , aCSF contained ( in mM ) : 119 NaCl , 2 . 5 KCl , 1 Na2HPO4 , 26 NaHCO3 , 4 CaCl2 , 4 MgCl2 , 11 glucose , 0 . 002 2-chloroadenosine and 0 . 01 SR-95531 and glass pipettes were filled with intracellular solution containing ( in mM ) : 115 CsCH3SO3 , 20 CsCl , 10 HEPES , 2 . 5 MgCl2 4 Na2-ATP , 0 . 4 Na-GTP , 10 phosphocreatine , 0 . 1 spermine at pH 7 . 3 . Slices were maintained at 25°C throughout the recordings . LTP was induced by depolarization of the cell to −10 mV while stimulating the test pathway at 2 Hz for 100 s . The control pathway did not receive input during this period . Outside-out patches were pulled from GFP positive or negative CA1 cell bodies and patches were subjected to fast-exchange perfusion in HEPES-based aCSF ( see SCE ) containing 100 uM cyclothiazide , with or without 1 mM L-glutamate . In voltage-clamp mode , a 500 ms holding potential ramp from −100 mV to +100 mV was applied to patches . Recordings in the absence of glutamate were subtracted from those in the presence of glutamate and −60 mV , 0 mV and +40 mV current amplitudes were used to calculate rectification index as described above . Miniature EPSC recordings , digitized at 100 kHz , were subjected to noise analysis using a custom program running in MATLAB ( MathWorks ) ( supplied by Andrew Penn , University of Sussex; available on MATLAB File Exchange , ID: 61567; https://uk . mathworks . com/matlabcentral/fileexchange/61567-peaker-analysis-toolbox ) following ( Hartveit and Veruki , 2007 ) and ( Benke et al . , 2001 ) . Briefly , events were detected using a template-based search ( Pernía-Andrade et al . , 2012 ) , aligned by their point of steepest rise and peak scaled to account for differences in synaptic receptor number . Traces were filtered to those with a 10–90% rise time of less than 0 . 9 ms and subjected to visual inspection to eliminate obvious artifacts , overlapping mEPSCs or insufficient peak alignment . Correlations between peak amplitude , rise and decay times were analyzed to detect and eliminate cells with excessive electrical filtering . Following elimination of suboptimal events , only cells with at least 20 successful events were included for variance analysis . Variance vs amplitude plots were produced for binned decay phase data of mEPSCs ( 15 bins ) and were fitted with a parabolic curve with the equation:σ2 ( I ) =iI−I2N+σb2 from which single-channel current ( i ) could be calculated , being proportional to the initial gradient of the parabolic curve . Single-channel conductance is related to current by the equation;γ=i ( Vm−Erev ) where membrane potential ( Vm ) and reversal potential ( Erev ) were −60 mV and 0 mV respectively . To visualize dendritic spines , 1 mg/ml Lucifer Yellow was added to the intracellular solution . Cells were maintained in a whole-cell configuration for 10 min before live imaging on an inverted Leica SP8 confocal microscope in SCE extracellular solution . Z-stacks of 50 μm regions of secondary dendrite were imaged using a 63X oil-immersion objective , deconvolved ( Huygens Professional ) , and segmented ( Imaris ) , before manual counting of spines . Hippocampal cultures were cotransfected ( 1:1 ) with pN1-mCherry ( Clontech ) and SEP-GluA2 or SEP-GluA2 ΔNTD and imaged in aCSF containing ( in mM ) : 150 NaCl , 2 . 5 KCl , 2 MgCl2 , 2 CaCl2 , 20 HEPES , 10 Glucose at pH 7 . 3 in a heated chamber at 37°C . Images were acquired on a Leica SP8 confocal microscope using a 63X objective lens at 30 s intervals . Photobleaching was achieved by repetitive xy scanning of the region of interest at high laser intensity . Fluorescence during bleaching was monitored to ensure steady state complete bleaching was achieved and bleaching parameters were constant for all samples and repetitions . Analysis was conducted using Image J ( Schneider et al . , 2012 ) . Photobleaching due to image acquisition was corrected by normalization to non-photobleached spines or dendrites , distant to a bleached spine . HEK293T cells ( ATCC Cat# CRL-11268 , RRID:CVCL_1926 , Lot 58483269: identity authenticated by STR analysis , mycoplasma negative ) were co-transfected with pN1-EGFP and AMPAR constructs using Effectene ( QIAGEN; Germany ) . Two days post-transfection , cells were washed in phosphate buffered saline ( PBS ) and incubated with AF647 conjugated primary antibody ( anti-myc 9E10 , Santa Cruz Biotechnology; Dallas TX , RRID:AB_627268 ) for 30 mins on ice in PBS containing 10% fetal bovine serum ( FBS ) . Antibody was removed and cells were washed further in PBS before resuspension in PBS containing 10% FBS and 1:1000 DAPI . Flow cytometry was performed using a LSR II flow cytometer ( BD; Franklin Lakes , NJ ) . AF647 fluorescence was quantified and represents construct surface expression . Cells either positive for DAPI fluorescence or negative for EGFP fluorescence were discarded from analysis as dead or untransfected . AF647 fluorescence of untransfected cells was measured and subtracted during quantifications of surface expression . All data are presented as Mean ± Standard Error of the Mean ( SEM ) . With two-sample comparisons , paired or unpaired Student’s t-tests are applied as appropriate . For multiple sample comparisons , One-way ANOVA with a Tukey’s multiple comparisons test was used .	Neurons send signals via electrical impulses that are transmitted between cells by small molecules known as neurotransmitters . The information is passed from neuron to neuron at specialized points of contact termed synapses . On release of neurotransmitters from the first neuron , the molecules attach to ‘docking stations’ called receptors on the next neuron , referred to as the postsynaptic cell . One of these receptors , the AMPA receptor , transmits signals by binding to a neurotransmitter called glutamate . Previous research has shown that in order to bind glutamate effectively , these receptors need to be trapped and anchored at the correct location at the synapse . This trapping mechanism controls the number of receptors present , which strengthens the synapse , and ultimately mediates learning and memory . However , it is still not clear how AMPA receptor trapping is achieved . To investigate this question , Watson et al . examined how AMPA receptors ( and mutant forms of the receptor ) affect the communication between neurons using brain slices from mice . The experiments show that an external segment of the AMPA receptor called the N-terminal domain ( or NTD for short ) is a key element for receptor anchoring at the postsynapse . The AMPA receptor is made out of four different subunits; when the NTD portion was removed from one specific subunit , fewer receptors were anchored correctly at the postsynapse . When the NTD was removed from another subunit , it completely prevented the synapse from learning . Therefore , the NTD brings about subunit-selective anchoring of the AMPA receptor , which affects the ability of the synapse to transmit signals . Important next steps would be to identify the proteins that interact with the NTD and how this specific anchoring affects the strength of the synapse . Another key step will be to understand what mechanisms control the number of AMPA receptors at synapses , to ultimately enable learning .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2017	Synaptic transmission and plasticity require AMPA receptor anchoring via its N-terminal domain
Reef-building corals depend on intracellular dinoflagellate symbionts that provide nutrients . Besides sugars , the transfer of sterols is essential for corals and other sterol-auxotrophic cnidarians . Sterols are important cell components , and variants of the conserved Niemann-Pick Type C2 ( NPC2 ) sterol transporter are vastly up-regulated in symbiotic cnidarians . Types and proportions of transferred sterols and the mechanism of their transfer , however , remain unknown . Using different pairings of symbiont strains with lines of Aiptasia anemones or Acropora corals , we observe both symbiont- and host-driven patterns of sterol transfer , revealing plasticity of sterol use and functional substitution . We propose that sterol transfer is mediated by the symbiosis-specific , non-canonical NPC2 proteins , which gradually accumulate in the symbiosome . Our data suggest that non-canonical NPCs are adapted to the symbiosome environment , including low pH , and play an important role in allowing corals to dominate nutrient-poor shallow tropical seas worldwide . Many plants and animals cultivate symbioses with microorganisms for nutrient exchange . Cnidarians , such as reef-building corals and anemones , establish an ecologically critical endosymbiosis with photosynthetic dinoflagellate algae ( Douglas , 2010 ) ( family Symbiodiniaceae ) ( LaJeunesse et al . , 2018 ) . Their symbionts reside within endo/lysosomal-like organelles , termed symbiosomes , and transfer photosynthetic products to their hosts ( Muscatine , 1990; Yellowlees et al . , 2008 ) . In addition to sugars that mostly provide energy , recent studies hint at the importance of the transfer of various lipids including sterols ( Crossland et al . , 1980; Battey and Patton , 1984; Revel et al . , 2016 ) . Sterols are essential building blocks for the cell membrane and endomembrane systems , in the form of cholesterol and other sterol variants . Cnidarians are sterol auxotrophs ( Baumgarten et al . , 2015; Gold et al . , 2016 ) that must acquire these essential compounds from diet and/or symbionts ( Goad , 1981 ) . In line with this , non-canonical variants of the conserved cholesterol transporter Niemann-Pick Type C2 ( NPC2 ) are among the most up-regulated genes in symbiotic Exaiptasia pallida ( commonly Aiptasia ) and Anemonia viridis anemones ( Dani et al . , 2014; Lehnert et al . , 2014; Kuo et al . , 2010; Ganot et al . , 2011; Wolfowicz et al . , 2016 ) . Dinoflagellates synthesize various sterols , many of which are found in symbiotic cnidarians ( Bohlin et al . , 1981; Withers et al . , 1982; Ciereszko , 1989 ) ; however , the specific combinations of transferred sterols , as well as the mechanism of this transfer remain unknown . To what extent is the specific mix of transferred sterols controlled by the host , symbiont , or both – reflecting physiological relevance – and how is such selective transport achieved ? To answer these questions , we took advantage of the availability of distinct strains of Symbiodiniaceae symbionts with different and complex sterol compositions ( Bohlin et al . , 1981; Withers et al . , 1982; Ciereszko , 1989 ) , and of various hosts . Besides the coral Acropora digitifera , we investigated different host lines of the symbiotic anemone Aiptasia , an emerging model system for coral-algal symbiosis ( Tolleter et al . , 2013; Neubauer et al . , 2017 ) . We used gas chromatography/mass spectrometry ( GC/MS ) to semi-quantitatively profile relative sterol abundances in three compatible symbiont strains ( Xiang et al . , 2013; Hambleton et al . , 2014 ) , and related this to sterol abundances in the coral and in three Aiptasia laboratory lines ( Grawunder et al . , 2015 ) , with or without symbionts ( Figure 1 , Figure 1—source data 1 ) . First , to validate our assay and to show that algal sterols are indeed transferred to host tissue , we determined the host sterol composition without symbionts ( aposymbiotic ) , in symbiosis with recent dietary input ( two weeks since last feeding , ‘intermediate’ ) , and in symbiosis with essentially no dietary input ( five weeks since last feeding , ‘symbiotic’ ) . For the Aiptasia F003 host line , this revealed a gradual transition from an initial aposymbiotic , food-derived cholesterol profile to a cholesterol-reduced , algal sterol-enriched symbiotic profile that was also found in the symbiont-free eggs ( and is thus present in host tissue ) ( Figure 1A ) . We also compared the sterol composition of coral symbiotic polyps collected from the wild to that of their symbiont-free eggs , which again proved nearly identical sterol compositions ( Figure 1A ) and unambiguously revealed symbiont-to-host tissue transfer . Taken together , this suggests that symbiont-derived sterols can functionally replace dietary cholesterol without any further chemical conversion by the host . Moreover , the sterol content of the hosts is highly plastic , and sterols are used flexibly as they become available from food and/or symbionts . We next focused on the sterol compositions in different symbiont-host pairings , to determine how these would change upon switching of either symbiont or host line . To this end , we investigated the same Aiptasia line CC7 hosting distinct symbionts ( SSA01 or SSB01 , see Materials and methods ) with different symbiont profiles; and the same symbiont ( SSB01 ) in two distinct host lines ( CC7 and H2 ) , as well the symbiont CCMP2466 similar to that in Acropora . We found that Aiptasia CC7 hosting Symbiodiniaceae strain SSA01 contained a large proportion of stigmasterol-like sterol ( dark blue , Figure 1B ) when compared to campesterol ( light blue , Figure 1B ) . In contrast , the same Aiptasia line hosting strain SSB01 contained minimal stigmasterol-like derivatives compared to campesterol , as well as the unique sterol gorgosterol ( light blue and pink , respectively , Figure 1B ) , characterized by an unusual cyclopropyl group ( Ciereszko , 1989 ) . ( Figure 1—figure supplement 1 ) . A very similar sterol profile was observed when the same symbiont ( SSB01 ) infected the H2 host line , indicating that the host sterol profile was largely symbiont-driven . Likewise , in Aiptasia line F003 hosting both SSA01 and SSB01 , the sterol proportions reflect both symbionts: a dominance of stigmasterol-like sterol ( reflecting SSA01 ) together with gorgosterol ( reflecting SSB01 ) ( Figure 1A ) . We also compared the sterol profile of Acropora colonies collected from the wild to that of a closely related symbiont CCMP2466 in laboratory culture and found a strong enrichment for gorgosterol and campesterol at the expense of stigmasterol-like sterols – highly reminiscent of the trend previously observed in the SSB01/CC7 and SSB01/H2 pairings ( Figure 1A ) . We thus observed two major patterns of sterol transfer in our symbiont-host combinations – one enriching for stigmasterol-like sterols ( combinations SSA01/CC7 and SSA01 +SSB01/F003 ) , and another one enriching for gorgosterol and campesterol ( combinations SSB01/CC7; SSB01/H2; and CCMP2466/Acropora ) . This suggests that selective sterol transfer and/or accumulation by the host may occur . Moreover , symbionts may change their sterol synthesis profile as symbiotic vs . free-living cells . To address this , we separated anemone homogenates by centrifugation into symbiont-enriched ( pellet , although substantial host tissue remained , Figure 1—figure supplement 2A and Figure 1—source data 1 ) and symbiont-depleted ( supernatant ) fractions , for which the sterol profiles could be directly compared to free-living symbionts cultured under similar conditions . This revealed that certain sterols were absent in symbiont-enriched pellets yet present in symbiont cultures ( Figure 1C , Figure 1—figure supplement 2B ) . For example , stigmasterol/gorgosterol-like ( dark purple ) and ergost-tetrol-like sterol ( light purple ) are proportionally highly abundant in cultured symbionts , yet are basically absent in all pellet samples ( Figure 1C , Figure 1—figure supplement 2B ) . This suggests that synthesis of at least some sterols changes in residence vs . in culture , providing further support that the symbiont has a major influence on which specific composition and proportion of the sterols are transferred during symbiosis . Further , cultured symbionts exhibited some degree of plasticity of sterol profiles under various culturing conditions ( e . g . SSA01 in Figure 1A vs . Figure 1C ) . To elucidate possible molecular mechanisms how symbiont-hosting cells may influence sterol transfer from the symbiont , we focused on non-canonical members of the highly conserved NPC2 protein family ( Dani et al . , 2014; Lehnert et al . , 2014 ) . The current hypothesis in the field is that non-canonical NPC2s may specifically facilitate transfer of symbiont-produced sterols in cnidarian-algal symbiosis ( Revel et al . , 2016; Baumgarten et al . , 2015; Wolfowicz et al . , 2016; Dani et al . , 2017 ) . However , NPC2s may serve other purposes , for example signaling ( Baumgarten et al . , 2015; Dani et al . , 2017 ) , and mechanistic analyses of NPC2 function are lacking . To characterize them further , we first compared the genomic complement of NPC2 homologues in symbiotic cnidarians to that of non-symbiotic metazoans , uncovering several previously unidentified homologues in the reef-building corals and other taxa ( asterisks , Figure 2A , Supplementary file 1 ) . A Bayesian tree reconstruction placed all canonical NPC2 family members ( identified by three shared introns ) on a large multifurcation , and all previously and newly identified non-canonical NPC2 ( identified by the absence of introns due to retrotransposition [Dani et al . , 2014] ) to a basal position , most likely attracted by the Capsaspora outgroup NPC2s . This indicated higher sequence divergence of non-canonical NPC2s; and in line with this , they contain only around half as many residues under negative ( purifying ) selection ( 35 to 61 ) as canonical NPC2s and twice as many residues under positive ( diversifying ) selection ( 12 to 5 ) ( Figure 2B ) . Our analysis also revealed that non-canonical NPC2 homologues are confined to cnidarians within the anthozoan class , as they did not appear in the earlier-branching sponge Amphimedon nor in the hydrozoans Hydra magnipapillata and Hydractinia echinata . Notably , the occurrence of non-canonical NPC2s appeared to correlate with symbiotic state: the symbiotic anthozoans ( Aiptasia , Acropora , Montastrea ) have several non-canonical NPC2 homologues ( 3 , 3 , and 2 , respectively ) . In contrast , the non-symbiotic anemone Nematostella displays evolutionary traces of a single non-canonical NPC2 , which either failed to expand or underwent higher loss ( Figure 2A ) . We next investigated the expression of all Aiptasia NPC2s in vitro and in vivo ( Figure 3 ) . As determined by qPCR and Western blotting using custom-made antibodies ( Figure 3—figure supplement 1 ) , two of the three non-canonical NPC2 homologues displayed substantially higher expression at the transcript and protein levels in symbiotic but not aposymbiotic animals ( closed blue symbols; Figure 3A+B ) . The third non-canonical NPC2 homologue was highly expressed in both symbiotic and aposymbiotic animals , yet more so in symbiotic animals . Conversely , canonical NPC2s were highly expressed in both symbiotic and aposymbiotic animals ( closed red symbols ) . Likewise , the non-symbiotic anemone Nematostella exhibited ubiquitously high expression of canonical NPC2 genes ( open red symbols ) , whereas the non-canonical NPC2 gene was highly expressed only upon feeding ( open blue symbols ) . Aposymbiotic embryos of the symbiotic coral Acropora , as well as Nematostella embryos , contained maternally provided canonical NPC2 transcripts , suggesting that these are required for development ( Figure 3—figure supplement 2 ) . Notably , several canonical NPC2s in Aiptasia ( XM_021046710 , XM_021041174 ) and Nematostella ( XM_001635452 ) may be ‘in transition’ to becoming non-canonical: they were expressed at intermediate abundances between the two groups , and they responded to symbiosis ( Aiptasia ) or feeding ( Nematostella ) ( red square and triangles , Figure 3A + 3B ) . Some of their intron/exon structures reflected those of the non-canonical group ( red triangles , Figure 2A ) . Immunofluorescence analysis revealed that the non-canonical NPC2s decorated intracellular symbionts in Aiptasia in vivo ( Figure 3C + D ) , consistent with previous data in Anemonia viridis ( Dani et al . , 2014; Dani et al . , 2017 ) . The NPC2 signal appears to be restricted to the symbiosome and absent from the cytoplasm of the symbiont-containing cell ( Figure 3—figure supplement 3 ) . We noted that non-canonical NPC2s decorate some but not all symbionts ( Dani et al . , 2014; Dani et al . , 2017 ) , suggesting that at any given time , symbiosomes are a dynamic group of specialized organelles . To gain further insight into the NPC2-decorated symbiosome dynamics , we measured the spatio-temporal regulation of non-canonical NPC2s in Aiptasia larvae establishing symbiosis ( ‘infection’ ) with Symbiodiniaceae strain SSB01 ( Hambleton et al . , 2014; Bucher et al . , 2016 ) . Indeed , non-canonical NPC2 slowly decorated intracellular symbionts over time ( Figure 3E , Figure 3—figure supplement 4 ) . This localization ranged from weak ‘grainy’ patterns to stronger ‘halos’ around symbionts ( arrows , Figure 3C ) . We quantified infection rates , symbiont load of individual larvae , and non-canonical NPC2 signal intensity ( Figure 3F , Figure 3—figure supplement 5 ) . We found that infection rates remained steady after removal of symbionts from the environment , whereas the proportion of larvae showing non-canonical NPC2 signal continued to increase to eventually include the majority of infected larvae ( Figure 3F ) . Concordantly , the proportion of symbionts within each larva surrounded by NPC2 signal also increased over time , as did the signal strength ( Figure 3F ) . Finally , infected larvae displaying any NPC2 signal generally contained a higher symbiont load than their infected , unlabelled counterparts ( Figure 3—figure supplement 5 ) . Thus , non-canonical NPC2 is increasingly expressed and recruited to symbionts over time , suggesting that non-canonical NPC2 function becomes important primarily once symbiosomes become ‘mature’ . As a first step towards elucidating NPC2 function during symbiosis , we investigated the effect of global sterol transport inhibition by treating symbiotic and aposymbiotic adult Aiptasia with the drug U18666A , a competitive inhibitor of the NPC2 binding partner NPC1 that is required for efficient cholesterol egress from lysosomes ( Liscum and Faust , 1989; Cenedella , 2009; Vance , 2010; Lu et al . , 2015 ) . Because of the profound effect of this drug on all cells and thus anemone physiology , severe effects are to be expected . Accordingly , we found that both symbiotic and aposymbiotic anemones appear to lose tissue and shorten their tentacles in a dose- and duration-dependent manner . However , symbiotic anemones showed such effects on host physiology faster than their aposymbiotic counterparts ( Figure 3G , Figure 3—figure supplement 6 ) . Moreover , symbiont density decreased in response to U18666A treatment ( Figure 3H ) . We observed similar effects with A . digitifera juvenile primary polyps stably hosting Symbiodiniaceae strain SSB01 when exposed to increasing concentrations of U18666A ( Figure 3—figure supplement 7 ) . This suggests that inhibition of sterol transport affects symbiosis stability and may lead to loss of symbionts ( ‘bleaching’ ) . Further , the disruption of global sterol transport compromises host tissues in all cases , emphasizing the importance of sterols in tissue homeostasis . To test sterol-binding properties of Aiptasia NPC2 , we compared the most conserved canonical NPC2 to the non-canonical NPC2 most up-regulated upon symbiosis ( XM_021041171 to XM_021052404 , respectively ) . We used lipidomics to quantify lipids bound by immunoprecipitated native or recombinant NPC2s ( Figure 4 , Figure 4—figure supplement 1 ) ( after Li et al . , 2010 ) . Recombinant proteins were expressed in HEK 293T cells , after which cell lysates were mixed with Symbiodiniaceae SSB01 homogenates at either neutral conditions ( pH 7 ) or acidic conditions reflecting the lysosome/symbiosome ( pH 5 ) . Under both conditions , canonical and non-canonical NPC2:mCherry fusion proteins bound symbiont-produced sterols significantly above the background levels of the control , mCherry alone ( Figure 4A ) . The relative proportions of bound sterols generally exhibited equilibrium levels with the corresponding symbiont homogenate ( Figure 4B ) . To validate sterol binding by non-canonical NPC2 in vivo , we also immunoprecipitated the native non-canonical NPC2 and bound sterols directly from homogenates of symbiotic Aiptasia . Again , we detected symbiont-produced sterols above background levels , validating our heterologous system and indicating that these proteins bind sterols in vivo during symbiosis ( Figure 4C ) . These data indicate that , despite their evolutionary divergence , both types of Aiptasia NPC2s have the conserved function of binding sterols in lysosomal-like environments . Although we cannot rule out subtle differences in sterol binding dynamics between the two proteins , our results suggested no differential binding between canonical and non-canonical NPC2s , consistent with the observations that the sterol ligand and the residues lining the binding cavity tolerate considerable variations ( Xu et al . , 2007; Liou et al . , 2006 ) . Corroborating this , we were unable to detect any difference in the differential expression of canonical and non-canonical NPC2s between aposymbiotic and symbiotic state in three symbiont-Aiptasia pairings ( Figure 4D ) . With data suggesting both NPC2 types can bind symbiont-produced sterols , we were therefore left with the question: what is the functional advantage of localizing non-canonical NPC2s specifically in the symbiosome ? The mature symbiosome , where non-canonical NPC2 appears to function , remains poorly understood; however , extreme acidity appears to be a unique characteristic of these specialized cellular compartments . Whereas lumenal pH of classic lysosomes can range from 4 . 7 to 6 ( Johnson et al . , 2016 ) , recent work indicates that mature symbiosomes in steady-state symbiosis are even more acidic ( pH ~4 ) to promote efficient photosynthesis ( Barott et al . , 2015 ) . We therefore sought to compare the stability/solubility of representative canonical and non-canonical NPC2s at different pH ( Figure 4E + F , Figure 4—figure supplement 2 ) . Interestingly , the patterns of extracted soluble proteins vary among NPC2s: canonical NPC2 appears in one predominant form at both pH’s ( Figure 4E , red arrowhead ) , whereas one of the symbiosis-responsive non-canonical NPC2s ( XM_021052412; Figure 3A + B ) always appears in two forms in both conditions ( Figure 4E , arrowheads and asterisks ) . Strikingly , the pattern for the other symbiosis-specific non-canonical NPC2 ( XM_021052404; Figures 3A–G and 4A–C ) is distinct between pH 7 and pH 5 , with a consistently occurring additional band at higher pH ( Figure 4E , blue arrow ) . Although we cannot rule out that the additional bands reflect degradation products , we favor the interpretation that they most likely represent distinct glycoforms , which also occur in vivo ( Figure 3B ) . When quantifying the protein variant common to all samples ( Figure 4E , arrowheads ) , we found that at pH 5 , the non-canonical NPC2s were consistently more abundant in the soluble fraction than the canonical counterpart ( Figure 4F ) . For all proteins tested , the ratio of the predominant soluble protein variant at pH 7 to that at pH 5 was always >1 , indicating more solubility at pH 7 . However , the ratio was higher for Aiptasia canonical NPC2 than for the non-canonical NPC2s , indicating that the former is relatively less soluble at pH 5 ( Figure 4F ) . Taken together , symbiosis-responsive non-canonical NPC2 appears to be more soluble/stable than canonical NPC2 at a lower pH , likely characteristic of the symbiosome . In line with this , all Aiptasia non-canonical NPC2 proteins harbored glycosylation sites and a glycine followed by a histidine residue ( Figure 2B ) , which may contribute to protein stability in acidic environments ( Rudd et al . , 1994; Hanson et al . , 2009; Culyba et al . , 2011 ) . However , pH-dependent protein stability is difficult to predict and functional experiments are required to determine whether such motifs ( or others ) play a role for the adaptation to the symbiosome or not . In summary , our data reveal that the transfer of complex mixtures of symbiont-derived sterols is a key feature of anthozoan photosymbiosis ( Figure 4G ) , whereby the specific composition and proportion of transferred sterols appears to be under the control of both symbiont and host . While the non-canonical NPC2 sterol-binding proteins are part of the machinery transferring sterols from symbiont to host , they do not contribute to host sterol selection by differential expression or differential binding . Instead , our assays reveal the possibility of an increased tolerance to acidic conditions of non-canonical NPC2s and their late accumulation in the symbiosome , consistent with gradual enrichment upon increasing symbiosome acidification . We propose that whereas ubiquitously expressed canonical NPC2 homologues are ‘workhorses’ in sterol trafficking throughout the host , non-canonical NPC2s are spatiotemporally regulated to accumulate as the symbiosome matures , developing into a unique compartment optimized to promote the interaction and communication of the symbiotic partners ( Figure 4G ) . This allows symbiotic cnidarians to flexibly use symbiont-produced sterols , with reef-building corals nearly fully substituting these for prey-derived cholesterol , supporting survival in nutrient-poor environments . More broadly , our findings indicate that carbon acquisition by lipid transfer , similar to other symbioses ( Keymer et al . , 2017 ) , is a major driver of coral-algal symbiotic relationships as a means to adapt to various ecological niches by efficient exploitation of limited resources . Samples were extracted with a modified Bligh-Dyer method: briefly , either 300 µl aqueous Aiptasia or Nematostella homogenate was added to 750 µl HPLC-grade methanol , or 300 µl ultra-pure water was added to the Acropora sample already in 750 µl methanol or ethanol . After shaking at 70°C for 45 min , the mixture was extracted with 375 µl HPLC-grade chloroform and 300 µl ultra-pure water and centrifugation . The dried organic phase was then saponified with 500 µl of 5% KOH in a 9:1 methanol:water solution and incubating at 68°C for 1 hr . The mixture was then extracted with water and chloroform followed by centrifugation . Lipids in the dried organic phase were derivatized to trimethylsilyl ethers with 25–40 µl MSTFA ( #69479 , Sigma Aldrich ) at 60°C for 0 . 5–1 hr and immediately analysed . 1 µl of each mixture was injected into a QP2010-Plus GC/MS ( Shimadzu ) and with a protocol ( adapted from Schouten et al . , 1998 ) as follows: oven temperature 60°C , increase to 130°C at 20 °C/min , then increase to 300°C at 4 °C/min and hold for 10 min . Spectra were collected between m/z 40 and 850 and were analysed in GCMS PostRun Analysis Software ( Shimadzu ) by comparison to the National Institute of Standards and Technology 2011 database . Relative sterol composition as percent of total sterols were calculated from integrated peak intensity on the total ion chromatograph for each sample . Antibodies were raised against the peptides K-YGIDVFCDEIRIHLT ( XM_021052412 ) , K-AKNDIFCNSIPFNLV ( XM_021052404 ) , and K-VQNNVLCGEVTLTLM ( XM_021052381 ) coupled to the adjuvant keyhole limpet hemocyanin in rabbits ( BioScience GmbH ) . Antibodies were affinity-purified from the antisera using the synthetic peptides ( INTAVIS Bioanalytical Instruments AG ) coupled to NHS-Activated Sepharose Fast Flow 4 ( 17090601 , GE Health Care Life Sciences ) according to the manufacturer’s protocols . In dot blots , peptides dissolved in DMSO or water were spotted onto nitrocellulose membranes and allowed to dry 1 hr in a dessicant chamber . Blots were blocked in 5% milk PBS-T for 2 . 5 hr at RT and then incubated at 4°C overnight with non-canonical NPC2 antibodies diluted in 5% milk PBS-T as follows: ( XM_021052412 at 1:1000 , XM_021052404 at 1:5000 , and XM_021052381 at 1:500 ) . Blots were then incubated with HRP-coupled anti-rabbit antibody and further processed as described below for ‘Western blots’ . Two aposymbiotic or symbiotic adult Aiptasia polyps were homogenized in buffer A with 2X Halt Protease Inhibitor Cocktail ( 78430 , Thermo Fisher Scientific ) and then sonicated on ice ( Sonifier 250 , Branson Ultrasonics ) with two rounds of 25 pulses at duty cycle 40% , output control 1 . 8 . From cultured Symbiodiniaceae strain SSB01 , 1 . 2 × 107 cells were collected by gentle centrifugation . After addition of buffer A and glass beads ( 425–600 µm ) , cells were disrupted by vortexing six times for 1 min each , with 1 min on ice in between each , then further disrupted by passage through a G23 needle . All homogenates were then centrifuged at 20 , 000xg for 10 min at 4°C , and three sets of identical volumes of the supernatants were resolved on a 12% Tricine-SDS-Page gel and transferred by Western blot onto nitrocellulose membranes . Membranes were blocked for 1 hr in 5% milk PBS-T and then incubated with antibodies raised against three different non-canonical Aiptasia NPC2s ( XM_021052404 at 1:4000 , XM_021052412 at 1:1000 and XM_021052381 at 1:500 ) in 5% milk PBS-T at 4°C overnight , followed by incubation with HRP-coupled anti-rabbit ( Jackson ImmunoResearch ) at 1:10000 in 5% milk PBS-T at RT for 1 hr , and then detection with ECL ( GERPN2232 , Sigma-Aldrich ) and imaging on ECL Imager ( ChemoCam , Intas ) . For peptide-blocked controls , 40 µg of homogenate supernatant per lane was resolved on a 10% Tris-tricine-SDS-Page gel and transferred and blocked as above . Antibodies were diluted in 5% milk PBS-T ( XM_021052404 at 1:2000 , XM_021052412 at 1:500 and XM_021052381 at 1:500 ) and the corresponding immunogenic peptides solubilized in DMSO or PBS at 0 . 5 mg/ml - 1 mg/ml were added at the indicated peptide:antibody ( mass:mass ) ratios . The peptide-antibody mixtures were rotated overnight at 4°C and then incubated with the blots at 4°C for approx . 60 hr , after which blots were incubated with anti-rabbit secondary antibody and processed as above . Blots were then re-blocked , incubated with anti-alpha-tubulin antibody ( 1:1000 , T9026 , Sigma-Aldrich ) , then HRP-coupled anti-mouse ( 1:10000 , Jackson ImmunoResearch ) , and imaged as above . Supernatants of recombinant NPC2 proteins in 1 ml Buffer A or B were obtained as described above . Equal volumes of supernatant were mixed with loading dye and resolved by SDS PAGE and Western blotting as described for ‘Immunoprecipitation’ . Quantification was performed in Fiji ( Schindelin et al . , 2012 ) : for each band , the integrated density ( ID ) in a rectangular region-of-interest ( ROI ) around the band was calculated , less the background ( ID of the same ROI above the band ) . Fixed larvae in PBS at 4°C were permeabilized in PBT for 2 hr at RT . Samples were then incubated in blocking buffer ( 5% normal goat serum and 1% BSA in PBT ) overnight at 4°C and then with primary antibody diluted in block buffer for 4 hr at RT at the following concentrations: 4 . 5 µg/ml ( XM_021052404 ) , 1 . 5 µg/ml ( XM_021052412 ) , and 2 µg/ml ( XM_021052381 ) . Samples were then washed twice for 5 min with PBT at RT , twice for approx . 18 hr at 4°C , then incubated with secondary antibody ( goat anti-rabbit IgG-Alexa488; ab150089 , Abcam ) diluted to 4 µg/ml in block buffer for approx . 5 hr at RT . Samples were then washed with PBT three times for 5 min each at RT , then approx . 18 hr at 4°C . When phalloidin staining was included , samples were then washed with 1% BSA in PBS and incubated with Phalloidin-Atto 565 ( 94072 , Sigma-Aldrich ) in 1% BSA in PBS overnight at 4°C . Samples were then incubated with Hoechst 33258 at 10 µg/ml in PBT for 1 hr at RT , washed 3x with PBT for 5 min each , and then washed into PBS at 4°C overnight . PBS was replaced with 95% glycerol with 2 . 5 mg/ml DABCO , and the larvae were mounted for microscopy . In peptide-blocked controls , the corresponding immunogenic peptides dissolved in PBS or DMSO at 0 . 5 mg/ml – 1 mg/ml were added to diluted primary antibodies ( XM_021052412 at 1:200 [1 . 5 µg/ml] , XM_021052404 at 1:750 [0 . 6 µg/ml] , and XM_021052381 at 1:200 [2 µg/ml] ) and rotated at 1 hr at RT before being added to samples , which were then processed as described . Symbiotic and aposymbiotic Aiptasia polyps were allowed to attach for 2 d in 6-well culture plates before exposure to U18666A ( U3633 , Sigma Aldrich ) in DMSO at the indicated concentrations in FASW; final percentage of DMSO was <0 . 05% . Polyps were cultured at 26°C at 12:12 L:D and photographed daily , followed by wash and drug re-addition . Symbiont density per anemone was quantified by homogenization in 200 µl ultrapure water with 0 . 01% SDS using a 23G needle and 1 ml syringe , after which samples were quantified for cells by visual particle counter ( TC20 , BioRad ) and for total protein by the Pierce BCA Protein Assay Kit ( 23227 , Thermo Fisher Scientific ) . Acropora polyps hosting Symbiodiniaceae SSB01 were exposed to U18666A as described , except that they were cultured in FNSW . Confocal microscopy of NPC2 immunofluorescence was performed using a Leica SP8 system with an HC PL Apo CS2 63x/1 . 30 GLYC objective . Hoechst was excited at 405 nm and detected at 410–501 nm , and algal autofluorescence was excited at 633 nm and detected at 645–741 nm . In a second sequential scan , Alexa-488 ( secondary antibody ) was excited at 496 nm and detected at 501–541 nm . Z-stacks were collected with a step size of 0 . 5 µm and 3x line averaging . A zoom factor of 5 or , for whole larvae , 1 . 33 , was used , and a pinhole of 1 Airy unit . Quantification and imaging NPC2 IF over a time-course was carried out using a Nikon Eclipse Ti epifluorescence compound microscope with a Plan Apo λ 40x objective , Sola light source , and GFP filter set . Images were captured with a Nikon DS-Qi2 with an exposure time of 1 s . Fluorescence microscopy of Aiptasia adults was carried out using a Nikon SMZ18 fluorescence stereoscope with a 0 . 5X objective; endogenous autofluorescence of symbiont photosynthetic antennae was visualized with a Texas Red filter set , and images were captured at magnification 15X with an Orca-Flash4 . 0 camera ( C11440 , Hamamatsu ) at 300 ms exposure using Nikon Elements software and processed in Fiji ( Schindelin et al . , 2012 ) . Acropora polyps were photographed as described ( Wolfowicz et al . , 2016 ) , and fluorescence was quantified in Fiji ( Schindelin et al . , 2012 ) as total fluorescence in the polyp area minus adjacent background . In GC/MS-based sterol profiling , shown in Figure 1A are representatives of n = 3 ( Aiptasia , SSB01 ) or n = 2 ( Acropora , SSA01 , CCMP2556 ) samples each , shown in Figure 1B are representatives of n = 3 samples each , and shown in Figure 1C and Figure 1—figure supplement 2 are averages ( A ) and representatives ( B ) of n = 2 samples each . In gene expression analyses by qPCR ( Figure 3A ) , shown are average values of 6 samples per condition , six animals per sample , each in technical duplicate for Aiptasia . For Nematostella ( Figure 3—figure supplement 2 ) , shown are average values of 2 animals per sample , two samples per condition , each in technical duplicate . For Acropora ( Figure 3—figure supplement 2 ) , shown are averages of two biological replicates , each in technical duplicate . In NPC2 immunofluorescence in Aiptasia larvae ( Figure 3F ) , shown is a representative of two independent experiments , each with triplicate samples of >50 larvae per time-point . In sterol-blocking U18666A pharmacological experiments , shown are representative images of n = 3 polyps per anemone type and condition , with all anemones shown in Figure 3—figure supplement 6; symbiotic representatives are from one of three replicate experiments ( Figure 3G + H ) . Quantification of symbiont density ( Figure 3H ) in n = 3 anemones per condition , each in technical duplicate . Shown in Figure 3—figure supplement 7 are representative images of n = 5 polyps across duplicate wells ( n = 4 for 10 µM ) . In immunoprecipitation-lipidomics experiments , shown are averages of duplicate samples , with representative experiments shown of two ( Figure 4C ) or three replicate experiments ( Figure 4A + B ) . For experiments assessing soluble NPC2 at different pHs , ratios of ID at pH 7 divided by that at pH 5 were calculated from duplicate-loaded bands per protein per pH condition from a single blot , from three ( canonical NPC2 XM_021041171; non-canonical NPC2 XM_021052404 ) or six ( non-canonical NPC2 XM_021052412; crmCherry alone ) replicate experiments ( Figure 4F ) . Shown in Figure 4E are one pair of treatments in a representative experiment , from the aforementioned number of replicate experiments .	Coral reefs are the most biodiverse marine ecosystems on our planet . Their immense productivity is driven by friendly relationships , or symbioses , between microbes called algae and the corals . Related organisms , such as anemones , also rely on these close associations . The algae use energy from sunlight to make sugars , cholesterol and other molecules that they supply to their host . In exchange , the host’s cells provide homes for the algae inside specialist , acidic structures called symbiosomes . Corals and anemones particularly need cholesterol and other ‘sterol’ molecules from the algae , because they are unable to create these building blocks themselves . In mammals , a protein known as Niemann-Pick Type C2 ( NPC2 ) transports cholesterol out of storage structures into the main body of the cell . Corals and anemones have many different , ‘atypical’ NPC2 proteins: some are produced more during symbiosis , and these are mainly found in symbiosomes . However , it was not known what role these NPC2 proteins play during symbioses . Here , Hambleton et al . studied the symbioses that the anemone Aiptasia and the coral Acropora create with different strains of Symbiodiniaceae algae . The experiments found that the strain of algae dictated the mixture of sterols inside their hosts . The hosts could flexibly use different mixes of sterols and even replace cholesterol with other types of sterols produced by the algae . Atypical NPC2 proteins accumulated over time within the symbiosome and directly bound to cholesterol and various sterols the way other NPC2 proteins normally do . Further experiments suggest that , compared to other NPC2s , atypical NPC2 proteins may be better adapted to the acidic conditions in the symbiosome . Taken together , Hambleton et al . propose that atypical NPC2 proteins may play an important role in allowing corals to thrive in environments poor in nutrients . The first coral reefs emerged over 200 million years ago , when the Earth still only had one continent . Having built-in algae that provide the organisms with nutrients is thought to be the main driver for the formation of coral reefs and the explosion of diversity in coral species . Yet these ancient relationships are now under threat all around the world: environmental stress is causing the algae to be expelled from the corals , leading to the reefs ‘bleaching’ and starving . The more is known about the details of the symbiosis , the more we can understand how corals have evolved , and how we could help them survive the crisis that they are currently facing .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "evolutionary", "biology", "cell", "biology" ]	2019	Sterol transfer by atypical cholesterol-binding NPC2 proteins in coral-algal symbiosis
Developmental signaling pathways often activate their own inhibitors . Such inhibitory feedback has been suggested to restrict the spatial and temporal extent of signaling or mitigate signaling fluctuations , but these models are difficult to rigorously test . Here , we determine whether the ability of the mesendoderm inducer Nodal to activate its inhibitor Lefty is required for development . We find that zebrafish lefty mutants exhibit excess Nodal signaling and increased specification of mesendoderm , resulting in embryonic lethality . Strikingly , development can be fully restored without feedback: Lethal patterning defects in lefty mutants can be rescued by ectopic expression of lefty far from its normal expression domain or by spatially and temporally uniform exposure to a Nodal inhibitor drug . While drug-treated mutants are less tolerant of mild perturbations to Nodal signaling levels than wild type embryos , they can develop into healthy adults . These results indicate that patterning without inhibitory feedback is functional but fragile . Feedback inhibition is a common feature of developmental pathways across phyla ( Freeman , 2000; Freeman and Gurdon , 2002; Meinhardt , 2009; Piddini and Vincent , 2009; Ribes and Briscoe , 2009; Rogers and Schier , 2011 ) . Feedback inhibitors contribute to patterning processes in tissues ranging from the mouse neural tube ( Ribes and Briscoe , 2009 ) and the zebrafish hindbrain ( White and Schilling , 2008 ) , to the Drosophila wing ( Gerlitz and Basler , 2002; Piddini and Vincent , 2009; Zeng et al . , 2000 ) and eye ( Freeman , 1997 ) . Consistent with a general requirement for feedback control in development , inactivation of feedback inhibitors often results in disastrous patterning defects . However , directly testing the role of feedback per se has remained challenging: Eliminating inhibitors removes feedback , but it also increases signaling levels . Experiments that decouple inhibitor activation and inhibition while maintaining near-normal signaling levels are therefore required to unambiguously test the role of feedback in developmental patterning . Inhibitory feedback has been invoked to explain the sophisticated spatiotemporal control and robust performance observed in patterning circuits ( Ribes and Briscoe , 2009 ) . Inhibitory feedback has been suggested to turn off pathway activity when it is no longer needed , and to regulate spatial profiles of pathway activation ( Barkai and Shilo , 2009; Ben-Zvi et al . , 2008; Dessaud et al . , 2007; Freeman , 2000; Gerlitz and Basler , 2002; Golembo et al . , 1996; Lecuit and Cohen , 1998; Piddini and Vincent , 2009; Ribes and Briscoe , 2009; Schilling et al . , 2012; Shiratori and Hamada , 2006; van Boxtel et al . , 2015 ) . Additionally , theoretical considerations suggest that negative feedback could enable the embryo to adjust signaling levels in response to unexpected perturbations or biochemical fluctuations ( Barkai and Shilo , 2009; Eldar et al . , 2003; Lander et al . , 2009 ) . The Nodal/Lefty system has become a paradigm for feedback inhibition in development ( Chen and Schier , 2002; Duboc et al . , 2008; Freeman , 2000; Hamada , 2012; Kondo and Miura , 2010; Meinhardt , 2009; Meno et al . , 1999; Nakamura et al . , 2006; Rogers and Schier , 2011; Schier , 2009; Shen , 2007 ) . Nodal is a TGFβ superfamily ligand that induces the phosphorylation and subsequent nuclear localization of the transcription factor Smad2 in target cells . In early embryos , Nodal signaling induces mesendodermal fates through graded activation of target genes , with higher signaling intensities biasing cells toward endoderm . Nodal signaling intensity is shaped by the interplay between Nodal ligands and Leftys , secreted signaling inhibitors that prevent Nodal from binding to its receptors ( Chen and Shen , 2004; Cheng et al . , 2004 ) . Studies of mouse lefty mutants and zebrafish lefty morphants reveal that loss of inhibition results in expanded domains of Nodal target expression , increased mesendodermal specification , and embryonic lethality ( Agathon et al . , 2001; Chen and Schier , 2002; Feldman et al . , 2002; Meno et al . , 1999; van Boxtel et al . , 2015 ) . Antagonism by Lefty is thus important for preventing overactive Nodal signaling during mesendodermal patterning . Lefty production is coupled to Nodal signaling , forming a negative feedback loop that is conserved from sea urchins to humans . For example , in zebrafish , lefty1 and lefty2 are induced by endogenous Nodal signaling at the blastoderm margin , expression of Nodal can drive ectopic lefty production , and loss of Nodal signaling abolishes expression of lefty ( Figure 1A , B ) ( Meno et al . , 1999 ) . Despite the ubiquity of this motif , the functions provided by coupling Nodal activation to inhibition remain unclear . Several roles for Lefty feedback have been suggested . First , Nodal and Lefty were proposed to form a reaction-diffusion patterning system that regulates both mesendoderm formation and left/right patterning in zebrafish ( Chen and Schier , 2002; Kondo and Miura , 2010; Meinhardt , 2009; Müller et al . , 2012; Schier , 2009; Shen , 2007; Shiratori and Hamada , 2006 ) . Notably , Nodal and Lefty fulfill the key biophysical requirements of a classical reaction-diffusion system: Both Nodal ligands ( Cyclops and Squint ) act as short-range ( Cyclops ) and mid-range ( Squint ) activators that induce their own expression as well as that of Lefty1 and Lefty2 , which act as long-range , highly mobile inhibitors ( Chen and Schier , 2002; 2001; Feldman et al . , 2002; Meno et al . , 1999; Müller et al . , 2012 ) . Second , Lefty was argued to temporally restrict Nodal signaling by creating a ‘window’ of signaling competence ( van Boxtel et al . , 2015 ) . In this model , Nodal signaling proceeds until sufficient Lefty accumulates to shut down further signaling . Third , theoretical studies suggest that inhibitory feedback has the potential to mitigate fluctuations in signaling ( Lander et al . , 2009 ) . Deleterious increases or decreases in Nodal signaling could therefore be offset by adjustments in Lefty-mediated inhibition , ensuring robust development in the face of variation in the external environment or expression of pathway components . To understand the role of inhibitory feedback in the Nodal/Lefty patterning system , we created embryos in which Nodal inhibition was decoupled from Nodal signaling . We found that inhibitory feedback mitigates signaling perturbations but is dispensable for development . To determine the consequences of removing feedback inhibition , we first used TALENs to generate null lefty1 and lefty2 alleles in zebrafish ( Figure 1C–T ) ( Bedell et al . , 2012; Reyon et al . , 2012; Sander et al . , 2011; Sanjana et al . , 2012 ) . lefty1a145 contains a 13-base-pair deletion that removes the translational start site and part of the predicted signal sequence ( Figure 1C ) , and lefty2a146 contains an 11-base-pair deletion that results in a stop codon after amino acid 18 ( Figure 1D ) . In contrast to wild type lefty mRNA , mutant lefty mRNAs were unable to induce Nodal loss-of-function phenotypes when injected into zebrafish embryos ( Figure 1E–J” ) . lefty1-/- and lefty2-/- single mutants were viable and exhibited no or only minor increases in mesendodermal gene expression ( Figure 1N , O , T , Figure 3—figure supplement 1E , F ) , consistent with their overlapping early expression domains ( Figure 3—figure supplement 1A , B ) . lefty1-/- but not lefty2-/- mutants had heart laterality defects ( Figure 1—figure supplement 1 ) , a hallmark of abnormal Nodal signaling ( Bakkers et al . , 2009 ) and reflecting their distinct spatial expression patterns during left-right patterning ( Bisgrove et al . , 1999 ) . A single functional lefty allele was sufficient for viability ( Figure 1Q , R , T ) , but lefty1-/-;lefty2-/- double mutants exhibited severe patterning defects , including loss of heart , eyes , and tail ( Figure 1S , Figure 1—figure supplement 1A ) . At the level of tissue patterning , lefty1-/-;lefty2-/- mutant embryos had expanded pSmad2 signaling gradients ( Figure 2 ) prior to and during gastrulation: Signaling gradients had higher amplitudes and longer ranges in double mutants compared to wild type embryos ( Figure 2C ) . Consistent with the expansion of the pSmad2 signaling gradient , lefty1-/-;lefty2-/- mutant embryos exhibited expanded expression of mesendodermal genes by gastrulation stages , whereas single mutants exhibited no or relatively minor upregulation ( Figure 3 , Figure 3—figure supplement 1E , F ) . Expanded mesendoderm has also been reported in lefty double morphants ( Agathon et al . , 2001; Chen and Schier , 2002; Feldman et al . , 2002; van Boxtel et al . , 2015 ) , but lefty1-/-;lefty2-/- mutants differ in several aspects from morphants . For example , lefty morphants display disrupted gastrulation and do not survive past 24 hr , whereas lefty double mutants successfully gastrulate and survive past 24 hr ( Figure 1—figure supplement 2 ) . These differences are not caused by compensation in lefty1-/-;lefty2-/- mutants but by off-target morpholino effects: lefty1-/-;lefty2-/- mutants injected with lefty1/2 morpholinos also display disrupted gastrulation and die by 24 hr ( Figure 1—figure supplement 2 ) . Together , our results demonstrate overlapping roles for lefty1 and lefty2 in restricting mesendoderm formation , but unique roles in left-right patterning . The patterning defects in lefty double mutants show a requirement for reduction of Nodal activity but do not establish a requirement specifically for inhibitory feedback . It is possible that a reduction in Nodal signaling by means other than inhibitory feedback could support patterning . For example , reducing nodal gene dosage could suppress lefty double mutant defects ( Chen and Schier , 2002; Feldman et al . , 2002; Meno et al . , 1999 ) . In support of this hypothesis , mutations in the Nodal genes squint or cyclops suppressed multiple aspects of the lefty1-/-;lefty2-/- mutant phenotype ( Figure 4 ) . cyclops-/-;lefty1-/-;lefty2-/- ( Figure 4C’’ ) and squint-/-;lefty1-/-;lefty2-/- ( Figure 4F’’ ) mutants formed eyes and full-length tails , structures missing in lefty1-/-;lefty2-/- mutants ( Figure 4A’’ , D’’ , M , N ) . Moreover , upregulation of mesendodermal gene expression was suppressed in squint-/-;lefty1-/-;lefty2-/- mutants compared to lefty1-/-;lefty2-/- mutants ( Figure 4G–L’’’ ) . Although these triple mutants are not viable , two functional nodal alleles are sufficient to generate mesendodermal gene expression patterns that are similar to those observed in wild type embryos with four nodal alleles and four lefty alleles . In previous studies , removal of squint , but not cyclops , partially suppressed defects in lefty double morphants ( Chen and Schier , 2002; Feldman et al . , 2002 ) , but our results indicate that Lefty inhibits both Squint and Cyclops . The failure to fully rescue development may reflect an inability to precisely modulate Nodal dosage with this genetic approach , but reduction of Nodal dosage can partially rescue development in the complete absence of Lefty-mediated inhibition . Nodal signaling induces expression and secretion of Lefty at the embryo margin ( Meno et al . , 1999 ) ( Figure 1A , B , Figure 3—figure supplement 1A , B ) . To test the importance of the spatial coupling of lefty expression and Nodal signaling , we asked whether lefty needs to be induced where the Nodal pathway is active . We generated clones that expressed lefty or lefty-gfp ectopically in lefty1-/-;lefty2-/- mutant embryos , independent of Nodal signaling and outside of the endogenous lefty expression domain ( Figure 5A , B , Figure 5—figure supplements 1 and 2 , Figure 1B , Figure 3—figure supplement 1A , B ) . We injected lefty mRNA into lefty1-/-;lefty2-/- mutant embryos and transplanted around 50 cells from these donor embryos into the animal pole of host lefty1-/-;lefty2-/- mutant embryos at sphere stage , when lefty expression normally commences ( Figure 5A , E , F , Figure 3—figure supplement 1A , B ) . Strikingly , some of the lefty1-/-;lefty2-/- mutant hosts were rescued to normal morphology ( Figure 5C–F’ , Figure 5—figure supplement 2 ) and developed into fertile adults ( see Figure 5 legend for quantification ) . Thus , an ectopic , Nodal-independent source of Lefty at the animal pole can replace endogenous , Nodal-induced Lefty at the margin . The ability of ectopic Lefty-expressing clones to rescue lefty1-/-;lefty2-/- mutants is consistent with the high diffusivity and long-range , extracellular distribution of Lefty protein ( Chen and Schier , 2002; Marjoram and Wright , 2011; Müller et al . , 2012 ) ( Figure 1B ) . In contrast to the requirement for spatially restricted Nodal signaling ( Figure 5—figure supplement 3 ) , the spatial coupling of lefty expression to Nodal signaling is not required for normal development . Induction of lefty expression by Nodal signaling couples pathway activation to inhibition . To test whether patterning can occur when Nodal pathway inhibition is spatially and temporally decoupled from Nodal activity , we attempted to replace the inhibitory activity of Lefty with a small molecule drug , SB-505124 , that selectively inhibits Nodal signaling by preventing ATP from binding to Nodal receptors ( Figure 6A ) ( DaCosta Byfield et al . , 2004; Fan et al . , 2007; Hagos and Dougan , 2007; Hagos et al . , 2007; van Boxtel et al . , 2015; Vogt et al . , 2011 ) . We exposed lefty1-/-;lefty2-/- mutants to low concentrations of the Nodal inhibitor drug starting at three developmental stages: ( 1 ) the 8 cell stage , 3 hr before the onset of nodal and lefty expression , ( 2 ) sphere stage , when nodal and lefty expression normally begins , and ( 3 ) shield stage , 2 . 5 hr after expression has commenced . Strikingly , exposure of lefty1-/-;lefty2-/- mutants to inhibitor drug resulted in phenotypically normal embryos that developed into fertile adults ( Figure 6B–I , Figure 6—figure supplements 1 and 2 ) . Different drug concentrations were required for rescue depending on the timing of treatment: Mutants exposed at earlier times were rescued by lower concentrations of Nodal inhibitor drug than mutants exposed at later times ( Figure 6I , Figure 6—figure supplements 1 and 2 ) . Notably , even drug treatment starting hours before or after the onset of nodal and lefty expression rescued lefty1-/-;lefty2-/- mutants ( Figure 6D , F , I , Figure 6—figure supplements 1 and 2 ) . These results show that highly specific timing or progressively increasing levels of inhibition during embryogenesis are not required for mesendoderm development . To determine the mechanism by which Nodal inhibitor-treated mutants are rescued , we analyzed Smad2 phosphorylation and mesendodermal gene expression . Interestingly , despite their eventual rescue , lefty1-/-;lefty2-/- mutants that were exposed to inhibitor before the onset of Nodal signaling not only showed reduced Smad2 phosphorylation compared to double mutants , but initially displayed reduced Smad2 phosphorylation and endodermal gene expression compared to wild type ( Figure 6J–O’’ , Figure 2—figure supplement 1 , Figure 6—figure supplement 3A , B ) . The mechanism by which premature inhibitor exposure rescues development therefore involves a delay in full activation of Nodal signaling . Highlighting the sensitivity to Nodal inhibitor drug concentration , lefty double mutants exposed to excess or sub-rescuing doses exhibited diminished or expanded pSmad2 activity gradients , respectively , with corresponding Nodal loss- or gain-of-function phenotypes ( Figure 6—figure supplements 1 and 4 ) . Mutants exposed to inhibitor drug after the onset of Nodal signaling exhibited excess mesendodermal gene expression until ~5 hr post-exposure ( Figure 6P–U’’ , Figure 6—figure supplement 3E , F ) , but they ultimately developed normally ( Figure 6F , I , Figure 6—figure supplements 1 and 2 ) . Together , these results demonstrate that the precise spatial and temporal coupling of lefty expression and Nodal activity is not essential for normal development . The rescue of lefty mutants with a Nodal inhibitor drug shows that inhibitory feedback is not a requirement for mesendoderm patterning . This result leaves open the possibility that inhibitory feedback enhances developmental robustness by mitigating signaling fluctuations ( Lander et al . , 2009 ) . Perturbations that decrease Nodal signaling might be corrected by a compensatory decrease in activation of Lefty , thus restoring the activator/inhibitor balance . To test this model , we challenged wild type embryos with exogenous Nodal pathway activation or inhibition . Injecting low levels of lefty1 mRNA into wild type embryos dramatically reduced lefty2 transcript abundance ( Figure 7A , C , B , D ) , but left expression of the mesoderm marker noto relatively intact ( Figure 7A’ , C’ , B’ , D’ ) . Conversely , increasing signaling by injecting mRNA encoding constitutively-active smad2 ( CA-smad2 , ( Baker and Harland , 1996; Dick et al . , 2000; Gritsman et al . , 1999; Müller et al . , 1999 ) ) resulted in a marked increase in lefty2 expression ( Figure 7E , G , F , H ) , but unchanged noto expression ( Figure 7E’ , G’ , F’ and H’ ) . Wild type embryos thus appear to compensate for perturbed Nodal signaling by sensitively adjusting lefty levels . This model makes a key prediction: If Lefty feedback allows the embryo to correct signaling perturbations , embryos with compromised feedback should be more sensitive to Nodal signaling challenges . To test this hypothesis , we assessed developmental outcomes after manipulating Nodal signaling levels in wild type embryos and in drug-treated lefty1-/-;lefty2-/- mutants , in which inhibition is decoupled from signaling ( Figure 6A ) . Wild type embryos challenged with a small dose of lefty1 mRNA exhibited normal phenotypes at 24 hpf ( Figure 8A , B ) as well as normal mesendodermal gene expression at 50% epiboly and shield stages ( Figure 8H–I’ ) . In contrast , challenging feedback-compromised embryos with the same dose of lefty1 mRNA led to markedly decreased noto expression ( Figure 8I , K , M , I’ , K’ , M’ ) , and 24 hpf phenotypes resembling partial loss-of-function Nodal mutants ( Figure 8D , E ) ( Feldman et al . , 1998; Gritsman et al . , 1999 ) . Challenging with a modest increase in Nodal signaling revealed a similar difference in sensitivity . Wild type embryos tolerated a small dose of CA-smad2 mRNA ( Figure 8C ) , while drug-rescued lefty double mutants developed severe phenotypic defects in response to the same treatment ( Figure 8F ) , although changes in early mesendodermal gene expression were relatively minor ( Figure 8N–S’ ) . These results reveal that Lefty-mediated inhibitory feedback can mitigate aberrant fluctuations in Nodal signaling levels ( Figure 8G ) . The results in this study show that inhibitory feedback in the Nodal/Lefty system stabilizes Nodal signaling but is not essential for mesendoderm patterning and viability . The rescue of lefty mutants by ectopic lefty expression ( Figure 5 , Figure 5—figure supplements 1 and 2 ) and exposure to Nodal inhibitor drug ( Figure 6 , Figure 6—figure supplements 1–4 , Figure 2—figure supplement 1 ) is consistent with the high diffusivity of Lefty measured previously ( Müller et al . , 2012 ) , but is surprising in light of the functions assigned to inhibitory feedback . Specifically , inhibitory feedback has been implicated in ( 1 ) shutting down pathway activity at the appropriate time to generate a pulse or window of signaling , ( 2 ) shaping spatial signaling profiles , ( 3 ) acting as part of self-organizing reaction-diffusion systems , and ( 4 ) mitigating fluctuations in signaling activity . Below , we discuss our results in the context of these models . First , inhibitory feedback has been suggested to turn off signaling activity when it is no longer needed ( Dessaud et al . , 2007; Freeman , 2000; Golembo et al . , 1996; Ribes and Briscoe , 2009; Shiratori and Hamada , 2006 ) . For Nodal-mediated patterning , it has been proposed that progressively increasing Lefty levels shut down Nodal signaling at the onset of gastrulation ( van Boxtel et al . , 2015 ) . However , we find that Nodal signaling is already increased in lefty double mutants by sphere stage , suggesting an earlier role for Lefty ( Figure 2 ) . Moreover , Lefty can be replaced by an inhibitor drug added as early as the 8 cell stage ( Figure 6 , Figure 6—figure supplements 1–4 , Figure 2—figure supplement 1 ) , demonstrating that mesendoderm patterning can proceed without progressively increasing inhibition and without temporally precise feedback inhibition . Our results do not rule out that Lefty accumulation shuts down Nodal signaling during normal development , but they do indicate that this is not an absolute requirement for patterning . Second , inhibitory feedback has been implicated in shaping the spatial profile of pathway activity ( Barkai and Shilo , 2009; Ben-Zvi et al . , 2008; Freeman , 2000; Gerlitz and Basler , 2002; Lecuit and Cohen , 1998; Piddini and Vincent , 2009; Ribes and Briscoe , 2009; Schilling et al . , 2012; van Boxtel et al . , 2015 ) . In drug-rescued lefty1-/-;lefty2-/- mutants , Nodal activity gradients and mesendodermal gene expression patterns initially differed from wild type ( Figure 6J–U’’ , Figure 2—figure supplement 1 , Figure 6—figure supplements 3 and 4 ) . However , rescued mutants developed into fertile adults ( Figure 6H , Figure 6—figure supplement 2 ) , demonstrating that precise wild type activity gradients and mesendodermal gene expression patterns during early embryogenesis are not essential for germ layer development . Inhibitory feedback may subtly shape gene expression patterns , but the patterns achieved without feedback are a suitable starting point for successful development . Third , the dispensability of inhibitory feedback in the Nodal/Lefty system raises questions about the role of Nodal and Lefty as an activator/inhibitor pair in self-organizing reaction-diffusion models . Although Nodal and Lefty fulfill the key regulatory and biophysical requirements of a short-to-mid-range autocatalytic activator and a long-range feedback inhibitor ( Hamada , 2012; Meno et al . , 1999; Müller et al . , 2012; Schier , 2009 ) , our finding that development can be normal without inhibitory feedback indicates that this system does not require the ability to form self-organizing reaction-diffusion patterns . Instead , it is conceivable that the pre-patterning of the early embryo by maternal factors and the local activation of Nodal eliminate the need for self-organizing pattern generation by the Nodal/Lefty circuit . In this scenario , Nodal and Lefty may have constituted a reaction-diffusion activator/inhibitor pair in ancestral organisms but , through the addition of other regulatory layers , mesendoderm patterning lost the requirement for inhibitory feedback . Finally , feedback inhibition has been implicated in buffering fluctuations in pathway activity ( Barkai and Shilo , 2009; Eldar et al . , 2003; Lander et al . , 2009 ) . Feedback may be required to optimize inhibitor levels , as suggested by the narrow range ( ~2 fold ) of inhibitor concentrations that rescue Lefty loss ( Figure 6I , Figure 6—figure supplements 1 and 2 ) . Indeed , the adjustment of lefty expression in response to slight alterations in Nodal signaling ( Figure 7 ) and the failure of feedback-decoupled embryos to cope with perturbations in Nodal signaling ( Figure 8 ) support this idea . The ability to dynamically adjust pathway activity may allow the embryo to create reliable patterns in the face of endogenous signaling fluctuations and uncertain environmental conditions . We note , however , that Lefty feedback does not protect the embryo against all perturbations: Drug-rescued lefty mutants actually fared better than wild type embryos when challenged with injection of squint mRNA ( Figure 8—figure supplement 1 ) . Dissecting why Lefty feedback corrects some perturbations but not others will provide a window into the mechanisms and limits of robust patterning . Our results have implications not only for the roles of feedback inhibition during development , but also demonstrate the feasibility of preventing patterning defects with small molecule drug exposure . Although suggested applications to human embryos might currently seem fanciful and would be challenging and fraught with ethical concerns , embryos bearing compromised patterning circuits could be identified by sequencing a single embryonic cell , and birth defects could be prevented by exposure to the appropriate small molecule . More generally , our study adds a new facet to recent revisions of classical patterning models ( Alexandre et al . , 2014; Chen et al . , 2012; Dominici et al . , 2017; Dubrulle et al . , 2015; Varadarajan et al . , 2017 ) . For example , a tethered form of Wingless can replace endogenous Wingless , challenging models in which a gradient of diffusing Wingless is indispensable for tissue patterning ( Alexandre et al . , 2014 ) . In the same vein , our observations challenge models in which inhibitory feedback is an absolute requirement for patterning and viability , but support the idea that inhibitory feedback enhances robustness by stabilizing signaling during development . Mutations in lefty genes were induced using TALENs ( Bedell et al . , 2012; Sander et al . , 2011; Sanjana et al . , 2012 ) . The lefty1 TALEN pair was generated using the FLASH assembly kit ( Reyon et al . , 2012 ) ; the lefty2 TALEN pair was generated using the TALE Toolbox ( Sanjana et al . , 2012 ) . lefty TALENs target sites: lefty1 TALEN L: tcctgcaccttgaaaaga lefty1 TALEN R: tgcgcaaaggaggcacgc lefty2 TALEN L: ttcatccagctgttcatttt lefty2 TALEN R: tgctggaatccctgtgtgag Embryos from an incross of the TLAB wild type strain were injected at the one-cell stage with 300–450 pg mRNA encoding each TALEN pair . Injected fish were grown using standard fish husbandry protocols and fin clipped as adults . Genomic DNA was generated from fin material using the Hot Shot method ( Meeker et al . , 2007 ) . To identify animals carrying mutations , PCR using primers flanking the target sites was carried out and the resulting amplicons were re-annealed and digested with mismatch-cleaving T7 endonuclease I ( NEB ) ( Mussolino et al . , 2011 ) . PCR products from positive animals were cloned using a TOPO TA kit ( Life Technologies ) and sequenced . Positive animals were outcrossed and progeny were sequenced and tested for germline transmission . Primers flanking lefty TALEN target sites: lefty1 forward primer: catgtatcaccttccctctgatgtc lefty1 reverse primer: gcattagcctatatgttaacttgcac lefty2 forward primer: tacttatcaacatgagcatcaatgg lefty2 reverse primer: gaattgtgcataagtaacccacctg Genomic DNA was generated using the Hot Shot method ( Meeker et al . , 2007 ) . lefty1: The 13-base-pair deletion in lefty1a145 destroys a PshAI restriction site . To genotype the lefty1 locus , PCR amplicons were generated using primers flanking the deletion and subsequently digested with PshAI endonuclease ( NEB ) . Genotyping primers were identical to the lefty1 forward/reverse primers described above . Complete digestion by PshAI indicates that both alleles are wild type , partial digestion indicates heterozygosity , and failure to digest indicates homozygosity for the lefty1a145 mutation . lefty2: The 11-base-pair deletion in lefty2a146 was detected using a mutant-specific forward primer that spans the deletion . A forward primer specific to the wild type allele was also designed , as well as a reverse primer that is fully complementary to both alleles . To genotype the lefty2 locus , PCR was carried out using either the wild type- or mutant-specific forward primer and the common reverse primer . A band with the wild type- but not mutant-specific primer indicates that both alleles are wild type , bands with both primer sets indicate heterozygosity , and a band with the mutant- but not wild type-specific primers indicates homozygosity for the lefty2a146 mutation . Optimal PCR conditions: Taq polymerase , 25 cycles , 57°C annealing temperature . lefty2 wild type genotyping forward primer: cattttgaccacagcgat lefty2 mutant genotyping forward primer: gttcattttgaccactcac The common reverse primer was identical to lefty2 reverse primer described above . squint: The squintcz35 allele has a ~ 1 . 9 kb insertion in exon 1 , and was detected as in ( Feldman et al . , 1998 ) . cyclops: The cyclopsm294 mutation destroys an AgeI restriction site , and was detected as in ( Sampath et al . , 1998 ) . To generate mRNA from all constructs , plasmids were linearized with NotI-HF endonuclease ( NEB ) and purified using a Qiagen PCR clean-up kit . Capped mRNA was generated from linearized plasmid using an SP6 mMessage mMachine kit ( Ambion ) and purified with a Qiagen RNeasy kit . After purification , mRNA was quantified using a NanoDrop spectrophotometer ( ThermoFisher ) and diluted to the appropriate concentration . For microinjections , a micrometer was used to adjust the drop volume to 0 . 5 nl . Depending on the concentration of the injection mix , a total volume of 1–2 nl was injected per embryo . Lefty1-GFP , Lefty2-GFP , and untagged Lefty1 and Lefty2 constructs used in Figures 5 and 7 were identical to those in ( Müller et al . , 2012 ) . These constructs lack endogenous UTRs and contain the consensus Kozak sequence gccacc immediately preceding the start codon . Allele activity experiments: To determine whether mutant lefty alleles retain Nodal inhibitory activity , wild type and mutant lefty mRNA was injected into wild type embryos and Nodal loss-of-function phenotypes were assessed ( Figure 1 ) . The 13 bp lefty1a145 mutation removes part of the endogenous Kozak sequence ( gaaaag ) . Therefore , lefty1 constructs containing this endogenous Kozak were generated , rather than the consensus Kozak sequence gccacc as in ( Müller et al . , 2012 ) . Primers with either the endogenous Kozak sequence ( for the wild type construct ) or the truncated endogenous sequence and deleted region of coding sequence ( for the mutant construct ) were designed , and the Lefty1 construct from ( Müller et al . , 2012 ) was used as a PCR template . The resulting fragments were cloned into BamHI and XhoI sites in pCS2 ( + ) . Both constructs lack endogenous UTRs . The lefty2 wild type construct was the same used in ( Müller et al . , 2012 ) and in the transplantation experiments in Figure 5 . The lefty2a146 construct was made by generating cDNA from lefty2 homozygous embryos , amplifying the mutant lefty2 coding sequence , and cloning the resulting fragment into ClaI and XhoI in the pCS2 ( + ) vector . In addition to the 11 bp deletion , the mutant construct contains three silent SNPs at position 184 ( T->C ) , 932 ( A->C ) , and 943 ( T->A ) . Both constructs lack endogenous UTRs and contain the consensus Kozak sequence gccacc immediately preceding the start codon . Nodal overexpression experiment: The construct used to generate squint mRNA in the Nodal overexpression experiment ( Figure 8—figure supplement 1 ) was identical to that used in ( Müller et al . , 2012 ) . This construct lacks endogenous UTRs and contains the consensus Kozak sequence gccacc immediately preceding the start codon . The protocol was modified from ( van Boxtel et al . , 2015 ) . Briefly , embryos were fixed in 4% formaldehyde ( in 1x PBS ) overnight at 4°C , washed in PBST ( 1x PBS + 0 . 1% ( w/v ) Tween 20 ) , manually deyolked , dehydrated in a MeOH/PBST series ( 25% , 50% , 75% , and 100% MeOH ) , and stored at −20°C until staining . To prepare for staining , embryos were rehydrated in a MeOH/PBSTr ( 1x PBS + 1% ( w/v ) Triton X-100 ) series ( 75% , 50% , and 25% MeOH ) , washed 3x in PBSTr , and incubated for 20 min in ice-cold acetone . Embryos were then washed 3x in PBSTr , incubated in antibody binding buffer ( PBSTr +1% ( v/v ) DMSO ) for two hours at room temperature , then incubated overnight at 4°C with a 1:1000 dilution of α-pSmad2 antibody ( Cell Signaling Technology #8828 , Danvers , MA , USA ) in antibody binding buffer . After primary treatment , embryos were washed 6x in PBSTr , incubated in antibody binding buffer for 30 min at room temperature , and incubated for two hours at room temperature with a 1:2000 dilution of goat α-rabbit Alexa 647 conjugate ( ThermoFisher A-21245 ) in PBSTr +1% ( v/v ) DMSO . Embryos were then washed 6x in PBSTr , 3x in PBS and incubated with 200 nM Sytox green in PBS for 30 min at room temperature . Finally , embryos were washed 3x in PBS and dehydrated in a MeOH/PBS series ( 50% and 100% MeOH ) . Stained embryos were stored at −20°C in 100% MeOH until imaging . Embryos were mounted in agarose and cleared with 2:1 benzyl benzoate:benzyl alcohol ( BBBA ) ( Yokomizo et al . , 2012 ) . Briefly , a dehydrated embryo was dropped into molten low-melting point agarose ( 1% ( w/v ) in H2O ) , transferred onto a coverglass and oriented manually . All embryos ( other than sphere stage ) were mounted ‘margin down’ ( i . e . with the animal/vegetal axis parallel to the coverglass , see below for rationale ) . Shield stage embryos were rolled to ensure that the dorsal/ventral axis was parallel to the coverglass . Sphere stage embryos were mounted with the animal/vegetal axis perpendicular to the coverglass ( animal pole facing up ) . The agarose drop was then dehydrated with three washes of 100% MeOH , two washes of 50:50 ( v/v ) BBBA:MeOH , and 3 washes of BBBA . Cleared embryos were then sealed to a microscope slide using fast wells reagent reservoirs ( Grace Bio-Labs ) . Imaging was performed on Sytox Green and Alexa 647 channels using an LSM 700 confocal microscope ( 20x air objective , 0 . 5 NA ) . Image stacks extended from the embryo margin ( adjacent to the coverglass ) to beyond the center of the embryo . Z-planes were spaced at 2 μm intervals . Quantification of pSmad2 and Sytox green staining intensity in laterally-mounted embryos was performed using the ten z-slices surrounding the center of the embryo axis ( i . e . five slices above and five slices below the embryo center ) . This region was chosen to minimize artifacts due to light scattering , which causes a decrease in apparent fluorescence intensity in deeper tissue planes . Our procedure—looking only at slices close to the ‘central plane’ of the embryo—allows the entire gradient to be sampled within each slice , and ensured that all data were taken from planes within a narrow range of imaging depths , effectively controlling for signal drop-off with imaging depth . Nuclei were segmented from the Sytox Green channel images using a custom pipeline implemented in MATLAB ( Source code 1 ) . Briefly , out-of-plane background signal was approximated by blurring adjacent z-slices ( i . e . the slice above and below the plane being segmented ) with a Gaussian smoothing kernel and summing . This background was subtracted from the segmentation image , and preliminary boundaries for nuclei were identified by adaptive thresholding ( http://homepages . inf . ed . ac . uk/rbf/HIPR2/adpthrsh . htm ) of the resulting image . Spurious objects were discarded by morphological filtering ( based on object size ) . Final segmentation boundaries were defined after manual checking and correction with a custom MATLAB script ( Source code 1 ) . The fluorescence intensity of each segmented nucleus was defined as the mean intensity of its constituent pixels . The distance of each nucleus from the margin was defined along a curved embryo contour ( Figure 2—figure supplement 1 ) . This contour was defined by 1 ) projecting all segmented nuclei centroids onto a single z-plane , 2 ) creating a full embryo ‘mask’ by filling a convex hull containing all of these points , 3 ) identifying the left and right margin boundaries as the points of maximum curvature on the convex hull , 4 ) taking the distance transform of the embryo mask , and 5 ) stepping along the ‘valley’ of the distance transform that connects the left and right margins ( as defined above ) . This rough contour was then smoothed using a Savitsky-Golay filter to yield the final contour . The position of each nucleus was then projected onto the contour , and the distance from the margin ( as plotted in Figure 2C and Figure 2—figure supplement 1 ) was determined as the distance to the closest margin along this curve . Embryos were fixed overnight at 4°C with 4% formaldehyde in PBS . In situ hybridization was carried out as in ( Thisse and Thisse , 2008 ) and representative embryos were imaged in 2:1 benzyl benzoate:benzyl alcohol ( BBBA ) with a Zeiss Axio Imager . Z1 microscope . When genotyping was necessary , genomic DNA was generated after imaging as in ( Meeker et al . , 2007 ) and used in the described genotyping assays . The Nodal inhibitor SB-505124 ( S4696 , Sigma-Aldrich ) ( DaCosta Byfield et al . , 2004; Fan et al . , 2007; Hagos et al . , 2007; Hagos and Dougan , 2007; van Boxtel et al . , 2015; Vogt et al . , 2011 ) was dissolved in DMSO to a generate a stock at 10 mM and stored at 4°C . Fresh dilutions were made the same day experiments were carried out . Some batch-to-batch variability occurred , as well as a slight decrease in efficacy of the stock over time . The precise rescuing drug concentration must therefore be empirically determined for each stock of SB-505124 .	During animal development , a single fertilized cell gives rise to different tissues and organs . This ‘patterning’ process depends on signaling molecules that instruct cells in different positions in the embryo to acquire different identities . To avoid mistakes during patterning , each cell must receive the correct amount of signal at the appropriate time . In a process called ‘inhibitory feedback’ , a signaling molecule instructs cells to produce molecules that block its own signaling . Although inhibitory feedback is widely used during patterning in organisms ranging from sea urchins to mammals , its exact purpose is often not clear . In part this is because feedback is challenging to experimentally manipulate . Removing the inhibitor disrupts feedback , but also increases signaling . Since the effects of broken feedback and increased signaling are intertwined , any resulting developmental defects do not provide information about what feedback specifically does . In order to examine the role of feedback , it is therefore necessary to disconnect the production of the inhibitor from the signaling process . In developing embryos , a well-known signaling molecule called Nodal instructs cells to become specific types – for example , a heart or gut cell . Nodal also promotes the production of its inhibitor , Lefty . To understand how this feedback system works , Rogers , Lord et al . first removed Lefty from zebrafish embryos . These embryos had excessive levels of Nodal signaling , did not develop correctly , and could not survive . Bathing the embryos in a drug that inhibits Nodal reduced excess signaling and allowed them to develop successfully . In these drug-treated embryos , inhibitor production is disconnected from the signaling process , allowing the role of feedback to be examined . Drug-treated embryos were less able to tolerate fluctuations in Nodal signaling than normal zebrafish embryos , which could compensate for such disturbances by adjusting Lefty levels . Overall , it appears that inhibitory feedback in this patterning system is important to compensate for alterations in Nodal signaling , but is not essential for development . Understanding the role of inhibitory feedback will be useful for efforts to grow tissues and organs in the laboratory for clinical use . The results presented by Rogers , Lord et al . also suggest the possibility that drug treatments could be developed to help correct birth defects in the womb .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2017	Nodal patterning without Lefty inhibitory feedback is functional but fragile
Unraveling the genetic susceptibility of complex diseases such as chronic kidney disease remains challenging . Here , we used inbred rat models of kidney damage associated with elevated blood pressure for the comprehensive analysis of a major albuminuria susceptibility locus detected in these models . We characterized its genomic architecture by congenic substitution mapping , targeted next-generation sequencing , and compartment-specific RNA sequencing analysis in isolated glomeruli . This led to prioritization of transmembrane protein Tmem63c as a novel potential target . Tmem63c is differentially expressed in glomeruli of allele-specific rat models during onset of albuminuria . Patients with focal segmental glomerulosclerosis exhibited specific TMEM63C loss in podocytes . Functional analysis in zebrafish revealed a role for tmem63c in mediating the glomerular filtration barrier function . Our data demonstrate that integrative analysis of the genomic architecture of a complex trait locus is a powerful tool for identification of new targets such as Tmem63c for further translational investigation . The analysis of the genetic basis of common diseases remains challenging due to their complex pathogenesis and genetic heterogeneity in human populations ( Deng , 2015; Glazier et al . , 2002; McCarthy et al . , 2008 ) . This applies also to elevated blood pressure ( BP ) or hypertension , as recent meta-analyses of genome-wide association studies ( GWAS ) identified more than 100 gene loci associated with BP ( Hoffmann et al . , 2017; Warren et al . , 2017 ) . The effect size of the identified gene loci , however , is in general rather modest and less than 4% of the variance of BP phenotypes can be explained by these loci ( Hoffmann et al . , 2017; Warren et al . , 2017 ) , while around 40% to 50% of the variability appears heritable ( Levy et al . , 2007; Miall and Oldham , 1963 ) . Considerable evidence supports a major role of the kidney in BP regulation , and for the renal damage such as albuminuria development as a consequence of long-term BP elevation ( Coffman and Crowley , 2008; Mancia et al . , 2013 ) . Interestingly , genetic risk scores deploying BP genetic variants predict also hypertensive target organ damage in the heart , cerebral vessels , and eye , while only little evidence exists for an effect on kidney damage ( Ehret et al . , 2016 ) . In this regard , several inbred hypertensive rat models provide valuable complementary tools to uncover the genetics of kidney damage in hypertension ( Schulz and Kreutz , 2012; Yeo et al . , 2015 ) . These hypertensive models belong to the large panel of inbred strains that have been generated for a range of physiological and disease phenotypes by selective breeding ( Atanur et al . , 2013 ) . Of interest , compensatory alleles that protect against hypertensive organ damage have also been inadvertently co-selected in this process and explain the impressive resistance of some hypertensive strains against target organ damage ( Jacob , 2010; Jeffs et al . , 1997; Rubattu et al . , 1996; Schulz and Kreutz , 2012; Yeo et al . , 2015 ) . Quantitative trait loci ( QTL ) mapping strategies took advantage of these findings by intercrossing hypertensive strains with contrasting kidney damage phenotypes ( Schulz and Kreutz , 2012 ) . It remains challenging , however , to identify the molecular changes at a complex trait QTL that account for gene-gene or gene-environment interactions , incomplete penetrance , and epigenetic inheritance ( Buchner and Nadeau , 2015; Deng , 2015 ) . Here , we successfully overcome the obstacles of QTL mapping in rodents by analyzing two inbred hypertensive rat strains with contrasting kidney damage with albuminuria phenotypes . We used the Munich Wistar Frömter ( MWF ) rat as a suitable hypertensive model system to target the genetic basis of the albuminuria phenotype in contrast to the spontaneously hypertensive rat ( SHR ) representing a model that is protected against albuminuria development ( Schulz and Kreutz , 2012; van Es et al . , 2011 ) . Following initial QTL mapping in intercrosses of contrasting hypertensive strains ( Schulz and Kreutz , 2012 ) , we successfully combined conventional fine mapping using congenic strains with characterization of the genomic architecture in sub-QTLs by targeted next-generation sequencing ( NGS ) , and compartment-specific RNA sequencing ( RNA-Seq ) analysis . This comprehensive approach allowed us to prioritize transmembrane protein Tmem63c as a novel positional target for albuminuria development among several candidates . Functional relevance of Tmem63c was supported by allele-specific and BP-independent differential glomerular expression during onset of albuminuria in allele-specific minimal congenic rat lines derived from the contrasting rat models . Loss of glomerular TMEM63C expression in podocytes of patients with focal segmental glomerulosclerosis ( FSGS ) provided strong evidence for translational relevance . Loss-of-function studies in zebrafish induced a glomerular filtration barrier ( GFB ) defect compatible with the albuminuria phenotype , which was rescued upon co-injection of zebrafish tmem63c mRNA or rat Tmem63c mRNA , showing not only the specificity of the observed knockdown phenotype , but also conservation of the gene’s function across species . Ultrastructural analysis by electron microscopy , demonstrated severe morphological defects including podocyte damage with foot process effacement in tmem63c-deficient embryos . Altogether , our findings reveal TMEM63C being integral to podocyte physiology , and identified its potential role in glomerular renal damage in patients with chronic kidney disease . Our data demonstrate , despite the difficulties of QTL analysis in experimental models such as inbred strains , the feasibility to identify novel targets by combining conventional congenic substitution mapping with integrative analysis of genomic architecture of identified susceptibility loci and functional studies . We previously confirmed the pivotal role of a major albuminuria QTL on rat chromosome 6 ( RNO6 ) in the MWF model by generating a consomic MWF-6SHR strain , which carries RNO6 from the contrasting albuminuria-resistant SHR strain in the MWF genetic background ( Figure 1A–D ) ( Schulz et al . , 2007 ) . In addition , MWF rats inherit a deficit in nephron ( and glomeruli ) number , which represents a predisposition for the development of both hypertension and kidney damage ( Wang and Garrett , 2017 ) . We performed congenic substitution mapping for both albuminuria and glomerular density phenotypes by generating eight congenic lines by introgression of nested chromosomal fragments from SHR onto the MWF genetic background , and compared the renal phenotypes between congenic lines and the parental MWF strain ( Figure 1E , F ) . We successfully narrowed the original region identified by QTL mapping , spanning about 55 Mb between genetic markers D6Rat106 and D6Rat9 ( Schulz et al . , 2003 ) , to a smaller interval comprising 4 . 9 Mb between D6Mgh4 and D6Rat81 ( Figure 1E , F ) . The comparison between the two informative congenic lines MWF . SHR- ( D6Rat1-D6Mgh4 ) and MWF . SHR- ( D6Rat1-D6Rat81 ) revealed that 95% of the difference in albuminuria and 89% of the difference in glomerular density observed between the MWF and consomic MWF-6SHR strains is attributable to this interval ( Figure 1E , F ) . Subsequently , we set out to analyze the BP phenotype by direct intra-arterial measurements in the two congenic lines in comparison to the MWF and SHR strains . This analysis revealed similar mean arterial BP values in the two congenic lines and parental MWF strain ( Figure 2 ) . This findings clearly indicate the UAE difference between MWF . SHR- ( D6Rat1-D6Mgh4 ) and MWF . SHR- ( D6Rat1-D6Rat81 ) are not attributable to BP differences ( Figure 2 ) . Thus , we dissected away a role of this region for BP regulation and show that both the albuminuria and glomerular density phenotype co-localize in the same refined locus ( sub-QTL ) , supporting further exploration of this region as an independent candidate region for kidney damage . Direct comparison of glomerular density ( Figure 1G–J ) with the total glomerular number as previously estimated by the physical fractionator method ( Figure 1K ) ( Schulz et al . , 2007 ) confirmed the nephron deficit in MWF compared to SHR ( Figure 1J , p<0 . 0001 ) . We further demonstrated that this phenotype maps also to RNO6 ( Figure 1F ) when determined by glomerular density analysis in kidney sections ( Figure 1G–J ) . We employed targeted NGS on the region ranging from nucleotide positions 105 , 780 , 000 to 111 , 425 , 000 on RNO6 . After genotype calling and comparing genotypes of consomic and congenic rats , we refined coordinates of the kidney damage locus from 106 , 400 , 000 bp to 111 , 360 , 000 bp ( Figure 3A , B ) . This region contains 75 predicted protein-coding genes . In comparison to the reference genome ( Rattus norvegicus , ENSEMBL rn6 . 0 ) ( Yates et al . , 2016 ) , we identified 5 , 158 SNPs and 1893 small insertions and deletions ( INDELs ) in MWF , and 5326 single nucleotide polymorphisms ( SNPs ) and 1804 INDELs in SHR ( Figure 3—figure supplement 1 ) . Direct comparison between MWF and SHR revealed that both strains differ for 5376 SNPs and for 1613 INDELs ( Figure 3—figure supplement 1 ) , showing a remarkable pattern of stretches of variants either coming from the MWF or the contrasting SHR strain ( Figure 3A ) . As selective sweeps have recently been reported as a consequence of inbreeding in rats ( Atanur et al . , 2013 ) , we performed a formal analysis of selective processes such as directional selection or balancing selection , genetic hitchhiking , or introgression using the Tajima’s D statistics ( Tajima , 1989 ) . We observe such traces of selection with the majority of Tajima’s D values exceeding the generally accepted threshold of D > 2 ( Figure 3C ) . Functional annotation using PROVEAN scores identified eight potentially deleterious non-synonymous variants in five genes ( Table 1 ) . In addition , in the contrasting SHR reference strain one frameshift deletion was detected in the gene encoding neuroglobin ( Ngb ) . However , this deletion is also present in the entire clade of SHR-related rat strains derived from one ancestor including other strains with normal urinary albumin excretion and normal kidney function ( Table 2 ) ( Atanur et al . , 2013 ) . Consequently , this frameshift deletion is not to be considered involved in the kidney damage phenotype and was not further pursued . Thus , we identified no obvious single candidate by NGS analysis in the sub-QTL . In order to identify other positional candidates at the sub-QTL on RNO6 , we next embarked on RNA-Seq analysis in isolated glomeruli from MWF and SHR to assess global mRNA transcription patterns in the target compartment . We performed gene-based differential expression analysis using Cuffdiff and DESeq2 software tools ( Figure 3D , E ) . After correcting for multiple testing , 1838 genes were assigned a p-value<0 . 05 for Cuffdiff analysis and 1841 genes for DESeq2 , respectively , yielding a total set of 2454 unique differentially expressed genes . When filtering the results for those genes residing in the candidate region , we identified a total of 10 genes to be significantly differentially expressed between MWF and SHR ( Table 3 , Figure 3D , E ) at significant p-values . These genes were taken forward to validation by quantitative real-time PCR ( qPCR ) analysis in isolated glomeruli obtained from the two parental and MWF-6SHR consomic animals during onset of albuminuria occurring between 4 and 8 weeks of age ( Figure 4A ) . This analysis revealed that from the nine genes , which could be analyzed , only Tmem63c , showed consistent and allele-dependent differential expression during the crucial time window ( Figure 4A , B ) . RNA-Seq and qPCR analysis indicated a significant 2 . 5- to 3-fold upregulation of Tmem63c mRNA expression in isolated glomeruli in the MWF model , which was abolished by transfer of RNO6 from SHR onto the MWF genetic background in the corresponding MWF-6SHR consomic line ( Figure 4A ) . Moreover , comparison of Tmem63c mRNA expression between the two informative congenic lines MWF . SHR- ( D6Rat1-D6Rat81 ) and MWF . SHR- ( D6Rat1-D6Mgh4 ) confirmed an allelic ( cis ) regulation of Tmem63c mRNA expression in MWF and SHR , and its association with albuminuria ( Figure 4B ) . Further evaluation of the NGS data for Tmem63c in MWF and SHR revealed no significant sequence variants , with the exception of one detected variant in intron 18 ( ENSRNOT00000015571 ) at 111 , 101 , 251 bp at a potential splice site position . However , when analyzing our RNA-Seq data concerning differential exon usage in Tmem63c , we found no significant difference in exon usage between both parental strains . We then set out to perform immunohistochemistry analysis of TMEM63C in MWF kidney . This revealed TMEM63C expression in a podocyte-specific pattern ( Figure 4C ) . In contrast to the clearly elevated mRNA expression in isolated glomeruli , this analysis indicated no elevated glomerular protein expression in MWF , and only somewhat lower TMEM63C protein expression in podocytes in MWF compared to SHR at onset of albuminuria at 8 weeks of age ( Figure 4C ) . Aging MWF rats develop histopathological changes similar to those observed in patients with FSGS ( Remuzzi et al . , 1992; Schulz and Kreutz , 2012 ) . Podocyte injury with the development of glomerular proteinuria represents a pivotal hallmark of FSGS ( D'Agati et al . , 2011; Lim et al . , 2016 ) . Thus , we explored TMEM63C expression in patients with FSGS and healthy controls to evaluate its potential role for human kidney damage ( Yu et al . , 2016 ) . This analysis demonstrated that TMEM63C is expressed in podocytes of all glomeruli in healthy controls ( Figure 5A , E ) , while patients with FSGS exhibit a significant decrease of TMEM63C expression ( Figure 5B–E ) with a global loss of glomerular TMEM63C in the majority of patients analyzed ( Figure 5D–F ) . In addition to TMEM63C expression , we analyzed the expression of nephrin protein as a pivotal component of the slit diaphragm of the GFB ( Figure 5G–J ) ( Kestilä et al . , 1998 ) . Nephrin expression was also significantly reduced in patients with FSGS ( Figure 5H , I ) , which is in accordance with previously published results ( Kim et al . , 2002 ) . Moreover , we observed a shift from the normal linear staining pattern to a granular staining pattern as reported ( Figure 5H , J ) ( Doublier et al . , 2001; Wernerson et al . , 2003 ) . We further analyzed the effect of reduced TMEM63C expression in human podocytes in culture using small interfering RNA ( siRNA ) methodology ( Figure 6A ) . We found significantly impaired cell viability in response to TMEM63C downregulation ( Figure 6B ) . In addition , reduction of TMEM63C expression by siRNA decreased pro-survival signaling in human podocytes as indicated by reduced levels of pAKT ( Figure 6C ) , and increased pro-apoptotic transition of cytochrome C from mitochondria to the cytoplasm ( Figure 6D ) . To assess the functional role of tmem63c for albuminuria development , we utilized the transgenic zebrafish line Tg[fabp10a:gc-EGFP] ( Zhou and Hildebrandt , 2012 ) . This model expresses a vitamin D binding protein tagged with enhanced green fluorescent protein ( gc-EGFP ) in the liver , from which it is released into the blood stream and circulates under the normal conditions in the blood ( Figure 7A , G ) . Upon GFB damage , gc-EGFP leaks through the glomerular filtration barrier , indicated by a marked decrease in fluorescence in the trunk vasculature of Tg[fabp10a:gc-EGFP] embryos mimicking an albuminuria-like phenotype ( Figure 7A ) . To reduce tmem63c levels in developing zebrafish embryos , we used the morpholino knockdown technology as well as CRISPR/Cas9-mediated somatic mutagenesis ( Bassett et al . , 2013; Burger et al . , 2016 ) ( Figure 7B ) . In both morpholino-injected embryos and crispants ( CRISPR/Cas9-mediated somatic mutants ) , loss of tmem63c did not result in any visible developmental malformations apart from mild pericardial edema ( Figure 7C–F ) . Knockdown of tmem63c using morpholino technology resulted in a significant decrease in gc-EGFP fluorescence in the trunk vasculature at 120 hr post fertilization ( hpf ) ( Figure 7H ) . We corroborated this finding in tmem63c crispants ( Figure 7J , Figure 7—figure supplement 1D ) as well as by using another splice-blocking morpholino ( Figure 7—figure supplement 1E–I ) ; both experiments showed a similar albuminuria-like phenotype . To verify the specificity of the observed phenotype rescue experiments were carried out by co-injection of zebrafish tmem63c mRNA with tmem63c sgRNA/Cas9 complexes and rat Tmem63c mRNA ( mRNA sequence identity vs . zebrafish = 65 . 82% ( Clustal 2 . 1 ) , Figure 7—figure supplement 2 ) with tmem63c ATG-MO , respectively . For both cases similarly , the albuminuria-like phenotype could be specifically rescued proving knockdown specificity on the one hand and functional conservation of tmem63c across species on the other hand ( Figure 7I , K and L ) . Our data indicate that tmem63c may regulate the GFB integrity . To understand the possible functional changes in GFB in more detail , we deployed electron microscopy to visualize the GFB ultrastructure in embryos with reduced tmem63c levels . We observed significant changes in podocyte foot process morphology manifested by foot process effacement ( Figure 8A–C ) . Quantitative analysis in tmem63c crispants revealed a significant increase in the foot process width compared to uninjected controls and Cas9-controls ( Figure 8D ) with concomitant significant decrease of the number of slit diaphragms per µm glomerular basement membrane ( GBM ) ( Figure 8E ) . To analyze , whether the observed albuminuria-like phenotype upon tmem63c-deficiency is associated with the loss of podocytes , we utilized confocal microscopy to image glomeruli of Tg ( wt1b:EGFP ) embryos . In this analysis , tmem63c crispants ( tmem63c ex2-sgRNA ) showed a widened Bowman´s space and increased glomerular volumes compared to uninjected controls ( 133457 ± 59547 µm3 vs 67067 ± 21933 µm3 , p=0 . 04 ) as quantified by 3D surface reconstruction . In addition , we observed dilated capillary loops in the crispants ( Figure 8F–H ) . Quantification of podocyte cell number in embryos with reduced tmem63c levels revealed no changes in absolute cell number ( Figure 8I , Videos 1–3 ) , while relative podocyte cell number normalized to the total glomerular volume was significantly decreased compared to uninjected controls ( Figure 8J , Videos 1–3 ) . Collectively , our data indicate the conserved role for tmem63c in GFB function between fish , rodents , and humans . Previous genetic analysis in the fawn-hooded hypertensive ( FHH ) rat model led to the identification of naturally occurring genetic variants in RAB38 , member RAS oncogene family ( Rab38 ) , and shroom family member 3 ( Shroom3 ) that successfully complemented studies in humans supporting their role for albuminuria ( Rangel-Filho et al . , 2013; Yeo et al . , 2015 ) . Interestingly , the genetic variant in Rab38 was linked to altered tubular ( Rangel-Filho et al . , 2013; Teumer et al . , 2016 ) , and the variant in Shroom3 to altered glomerular ( Yeo et al . , 2015 ) albumin handling . Thus , animal models have been proven useful not only for the explanation of missing heritability ( Chatterjee et al . , 2013 ) , but also for the elucidation of the differences between glomerular and tubular origins of albuminuria ( Rangel-Filho et al . , 2013; Yeo et al . , 2015 ) . Currently , in GWAS meta-analyses in general population cohorts , only cubilin ( CUBN ) has been significantly associated with albuminuria ( Böger et al . , 2011 ) . In addition , variants in HS6ST1 and near RAB38/CTSC were implicated in albuminuria in patients with diabetes ( Teumer et al . , 2016 ) . Nevertheless , in parallel to the findings obtained in BP GWAS ( Hoffmann et al . , 2017; Warren et al . , 2017 ) , only a small fraction of the estimated heritability of albuminuria can be attributed to the identified genes ( Langefeld et al . , 2004; Teumer et al . , 2016 ) . The MWF rat , in which we identified at least 11 albuminuria QTL , highlights the polygenic nature of complex traits such as kidney damage with albuminuria in hypertension ( Schulz and Kreutz , 2012 ) . Here , we focused on a major albuminuria QTL ( Schulz and Kreutz , 2012; Schulz et al . , 2003 ) and refined the candidate region to a sub-QTL comprising 4 . 9 Mb . We confirmed that both the albuminuria and the nephron deficit phenotype map to the same genomic region supporting a genetic link between albuminuria and embryonic/fetal nephron development ( Wang and Garrett , 2017 ) . In addition , the informative congenic lines with differential albuminuria development showed similar BP values by which we showed that this genomic region affects albuminuria development independently from BP changes . In the targeted NGS analysis , multiple non-deleterious variants but no clear candidate was identified . However , significant signs of selective sweeps in this region , potentially leading to an enrichment of multiple non-deleterious alleles due to genetic hitchhiking , were detected . This is in agreement with a recent systematic genome sequencing study of laboratory inbred rat strains indicating that private single-nucleotide variants are highly concentrated in a small number of discrete regions of the genome ( Atanur et al . , 2013 ) . The authors of this report hypothesized that variants that are unique to a single strain reside within these regions because many of these regions were positively selected in the initial phenotype-driven derivation of these strains . Surprising - and challenging however - remains our observation that no obvious causative variant for the kidney damage phenotype was observed in our targeted NGS analysis . Notwithstanding , complex traits can also be a consequence of gene expression changes resulting in dysregulation of physiological pathways , a concept which becomes increasingly recognized as RNA sequencing is employed to distinguish tissue-specific gene expression patterns in a variety of complex diseases ( Joehanes et al . , 2017; Kirsten et al . , 2015 ) . We therefore complemented the DNA re-sequencing by compartment-specific RNA-Seq analysis which is rational , because MWF rats develop early glomerular changes that precede albuminuria development and show early podocyte injury in parallel with the onset of albuminuria ( Ijpelaar et al . , 2008 ) . We identified Tmem63c as a positional candidate based on its mRNA expression pattern with differential glomerular expression in allele-specific rat models . Immunohistochemistry analysis in MWF kidneys identified TMEM63C protein in podocytes , and in contrast to the observed marked glomerular mRNA upregulation only a modest downregulation . Bioinformatics analysis of our NGS and RNA-Seq data has not yet provided a potential explanation for this discrepancy , for example interstrain differences related to exon usage or other events shown to mediate protein expression control on the post-transcriptional level . The relationship between protein levels and their coding transcripts is , however , rather complex and factors such as spatial and temporal variations of mRNAs , as well as the local availability of resources for protein biosynthesis , strongly influence this relationship ( Liu et al . , 2016 ) . Spatial variation in expression regulation during renal injury development could be specifically important in podocytes due to their unique ultrastructural and molecular anatomy ( Endlich et al . , 2017 ) . To explore the potential clinical relevance of TMEM63C for kidney damage , we selected patients with FSGS because they mirror the disease pattern observed in the MWF model ( D'Agati et al . , 2011; Lim et al . , 2016; Yu et al . , 2016 ) . Importantly , patients with FSGS exhibit a significant decrease of TMEM63C protein levels in podocytes with a global loss of glomerular expression in the majority of patients . Furthermore , the loss of TMEM63C was associated with a decrease and altered granular staining pattern of nephrin protein in glomeruli of FSGS patients as previously reported ( Doublier et al . , 2001; Wernerson et al . , 2003 ) . The latter has been previously shown to correspond to the degree of foot process effacement ( Wernerson et al . , 2003 ) . While our data cannot establish a functional link between changes in nephrin and TMEM63c in FSGS , they nonetheless demonstrate a concomitant deficiency of both proteins in this setting . To further validate Tmem63c , we tested its functional relevance for albuminuria development in zebrafish ( Danio rerio ) . Our data confirm that in both tmem63c morphants and crispants , loss of tmem63c leads to a phenotype indicative of a GFB defect . Importantly , rescue of the observed phenotypes in zebrafish with rat Tmem63c mRNA pointed to the functional conservation across species . Ultrastructural analysis in zebrafish embryos demonstrated the role of tmem63c for the GFB integrity by revealing overt effacement of podocyte foot processes upon gene deficiency . Further structural analysis by confocal microscopy revealed dilated capillary loops and enlarged glomerular volumes with a reduction of relative podocyte number in relation to glomerular volume ( podocyte density ) in tmem63c-deficient embryos . Thus , our data reveal an important role of tmem63c for normal development of capillary structure and podocyte function of the glomerulus in zebrafish . Tmem63c belongs to the TMEM ( transmembrane protein ) gene family comprising more than 300 different proteins with about 580 transcript variants ( Wrzesiński et al . , 2015 ) . These proteins are predicted components of cellular membranes . However , the function of the majority of TMEM proteins - including Tmem63c - is currently unclear . There are no previous reports that demonstrate TMEM63C expression in the kidney , while assessment of the human protein atlas database indicates TMEM63C expression in human kidney in both the glomerular and tubular compartments ( Lundberg et al . , 2010 ) ; ‘Tissue expression of TMEM63C - Staining in kidney’ - The Human Protein Atlas , 2019; Uhlén et al . , 2015 ) . Tmem234 , another member of the TMEM family , may represent a component of the basal membrane of podocytes and has recently been associated with proteinuria development in zebrafish ( Rodriguez et al . , 2015 ) . The mechanism by which changes in Tmem63c expression cause podocyte damage and thus contribute to the final common pathway of injury in FSGS ( De Vriese et al . , 2018 ) remains however unclear and cannot be elucidated by our current analysis . Accordingly , our data do not allow to describe possible interactions between TMEM63c and the heterogeneous group of known causative FSGS genes . Nevertheless , previous reports suggested that Tmem63c and two other genes of the tmem63-family , i . e . Tmem63a and Tmem63b , are mammalian homologues of the hyperosmolarity-activated cation channel proteins AtCSC1 and its paralogue OSCA1 in plants ( Hou et al . , 2014; Zhao et al . , 2016 ) . Based on these reports , it appeared tempting to speculate that Tmem63c may be involved in mechanosensing in podocytes providing thus a functional basis for interactions with cytoskeleton genes or slit pore proteins , for example nephrin , in podocytes . However , the potential role of Tmem63c as a hyperosmolarity-activated cation channel remains controversial , since the activation by hyperosmolarity was not confirmed in a more recent study ( Murthy et al . , 2018 ) . Moreover , the latter report indicated that Tmem63c is phylogenetically divergent from Tmem63a and Tmem63b and lacks their function to induce stretch-activated ion currents . Consequently , further studies are needed to characterize the functional role of Tmem63c and to explore its potential to influence podocyte function as a novel target for therapeutic intervention ( Endlich et al . , 2017; Forst et al . , 2016; Wieder and Greka , 2016 ) . In contrast to our experiments in zebrafish , in the explorative analysis in biopsies of FSGS patients we cannot differentiate between a primary ( causative ) or secondary effect of TMEM63C expression loss in patients . Nevertheless , taken together with our experimental analysis , the data clearly support Tmem63c as a novel candidate for further translational research on kidney damage . Male rats were obtained from our MWF/Rkb ( RRID:RGD_724569 , laboratory code Rkb , http://dels . nas . edu/ilar/ ) and SHR/Rkb ( RRID:RGD_631696 , laboratory code Rkb , http://dels . nas . edu/ilar/ ) colonies at the Charité – Universitätsmedizin Berlin , Germany . The consomic MWF-6SHR ( RRID:RGD_1641831 ) was previously described ( Schulz et al . , 2007 ) . Rats were grouped under conditions of regular 12 hr diurnal cycles with an automated light switching device and climate-controlled conditions at a room temperature of 22°C . The rats were fed a normal diet containing 0 . 2% NaCl and had free access to food and water . A panel of eight congenic rat lines MWF . SHR- ( D6Rat1-D6Rat30 ) , MWF . SHR- ( D6Rat1-D6Rat106 ) , MWF . SHR- ( D6Rat1-D6Mit8 ) , MWF . SHR- ( D6Rat1-D6Rat121 ) , MWF . SHR- ( D6Rat1-D6Mgh4 ) , MWF . SHR- ( D6Rat1-D6Rat81 ) , MWF . SHR- ( D6Rat1-D6Rat115 ) , and MWF . SHR- ( D6Rat1-D6Rat184 ) was generated by transfer of different nested SHR segments onto the MWF background . For this procedure , male and female rats of the MWF-6SHR breeding , that were homozygous for all MWF chromosomes except RNO6 and heterozygous for RNO6 , were intercrossed ( Schulz et al . , 2003 ) . All experimental work in rat models was performed in accordance with the guidelines of the Charité-Universitätsmedizin Berlin and the local authority for animal protection ( Landesamt für Gesundheit und Soziales , Berlin , Germany ) for the use of laboratory animals . The registration numbers for the rat experiments are G 0255/09 and T 0189/02 . Urinary albumin excretion was measured as reported ( Kreutz et al . , 2000 ) . Direct intra-arterial BP measurements were performed in awake male rats at 14 weeks of age as previously described ( Kreutz et al . , 1995; Schulz et al . , 2010 ) . For determination of glomerular density , animals were sacrificed under ketamine-xylazine anesthesia ( 87 and 13 mg/kg body wt , respectively ) at week 4 . The right kidney was fixed in methacarn and embedded in paraffin . Tissue samples were cut into 5-μm-thick histological sections ( Figure 1G–I ) and stained with the periodic acid-Schiff ( PAS ) technique . Section analysis was performed by a photomicroscope Axiophot ( Zeiss ) and a digital camera system AxioCam MRc Rev . 3 FireWire ( Zeiss ) at a 10x magnification . Glomerular density was calculated using the formula n = G/FA ( D + T ) as reported ( ELIAS et al . , 1961; Lucas et al . , 1997 ) . Glomeruli in 20–25 fields for each sample were counted in the outer cortex zone ( Figure 1H , I ) . Glomerular diameter was calculated by the AxioVision release 4 . 8 . 2 software program ( Zeiss ) . This method was validated by comparison with the absolute nephron numbers as determined by the physical fractionator method in rat strains as previously reported ( Gundersen , 1986; Schulz et al . , 2007 ) . Based on the fine mapping results of the congenic MWF strains , a target candidate region of about 5 . 63 Mbp ( chr6:105 . 8–111 . 43 Mb , R . norvegicus , ENSEMBL rn6 . 0 ) ( Yates et al . , 2016 ) was defined for subsequent next-generation resequencing analysis . The solution-based SureSelectXT ( Agilent Technologies ) capture method was applied for custom target enrichment of the defined region according to the manufacturer’s instructions starting with 3 μg genomic rat DNA of MWF , SHR , MWF-6SHR , MWF . SHR- ( D6Rat1-D6Mgh4 ) , and MWF . SHR- ( D6Rat1-D6Rat81 ) ( n = 3 , each ) , uniquely labelled by index tags . Library quality control and final quantification for subsequent pooling of the 16 sequencing libraries was performed using the 2100 Bioanalyzer instrument ( Agilent Technologies ) . The pooled library was paired end sequenced ( 2 × 76 cycles plus index read ) on a MiSeq system ( Illumina ) using the MiSeq reagent kit v3 ( Illumina ) . The CASAVA software package v1 . 8 . 2 ( Illumina ) was used for demultiplexing of the sequencing reads and conversion to fastq data for further analysis . The resulting high-quality reads for identification of SNPs and short INDELs were mapped to the reference genome according to the annotation release Rat Genome Sequencing Consortium ( RGSC ) genome assembly v6 . 0 using BWA software , version 0 . 6 . 2 ( Li and Durbin , 2009 ) . Mapped reads were processed and calling of SNPs and short INDELs were performed by the GATK pipeline , version 2 . 8 ( Van der Auwera et al . , 2013 ) . Effects of the genomic variations were evaluated with the SnpEff software tool , version 3 . 3 ( Cingolani et al . , 2012 ) . Impairment of protein function by common exonic variants in MWF and SHR were analysed using the PROVEAN algorithm ( Choi et al . , 2012 ) . A PROVEAN Score <2 . 5 was considered significant for genes . The Tajima’s D statistic ( Tajima , 1989 ) was used to test for signatures of selection in the region of interest utilizing the vcftools software ( Vs . 0 . 1 . 13 ) ( Danecek et al . , 2011 ) . Different protocols were used for isolation of glomeruli from male rats at 4 and 8 weeks of age , due to the different body size . Rats were anesthetized with ketamine-xylazine ( 87 and 13 mg/kg body weight , respectively ) . In 4-week-old rats , the abdominal artery was catheterized and kidneys were perfused with 10 ml 1x phosphate buffered saline ( PBS ) and subsequently with 20 ml ferrous solution ( 12 . 5 g ferric oxide ( Iron ( II/III ) powder <5 micron , 98%; Sigma- Aldrich Chemie GmbH ) suspended in 1000 ml 1x PBS . Kidneys were removed , decapsulated and passed through a 125 μm steel sieve ( Retsch GmbH ) with 1x PBS . The glomeruli containing ferrous particles were gathered by a magnet , snap-frozen and stored at −80°C . Kidneys of 8-week-old rats were removed , decapsulated and passed through a 125 μm steel sieve with 1x PBS . The filtrate was put on a 71 μm steel sieve ( Retsch GmbH ) to separate glomeruli from the flow-through . Glomeruli were washed off the sieve with 1x PBS , centrifuged , immediately snap-frozen and stored at −80°C . RNA sequencing ( RNA-Seq ) was performed in glomerular RNA of male MWF and SHR rats at week 4 ( n = 3 , each ) . The NEBNext Poly ( A ) mRNA magnetic isolation module followed by library preparation using NEBNext Ultra RNA Library Prep Kit for Illumina ( New England BioLabs ) was applied on 1 μg total RNA to generate a cDNA library for subsequent paired end ( 80 cycles ) sequencing on the NextSeq 500 system ( Illumina ) using v2 chemistry yielding in about 415M single reads . RNA and library quality control was performed using the Bioanalyzer RNA 600 Nano and High-Sensitivity DNA Analysis Kit ( Agilent Technologies ) , respectively . The KAPA Library Quantification Kit ( Kapa Biosystems ) was used for library quantification . Initial quality control of the raw data was performed using Cutadapt version 1 . 9 ( Martin , 2012 ) program . Raw reads were quality trimmed ( minimal base quality: 25 , minimal read length after trimming: 70 nt ) , adapter sequences were removed from reads 3’ ends . TopHat2 version 2 . 1 . 0 ( Trapnell et al . , 2009; Trapnell et al . , 2010 ) software tool together with Bowtie2 aligner version 2 . 2 . 3 ( Langmead and Salzberg , 2012 ) was used for read mapping against ENSEMBL rn6 . 0 reference assembly ( Yates et al . , 2016 ) . After the reads have been mapped to the reference genome , the Cufflinks version 2 . 2 . 1 ( Trapnell et al . , 2010 ) program together with Ensembl ( release 81 ) gene annotation ( Aken et al . , 2016 ) , baw093 ) were used to assemble transcripts and estimate their abundances . Differential expression analysis was performed using both Cuffdiff version 2 . 2 . 1 software package ( Trapnell et al . , 2010 ) and DESeq2 R package version 1 . 12 . 4 ( Love et al . , 2014 ) . Genes having absolute fold change value <1 . 5 were excluded from further analysis . Genes were considered significantly differentially expressed if the corresponding adjusted p-value was less than 0 . 05 . First-strand cDNA synthesis was carried out on 2 µg of total RNA using the First Strand cDNA Synthesis Kit ( Fermentas Life Sciences ) following the manufacturer’s protocol . Isolated glomeruli preparations of rat strains were analyzed at week 4 and week 8 . qPCR of each gene was performed in a 7000 Real-Time PCR System ( Applied Biosystems ) with version 1 . 2 . 3 software or a 7500 Fast Real-Time PCR System with version 2 . 0 . 6 software ( Applied Biosystems ) using the comparative quantification cycle method as reported ( Fast SYBRGreen Master Mix or Power SYBR Green PCR Master Mix; Applied Biosystems ) ( Schulz et al . , 2008 ) . Primers are listed in Figure 4—source data 2 . Normalization of expression data was done by the reference gene hydroxymethylbilane synthase ( Hmbs ) ( Schulz et al . , 2008 ) . For all analyses , three technical replicates of each animal/experiment were performed . Genes with low mRNA expression levels were only considered when the quantification cycles ( Cqs ) were ≥30 and the Cqs of the no-template controls were at least 5 Cqs delayed . Acyl-CoA thioesterase 3 ( Acot3 ) demonstrated low expression levels and was therefore not analyzed . For determination of protein expression of TMEM63C an anti-TMEM63C antibody ( epitope: GLRGFARELDPAQFQEGLE , custom antibody production: Perbio Science Germany ) was generated . The epitope does not cross react with TMEM63A or TMEM63B or other genes . For Wilms tumor 1 ( WT1 ) protein expression analysis , we used a rabbit anti-WT1 antibody ( Santa Cruz ) . For nephrin protein expression analysis , we used a rabbit anti-nephrin antibody ( Abcam ) . Paraffin embedded rat kidney sections and human biopsy samples were cut at 4 µm and incubated with the anti-TMEM63C antibody ( 1:1600 for rat tissue , 1:800 for human biopsies ) , the anti-WT1 antibody ( 1:500 ) or the anti-nephrin antibody ( 1:750 ) . Rabbit IgG negative control fraction was used as a negative control in the same concentration as the primary antibody . Goat anti-rabbit EnVision HRP conjugate ( Dako ) was used as secondary antibody . The staining was visualized using diaminobenzidine as the chromogen and counterstained with haematoxylin . Consecutive slides of MWF and SHR kidney sections stained for TMEM63C and WT1 were evaluated to determine TMEM63C co-localization with podocytes . TMEM63C protein level in glomeruli was analyzed using ImageJ analysis . Renal biopsy samples of patients with FSGS ( Table 4 ) were collected from the archive of the Department of Pathology of the Leiden University Medical Center ( LUMC ) . Demographic data and laboratory data at time of biopsy were retrospectively retrieved from the patients’ medical records or pathology reports following the good practice guidelines of the LUMC . All biopsy samples were handled and analyzed anonymously in accordance with the Dutch National Ethics Guidelines ( Code for Proper Secondary Use of Human Tissue , Dutch Federation of Medical Scientific Societies ) . This study is in agreement with the Declaration of Helsinki and the Department of Health and Human Services Belmont Report and the use of the patient biopsies was approved by the medical ethical committee of the LUMC ( registration number G16 . 110 ) . Samples obtained from Eurotransplant donors that were unsuited for transplantation because of technical problems , were used as healthy controls . All sections were scored separately by two observers for TMEM63C intensity as well as for percentage of glomeruli with loss of TMEM63C staining . Each case was given a score for TMEM63C staining intensity: high intensity in >50% of glomeruli , intermediate intensity in >50% of glomeruli , low intensity in >50% of glomeruli or no TMEM63C staining present in >50% of glomeruli . Secondly , sections were scored based on the percentage of glomeruli with loss of TMEM63C expression in podocytes . Per case , the percentage of glomeruli with 1 ) no loss 2 ) <25% loss 3 ) 25–50% loss or 4 ) >50% loss of TMEM63C expression in podocytes was determined . For nephrin staining analysis , each glomerulus was scored based on staining pattern ( linear or granular ) and loss of staining ( no loss , segmental loss or global loss ) . All biopsy samples were handled and analyzed anonymously in accordance with the Dutch National Ethics Guidelines ( Code for Proper Secondary Use of Human Tissue , Dutch Federation of Medical Scientific Societies ) and in agreement with the Declaration of Helsinki and the Department of Health and Human Services Belmont Report . The use of the patient biopsies was approved by the medical ethical committee of the LUMC . Immortalized human podocytes ( RRID:CVCL_W186 , a kind gift from Professor Moin Saleem , MA , Academic and Children's Renal Unit , University of Bristol , Bristol , UK ) ( Saleem et al . , 2002 ) were used as described ( Eisenreich et al . , 2016 ) . The cell line has previously been authenticated ( Saleem et al . , 2002 ) and we have confirmed this by expression of podocyte specific markers such as podocin and synaptopodin as recently reported ( Eisenreich et al . , 2016 ) . The cell line tested negative for mycoplasma contamination . Before transfection , human podocytes were starved with FBS-free RPMI 1640 medium overnight . Transfection of cells was performed using 200 nM of TMEM63C-specific siRNAs ( siTMEM63C; Sigma-Aldrich Chemie GmbH ) or non-sense control siRNAs ( siControl; Sigma-Aldrich Chemie GmbH ) as well as Lipofectamine 2000 ( Life Technologies GmbH ) . The transfection efficacy of 25% in human podocytes was experimentally determined earlier ( Eisenreich et al . , 2016 ) . Western blot analyses were done as described earlier ( Eisenreich et al . , 2016; Langer et al . , 2016 ) . For detection , specific antibodies against TMEM63C ( Thermo Fisher Scientific ) , GAPDH ( Calbiochem ) , protein kinase B ( AKT; Merck Chemicals GmbH ) , and phospho-AKT ( Ser473 , pAKT; Merck Chemicals GmbH ) were used . Quantification of Western blot analyses were done using Gel-Pro Analyzer software version 4 . 0 . 00 . 001 ( Media Cybernetics ) . The cytochrome C releasing apoptosis assay kit ( BioVision Inc ) was used following the manufacturer’s protocol as described previously ( Eisenreich et al . , 2016 ) . In brief , 1 × 104 cells per well were transfected for 48 hr with siTMEM63C or siControl , respectively . After that , cells were lysed and the cytosolic fraction was separated from the mitochondrial fraction . Comparative Western blot analyses of these fractions using a cytochrome C-specific antibody were performed to determine pro-apoptotic translocation of cytochrome C from mitochondria into cytosol . The calcein AM ( acetoxymethyl ) cell viability kit ( Trevigen Inc ) was used as earlier described following the manufacturer’s protocol ( Eisenreich et al . , 2016 ) . In brief , 1 × 104 cells per well were transfected for 48 hr with siTMEM63C or siControl , respectively . Then , human podocytes were washed and incubated with calcein AM working solution for 30 min . Fluorescence was measured at 490 nm excitation and 520 nm emission . Zebrafish were bred , raised and maintained in accordance with the guidelines of the Max Delbrück Center for Molecular Medicine and the local authority for animal protection ( Landesamt für Gesundheit und Soziales , Berlin , Germany ) for the use of laboratory animals , and followed the ‘Principles of Laboratory Animal Care’ ( NIH publication no . 86–23 , revised 1985 ) as well as the current version of German Law on the Protection of Animals . Injection droplets of approximately 1 nl were injected into one-cell stage zygotes of the zebrafish wild type hybrid strain AB/Tülf and the transgenic lines Tg ( fabp10a:gc-EGFP ) ( Zhou and Hildebrandt , 2012 ) and Tg ( wt1b:GFP ) ( Bollig et al . , 2009; Perner et al . , 2007 ) . Morpholinos ( MO ) of the following sequences were synthesized by Gene Tools LLC Philomath: tmem63c ATG-MO 5’-CAGGCCAGGACTCAAACGCCATTGC-3’ , tmem63c ex2-sdMO 5'-TGTTATCATAGATGATGTACCAGCC-3' , and standard control oligo ( Control-MO ) 5’-CCTCTTACCTCAGTTACAATTTATA-3’ . Tmem63c ATG-MO was used at a final concentration of 0 . 3 mM ( Figure 7B ) , tmem63c ex2-sdMO was used at a final concentration of 0 . 5 mM ( Figure 7—figure supplement 1 ) . sgRNA targeting exon 2 ( Figure 7B ) was generated as described ( Bassett et al . , 2013; Burger et al . , 2016 ) using the ex2-sgRNA forward primer with CRISPR target site underlined: GAAATTAATACGACTCACTATAGGACGTCAGGAGTTTCCTGAGTTTTAGAGCTAGAAATAGC and the invariant reverse primer: AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC . PCR product was purified with GeneJET Gel Extraction Kit ( Thermo Fisher Scientific , respectively ) . sgRNA was transcribed using the MEGAscript T7 Kit ( Ambion ) and extracted with RNeasy Mini Kit ( Qiagen ) according to the manufacturer’s protocol . sgRNA was diluted to a final concentration of 159 . 6 ng/μl or 250 ng/µl , respectively using water and 1 M KCl ( final concentration: 300 mM ) and co-injected with Cas9-Protein of 600 ng/μl final concentration as described ( Burger et al . , 2016; Gagnon et al . , 2014 ) . To determine the efficiency of sgRNA-mediated mutagenesis crispants alleles were analyzed as described ( Figure 7—figure supplement 1D ) ( Burger et al . , 2016 ) . The following primers ( BioTez Berlin-Buch GmbH ) were used to amplify the genomic region flanking the CRISPR target site; forward: CAAATGGTGAACACTTGTGAATC , reverse: CTGCGGTTTACTGCGGAGATG . Computational sequence analysis was performed using CrispR Variants ( Lindsay et al . , 2016 ) . For an injection control Cas9 was diluted to a final concentration of 600 ng/µl using water and 1 M KCl ( final concentration: 300 mM; Cas9-Control ) . For efficiency analysis of the tmem63c ex2-sdMO a reverse transcriptase ( RT ) -PCR was carried out ( Figure 7—figure supplement 1F ) . At 24 hpf , RNA from 50 pooled embryos was isolated using Trizol Reagent ( Invitrogen ) ; DNase I digestion was performed using the RNAse-free DNase set ( Qiagen ) and samples were purified using the RNeasy Mini Kit ( Qiagen ) according to the manufacturer’s protocol . After determination of RNA quality and quantity , equal amounts of mRNA for each group analyzed were transcribed to cDNA using First strand cDNA synthesis kit ( Thermo Fisher ) according to the manufacturer’s protocol . We amplified tmem63c from cDNA using DreamTaq DNA Polymerase ( Thermo Fisher ) with the following primers: forward: CTGATGGAGGAGAACAGCACGG , reverse: ATACAGCAGAGCGAAGATACTGTG . Eucaryotic elongation factor 1 alpha 1 , like 1 ( eef1a1l1 ) was used as a loading control and amplified using the following primers: forward: TGGAGACAGCAAGAACGACC , reverse: GAGGTTGGGAAGAACACGCC . Primers ( BioTez Berlin-Buch GmbH ) for the In-Fusion HD Cloning Kit ( Takara ) were designed using the web tool provided by TaKaRa ( TaKaRa ) ( TaKaRa , 2018 ) ; forward: GCTTGATATCGAATTCATGGCGTTTGAGTCCTGGCCTGC , reverse: CGGGCTGCAGGAATTCTCACTGAAAAGCCACCGGACTG . tmem63c cDNA was amplified using Phusion High-Fidelity DNA polymerase ( Thermo Fisher Scientific ) . The pBluescript II SK ( + ) vector was linearized using EcoRI FD enzyme . The tmem63c ORF was cloned into the pBluescript II SK ( + ) vector using the In-Fusion HD Cloning Kit ( Takara ) . The tmem63c cDNA was sequence-verified using the common T7 forward and M13 reverse primers . For sequencing of the whole ORF , an additional primer was used , GTGCAGAAACTAATGAAGCTGG , located at 822–844 bp starting from the beginning of the ORF . For in vivo rescue experiments tmem63c cDNA ( Danio rerio ) was linearized using ApaI FD ( Thermo Fisher Scientific ) and purified using GeneJET Gel Extraction Kit ( Thermo Fisher Scientific ) . In vitro transcription of capped RNA and following TurboDNase treatment were performed using mMessage mMachine T7 Kit ( Ambion ) . For poly-A-tailing , the Poly ( A ) -tailing Kit ( Ambion ) was used , followed by RNA extraction using RNeasy Mini Kit ( Qiagen ) . The mRNA was diluted to a concentration of 100 ng/μl and injected into one-cell stage zygotes . For in vivo rescue experiments , mRNA of a concentration of 100 ng/µl and ex2-sgRNA of a concentration of 159 . 6 ng/µl were subsequently injected into the same one-cell stage zygotes . For in vivo rescue of the tmem63c ex2-sdMO-mediated knockdown , mRNA with a concentration of 100 ng/µl and ex2-sdMO with a concentration of 0 . 5 mM were subsequently injected into the same zygote at one- or one-to-four cell stage , respectively . Tmem63c cDNA ( Rattus norvegicus ) was synthesized by Thermo Fisher Scientific using their GeneArt Gene synthesis service . For in vivo rescue experiments , Tmem63c cDNA was linearized using XbaI FD ( Thermo Fisher Scientific ) and purified using GeneJET Gel Extraction Kit ( Thermo Fisher Scientific ) . In vitro transcription of capped RNA followed by TurboDNase treatment were performed using mMessage mMachine T7 Kit ( Ambion ) . For poly-A-tailing , the Poly ( A ) -tailing Kit ( Ambion ) was used , followed by RNA extraction by RNeasy Mini Kit ( Qiagen ) . The mRNA was diluted to a concentration of 100 ng/μl and injected into one-cell stage zygotes . For in vivo rescue of the tmem63c ATG-MO-mediated knockdown , mRNA with a concentration of 100 ng/µl and ATG-MO with a concentration of 0 . 3 mM were subsequently injected into the same zygote at one- or one to four cell-stage , respectively . To assess the functionality of the GFB , the gc-EGFP fluorescence in the trunk vasculature of Tg ( fabp10a:gc-EGFP ) embryos were evaluated at 120 hpf by epifluorescence microscopy . For CRISPR-Cas9-mediated somatogenesis of tmem63c the above described sgRNA was used in a concentration of 159 . 6 ng/µl . Each embryo was visually assigned to the ‘fluorescent group’ , ‘deficient-fluorescent group’ , or ‘crippled/dead’ and their number quantified . Due to the heterogeneous genotype of the used transgenic Tg ( fabp10a:gc-EGFP ) zebrafish families , the ‘deficient-fluorescent group’ included the embryos with reduced fluorescence in the trunk as well as embryos that did not carry the transgene ( Figure 7—figure supplement 1A–C ) . The percentage of injected embryos was normalized to the percentage of the control group for each category . Quantifications were performed for at least three individual injections . Tg ( fabp10a:gc-EGFP ) embryos at 120 hpf were fixed in 4% formaldehyde/0 . 5% glutaraldehyde ( EM-grade ) in 0 . 1 M phosphate buffer for 2 hr at RT . For knockdown analysis embryos were injected with tmem63c ex2-sgRNA of a concentration of 250 ng/µl to enhance the observed phenotype ( Figure 7E , F ) . Prior to analysis embryos were sorted for a clear knockdown phenotype . Samples were stained with 1% OsO4 for 2 hr , dehydrated in a graded ethanol series and propylene oxide and embedded in Poly/BedR 812 ( Polysciences , Eppelheim , Germany ) . Ultrathin sections were contrasted with uranyl acetate and lead citrate . Sections were examined with a FEI Morgagni electron microscope and a Morada CCD camera ( EMSIS GmbH , Münster , Germany ) . Image acquisition and quantification of podocyte foot process width and number of slit diaphragms per µm GBM was performed with the iTEM software ( EMSIS GmbH , Münster , Germany ) . Tg ( wt1b:EGFP ) embryos at 96 hpf were fixed in PEM buffer containing 4% formaldehyde and 0 . 1% Triton-X 100 for 2 hr at RT or overnight at 4°C . For knockdown analysis embryos were injected with tmem63c ex2-sgRNA with a concentration of 250 ng/µl to enhance the observed phenotype . Nuclei were stained using 4′ , 6-Diamidin-2-phenylindol ( DAPI , Sigma Aldrich , stock solution 1 mg/ml diluted 1:2000 in PBS ) overnight at 4°C . After removal of the yolk and mounting in 0 , 7% low-melting agarose , the kidneys of whole-mount fixed embryos were imaged using a Zeiss LSM 710 or LSM 700 microscope with a LD C-Apochromat 40 x NA1 . 1 water objective and ZEN 2 . 1 software by sequentially acquiring confocal z-stacks of the GFP ( 488 nm laser , emission 495–550 nm ) and the DAPI signal ( 405 nm laser , emission 420–480 nm ) with a pixel size of 102 . 4 nm . Care was taken to apply identical settings to all samples and not to oversaturate pixels . Quantification of podocyte cell number and glomerular volume was done using Imaris version 9 . 21 software ( RRID:SCR_007370 , Bitplane AG , Zurich , Switzerland ) . A 3D surface covering the total glomerular volume was manually edited by tracing the outlines of EGFP-positive cells for every second section of the z-stack . EGFP-positive cells of the glomerulus were included , while cells of the pronephric ducts were excluded . For quantification of podocyte cell number , the DAPI channel was masked with the EGFP channel using Fiji software ( RRID:SCR_002285 ) ( Schindelin et al . , 2012 ) to include DAPI+/EGFP+ cells only , thus representing nuclei of podocyte cells . Subsequently , a spot segmentation of the DAPI channel was performed . Estimated spot diameter was 4 µm . Spots were filtered for a minimum intensity of the EGFP channel and by using the Imaris quality filter for the occurrence of unspecific spots not matching the EGFP signal . Spots located outside the glomerular surface were manually deleted . Data are presented as mean ±SD for normally distributed data or median ( 25% percentile – 75% percentile , that is interquartile range [IQR] ) for non-normally distributed data with the indicated number of experiments . Normal distribution was determined using the Shapiro-Wilk test . For identification of outliers , Grubbs’ outliers test ( α = 0 . 05 ) was performed . Where appropriate , sample size calculations were performed by the power analyses program G*Power according to Cohen ( Cohen , 1988; Faul et al . , 2009 ) . Differences between experimental rat and zebrafish groups were analyzed using One-way ANOVA with post-hoc Bonferroni’s multiple comparisons test and non-parametric Mann-Whitney-U or Kruskall-Wallis test with Dunn’s multiple comparisons post-hoc test , when appropriate . For the analysis shown in Figure 7L and Figure 7—figure supplement 1I Gaussian distribution was assumed due to the number of embryos categorized . Differences between FSGS patients and controls were analyzed using the Linear-by-Linear association test and the Mann-Whitney U test . Differences in human cultured podocytes were analyzed using two-tailed Student’s t-test . Statistical analysis was performed using SPSS and GraphPad Prism 6 software 6 . 00 ( RRID:SCR_002798 , GraphPad Software , La Jolla , CA ) . p values < 0 . 05 were considered as statistically significant . The genomic and transcriptomic data from this publication have been deposited to the NCBI ( https://www . ncbi . nlm . nih . gov/ ) curated repositories , GEO , and SRA , and assigned the identifier SubmissionID: SUB2950675 and BioProject ID: PRJNA398197 ( DNA-Seq ) and accession GSE102546 ( RNA-Seq ) .	The human kidneys filter the entire volume of the blood about 300 times each day . This ability depends on specialized cells , known as podocytes , which wrap around some of the blood vessels in the kidney . These cells control which molecules leave the blood based on their size . Normally large molecules like proteins are blocked , while smaller molecules including waste products , toxins , excess water and salts pass through into the urine . If this filtration system is damaged , by high blood pressure , for example , it can lead to chronic kidney disease . A hallmark of this disease , often called CKD for short , is high levels of the protein albumin in the urine . Previous studies involving rats with high blood pressure have found several regions of the genome that contribute to high levels of albumin in the urine , including one on chromosome 6 . However , this region contains several genes and it was unclear which genes affected the condition . Schulz et al . set out to narrow down the list and find specific genes that might contribute to elevated albumin in the urine of rats with high blood pressure . This search identified the gene for a protein called TMEM63c as a likely candidate . This protein spans the outer membrane of podocyte cells . Analysis of kidney biopsies showed that patients with chronic kidney disease also had low levels of this protein in their podocytes . Further experiments , this time in zebrafish , showed that reducing the activity of the gene for tmem63c led to damaged podocytes and a leakier filter in the kidneys . The results suggest that this gene plays an important role in the integrity of the kidneys filtration barrier . It is possible that faulty versions of this gene are behind some cases of chronic kidney disease . If this proves to be the case , a better understanding of the role of this gene may lead to new treatments for the condition .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "genetics", "and", "genomics" ]	2019	Analysis of the genomic architecture of a complex trait locus in hypertensive rat models links Tmem63c to kidney damage
Malaria inflicts an enormous burden on global human health . The emergence of parasite resistance to front-line drugs has prompted a renewed focus on the repositioning of clinically approved drugs as potential anti-malarial therapies . Antibiotics that inhibit protein translation are promising candidates for repositioning . We have solved the cryo-EM structure of the cytoplasmic ribosome from the human malaria parasite , Plasmodium falciparum , in complex with emetine at 3 . 2 Å resolution . Emetine is an anti-protozoan drug used in the treatment of ameobiasis that also displays potent anti-malarial activity . Emetine interacts with the E-site of the ribosomal small subunit and shares a similar binding site with the antibiotic pactamycin , thereby delivering its therapeutic effect by blocking mRNA/tRNA translocation . As the first cryo-EM structure that visualizes an antibiotic bound to any ribosome at atomic resolution , this establishes cryo-EM as a powerful tool for screening and guiding the design of drugs that target parasite translation machinery . Malaria is responsible for an estimated 627 , 000 annual deaths worldwide , with the majority of victims being children under 5 years of age ( WHO , 2012 ) . At present there is no licensed malaria vaccine and parasites have developed resistance to all front-line anti-malarial drugs . As such , there is an urgent need for novel therapeutics that can be used as monotherapies or as partner drugs for combinatorial regimes ( Kremsner and Krishna , 2004 ) . An alternative to novel candidates is the repurposing or repositioning of clinically approved drugs that can be used in combination with known anti-malarials , such as chloroquine , antifolates , and artemisinin , to increase their useable lifespan by reducing resistance ( Grimberg and Mehlotra , 2011 ) . The etiological agents for malaria are a family of unicellular protozoan pathogens of the genus Plasmodium . The parasite has a complex two-host lifecycle with a sexual stage occurring in the mosquito vector and an asexual stage in the human host . It is during the asexual blood stage that disease symptoms in humans first appear , including those associated with severe malaria , and it is often at this stage that the need for clinical intervention becomes apparent ( Miller et al . , 2002 ) . Much of malaria pathology is the result of exponential growth of the parasite within erythrocytes , and given the critical role that protein synthesis plays in this , the translational machinery is an attractive drug target . Protein translation in the parasite is focused on three centers ( Jackson et al . , 2011 ) : the cytoplasmic ribosome , responsible for the vast majority of protein synthesis , and organellar ribosomes of the mitochondrion and non-photosynthetic plastid , termed the apicoplast ( McFadden et al . , 1996 ) . In addition , and unusually for a eukaryotic cell , Plasmodium species have two distinct types of cytoplasmic ribosome that differ in their ribosomal RNA ( rRNA ) composition . These are expressed at different stages of the lifecycle , one predominantly in the mosquito vector and the other in the mammalian host , with evidence that both can occur simultaneously for limited periods ( Waters et al . , 1989 ) . Antibiotics known to target the apicoplast ribosome , such as the macrolide azithromycin , demonstrate a delayed-death effect , whereby treated parasites die in the second generation of drug exposure , and therefore have slow clinical onset ( Dahl and Rosenthal , 2007 ; Goodman et al . , 2007 ) . However , because anti-malarial treatment at the blood-stage requires rapid intervention , we focused on the dominant , blood stage-specific cytoplasmic ribosome from the most virulent form of Plasmodium , P . falciparum ( Pf80S ) ( Waters et al . , 1989 ) , as inhibition of cytosolic translation would be expected to be direct and fast-acting . Pf80S is both a candidate for development of novel therapeutics that target specific differences between itself and its counterpart in the human cytosol , and also for repurposing of anti-protozoan inhibitors , such as emetine ( Matthews et al . , 2013 ) . In this present study , we solved the structure of Pf80S–emetine complex at 3 . 2 Å resolution and built a fully-refined all-atom model . This represents , to our knowledge , the first structure of an entire eukaryotic ribosome at atomic resolution solved by electron cryo-microscopy ( cryo-EM ) . Pf80S has a broad distribution of Pf-specific elements across its surface , with particularly long rRNA expansion segments ( ESs ) in the small subunit . The atomic structure of Pf80S in complex with emetine reveals the molecular basis of this clinically relevant anti-protozoan translation inhibitor . In doing so , we establish cryo-EM as a powerful tool for structure-based drug design . Cytoplasmic ribosomes were isolated from the 3D7 strain of P . falciparum parasites maintained in human erythrocytes ( Figure 1A , B ) . Limitations in parasite culture volume , yielding ∼1010 parasitized red blood cells and low yield of sample material ( 1 g of parasites yielded 0 . 35 mg Pf80S ) , precluded an ability to crystallize Pf80S to solve the structure by conventional X-ray crystallography . We therefore exploited recent advances in direct electron detection and statistical image processing ( Bai et al . , 2013; Allegretti et al . , 2014 ) to determine the structure by cryo-EM at an overall resolution of 3 . 2 Å ( Figure 1C–E , Figure 1—figure supplement 1 ) . 10 . 7554/eLife . 03080 . 003Figure 1 . Cryo-EM data and processing . ( A ) Sucrose gradient purification of Pf80S ribosomes . ( B ) Representative electron micrograph showing Pf80S particles . ( C ) Fourier Shell Correlation ( FSC ) curves indicating the overall resolutions of unmasked ( red ) , Pf40S masked ( green ) and Pf60S masked ( blue ) reconstructions of the Pf80S–emetine complex . ( D ) Representative density with built models of a β-strand with well-resolved side chains ( left ) , an RNA segment with separated bases ( middle ) , and a magnesium ion ( green sphere ) that is coordinated by RNA backbone phosphates . ( E ) Density maps colored according to local resolution for the unmasked Pf80S ( left ) and masked Pf40S and Pf60S subunits ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 00310 . 7554/eLife . 03080 . 004Figure 1—figure supplement 1 . FSC curves between the final refined atomic model and the reconstructions from all particles ( black ) ; between the model refined in the reconstruction from only half of the particles and the reconstruction from that same half ( FSCwork , red ) ; and between that same model and the reconstruction from the other half of the particles ( FSCtest , green ) , for Pf40S ( A ) and Pf60S ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 004 Protein side chains and RNA bases were clearly resolved in our maps ( Figure 1D ) . The use of model building and refinement tools that were adapted from X-ray crystallography ( Amunts et al . , 2014 ) led to a near-complete atomic model with excellent geometrical properties ( Figure 2A , B; Table 1 ) . The ribosome model comprises the large ( Pf60S ) and small subunit ( Pf40S ) with a total of 74 proteins ( Tables 2 and 3 ) as well as the 5S , 5 . 8S , 18S , and 28S rRNAs and a tRNA bound at the E-site . The head region of Pf40S has weaker density than the rest of the ribosome due to the inherent flexibility at the neck ( centered around h28 ) . This meant that eS31 , located in the beak of the 40S head ( Rabl et al . , 2011 ) , could not be positioned accurately , and has therefore been omitted from the final model . Using base-pair information extracted directly from the atomic model it was possible to revise secondary structure diagrams for P . falciparum rRNA ( Figure 2—figure supplements 1–3 ) , facilitating comparison with rRNA of other species . 10 . 7554/eLife . 03080 . 005Figure 2 . Structure of the Pf80S ribosome . Overview of Pf80S atomic model showing views facing ( A ) tRNA entry side and ( B ) tRNA exit side . rRNAs are shown in gray , proteins numbered according to Ban et al . ( 2014 ) . ( C and D ) Pf40S and Pf60S subunits are colored in yellow and blue respectively . Flexible regions are shown in red and at a resolution of 6 Å . Pf-specific expansion segments ( ESs ) relative to human ribosomes are labeled . Their numbering is as described for the human cytoplasmic ribosome ( Anger et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 00510 . 7554/eLife . 03080 . 006Figure 2—figure supplement 1 . Secondary structure of Pf18S rRNAs . Pf-specific ESs are highlighted in a labeled red box . Regions not built in the atomic model are colored in blue text . The secondary structure was modified from the CRW site ( Cannone et al . , 2002 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 00610 . 7554/eLife . 03080 . 007Figure 2—figure supplement 2 . Secondary structure of the 5′ half of Pf 28S rRNA . Pf-specific ESs are highlighted in a labeled red box . Regions not built in the atomic model are colored in blue text . The secondary structure was modified from the CRW site ( Cannone et al . , 2002 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 00710 . 7554/eLife . 03080 . 008Figure 2—figure supplement 3 . Secondary structure of the 3′ half of Pf28S rRNA . Pf-specific ESs are highlighted in a labeled red box . Regions not built in the atomic model are colored in blue text . The secondary structure was modified from the CRW site ( Cannone et al . , 2002 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 00810 . 7554/eLife . 03080 . 009Table 1 . Refinement and model statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 009Pf80S–emetineData collection Particles105 , 247 Pixel size ( Å ) 1 . 34 Defocus range ( μm ) 0 . 8–3 . 8 Voltage ( kV ) 300 Electron dose ( e− Å−2 ) 20Pf60SPf40SModel composition Non-hydrogen atoms124 , 50968 , 858 Protein residues6 , 2444 , 106 RNA bases3 , 4601 , 682 Ligands ( Zn2+/Mg2+/emetine ) 5/163/01/67/1Refinement Resolution used for refinement ( Å ) 3 . 13 . 3 Map sharpening B-factor ( Å2 ) −60 . 3−79 . 9 Average B factor ( Å2 ) 113 . 1143 . 2 Rfactor0 . 22940 . 257 Fourier Shell Correlation†0 . 860 . 854Rms deviations Bonds ( Å ) 0 . 0060 . 007 Angles ( ° ) 1 . 201 . 29Validation ( proteins ) Molprobity score2 . 45 ( 96th percentile ) 2 . 73 ( 95th percentile ) Clashscore , all atoms3 . 65 ( 100th percentile ) 4 . 23 ( 100th percentile ) Good rotamers ( % ) 90 . 086 . 0Ramachandran plot Favored ( % ) 90 . 485 . 4 Outliers ( % ) 2 . 44 . 2Validation ( RNA ) Correct sugar puckers ( % ) 97 . 397 . 5 Good backbone conformations ( % ) 71 . 170 . 0Rfactor = Σ\|\|Fobs\| − \|\|Fcalc\|/Σ\|Fobs\| . †FSCoverall = Σ ( Nshell FSCshell ) /Σ ( Nshell ) , where FSCshell is the FSC in a given shell , Nshell is the number of ‘structure factors’ in the shell . FSCshell = Σ ( Fmodel FEM ) / ( √ ( Σ ( \|F\|2model ) ) √ ( Σ ( F2EM ) ) . 10 . 7554/eLife . 03080 . 010Table 2 . Ribosomal proteins of the Pf40S subunitDOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 010Protein namesUniprot IDPlasmoDB IDChain IDBuilt residuesExtensions compared to humanTotal number of residueseS1RS3A_PLAF7PF3D7_0322900B24–233245–262262uS2RSSA_PLAF7PF3D7_1026800C10–204–263uS3Q8IKH8_PLAF7PF3D7_1465900D4–39; 65–78; 97–193; 207–216–221uS4Q8I3R0_PLAF7PF3D7_0520000E2–186–189eS4Q8IIU8_PLAF7PF3D7_1105400F2–258–261uS5Q8IL02_PLAF7PF3D7_1447000G39–262–272eS6Q8IDR9_PLAF7PF3D7_1342000H1–160; 170–213249–306306uS7Q8IBN5_PLAF7PF3D7_0721600I7–118; 128–195–195eS7Q8IET7_PLAF7PF3D7_1302800J3–190–194uS8O77395_PLAF7PF3D7_0316800K2–130–130eS8Q8IM10_PLAF7PF3D7_1408600L5–120; 161–213; 216–218154–163218uS9Q8IAX5_PLAF7PF3D7_0813900M6–143–144uS10Q8IK02_PLAF7PF3D7_1003500N21–118–118eS10Q8IBQ5_PLAF7PF3D7_0719700O11–89–137uS11Q8I3U6_PLAF7PF3D7_0516200P25–151–151uS12O97248_PLAF7PF3D7_0306900Q2–145–145eS12RS12_PLAF7PF3D7_0307100R22–78; 85–100; 111–13510–16141uS13Q8IIA2_PLAF7PF3D7_1126200S12–139–156uS14C0H4K8_PLAF7PF3D7_0705700T7–54–54uS15Q8IDB0_PLAF7PF3D7_1358800U3–151–151uS17O77381_PLAF7PF3D7_0317600V6–25; 36–161–161eS17Q8I502_PLAF7PF3D7_1242700W3–83; 97–110–137uS19C0H5C2_PLAF7PF3D7_1317800X21–95; 103–123–145eS19Q8IFP2_PLAF7PF3D7_0422400Y15–1681–19170eS21Q8IHS5_PLAF7PF3D7_1144000Z11–82–82eS24Q8I3R6_PLAF7PF3D7_051940013–122–133eS25Q8ILN8_PLAF7PF3D7_1421200235–42; 58–84; 97–102–105eS26O96258_PLAF7PF3D7_021780032–96–107eS27Q8IEN2_PLAF7PF3D7_130830047–82–82eS28Q8IKL9_PLAF7PF3D7_146130052–29; 37–66–67eS30RS30_PLAF7PF3D7_021920066–48–58eS31Q8IM64_PLAF7PF3D7_1402500–Not built–14910 . 7554/eLife . 03080 . 011Table 3 . Ribosomal proteins of the Pf60S subunitDOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 011Protein namesUniprot IDPlasmoDB IDChain IDBuilt residuesExtensions compared to humanTotal number of residuesuL2Q8I3T9_PLAF7PF3D7_0516900D2–248–260uL3Q8IJC6_PLAF7PF3D7_1027800E2–381–386uL4Q8I431_PLAF7PF3D7_0507100F6–395373–411411uL5Q8IBQ6_PLAF7PF3D7_0719600G8–51; 64–85; 92–106; 124–166–173uL6Q8IE85_PLAF7PF3D7_1323100H2–186–190eL6Q8IDV1_PLAF7PF3D7_1338200I9–151; 158–221110–118; 139–143; 174–182221eL8Q8ILL2_PLAF7PF3D7_1424400J40–46; 54–131; 147–28311–24;279–283283uL13Q8IJZ7_PLAF7PF3D7_1004000K1–201–202eL13Q8IAX6_PLAF7PF3D7_0814000L2–212134–141; 168–174215uL14Q8IE09_PLAF7PF3D7_1331800M8–139–139eL14Q8ILE8_PLAF7PF3D7_1431700N5–1501–18165uL15C6KT23_PLAF7PF3D7_0618300O2–148–148eL15C0H4A6_PLAF7PF3D7_0415900P2–205–205uL16Q8ILV2_PLAF7PF3D7_1414300Q2–101; 118–206–219uL18Q8ILL3_PLAF7PF3D7_1424100R5–126; 141–185; 189–250; 271–293–294eL18C0H5G3_PLAF7PF3D7_1341200U5–184–184eL19C6KSY6_PLAF7PF3D7_0614500T2–182–182eL20Q8IDS6_PLAF7PF3D7_1341200S2–187–184eL21Q8ILK3_PLAF7PF3D7_1426000V4–158–161uL22Q8IDI5_PLAF7PF3D7_1351400W4–154; 197–215–203eL22Q8IB51_PLAF7PF3D7_0821700X40–1364–18; 34–38139uL23Q8IE82_PLAF7PF3D7_1323400Y88–18813–34; 57–67190uL24O77364_PLAF7PF3D7_0312800Z2–122–126eL24Q8IEM3_PLAF7PF3D7_130910008–69–162eL27Q8IKM5_PLAF7PF3D7_146070012–126;132–146–146eL28Q8IHU0_PLAF7PF3D7_114250022–69; 77–82; 86–98; 103–119–127uL29Q8IIB4_PLAF7PF3D7_112490033–121–124eL29C6S3J6_PLAF7PF3D7_146030042–67–67uL30O97250_PLAF7PF3D7_0307200535–257–257eL30Q8IJK8_PLAF7PF3D7_101940068–105–108eL31Q8I463_PLAF7PF3D7_0503800715–88; 95–116–120eL32Q8I3B0_PLAF7PF3D7_090390082–126–131eL33Q8IHT9_PLAF7PF3D7_1142600935–1371–35140eL34Q8IBY4_PLAF7PF3D7_0710600a2–107–150eL36Q8I713_PLAF7PF3D7_1109900b2–27; 38–1065–10112eL37C0H4L5_PLAF7PF3D7_0706400c2–90–92eL38Q8II62_PLAF7PF3D7_1130100d2–31; 36–77–87eL39C0H4H3_PLAF7PF3D7_0611700e2–30; 38–51–51eL40Q8ID50_PLAF7PF3D7_1365900f1–51–52eL41C6S3G4_PLAF7PF3D7_1144300g3–391–1439eL43RL37A_PLAF7PF3D7_0210100 . 1h2–86–96eL44RL44_PLAF7PF3D7_0304400i2–96–104 Currently , high resolution structures of eukaryotic ribosomes have been solved using X-ray crystallography and are limited to just three structures; the individual subunits from a ciliated protozoan , Tetrahymena thermophila ( Klinge et al . , 2011; Rabl et al . , 2011 ) , and the complete 80S ribosome from the yeast Saccharomyces cerevisiae ( Ben-Shem et al . , 2011 ) . These models have been used to interpret lower resolution structures solved by cryo-EM of other species including the yeast Kluyveromyces lactis ( Fernandez et al . , 2014 ) , Drosophila melanogaster ( Anger et al . , 2013 ) , Trypanosoma brucei ( Hashem et al . , 2013 ) , as well as human ribosomes ( Anger et al . , 2013 ) and provide the basis of the nomenclature used for describing the structures . To examine overall architectural differences , we compared the model of Pf80S to yeast 80S ( Ben-Shem et al . , 2011 ) . Perhaps the largest difference is the absence of RACK1 ( Figure 1A , B ) , which associates with the head of the 40S in the vicinity of the mRNA exit channel ( Sengupta et al . , 2004; Rabl et al . , 2011 ) and has been identified in all eukaryotic ribosome structures solved to-date . RACK1 serves as a signaling scaffold that can recruit other proteins to the ribosome and may link the ribosome with signal transduction pathways , thus allowing translation regulation in response to stimuli . It has also been proposed that RACK1 promotes the docking of ribosomes at sites where local translation is required ( Nilsson et al . , 2004 ) . PfRACK1 is conserved with its human homolog with an identity of 60% . The binding site on the ribosome , which comprises eS17 , uS3 , and 18S rRNA helices h39 and h40 ( Figure 1A , B ) , also appears highly conserved ( Rabl et al . , 2011 ) . However , the C-terminus of uS3 is not resolved in our structure and probably only becomes ordered upon binding RACK1 . The absence of PfRACK1 as an integral member of the small subunit indicates either a different mode of interaction between the ribosome and PfRACK1 in Plasmodium compared to humans , or that under the culturing conditions used PfRACK1 is not expressed , or expressed in a form that does not interact with the ribosome . In yeast , RACK1 has been shown to be present in both a ribosome- and a non-ribosome-bound form dependent on growth conditions ( Baum et al . , 2004 ) . If the interaction between PfRACK1 and the Pf40S is weaker than in other organisms , the possibility that PfRACK1 dissociated during purification and grid preparation cannot be discounted . The yeast 80S structure was also solved in the presence of STM1 , a translation repressor protein , that binds to the head region of the 40S and blocks mRNA entry and binding of tRNA to the A- and P-sites ( Rabl et al . , 2011 ) . The human and D . melanogaster structures also co-purified with an STM1-like protein ( SERBP1 and VIG2 respectively ) ( Anger et al . , 2013 ) . Pf80S is not bound by a suppressor molecule , as also observed for the T . brucei structure ( Hashem et al . , 2013 ) , and hence reflects a ribosome capable of active translation . Pf80S co-purifies with a tRNA bound to the E-site . Although the density is not well resolved , presumably as a result of low and mixed occupancy , it could be interpreted by positioning a model of tRNAMet . The presence of tRNA helps to partially stabilise the L1 stalk near the elbow of the tRNA , however the stalk remains considerably flexible and is averaged out of the high-resolution reconstruction . Perhaps due to the absence of RACK1 and/or STM1 or the presence of an E-site tRNA , the head of Pf40S adopts an orientation with respect to the body that is different to the yeast structure , with uS11 at the beak of the small subunit displaced by more than 10 Å . The root mean square deviation ( RMSD ) of the two small subunits is 2 . 9 Å2 , however if the head and body are superimposed independently this improves to 1 . 0 Å2 and 1 . 5 Å2 respectively . The structure of Pf60S superimposes with the yeast 60S with a RMSD of 1 . 6 Å2 . The largest morphological differences in this subunit result from a cluster of rRNA helices ( ES7AL , ES15L , and ES7CL ) protruding at the solvent side . Given the potential of Pf80S as a drug target , we sought to describe its detailed structure in comparison to its direct counterpart in the human cytoplasm , where a 4 . 8 Å cryo-EM 80S structure represents the highest resolution solved to-date ( Anger et al . , 2013 ) . Therefore , all protein extensions and rRNA expansion segments ( ESs ) are annotated on the basis of comparison with human ribosomes . While the core of the Pf80S and human ribosome are conserved , the periphery of the ribosomes differs extensively in the nature and length of rRNA ESs and protein extensions . The constraints on rRNA expansion appear to be fewer than on protein extension , as rRNA contributes greater to the mass difference between species . Compared to human ribosomes , P . falciparum typically has shorter ESs , some of which are entirely absent in the large subunit ( ES7D-HL , ES9AL , ES10L , ES20L , ES30L ) ( Table 4 ) . The functions , if any , of many of these ESs are not well known . ES7E , which is highly conserved in vertebrates , is implicated in selenoprotein synthesis by binding the SBP2 protein that specifically recruits the selenocysteine-specific tRNA and elongation factor ( Kossinova et al . , 2014 ) . While P . falciparum does utilize selenocysteine , it is incorporated into very few proteins ( Lobanov et al . , 2006 ) and there is no homolog of SBP2 , providing a possible explanation for why ES7E is not present in Plasmodium . 10 . 7554/eLife . 03080 . 012Table 4 . Comparison of ESs in Pf80S and human cytoplasmic ribosomesDOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 012rRNAESHelixComparison between Pf80S and human ribosomes18SES2SShorter loop in Pf80SES3SAConservedBTruncated in Pf80SES13SConservedES6SAExpanded in Pf80SBTruncated in Pf80SCConservedDExpanded in Pf80SEConservedES7SExpanded in Pf80SES14SConservedES9SExpanded in Pf80SES10SExpanded in Pf80SES12SHelix truncated in Pf80S28SES3LConservedES4LConservedES5LConservedES7LATruncated in Pf80SBTruncated . Loop in Pf80S forms a novel interaction with eL14B1Pf-specific ESCPresentD–HAbsent from Pf80SES8LH28Expanded in Pf80SES9LAAbsent in Pf80SH30ConservedH31ConservedES10LAbsent in Pf80SES12LExpanded in Pf80SES15LATruncated in Pf80SES19LTruncated in Pf80SES20LAAbsent in Pf80SBConserved in Pf80SES26LExpanded in Pf80SES27LA–CNot present in Pf80S model , predicted divergence between Pf and human cytoplasmic ribosomesES30LAbsent in Pf80SES31LAConservedBExpanded in Pf80SCConservedES34LPf-specific ESES36LPf-specific ESES39LAConserved; preceding loop in Pf80S forms a short helix ( 3 base pairs ) with the 5′ end of the 5 . 8S rRNABConservedES41LConserved The largest Pf-specific ESs are concentrated in the 18S rRNA , with ES6S and ES9S being particularly extended ( Figure 2C , D; Figure 2—figure supplement 1 ) . These ESs , like those described in both the human ( Anger et al . , 2013 ) and Trypanosoma brucei ( Hashem et al . , 2013 ) ribosome structures , are highly flexible and , in our structure , are only partly visible using a map filtered at 6 Å ( Figure 2C , D ) . We have therefore not included these sections in our atomic model . ES10S is located at the top of the 40S head and has been partially built . P . falciparum ribosomes resemble those of T . brucei in that both have large ES6S and ES7S , although these are slightly larger in T . brucei ( Hashem et al . , 2013 ) . ES6S is in contact with ribosomal components that form part of the mRNA entry and exit sites and was therefore suggested as being involved in translation initiation ( Jenner et al . , 2012 ) . Recently , ES6/7S have been implicated in binding of the conserved translation initiation factor eIF3 based on superposition with a mammalian 43S complex ( Hashem et al . , 2013 ) . Almost 90 nucleotides of ES6AS are averaged out of our high-resolution reconstruction indicating this stalk is highly flexible , perhaps acting in a manner similar to the P stalk ( known as the L7/L12 stalk in prokaryotes ) by recruiting factors necessary for translation ( in this case eIF3 ) . The other large ES of the 18S rRNA , ES9S , is positioned at the head of the 40S . Given both the intrinsic mobility of the head and presumably the ES itself , there is no density for this ∼150 nucleotide Pf-specific element and the role it plays remains unclear . The sites of Pf-specific elements are broadly distributed across the solvent-accessible surface of the ribosome , although the region surrounding the exit tunnel is conserved ( Klinge et al . , 2011 ) and undecorated with ESs and protein extensions ( Figure 2C , D ) . The subunit interface and eukaryotic-specific bridges , which in addition to having structural roles help transmit information to coordinate activity during translation ( Ben-Shem et al . , 2011 ) , are generally highly conserved in Pf80S . There are a couple of examples of stabilizing interactions that are not observed in human ribosomes . Firstly , eL41 , the smallest ribosomal protein , bridges the two subunits ( Ben-Shem et al . , 2011 ) and has a 14-residue Pf-specific N-terminal extension that reaches into a pocket formed by 18S rRNA of the small subunit and tightly anchors the protein ( Figure 3A ) . Secondly , an additional small bridge ( ∼200 Å2 ) is formed between the platform of Pf40S and the region around the L1 stalk by the C-terminal helix extension of eL8 interacting with the C-terminal helix of eS1 ( Figure 3B ) . 10 . 7554/eLife . 03080 . 013Figure 3 . Details of Pf-specific protein extensions and rRNA ESs near the ( A and B ) subunit interface ( C ) P stalk and ( D ) the L1 stalk . Pf-specific elements are shown in red . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 013 Further ordered Pf-specific elements are concentrated near the L1 and P stalks of Pf60S . Directly above the P stalk , the Pf-specific ES7B1L forms a diverted part of ES7CL that is stabilized by several electrostatic interactions with a C-terminal helix extension of uL4 ( Figure 3C ) . Towards the back of the P-stalk , the C-terminal helix extension of eL14 caps the stem loop of ES7BL ( Figure 3C ) . On the opposite side of the ribosome , near the E-site tRNA , the Pf-specific stem loop ES34L is positioned directly above the L1 stalk ( Figure 3D ) . This ES appears to have caused a 60° rotation of the C-terminal helix of eL13 relative to its position in human ribosomes ( Figure 3D ) . The tip of the helix is displaced by ∼28 Å away from the L1 stalk and now stabilizes the interaction between ES34L and the loop of h22 . Since the L1 stalk is required for coordinating the movement of tRNAs and the P stalk is required for coordinating the movement of translation factors during the various steps of protein synthesis ( Gonzalo and Reboud , 2003 ) , the expanded mass around the stalks of Pf80S may have functional implications for translation in P . falciparum . The ability to determine atomic-resolution structures of Pf80S provides a platform for investigating the action of anti-malarial therapeutics that target the ribosome . The clinically used , broad-spectrum eukaryotic translation inhibitor emetine ( Figure 4A ) ( Grollman , 1968 ) , has been reported to act as a translocation inhibitor targeting the ribosome ( Jimenez et al . , 1977; Dinos et al . , 2004 ) , although its precise mode of action is unknown . Emetine is a natural product alkaloid from the plant Carapichea ipecacuanha , and an approved medicine for the treatment of amoebiasis ( Goodwin et al . , 1948 ) . Although its toxicity associated with chronic usage in humans has limited its clinical use against malaria in its current formulation ( Dempsey and Salem , 1966 ) , emetine does demonstrate potent antimalarial activity with a 50% inhibitory concentration ( IC50 ) of 47 nM against the blood stage of multidrug resistant strains of P . falciparum ( Matthews et al . , 2013 ) . Moreover , the immediate therapeutic effect it offers by rapid killing of blood stage parasites may warrant re-consideration of the use of emetine or its derivatives for short periods during acute malaria infection ( James , 1985 ) . 10 . 7554/eLife . 03080 . 014Figure 4 . Emetine binds to the E-site of the Pf40S subunit . ( A ) 2D chemical structure of emetine . ( B ) A 4 . 5 Å filtered difference map ( red density ) at 5 standard deviation overlaid with the Pf80S map filtered at 6 Å ( blue and yellow for Pf60S and Pf40S respectively ) , showing the emetine density at the E-site of the Pf40S . The emetine binding site in ( C ) empty and ( D ) emetine-bound structures , with ( E ) density for emetine alone at 3 . 2 Å . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 014 Incubation of purified Pf80S with a 1 mM emetine solution prior to cryo-EM grid preparation , led to a 3 . 2 Å resolution structure of the complex . Using soft masking , the resolution for the large subunit improved to 3 . 1 Å , with the small subunit at 3 . 3 Å ( Figure 1C ) . A difference map was calculated from the reconstructions with and without emetine and showed a single , continuous feature near the E-site of Pf40S with a shape and size congruent with a single emetine molecule when thresholded at 5 standard deviations , and with a maximum value of 11 standard deviations ( Figure 4B ) . At this position in our map , the density provided sufficient detail to confidently model the emetine molecule ( Figure 4C–E ) . The emetine binding pocket is formed at the interface between 18S rRNA helices 23 , 24 , 45 , and the C-terminus of uS11 ( Figure 5A ) . Comparison with the unliganded map showed that binding of emetine does not induce changes to the pocket ( Figure 4C , D ) . The benzo[a]quinolizine ring of emetine mimics a base-stacking interaction with G973 of h23 and its ethyl group forms a hydrophobic interaction with C1075 and C1076 of h24 , whereas the isoquinoline ring is stacked against the C-terminal Leu151 of uS11 ( Figure 5B , C ) . The interaction is stabilized by a hydrogen bond formed between the NH group of the isoquinoline ring in emetine and an oxygen atom on the backbone of U2061 of h45 ( Figure 5B , C ) . Although there is no high-resolution structure of the human cytoplasmic ribosome , comparison of the emetine binding site in Pf80S with the equivalent region in the 4 . 8 Å human structure ( Anger et al . , 2013 ) revealed that each of the core binding elements are conserved ( Figure 5—figure supplement 1 ) indicating that emetine likely binds to the cytoplasmic host ribosomes in the same way , potentially accounting for the observed cytotoxicity in humans . 10 . 7554/eLife . 03080 . 015Figure 5 . Molecular details of the emetine–ribosome interaction . ( A ) Overview of emetine at the binding interface formed by the three conserved rRNA helices and uS11 . h23 is in green , h24 in cyan , h45 in blue , uS11 in pink , and emetine in yellow . ( B ) 2D representation showing the interaction of emetine with binding residues . Substitution contour represents potential space for chemical modification of emetine . ( C ) Residues in physical contact with emetine . Hydrogen bond is indicated as dashes . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 01510 . 7554/eLife . 03080 . 016Figure 5—figure supplement 1 . Comparison of the emetine binding residues between Pf80S and human ribosomes . Human and Pf-specific elements are colored in yellow and cyan respectively , with Pf numbering . Emetine is in purple . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 016 The identified binding site is consistent with mutations of Arg149 and Arg150 of uS11 in Chinese hamster ovary ( CHO ) cells that have been found to confer resistance to emetine ( Madjar et al . , 1982 ) . At the emetine-binding pocket , h24 is sandwiched between the apexes of h23 and h45 . The C-terminus of uS11 adopts a long coil with seven basic residues ( residues 141–151; RKKSGRRGRRL ) , which form electrostatic interactions with the phosphate backbones of h45 , h23 and h24 , thereby stabilizing the conformation of this coil together with the 18S rRNA ( Figure 5A ) . This would explain the molecular basis for resistance whereby mutations of the C-terminal arginine residues of uS11 destabilize h23 and h45 , disrupting the binding pocket . The mode of binding of emetine resembles the way in which pactamycin , previously thought to be a unique class of antibiotic , binds to the bacterial 30S ( Brodersen et al . , 2000 ) . In both structures the guanine base at the tip of h23 ( G973 in Pf; G693 in bacteria ) forms a stacking interaction with the hydrophobic rings of either compound . Moreover , the two cytosine bases of h24 ( C1075 and 1076 in Pf; C795 and 796 in bacteria ) are each involved in drug binding ( Brodersen et al . , 2000; Figure 6 ) . The hydrogen bond to the backbone of h45 and the hydrophobic interaction with Leu151 of uS11 are specific to the Pf80S–emetine interaction . In the 30S-pactamycin complex , the last base of the E-site codon of the mRNA was displaced 12 . 5 Å compared to the native path of mRNA ( Brodersen et al . , 2000 ) thereby blocking mRNA/tRNA entry into the E-site during the translocation step of protein synthesis ( Dinos et al . , 2004 ) . Based on these structures , emetine appears to elicit its inhibitory effect by the same mechanism as pactamycin . 10 . 7554/eLife . 03080 . 017Figure 6 . Comparison with pactamycin . Superposition of emetine and pactamycin at the Pf40S emetine binding pocket . Emetine and pactamycin are shown in yellow and red respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03080 . 017 The resolution revolution in cryo-EM ( Kühlbrandt , 2014 ) is the product of a new generation of sensors that detect electrons directly ( without first converting to light ) and have improved quantum efficiencies . These cameras are fast enough to follow beam-induced movement of the particles caused by irradiation with electrons . Statistical movie processing can compensate for this movement allowing for structures to be solved at atomic precision . We have harnessed these technological advances to determine the first structure of a ribosome from a parasite at atomic resolution . Previously , structures of eukaryotic cytosolic 80S ribosomes at a similar resolution had only been possible using X-ray crystallography ( Ben-Shem et al . , 2011 ) . From the reconstruction of Pf80S–emetine complex at 3 . 2 Å , we determined a stereochemically accurate all-atom model using recent developments in model building , refinement , and validation ( Amunts et al . , 2014 ) . The structure of Pf80S further demonstrates the diversity of ribosome structures among eukaryotes , especially in terms of the location and nature of ESs at the periphery , while maintaining a conserved core . The observation of Pf-specific features could serve as the basis for exploring their functional relevance as an essential , first step towards finding efficacious and clinically safe anti-malarial drugs . An alternative to drug development against Pf-specific ribosomal elements is the repurposing of existing antibiotics as anti-malarials . By determining the structure of Pf80S in both a liganded and unliganded state , we were able to locate the binding site of the anti-protozoan inhibitor , emetine , using an unbiased difference map . That emetine and pactamycin share a binding pocket in eukaryotic ribosomes could not be predicted based on the chemical structures of the drug molecules only . Pactamycin itself has been shown to have potent antiprotozoal activity against both drug-susceptible and drug-resistant strains of P . falciparum ( Otoguro et al . , 2010 ) . Chemical modifications to pactamycin have yielded analogs that maintain antimalarial activity but with reduced cytotoxicity against mammalian cells ( Lu et al . , 2011 ) . Similarly , an emetine derivative , dehydroemetine , which differs by the presence of a double bond next to the ethyl group of benzo[a]quinolizine ring , exhibits less toxic effects than the parental compound while maintaining anti-parasitic properties ( Dempsey and Salem , 1966; Chintana et al . , 1986 ) . This suggests that compounds targeting the emetine/pactamycin binding site are amenable to optimization , potentially leading to drugs more suited to clinical use . The Pf80S–emetine structure reveals an edge centered on the ethyl group of the molecule that could be subjected to modification to increase the affinity of emetine for the binding pocket ( Figure 5B , labelled as the ‘contour edge’ ) . Although based on the similarity with the binding site in humans it is unlikely that emetine can be structurally modified to not bind the mammalian system , as demonstrated in the case of dehydroemetine modifications can reduce its cytotoxicity . Although the mechanism for such reduced cytotoxicity mediated by pactamycin and emetine analogs is not known , it may be possible that these derived compounds selectively target tumor/parasite cells that are rapidly dividing , whereby protein synthesis is more sensitive to drug action in these cells . As in the case of antibiotics repurposed as antitumor agents , there is a clinical role for eukaryotic antibiotics that target systems with differential rates of translation provided usage is carefully directed . In malaria , eukaryotic antibiotics , such as emetine , could be used in combination with the slow-acting , but more specific apicoplast-targeting antibiotics ( Dahl and Rosenthal , 2007 ) . This work demonstrates the power of contemporary cryo-EM for drug discovery . A drug , with a previously unknown binding site , can be visualized inside a macromolecular complex that is almost 10 , 000 times larger in molecular weight and at a level of detail comparable to that obtained by X-ray crystallography . By avoiding the need for crystallization one of the bottlenecks of solving a structure is bypassed . It allows structures to be solved from very small sample quantities , with sample heterogeneity improved through image processing . As such , cryo-EM is of particular use for solving the structures of macromolecules in their native state , isolated from pathogenic organisms where culturing large quantities is not possible . In summary , our cryo-EM analyses reveal the first structure of a ribosome from a parasite at atomic resolution , along with detailed insights into the molecular basis of a known anti-protozoan translation inhibitor . Finally , it demonstrates that cryo-EM offers an attractive route towards the development of new compounds that target macromolecules by facilitating structure–activity relationships in otherwise intractable biological systems . Wild-type 3D7 strain of P . falciparum parasites were maintained in human erythrocytes ( blood group O ) at a hematocrit of 4% with 10% Albumax . Saponin lysed parasite pellets were incubated with lysis buffer ( 20 mM Hepes , pH 7 . 4 , 250 mM KCl , 25 mM Mg ( CH3COO ) 2 , 0 . 15% Triton , 5 mM 2-mecaptoethanol ) at 4°C for 1 hr . Ribosomes were purified by ultracentrifugation initially with a sucrose cushion ( 20 mM Hepes pH 7 . 4 , 1 . 1 M sucrose , 40 mM KCH3COO , 10 mM NH4CH3COO , 10 mM Mg ( CH3COO ) 2 , and 5 mM 2-mecaptoethanol ) followed by a 10–40% sucrose gradient separation step using the same buffer . Aliquots of 3 μl of purified Pf80S at a concentration of ∼160 nM ( ∼0 . 5 mg/ml ) were incubated for 30 s on glow-discharged holey carbon grids ( Quantifoil R1 . 2/1 . 3 ) , on which a home-made continuous carbon film ( estimated to be ∼30 Å thick ) had previously been deposited . Grids were blotted for 2 . 5 s and flash frozen in liquid ethane using an FEI Vitrobot . For the empty Pf80S sample , grids were transferred to an FEI Titan Krios electron microscope that was operated at 300 kV . Images were recorded manually during two non-consecutive days on a back-thinned FEI Falcon II detector at a calibrated magnification of 135 , 922 ( yielding a pixel size of 1 . 03 Å ) . Defocus values in the final data set ranged from 0 . 7 to 3 . 9 µm . To prepare the Pf80S–emetine sample , purified Pf80S at 160 nM was incubated with a 1 mM solution of emetine in 20 mM Hepes pH7 . 4 , 40 mM KCH3COO , 10 mM NH4CH3COO , 10 mM Mg ( CH3COO ) 2 , and 5 mM 2-mecaptoethanol for 15 min at 25°C prior to blotting and freezing as described above . Pf80S–emetine grids were transferred to an FEI Tecnai Polara electron microscope that was operated at 300 kV . Images were recorded manually during two non-consecutive days on a back-thinned FEI Falcon II detector at a calibrated magnification of 104 , 478 ( yielding a pixel size of 1 . 34 Å ) . Defocus values in the final data set ranged from 0 . 8 to 3 . 8 µm . During the data collection sessions of both samples , all images that showed signs of significant astigmatism or drift were discarded . An in-house built system was used to intercept the videos frames from the detector at a rate of 17 s−1 for the Krios and 16 s−1 for the Polara microscope . We used RELION ( version 1 . 3-beta ) for automated selection of 126 , 727 particles from 1310 micrographs for the empty Pf80S sample; and 158 , 212 particles from 1081 micrographs for the Pf80S–emetine sample . Contrast transfer function parameters were estimated using CTFFIND3 ( Mindell and Grigorieff , 2003 ) . All 2D and 3D classifications and refinements were performed using RELION ( Scheres , 2012 ) . To discard bad particles , we used a single round of reference-free 2D class averaging with 100 classes for both data sets , and a single round of 3D classification with four classes for the Pf80S–emetine data set . The final refinement for the empty Pf80S and Pf80S–emetine sample contained 72 , 293 and 105 , 247 particles , respectively . A 60 Å low-pass filtered cryo-EM reconstruction of the yeast cytoplasmic 80S ribosome ( EMDB-2275 [Ben-Shem et al . , 2010] ) was used as an initial model for the 3D refinement . For the correction of beam-induced movements , we used statistical movie processing as described previously ( Bai et al . , 2013 ) , with running averages of five movie frames , and a standard deviation of 1 pixel for the translational alignment . To further increase the accuracy of the movement correction , we used the beta version of RELION-1 . 3 to fit linear tracks through the optimal translations for all running averages , and included neighboring particles on the micrograph in these fits . In addition , we employed a resolution and dose-dependent model for the radiation damage , where each frame is weighted with a different B-factor as was estimated from single-frame reconstructions . These procedures yielded maps with an overall resolution of 3 . 4 Å for the empty Pf80S and 3 . 2 Å for Pf80S–emetine . Reported resolutions are based on the gold-standard FSC = 0 . 143 criterion ( Chen et al . , 2013 ) and were corrected for the effects of a soft mask on the FSC curve using high-resolution noise substitution ( Chen et al . , 2013 ) . Soft masks were made by converting atomic models into density maps , binarizing those , and adding cosine-shaped edges . Prior to visualization , all density maps were corrected for the modulation transfer function ( MTF ) of the detector , and then sharpened by applying a negative B-factor ( Table 1 ) that was estimated using automated procedures ( Rosenthal and Henderson , 2003 ) . In order to locate emetine in the Pf80S–emetine reconstruction , we calculated a difference map between the reconstructions of empty Pf80S and Pf80S–emetine . To this purpose , the two MTF-corrected and B-factor sharpened maps were aligned with respect to each other using the ‘Fit in Map’ functionality in UCSF Chimera ( Pettersen et al . , 2004 ) , and the empty Pf80S map was re-interpolated on the Cartesian grid of the Pf80S–emetine map prior to subtraction of the maps in RELION . For visualization purposes , the resulting difference map was low-pass filtered at 4 . 5 Å and the threshold was set at 5 standard deviations as calculated within the area of the Pf80S ribosome ( Figure 4B ) . At this threshold , only one continuous U-shaped feature was visible . The highest difference density inside this feature extended to 11 standard deviations in the difference map . Local resolution variations in all reconstructions were estimated using ResMap ( Kucukelbir et al . , 2014 ) . Presumably due to unresolved structural heterogeneity the local resolution in the small ribosomal subunit was typically worse than in the large ribosomal subunit . Therefore , for the Pf80S–emetine structure , we performed two additional ‘focussed’ refinements , where we masked out the large or the small subunit at every iteration . This gave rise to two maps ( Figure 1E ) with improved density for either the small subunit ( at an overall resolution of 3 . 3 Å ) or the large ribosomal subunit ( at an overall resolution of 3 . 1 Å ) , and these maps were used for the refinement of the atomic model as described below . Ribosomal protein sequences from the 3D7 strain of P . falciparum were taken from PlasmoDB ( The Plasmodium Genome Database Collaborative , 2001 ) and used as template sequences to obtain homology models generated from I-TASSER ( Roy et al . , 2010 ) . Homology models were fitted into the reconstructed map of Pf80S using Chimera ( Pettersen et al . , 2004 ) . Each protein was then subjected to a jiggle-fit and extensively rebuilt with sidechains placed into the map density using Coot v . 0 . 8 ( Emsley et al . , 2010 ) . The sequences of the Pf80S rRNAs were obtained from PlasmoDB ( The Plasmodium Genome Database Collaborative , 2001 ) and aligned using Clustal Omega ( Sievers et al . , 2011 ) with the rRNA sequences extracted from the Saccharomyces cerevisae ( Sc ) 80S structure ( PDB ID: 3U5B and 3U5D ) ( Ben-Shem et al . , 2011 ) . Conserved regions without insertions or deletions were extracted from the yeast structure , mutated and renumbered . These conserved sections were then connected by de novo building of RNA . The complete rRNA was then manually rebuilt in Coot to optimize the fit to density . Building was aided by secondary structure predictions downloaded from the Comparative RNA Website ( Cannone et al . , 2002 ) . The model was refined using REFMAC v . 5 . 8 , which was modified for structures determined by cryo-EM ( Murshudov et al . , 2011; Amunts et al . , 2014 ) . The Pf80S atomic model was refined as separate 60S and 40S subunits in the two maps that were obtained for either subunit in the focused refinements of the cryo-EM reconstructions . Structure factors for the ( Fourier-space ) refinement in REFMAC were obtained by cutting out sections of the corresponding maps with a 3 Å radius from the center of each atom in the model , and structure factor phases were not altered during refinement . Throughout refinement , reference and secondary structure restraints were applied to the ribosomal proteins using the Sc80S structure as a reference model ( Nicholls et al . , 2012 ) . Base pair and parallelization restraints obtained using LIBG were also applied throughout refinement ( Amunts et al . , 2014 ) . The stereochemistry of the rRNA model was further improved using the ERRASER-PHENIX pipeline ( Chou et al . , 2013 ) . Ramachandran restraints were not applied during refinement to preserve backbone dihedral angles for validation . The R-factor and average overall Fourier shell correlation were monitored during refinement ( Table 1 ) and the final model was validated using MolProbity ( Chen et al . , 2010 ) . For cross-validation against over-fitting , we randomly displaced the atoms of our final model ( with an RMSD of 0 . 5 Å ) and performed a fully restrained refinement against a map that was reconstructed from only one of the two independent halves of the data that were used in our gold-standard FSC procedure . We then calculated FSC curves between the resulting model and the half-map against which it had been refined ( FSCwork ) , as well as the FSC curve between that model and the other half-map ( FSCtest ) . The observation that the FSCwork and FSCtest curves nearly overlap demonstrates the absence of overfitting of the model ( Figure 1—figure supplement 1 ) .	Each year , malaria kills more than 600 , 000 people , mostly children younger than 5 years old . Humans who have been bitten by mosquitoes infected with malaria-causing parasites become ill as the parasites rapidly multiply in blood cells . Although there are several drugs that are currently used to treat malaria , the parasites are rapidly developing resistance to them , setting off an urgent hunt for new malaria drugs . Developing new antimalarial medications from scratch is likely to take decades—too long to combat the current public health threat posed by emerging strains of drug-resistant parasites . To speed up the process , scientists are investigating whether drugs developed for other illnesses may also act as therapies for malaria , either when used alone or in combination with existing malaria drugs . Certain antibiotics—including one called emetine—have already shown promise as antimalarial drugs . These antibiotics prevent the parasites from multiplying by interfering with the ribosome—the part of a cell that builds new proteins . However , humans become ill after taking emetine for long periods because it also blocks the production of human proteins . Tweaking emetine so that it acts only against the production of parasite proteins would make it a safer malaria treatment . To do this , scientists must first map the precise interactions between the drug and the ribosomes in parasites . Wong et al . have now used a technique called cryo-electron microscopy to examine the ribosome of the most virulent form of malaria parasite . This technique uses very cold temperatures to rapidly freeze molecules , allowing scientists to look at molecular-level details without distorting the structure of the molecule—a problem sometimes encountered in other techniques . The images of the parasitic ribosome taken by Wong , Bai , Brown et al . show that emetine binds to the end of the ribosome where the instructions for how to assemble amino acids into a protein are copied from strands of RNA . In addition , the images revealed features of the parasitic ribosome that are not found in the human form . Drug makers could exploit these features to improve emetine so that it more specifically targets the production of proteins by the parasite and is less toxic to humans .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics" ]	2014	Cryo-EM structure of the Plasmodium falciparum 80S ribosome bound to the anti-protozoan drug emetine
Adult-born granule cells ( ABGCs ) are involved in certain forms of hippocampus-dependent learning and memory . It has been proposed that young but functionally integrated ABGCs ( 4-weeks-old ) specifically contribute to pattern separation functions of the dentate gyrus due to their heightened excitability , whereas old ABGCs ( >8 weeks old ) lose these capabilities . Measuring multiple cellular and integrative characteristics of 3- 10-week-old individual ABGCs , we show that ABGCs consist of two functionally distinguishable populations showing highly distinct input integration properties ( one group being highly sensitive to narrow input intensity ranges while the other group linearly reports input strength ) that are largely independent of the cellular age and maturation stage , suggesting that ‘classmate’ cells ( born during the same period ) can contribute to the network with fundamentally different functions . Thus , ABGCs provide two temporally overlapping but functionally distinct neuronal cell populations , adding a novel level of complexity to our understanding of how life-long neurogenesis contributes to adult brain function . Adult neurogenesis contributes to certain forms of hippocampus-dependent behavior and is associated with a number of neuro-psychiatric diseases ( Parent and Murphy , 2008; Deng et al . , 2010; Kheirbek et al . , 2012; Spalding et al . , 2013 ) . Recent data suggested that young ABGCs ( around 4 weeks old ) contribute to a discrete pattern separation function , whereas older cells ( 8 weeks or older ) are not necessary for this dentate gyrus-dependent function , therefore functionally different pools of granule cells provide unique plasticity to the hippocampal circuits ( Clelland et al . , 2009; Aimone et al . , 2010; Alme et al . , 2010; Sahay et al . , 2011a; Nakashiba et al . , 2012; Neunuebel and Knierim , 2012 ) . Current theories on adult neurogenesis are based on the provisional correlations between the two distinct physiological functions and age-dependent maturation of cellular ( including synaptic , biophysical and molecular ) properties ( Aimone et al . , 2006 , 2010; Sahay et al . , 2011b ) . This is supported by numerous observations showing that after their birth , ABGCs undergo a continuous maturation process , lasting for 8–10 weeks . ABGCs acquire neuronal properties including synaptic inputs and outputs , and capability of firing action potentials 3 to 4 weeks after their birth . Notably , ABGCs at this cellular age are highly excitable , show enhanced synaptic plasticity and are differently modulated by inhibition compared to ABGCs at the end of the maturation period ( Wang et al . , 2000; Schmidt-Hieber et al . , 2004; Laplagne et al . , 2006; Toni et al . , 2008; Mongiat et al . , 2009; Gu et al . , 2012; Marín-Burgin et al . , 2012; Vivar et al . , 2012; Dieni et al . , 2013 ) . However , it remains unknown how two populations emerge by a continuous maturation of the underlying cellular properties of ABGCs . How do individual ABGCs transform from ‘young’ to ‘old’ properties ? There are three testable possibilities . First , their functionally important properties may develop continuously ( Figure 1A ) . However , if this is the case , it may contradict the general notion that two distinct ABGC populations exist , and ABGCs would provide a functional continuum . The second possibility is that , as during their early maturation when becoming functionally integrated ( <4 weeks ) , ABGCs switch function according to a predetermined program ( Figure 1B ) . In this situation there are two clearly distinct populations and they would switch within a short temporal window at a predefined stage of their postmitotic life . Third , ABGCs may be susceptible to extrinsic cues allowing for a functional switch for an extended period ( Figure 1C ) . We tested these hypotheses by resolving the integrative properties of individual ABGCs because data-pooling could mask the differences between above hypotheses . Thus , we here analyzed if ( i ) all cellular properties develop concomitantly , ( ii ) if there are biophysical properties that allow the emergence of only two populations between 3 and 10 weeks after their birth , and ( iii ) how and when ABGCs switch function . Using this approach , we show that ABGCs consist of two functionally distinct populations during an extended period , between 3 and 9 weeks of age in rats , by being sensitive to distinct aspects of their inputs . 10 . 7554/eLife . 03104 . 003Figure 1 . Potential theoretical modes of postmitotic maturation of functional properties . Each colored line represents the age-dependent change of a theoretical parameter from individual ABGCs . ( A ) Gradual maturation of the properties results in widely distributed functional continuum . ( B ) Temporally predefined functional switch . ( C ) The functional switch occurs in an extended temporal window . DOI: http://dx . doi . org/10 . 7554/eLife . 03104 . 003 To analyze the cellular maturation of ABGCs , we compared a variety of intrinsic biophysical and input–output transformation properties at seven different age-groups ( 3–10 weeks after cells are born ) of individual birth-dated ABGCs in young adult rats using retroviral labeling ( Figure 2 , Zhao et al . , 2006 ) . The majority of the tested parameters ( including input resistance , membrane time constant , whole-cell capacitance , resting membrane potential , action potential threshold , peak dV/dt of the spikes , and maximal firing rate ) of individual ABGCs changed continuously with age and , consequently , the distribution of the data points from individual cells was wide , without the emergence of distinguishable populations ( Figure 3A–B , statistical values are indicated in the figures—see also Supplementary file 1 ) , reflecting the continuous maturation of these properties in accordance with previous observations ( Mongiat et al . , 2009; Marín-Burgin et al . , 2012 ) . 10 . 7554/eLife . 03104 . 004Figure 2 . Maturation of the biophysical and integrative properties of ABGCs . ( A ) The RFP and biocytin-labeled cells in the dentate gyrus ( left panels , d . p . i . : day after virus injection ) , spiny dendrites ( middle panels ) , and typical mossy fiber terminals in the stratum lucidum of the CA3 region ( right ) confirm granule cell identity . ( B ) Four representative RFP-expressing granule cells 34 , 47 , and 63 days after CAG-RFP virus labeling . The 63-day-old AGBCs were recorded from the same slice . ( C ) Average subthreshold voltage responses of the example cells to small ( −10 pA ) current steps . Input resistance ( Rin ) , membrane time constant ( τM ) , and resting membrane potential ( RMP ) of the cells are indicated . ( D ) Spike parameters of the example cells at lower current intensities ( dV/dt: maximal rate of rise , thr: action potential threshold ) . ( E ) Maximal firing rate of the four cells in response to square pulse current injection . ( F ) Responses of the cells to sinusoidal current injections with increasing amplitude ( Δ50 pA ) at 10 and 80 Hz . The traces are shown until the firing reached saturation . ( G ) Number of spikes generated in the example cells as a function of the peak amplitude of the injected sinusoid currents at the all tested frequencies . Gray symbols indicate values that were omitted from the analysis due to lack or saturation of spiking . Offset values describe the minimum input intensities to reach 50% spiking output . ( H ) Increments of the firing ( i . e . , the first derivative of the curves in panel F ) of the cells . These values were used for the calculation of the average slope ( as mean , ASL ) and the variance of firing ( as variance , VAR ) . Note that cells 1 and 3 have exceptionally large values at certain input intensity ranges indicating that these cells were more sensitive to certain input intensities . This characteristic is quantified by the large VAR value . DOI: http://dx . doi . org/10 . 7554/eLife . 03104 . 00410 . 7554/eLife . 03104 . 005Figure 3 . Adult born granule cells ( 3–10 week old ) can be divided into two distinct populations based on cell-to-cell differences in input–output transformation . ( A ) Left , input resistance of individual ABGCs with various ages ( gray crosses ) . Red and blue circles ( S- and L-group members , respectively ) highlight the values for the example cells shown in Figure 1 . N . L . : not labeled control cells . Right , probability distribution of the data set shows single peak ( single Gaussian fit: F = 0 . 0001 ) . ( B ) Monotonous probability distribution of membrane time constant , whole cell capacitance , resting membrane potential , maximal rate of rise of spikes , relative offset of the input–output curves , and action potential threshold data from the same set of cells as above . ( C ) Left , average slope ( top ) and variance of the slope ( bottom ) of the same individual ABGCs as above with various ages ( gray crosses ) . Blue lines indicate the lack of correlation between the gain of the input–output functions and the age of individual cells within the S-group ( linear fit , ASL: R2 = −0 . 029 , p = 0 . 89 , VAR: R2 = 0 . 01 , p = 0 . 257 ) . Right , two population emerges from the distribution of the average slope values of individual cells ( two peaks Gaussian , ASL: F = 0 . 0014 , VAR: F = 0 . 0001 ) . The centers of the two clusters and average distance values from the centers ( error bars ) are shown on the right ( K-means analysis , F < 10−9 , horizontal gray lines on the left panels indicate the separation by the K-means analysis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03104 . 005 In addition to these basic biophysical parameters , we measured the suprathreshold input–output functions in response to sinusoidal current injections to mimic temporally organized input patterns in physiologically relevant frequency ranges ( Figure 2F–H , Pernia-Andrade and Jonas , 2014 ) . During these protocols , due to the fluctuating membrane potential , the contribution of voltage-gated ionic channels to the firing is more relevant than in the case of square pulse injection . In contrast to basic membrane properties , the analysis of the gain of the input–output functions of the same cells revealed two significantly distinct populations using Gaussian fits and K-means analysis ( Figure 3C ) . The input–output function of the first group was characterized by a steep average slope ( ASL ) and highly variable ( measured as the variance of the slope , VAR; Figure 2F–H ) spike responses , suggesting that this cell population is exceptionally sensitive to certain narrow input strengths , and thus highly suited for disambiguating input–output functions at single cell level ( notice the out-of-average values on Figure 2H for the first and third cells , hereafter referred to as S-group , standing for Sensitive ) . Within the S-group the integrative parameters were independent of the actual age of the individual cells ( linear fits on Figure 3C ) demonstrating that similar cellular functionality is maintained throughout an extended period ( between 3–9 weeks ) unless the individual cell switched to the second integration mode . This second group of ABGCs responded with constantly and incrementally increasing , less sensitive spike output ( L-group; referring to Linear ) enabling them to linearly report various input strengths . A different parameter of the input–output functions , the offset did not divide the same ABGC samples into two populations , indicating that the functional separation of ABGCs is restricted to specific cellular properties ( Figure 3B ) . Altogether , the above results show that ABGCs form two functionally distinct populations during a long period of their maturation based on their different sensitivity to temporally organized inputs . The above analysis suggested that age alone does not directly determine the functionality of individual ABGCs . However , the probability of whether an individual cell behaved as a member of S- or L-groups shows age-dependence ( Figure 4A ) . Within all analyzed cells , the youngest ( 3 weeks old ) and the oldest ( 10 weeks old ) belonged to the S- or L-group , respectively; however , both functional groups were present with changing probabilities during the intermediate ages between 5 and 9 weeks of age . This observation was further supported by the parabolic distributions of the mean-variance plots of ASL and VAR ( Figure 4B , Supplementary file 1 ) . Thus , the continuously increasing probability of L-groups could mask the two functionally distinct groups of ABGCs when global population properties were analyzed as averages ( Mongiat et al . , 2009 ) . 10 . 7554/eLife . 03104 . 006Figure 4 . Independence of the output properties of individual ABGCs from age and input resistance . ( A ) Probability of the members of the clusters defined by the K-means analysis continuously shifts from S-group ( blue ) toward L-group ( red ) during maturation indicating the higher prevalence of ABGCs with shallow and invariable input–output function . ( B ) The population level functional switch is also suggested by the higher variance of the input–output parameters during the transition age period and low variance in the youngest and most matured populations ( parabolic fit , ASL: R2 = 0 . 611 , F ( ANOVA ) = 0 . 0035; VAR: R2 = 0 . 853 , F = 0 . 00017 ) . Numbers indicate the age of the data sets . ( C ) Correlation of the integrative parameters to the input resistance within the two functionally different groups ( red and blue symbols ) defined by K-means cluster analysis ( see Figure 3 ) . Gray circles indicate the four example cells from Figure 2 ( linear fits , ASL: R2 = 0 . 087 , p = 0 . 045 for S-group , R2 = 0 . 576 , p = 2 . 9 × 10−7 for L-group; VAR: R2 = 0 . 02 , p = 0 . 19 for S-group R2 = 0 . 467 , p = 2 . 6 × 10−6 for L-group ) . ( D ) Correlation between the normalized current needed to reach output that is half of the input frequency and input resistance of individual ABGCs belonging to the two functionally different groups ( R2 = 0 . 39 , p = 0 . 0001 for S-group; R2 = 0 . 607 , p = 2 × 10−8 for L-group; R2 = 0 . 682 , p < 10−8 for both groups ) . ( E ) Considering multiple membrane parameters of the individual cells ( resting membrane potential , membrane time constant , whole cell capacitance , input resistance , threshold , peak dV/dt , maximal firing rate ) for hierarchical cluster analysis ( Ward method with normalized values ) were not sufficient to predict the functional identity of the cells . Data from individual cells are aligned vertically ( lines in the upper panel and symbols in the middle and bottom panels ) . Thus , the order of the points along the X-axes is determined by the results of the cluster analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 03104 . 00610 . 7554/eLife . 03104 . 007Figure 4—figure supplement 1 . Coexistence of S- and L-functionalities among granule cells from non-labeled animals . ( A ) Non-labeled granule cells ( thus unknown cellular age ) were randomly recorded ( albeit in the lower half of cell layer of adult dentate ( P68–101 , not subject of virus injections ) ) . The peaks of the double Gaussian distributions in the birth-dated sample ( Figure 3C ) and in this new non-labeled sample were very similar ( ASL: 1 . 522 ± 0 . 056 and 2 . 389 ± 0 . 028 vs 1 . 611 ± 0 . 008 and 2 . 377 ± 0 . 013; VAR: 0 . 638 ± 0 . 046 and 1 . 621 ± 0 . 052 vs 0 . 616 ± 0 . 02 and 1 . 477 ± 0 . 09 ) and the K-means analyses also predicted similar centers and average distance from centers ( ASL: 1 . 461 ± 0 . 197 and 2 . 37 ± 0 . 219 vs 1 . 627 ± 0 . 182 and 2 . 409 ± 0 . 168; VAR: 0 . 637 ± 0 . 217 and 1 . 643 ± 0 . 223 vs 0 . 659 ± 0 . 163 and 1 . 536 ± 0 . 251 ) . ( B ) The observed unusual and group-specific dependence ( or lack of thereof ) of the slope on the input resistance was similar in birth-dated ABGCs ( see Figure 4C ) and in the non-representative sample ( unknown age ) from non-labeled cells . DOI: http://dx . doi . org/10 . 7554/eLife . 03104 . 00710 . 7554/eLife . 03104 . 008Figure 4—figure supplement 2 . The input–output transformation of individual cells is maintained in two stable states by complex mechanisms . ( A ) Lowering the recording temperature resulted in the elimination of the separation between individual granule cells ( non-labeled adult animals , see 'Materials and methods' and Figure Supplement 1 ) based on the ASL and VAR when the same granule cells were analyzed at 28–29°C after establishing its parameters in physiological temperature . With this intervention , practically the same mechanisms ( such as voltage-gated channels ) were available during the two recording conditions , but the cellular excitability is altered in a complex manner . For example , changes in the classical excitability parameters ( e . g . , Rin , threshold ) suggested increased excitability; whereas the capability to elicit high frequency firing is decreased . Left column graphs show the relative changes and absolute voltage shifts in the biophysical properties of the recorded cell . The right graph shows the change in the ASL during lower temperature plotted against the initial ASL value of individual cells . ( B ) When the excitability of the cells was challenged by proportionally enhancing calcium-dependent mechanisms by 4 mM extracellular calcium ( relative to the 2 mM control condition ) , the effect on the ASL did not depend on the initial values ( i . e . , there was no group specific effect ) ; even though the calcium elevation has similar effects on the general excitability parameters of the cells as the lowered temperature . ( C ) GIRK channel activation by ML297 ( 2 . 5 µM ) increased the ASL only in a subset of the recorded cells , which had low ASL during control conditions ( like the L-group cells ) . In spite of the similar group specific effects on ASL , GIRK activation , and lower temperature had largely opposite effects on the biophysical properties . ( D ) Application of GABAA and AMPA/kainate receptor antagonists , gabazine ( 5 µM ) and CNQX ( 10 µM ) slightly increased the ASL value . However , the effect was not depended on the initial state of the tested cells indicating that S- and L-group properties are established independent of the spontaneous synaptic activity . DOI: http://dx . doi . org/10 . 7554/eLife . 03104 . 008 Our data indicate that in 3–9 weeks old ABGCs the gain of the input–output properties is not exclusively determined by the input resistance . ABGCs within the S-group achieved similar input–output computations in spite of largely different input resistances . Strikingly , this becomes clear by the lack of correlations of the gain of individual S-group cells to their input resistance ( Figure 4C ) . However , the input–output function of ABGCs from the L-group showed the expected dependence on the input resistance of individual cells . In contrast to the gain , the offset of the input–output function of individual ABGCs showed a clear dependence on the input resistance across both cell populations ( Figure 4D ) . Thus , the independence of the gain of the input–output transformation from the continuously developing biophysical parameters at the level of individual cells allows for the emergence of only two temporally overlapping and functionally distinct populations within 5- to 9-weeks-old ABGCs . Importantly , independent experiments in which granule cells were recorded in non-labeled adult animals ( 'Materials and methods' ) , confirmed the coexistence of S- and L-functionalities among granule cells based on their ASL and VAR , and these individual cells showed similar correlations ( or lack thereof ) between their integrative functions and input resistance ( Figure 4—figure supplement 1 ) as in the case of the above birth-dated data set ( Figure 4C ) . Next , we tested whether all measured biophysical parameters of individual birth-dated ABGCs can cooperatively predict the functional separation at the level of the gain of their input–output functions . We performed cluster analysis of the recorded cells based on seven intrinsic parameters to define two groups ( Figure 4E ) . The two intrinsic parameter groups determined by cluster analysis did not match with the S- and L-group identity of the same individual cells . This latter result shows that consideration of multiple parameters by their arithmetical values is not sufficient to predict the functional separation of S- and L-groups . Thus , a complex and balanced interaction is probably behind the formation of the two states , as also suggested by additional experiments , in which the cellular excitability was altered by decreasing the temperature , adding extracellular calcium or activating background conductances ( Figure 4—figure supplement 2A–C ) . We also tested whether the two distinct input–output functions are due to their distinct synaptic drives ( Dieni et al . , 2013 ) by blocking GABAA and AMPA/kainate receptors . This intervention slightly increased the ASL value ( Figure 4—figure supplement 2D ) ; however , the effect was not dependent on the initial state of the tested cells indicating that S- and L-group properties are established largely independent of spontaneous synaptic activity . Here , we show that the gain of the input–output transformation of 3- to 10-weeks-old ABGCs exists at two functionally distinct states , allowing for the translation of similar excitatory drives into highly distinct action potential outputs in a manner that is not directly predicted by the cellular age alone ( refer to the alternative maturation hypotheses depicted in Figure 1A–C ) . This finding is in contrast to the continuous maturation hypothesis ( Figure 1A ) . Around the third postmitotic week , ABGCs represent a functionally homogeneous population ( S-group ) characterized by highly variable and sensitive output , which potentially underlies the effective disambiguation of input patterns because the output of S-group cells represent certain input ranges with exceptional efficacy . Importantly , this functional parameter is similar for an extended time period at the level of individual ABGCs , despite the continuous maturation of other biophysical parameters suggesting precise homeostatic tuning and complex interactions of the biophysical properties ( Marder and Goaillard , 2006 ) . However , between the fifth and ninth week ( under our experimental conditions ) , ABGCs switch function by losing their sensitivity to a particular input strength as their output incrementally reports a wide input range ( L-group ) . Thus , in the theoretical case of identical input strengths and patterns to S- and L-cells ( which allowed us to investigate the integrative properties of individual ABGCs in isolation ) , sensitivity of ABGCs in the S-group to certain input ranges , in opposed to linear reporting in cells of the L-group , allows a certain level of input disambiguation at the level of single granule cells . Furthermore , previously described distinct input rules ( Dieni et al . , 2013 ) may synergistically promote two distinct functions within the ABGC populations , if these rules are specifically associated with S- or L-group properties . Strikingly , our data indicate that ‘classmate’ cells ( born during the same period ) can contribute to the network with fundamentally different functions during an extended period after cells are born ( 5–9 weeks ) . Conversely , these data suggest that similar cellular functions can be served by ABGCs that were born at different periods during the animal's life . Therefore , the properties of individual cell rather than the cells' age determine how certain input strengths are computed; either by , disambiguating certain input combinations ( S-functionality ) or by representing all of its inputs by similar rate changes toward the downstream networks ( L-functionality ) . This observation challenges previous hypotheses on the plasticity provided by adult neurogenesis from being predominantly determined by the postmitotic cellular age and predicting similar functions of ABGCs born at the same time ( Aimone et al . , 2006 , 2010; Sahay et al . , 2011b ) . Moreover , environmental conditions that increase or reduce hippocampal neurogenesis may affect the relative contribution of newly generated granule cells to the S- or L-groups that may explain distinct behavioral consequences of altered neurogenesis ( Kempermann et al . , 1997; Gould and Tanapat , 1999; van Praag et al . , 1999 ) . 31- to 33-day-old male Wistar rats ( 95–135 g body weight ) were injected with a CAG-GFP or CAG-RFP Moloney murine leukemia virus vector ( Zhao et al . , 2006; Jessberger et al . , 2007 ) ( 0 . 8–1 µl ) using stereotaxically targeted ( 5 . 7–5 . 8 mm posterior , ± 4 . 4–4 . 5 mm lateral , and 5 . 6–6 mm ventral from bregma ) , conventional Hamilton syringe under ketamine/xylazine/pipolphen anesthesia ( 83/17/7 mg/body kg ) . Adult born granule cells were labeled along a broad longitudinal range ( 2–3 mm ) of the hippocampi . Note that we did not find cells in animals 3–10 weeks after virus injection , which had <3-week-old properties indicating the reliability and precision of the birth-dating method . Animals of this age were used because relatively large number of labeled ABGCs can be analyzed at a time of recording when the network properties of the dentate gyrus circuitry can be considered adult ( Laplagne et al . , 2006 ) . After the surgical procedure two or three siblings were housed together in large cages ( 75 cm × 35 cm ) equipped with a running wheel , toys and shelters until the electrophysiological experiments because running is known to increase the number of surviving adult-born neurons ( Kempermann et al . , 1997; Tashiro et al . , 2007; Dranovsky et al . , 2011 ) . For acute slice preparations , rats were deeply anaesthetized with isoflurane and slices ( 300–350 µm ) were cut in ice-cold artificial cerebrospinal fluid ( consisting of 85 mM NaCl , 75 mM sucrose , 2 . 5 mM KCl , 25 mM glucose , 1 . 25 mM NaH2PO4 , 4 mM MgCl2 , 0 . 5 mM CaCl2 , and 24 mM NaHCO3 ) . The orientation of cutting was perpendicular to the axis of the hippocampus at the level of virus injection . After the cutting , slices were kept at 32°C for 30 min and then at room temperature . During the recordings , slices were perfused with a solution containing 126 mM NaCl , 2 . 5 mM KCl , 26 mM NaHCO3 , 2 mM CaCl2 , 2 mM MgCl2 , 1 . 25 mM NaH2PO4 , and 10 mM glucose , 35–36°C . To reduce the instrumental capacitance ( including pipette capacitance ) , recording pipettes were pulled from thick glass ( i . d . 0 . 87; o . d . 1 . 5 mm ) . Pipette resistance was in the range of 5 . 5–9 . 5 MΩ , and the usual total instrumental capacitance was 9 . 5–11 pF , which was neutralized to the maximum obtainable level ( <2 pF remaining capacitance ) under current clamp conditions ( digitized at 50 kHz and low-pass filtered at 20 kHz ) . The intracellular solution contained 90 mM K-gluconate , 43 . 5 mM KCl , 1 . 8 mM NaCl , 1 . 7 mM MgCl2 , 50 µM EGTA , 10 mM HEPES , 2 mM Mg-ATP , 0 . 4 mM Na2-GTP , 10 mM phosphocreatine-disodium , and 8 mM biocytin ( pH 7 . 25 ) . Note that because of the relatively high chloride concentration in the intracellular solutions , differences in the cellular properties are unlikely due to age- , stage- , maturation-level-type specific chloride homeostasis ( Overstreet-Wadiche and Westbrook , 2006; Markwardt et al . , 2009 , 2011 ) . Note that the properties were tested while the spontaneous synaptic activity was left intact ( i . e . , no blockers were included during control conditions ) . The expression of GFP or RFP was verified usually by multiple criteria: match between the epifluorescence ( excitation at 490–510/540–580 nm , detection at 520LP/593–667 nm for GFP/RFP ) and Nomarski ( 900 nm ) differential interference contrast images ( Eclipse FN-1; Nikon , Japan ) , appearance of fluorescent signal in the recording pipette and washout of the intracellular labeling during the recording , and post hoc colocalization of fluorescent intracellular biocytin–and RFP/GFP signals . The epifluorescent illumination of slices was reduced as much as possible before and during the recordings in order to avoid any photo-damage of the labeled cells and usually did not last longer than few seconds above the cells . In majority of slices , in addition to birth-dated ABGCs , we also recorded non-labeled control cells , which located on the border of strata granulosum and moleculare , in order to provide controls for similar recording conditions across animals . Note that because virus injection took place always in P31–33 animals , their age at the time of recordings varied between P51 and P105 ( corresponding to recording of 20–72 days old ABGCs ) . Notably , no correlations were found between the age of the animals and cellular properties of non-labeled control cells . Semilunar granule cells , characterized by extremely low ( <90 MΩ ) input resistance and broad dendritic arbor , were excluded from the analysis . For post hoc anatomical processing , slices were fixed for a day in 0 . 1 M phosphate buffer containing 2% paraformaldehyde and 0 . 1% picric acid at 4°C . For visualizing the biocytin signal , sections were incubated overnight with Alexa Fluor 350-conjugated streptavidin ( 1:500; Invitrogen , Carlsbad , CA ) in 0 . 5% Triton X-100 and 2% NGS containing TBS buffer at 4°C . After washing and mounting in Vectashield ( Vector Laboratories , Burlingame , CA ) , the endogenous signal of the fluorescent protein was compared with the biocytin staining by using epifluorescent illumination ( DM2500; Leica , Germany ) . To reliably determine the potential correlations between the different intrinsic parameters , we collected data for each tested parameters from each analyzed cells ( 3–10 weeks old ) . ABGCs in the early phase of their maturation ( younger than 3 weeks ) were not analyzed because they are not yet fully integrated into the hippocampal network due to the lack of reliable high frequency spiking , which is the consequence of lower sodium channel densities . In order to measure the input–output characteristics of ABGCs during different stages of their maturation , we used sinusoidal current injection from theta to high gamma frequency bands ( 5 , 10 , 20 , 40 , 60 , 80 Hz , 50 pA increment from the holding level of 0 pA , tested in random order ) for 1 s and analyzed the number of the elicited action potentials . Application of sinusoidal current injections from resting membrane potential mimicked temporally correlated excitatory drives in functionally relevant frequency ranges , in opposed to square pulses , which would strongly recruit non-physiological mechanisms such as inactivation of voltage activated conductances . This type of mimicking of the excitatory drive to GCs is justified by the single supralinear integration zone of GCs ( i . e . , spike initiation in the axon initial segment [Krueppel et al . , 2011] ) . Furthermore , it has been reported that the amplitude of miniature excitatory events recorded at the somata does not increase further after the 3–4 weeks of ABGCs ( Mongiat et al . , 2009 ) . Using somatic current injection , thus , avoids the potential confounds introduced by short-term plasticity upon repetitive stimulation of the same fibers . We characterized the integrative properties of individual ABGCs by two reliable parameters , which measure the gain of the input–output functions: the average slope ( ASL ) and the variance ( VAR ) of the slope of input–output curves . The calculated parameters ( average slope , variance , and offset ) were weighted by the different frequencies using empirically determined correlations to obtain a pooled , frequency-independent data point from each recorded cells . Average slope ( ASL ) was calculated as the arithmetic mean of the first derivative of the input–output function and weighted by the square root of the frequency . Frequency-weighted variance of the gain of the firing ( VAR ) was calculated as the variance of the first derivative of the input–output function divided by the frequency . Thus , these two measures are sensitive to different aspects of the input–output function of a given cell and characterize individual cells with a single and reliable value . High ASL value suggests that the given cell is capable of large output changes in response to unitary input changes , whereas the large VAR highlights that the cell is more sensitive to a particular input intensity range . Importantly , the ASL and VAR values remained stable for individual cells provided stable membrane potential , input resistance , and capacitances values and well-compensated bridge balance . These parameters were strictly monitored in every recorded trace using a 50 ms long −50 pA step and manually corrected if necessary and recordings were excluded if the resting membrane potential changed more than 4 mV compared to the initially measured values . Input resistance was measured as the average steady-state voltage response to −10 pA current steps ( 30–100 traces excluding traces with large spontaneous events ) . Membrane time constant was fitted with single exponential on these traces between 2–100 ms both at the onset and the end of the current step . The maximum rate of rise ( peak dV/dt ) was measured on the first spike that was elicited using square pulse currents without post hoc filtering . Action potentials were defined as larger deflection in the first derivative of the recorded voltage trace than 20 mV/ms following post hoc low-pass filtering at 4 kHz . The maximum firing capability of the cells was challenged by 1 s long square current injections with increasing amplitude ( Δ20 pA ) until depolarization block was reached . Action potential threshold was measured as the voltage at 20 mV/ms of the dV/dt . The whole cell capacitance was measured in voltage clamp recordings using a −5 mV voltage step at −70 mV holding by measuring the integral area of the current response ( measured from the steady-state current level ) and divided by the voltage step amplitude . The offset of the input–output function was defined as the peak amplitude of the current waveform necessary to reach larger firing frequency than the half of the input frequency . For normalization , we weighted the values with the fourth root of the input frequencies . Correlations are characterized with adjusted R-square ( R2 ) . In an independent subset of experiments to provide evidence for the existence of S- and L-functionalities to test the potential underlying mechanisms , the properties of granule cells were tested in animals , which were not subject to virus injection . These were adult rats ( P68–101 ) kept with running wheels ( Figure 4—figure supplements 1–2 ) , and the granule cells were recorded mostly in the lower half of the cell layer . The same criteria were applied for these cells as in the case of the birth-dated ABGCs .	Remembering what happened on different occasions involves a process in the brain called pattern separation , which allows us to separate and distinguish our memories . One part of the brain where pattern separation occurs is called the dentate gyrus , which sits in the hippocampus—the brain region that is in charge of certain forms of learning and memory . Neurons called granule cells are thought to play a central role in hippocampal pattern separation . These cells , unlike the majority of nerve cells , can form at any time , and those that form in the mature brain are called adult born granule cells ( ABGCs ) . Although it usually takes 10 weeks for these cells to fully mature , they are capable of communicating with each other about 3–4 weeks after being generated . Previously , it had been reported that while young , 4-week-old ABGCs are required for pattern separation , slightly older ( 8 week old ) ABGCs are not . What intrinsic properties make ABGCs capable of contributing to pattern separation ? Is this property defined by the fate ( i . e . a predetermined program ) of the cell , or by the cell's experiences and activities ? To investigate these questions , Brunner et al . labeled ABGCs with a fluorescent tag when these neurons were born in adult male rats . Then , when the tagged cells were aged between 3 and 10 weeks old , the electrical properties of the labeled cells were measured from thin brain slices . Brunner et al . found that ABGCs respond to input signals with two different levels of sensitivity . The youngest cells ( 3–5 weeks old ) are exceptionally sensitive to a narrow range of input signal strengths , which is useful for pattern separation . The oldest investigated cells ( 10 weeks old ) , on the other hand , respond incrementally to a wide range of different input signal strengths . Under these experimental conditions , the cells changed how they respond to input signals some time between 5 and 9 weeks after being born . However , they either behaved like the youngest or like the oldest cells: no intermediate behavior was seen . Unexpectedly , the switch is not directly related to the age of the cells: cells born at the same time don't necessarily change behavior at the same time , and cells born at different times may behave similarly . Thus , Brunner et al . suggest that it is the experience of the cells , and not their fate , that determines how they help the dentate gyrus function during the investigated period .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "short", "report", "neuroscience" ]	2014	Adult-born granule cells mature through two functionally distinct states
Integrin activation is crucial for the regulation of leukocyte rolling , adhesion and trans-vessel migration during inflammation and occurs by engagement of myeloid cells through factors presented by inflamed vessels . However , endothelial-dependent mechanisms of myeloid cell recruitment are not fully understood . Here we show using an autoperfused flow chamber assay of whole blood neutrophils and intravital microscopy of the inflamed cremaster muscle that CD95 mediates leukocyte slow rolling , adhesion and transmigration upon binding of CD95-ligand ( CD95L ) that is presented by endothelial cells . In myeloid cells , CD95 triggers activation of Syk-Btk/PLCγ2/Rap1 signaling that ultimately leads to integrin activation . Excitingly , CD95-deficient myeloid cells exhibit impaired bacterial clearance in an animal model of sepsis induced by cecal ligation and puncture ( CLP ) . Our data identify the cellular and molecular mechanisms underlying the chemoattractant effect of endothelial cell-derived CD95L in induction of neutrophil recruitment and support the use of therapeutic inhibition of CD95’s activity in inflammatory diseases . Leukocyte recruitment comprises of a cascade with four major steps: slow rolling , firm adhesion , intraluminal crawling and trans-vessel migration ( Ley et al . , 2007 ) . Slow rolling and firm adhesion are mediated by selectin- and chemokine-induced integrin signaling . Selectin is expressed and presented to the vessel lumen by inflamed endothelial cells ( Zarbock et al . , 2011 ) . E-selectin engagement with PSGL-1 and CD44 ligands induces activation of the Src family kinases ( SFKs ) Hck , Fgr and Lyn ( Yago et al . , 2010 ) which then phosphorylate and activate immunoreceptor tyrosine-based activation motif ( ITAM ) -bearing adaptor protein Fc receptor common γ signaling chain ( FcRγ ) and DNAX activation protein of 12 kDa ( DAP12 ) ( Zarbock et al . , 2008 ) . These activated adaptor proteins recruit and phosphorylate spleen tyrosine kinase ( Syk ) , which in turn activates Bruton's tyrosine kinase ( Btk ) ( Mueller et al . , 2010; Yago et al . , 2010 ) . Btk further activates the phosphoinositide 3-kinase ( PI3K ) , phospholipase C γ2 ( PLCγ2 ) and p38 mitogen-activated protein kinase ( p38 MAPK ) pathways that mediate the integrin signaling to induce slow rolling . E-selectin- and integrin-mediated rolling crucially depend on Syk activation via binding to phosphorylated ITAM-domains ( Mueller et al . , 2010; Yago et al . , 2010; Zarbock et al . , 2007 ) . Interestingly , an ITAM-like motif was identified as a docking site for Src homology domain 2 ( SH2 ) -containing proteins in CD95 of neutrophils ( Daigle et al . , 2002 ) . In CD95L-stimulated myeloid cells , we identified Lyn as the major SFK that phosphorylates CD95’s tyrosine , thereby allowing recruitment of Syk , which via the PI3K/MMP9 pathway results in myeloid cell migration to the inflammatory site ( Letellier et al . , 2010 ) . However , the roles of CD95 in the initial cellular processes of myeloid cell recruitment , such as rolling and adhesion , remain unknown . CD95 ( Fas/Apo-1 ) was initially described as a death receptor mediating apoptosis via formation of the death-inducing signal complex ( DISC ) which further leads to activation of downstream caspases and apoptosis ( Peter et al . , 2007 ) . Interestingly , in T cells the apoptotic cascade is prevented via formation of CD44-ezrin-actin-CD95 signaling complexes ( Mielgo et al . , 2006 , 2007 ) . Likewise , in B cells the CD95-FADD interaction is prevented by binding of Btk to CD95 via its kinase and Pleckstrin homology ( PH ) domain ( Uckun , 1998; Vassilev et al . , 1999 ) . Taken together , we reasoned that upon stimulation with CD95L , CD95 might assemble a signaling complex to induce integrin activation for myeloid cell rolling and adhesion . Here , we report that CD95 mediates slow rolling and adhesion of myeloid cells via activation of integrin through stimulation of Syk-Btk-PLCγ2 or Btk-PLCγ2 signaling pathways . CD95 in myeloid cells or CD95L in endothelial cells is required for myeloid cell recruitment in in vivo animal models of inflammation . Further , deletion of CD95 in myeloid cells impairs bacterial clearance in systemic inflammation . Collectively , our data demonstrate that endothelial cell-derived CD95L serves as a chemokine in induction of neutrophil slow rolling and adhesion via integrin activation during inflammation . In order to study the role of CD95 in leukocyte slow rolling we used a mouse autoperfused flow chamber assay ( Chesnutt et al . , 2006 ) . This assay has the advantage of allowing examination of rolling and adhesion of neutrophils from whole blood , thereby preventing isolation-induced activation ( Forsyth and Levinsky , 1990; Glasser and Fiederlein , 1990 ) . In addition , using the Lyz2<CreGFP> reporter mice , 89 ± 2% of the rolling cells in the flow chamber have been identified as neutrophils ( Chesnutt et al . , 2006 ) . Consistent with previous reports , the rolling velocity of neutrophils is significantly reduced on E-selectin+ICAM1-coated chamber as compared to the E-selectin-coated chamber ( Figure 1A , Figure 1—figure supplement 1A ) ( Jung and Ley , 1999; Chesnutt et al . , 2006; Zarbock et al . , 2007 ) . Additional intravenous tail ( i . v . ) injection of CD95L one hour prior to flow chamber assay significantly reduced the rolling velocity as compared to control counterparts ( Figure 1B , C; 1 . 55 ± 0 . 07 µm/s vs . 1 . 16 ± 0 . 03 µm/s ) . In order to exclude the possibility that CD95 mediates neutrophil slow rolling via the CD95-induced chemokine production , which has been reported in various cell types ( Park et al . , 2003; Guo et al . , 2005; Altemeier et al . , 2007; Miwa et al . , 1998 ) , mouse blood was perfused through the flow chambers coated with E-selectin , ICAM1 and CD95L . The rolling velocity of neutrophils in 50 μg/ml CD95L-coated chamber was significantly lower than in the control group ( Figure 1B , C; 1 . 55 ± 0 . 07 µm/s vs . 0 . 84 ± 0 . 09 µm/s ) . Flow chamber coated with 50 μg/ml CD95L showed the strongest effect on slow rolling as compared to other coating concentrations ( Figure 1—figure supplement 1B ) . Rolling cells in CD95L alone-coated or CD95L/ICAM1-coated chamber were not detectable which indicated that CD95L-induced slow rolling was E-selectin-dependent ( data not shown ) . In addition , CD95L i . v . injection or CD95L-coating increased the number of rolling cells in the flow chamber as compared to the control group ( Figure 1D ) . To further confirm that the CD95L-induced neutrophil slow rolling was specific to CD95 , we specifically deleted CD95 in myeloid cells ( Fas<f/f>::Lyz2<Cre> ) . Although CD95-deficient neutrophils rolled at a similar velocity as Fas<f/f>neutrophils ( Figure 1—figure supplement 1A ) , CD95L i . v . -injection or CD95L coating failed to reduce neutrophils’ rolling velocity in Fas<f/f>::Lyz2<Cre> mice ( Figure 1B ) . Interestingly , Fas<f/f>::Lyz2<Cre> mice showed significantly less rolling cells in CD95L-coated flow chamber or upon CD95L injection as compared to the WT mice under the same condition ( Figure 1D ) . Control experiments demonstrated that Fas<f/f>::Lyz2<Cre> mice exhibited less rolling cells in a flow chamber coated with E-selectin and ICAM1 than Fas<f/f>mice , however this was not significant ( Figure 1—figure supplement 1C ) . These results imply that CD95 might be important for the arrest of neutrophils . 10 . 7554/eLife . 18542 . 003Figure 1 . CD95 signaling in myeloid cells is involved in mediating slow rolling , adhesion and transmigration . ( A ) Representative time lapse pictures of neutrophil slow rolling in flow chamber . Arrows indicate the rolling cells . Scale bar: 50 μm . ( B ) Rolling velocity of WT or Fas<f/f>::Lyz2<Cre> neutrophils in flow chambers upon the stimulation of immobilized CD95L or soluble CD95L . Data are presented as mean ± SEM , n=3–4 . ( C ) Cumulative histogram shows the velocity of rolling neutrophils in flow chambers coated with E-selectin/ICAM1 , E-selectin/ICAM1/CD95L or E-selectin/ICAM1+soluble CD95L stimulation . ( D ) Number of WT or Fas<f/f>::Lyz2<Cre> rolling cells in flow chambers upon the stimulation of immobilized CD95L or soluble CD95L . Data are presented as mean ± SEM , n=3–4 . ( E ) Rolling velocity of neutrophils in flow chambers coated with E-selectin/ICAM1 in the presence of immobilized CD95L or anti-CD11a antibody . Data are presented as mean ± SEM , n=3 . ( F ) Representative reflected light oblique transillumination pictures of postcapillary venules of Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice 2 hr after TNF-α application . Demarcations on each side of the venule determine the areas in which extravasated leukocytes were counted . ( G–I ) Rolling velocity of leukocytes ( G ) and numbers of adherent leukocytes ( H ) in the inflamed cremaster muscle venules and numbers of transmigrated leukocytes ( I ) in inflamed cremaster muscle of Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice . Data are presented as mean ± SEM , n=6 . Statistical significance was evaluated by one-way ANOVA followed by Bonferroni multiple comparison post hoc test in ( B , C , D , E ) ( F=13 . 44 , p<0 . 0001 in B , F=37 . 37 , p<0 . 0001 in C , F=10 . 21 , p<0 . 0001 in D , F=4 . 40 , p=0 . 0135 in E ) and two-tailed unpaired Student's t test in ( G–I ) , p<0 . 05 , p<0 . 01 , *p<0 . 001 , n . s not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 00310 . 7554/eLife . 18542 . 004Figure 1—figure supplement 1 . Rolling velocity of WT or Fas<f/f>::Lyz2<Cre> neutrophils in different conditions . ( A ) Rolling velocity of neutrophils from WT , Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice in flow chambers coated with E-selectin or E-selectin /ICAM1 . n=3 . ( B ) Rolling velocity of neutrophils in flow chambers coated with E-selectin/ICAM1 and different concentration of CD95L . n=3 . ( C ) Number of Fas<f/f>::Lyz2<Cre±> rolling cells in flow chambers coated with E-selectin and ICAM1 . , n=3 . ( D ) Rolling velocity of neutrophils in flow chambers coated with E-selectin/ICAM1 in the presence of immobilized CD95L or anti-CD11b antibody . n=3–4 . ( E , F ) Rolling velocity of neutrophils in flow chambers coated with L-selectin/ICAM1/CD95L ( E ) or P-selectin/ICAM1/CD95L ( F ) . n=3–4 . Data are presented as mean ± SEM . One-way ANOVA followed by Bonferroni multiple comparison post hoc test in ( A , B ) ( F=3 . 462 , p=0 . 0161 in A , F=16 . 23 , p<0 . 0001 in B ) and two-tailed unpaired Student's t test in ( C ) , p<0 . 05 , p<0 . 01 , p<0 . 001 , n . s not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 00410 . 7554/eLife . 18542 . 005Figure 1—figure supplement 2 . TNFRs surface expression level of neutrophils from Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice in homeostasis and inflamed conditions . ( A–B ) TNFR1 and TNFR2 surface expression level of neutrophils from Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice in homeostasis . n=6 . ( C–D ) TNFR1 and TNFR2 surface expression level of neutrophils from Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice at 6 hr post CLP . n=6 . Data are presented as mean ± SEM , Two-tailed unpaired Student's t test in , p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 00510 . 7554/eLife . 18542 . 006Figure 1—figure supplement 3 . CD95L i . v . injection or deletion of CD95 in myeloid cells doesn’t influence the integrin level in neutrophils . ( A ) Flow cytometry plot of blood neutrophils . ( B–D ) Mice were i . v . injected with saline or CD95L ( 10 μg ) . One hour later , blood samples were stained with antibodies of neutrophil markers and integrin subunits and analyzed by flow cytometry . Neutrophils expression levels of integrin αL ( B ) , integrin αM ( C ) and integrin β2 ( D ) are presented as mean ± SEM , n=3 . ( E ) Scheme of CD95 deletion in myeloid cells of Fas<f/f>::Lyz2<Cre> mouse line . ( F ) Blood samples of Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice were stained with antibodies of neutrophil markers and CD95 levels in neutrophils were analyzed by flow cytometry . n=3 . ( G–J ) Blood samples of Fas<f/f> and Fas<f/f>::Lyz2<Cre>mice were stained with antibodies of neutrophil markers and integrin subunits and analyzed by flow cytometry . Neutrophils expression levels of integrin αL ( G ) , integrin αM ( H ) , integrin β2 ( I ) and CD44 ( J ) are presented as mean ± SEM , n . s . , not significant , n=3 . ( K–L ) Ratio of neutrophils ( CD11b+Ly6G+ ) , monocytes ( CD11b+CD115+ ) , T cells ( CD3 ) and B cells ( CD19 ) among CD45+ cells in blood of Fas<f/f> and Fas<f/f>::Lyz2<Cre>mice . n=3 . ( M ) Absolute number of neutrophils in blood of Fas<f/f> and Fas<f/f>::Lyz2<Cre>mice . n=6 . Data are presented as mean ± SEM , Two-tailed unpaired Student's t test in ( C , F , H , I , K , M ) , p<0 . 05 , *p<0 . 001 , n . s not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 006 More importantly , the effect of coated CD95L on neutrophil slow rolling was blocked by an integrin αL neutralizing antibody , anti-CD11a , indicating that CD95L-induced slow rolling was integrin αL-dependent ( Figure 1E ) . However , integrin αM neutralizing antibody , anti-CD11b , did not block CD95L-induced slow rolling ( Figure 1—figure supplement 1D ) . In order to examine whether CD95 is also involved in L- and P-selectin-mediated rolling , we performed the autoperfused flow chamber assay with chambers coated with L/P-selectin , ICAM1 and CD95L respectively . CD95L stimulation did not significantly impact the rolling velocity in L-selectin or P-selectin coated chambers ( Figure 1—figure supplement 1E , F ) . To further evaluate the effect of CD95-induced rolling and adhesion in vivo , we conducted intravital microscopy of the inflamed cremaster muscle from Fas<f/f> or Fas<f/f>::Lyz2<Cre> mice 2 hr after administration of tumor necrosis factor α ( TNF-α; 500 ng/mice intrascrotally , Figure 1F ) . It has been reported that >95% of all adherent and rolling leukocytes are neutrophils in this model ( Jung et al . , 1998 ) . Interestingly , the rolling velocity of leukocyte in Fas<f/f>::Lyz2<Cre> mice was not reduced ( Figure 1G ) , indicating a redundant function of TNF-α and CD95 in modulation of rolling velocity , similar to the redundancy previously reported in a model of traumatic brain injury in mice ( Bermpohl et al . , 2007 ) . Importantly , two studies showed that TNF was involved in neutrophil and T-cell adhesion via TNF-induced inside-out signaling ( Lauterbach et al . , 2008; Li et al . , 2016 ) . In order to clarify the redundant effect of TNF on CD95-deficiency , we stained neutrophils from blood of Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice for TNF receptors ( TNFR ) and observed that naïve Fas<f/f>::Lyz2<Cre> mice expressed higher levels of TNFR2 but similar levels of TNFR1 . However , at 6 hr after cecal ligation and puncture ( CLP ) , neutrophils from Fas<f/f>::Lyz2<Cre> mice had higher expression of TNFR1 but similar expression of TNFR2 ( Figure 1—figure supplement 2 ) . Thus , increased TNFR1 expression upon inflammation might compensate for the lack of CD95 . Consistent with our previous report that CD95 triggers transmigration of myeloid cells to the inflammatory site ( Letellier et al . , 2010 ) , the numbers of adherent cells and transmigrated cells in Fas<f/f>::Lyz2<Cre> mice were reduced as compared to Fas<f/f> mice ( Figure 1H , and I ) , Neutrophil slow rolling is mainly mediated by activation of lymphocyte function-associated antigen 1 ( LFA-1 , Integrin αLβ2 ) ( Chesnutt et al . , 2006; Zarbock et al . , 2007 ) . Therefore , integrin αL , αM and β2 surface expression levels on neutrophils were assessed by flow cytometry in whole blood of CD95L-injected and control mice . No significant difference between these groups could be detected ( Figure 1—figure supplement 3A–D ) . Similarly , in Fas<f/f>::Lyz2<Cre> mice ( Figure 1—figure supplement 3E ) , there was no difference in integrin αL and β2 expression between CD95-deficient neutrophils and Fas<f/f> neutrophils , but only an increase of integrin αM levels in CD95-deficient neutrophils ( Figure 1—figure supplement 3F–I ) . CD95-deficient neutrophils also expressed the same level of CD44 , the ligand of E-seletin and P-selectin , as the Fas<f/f> neutrophils ( Figure 1—figure supplement 3J ) . In addition , the ratio of neutrophils and monocytes increased in the blood of Fas<f/f>::Lyz2<Cre> mice , but T cells and B cells were not changed as compared to the control mice , and the absolute number of neutrophils in the blood of Fas<f/f>::Lyz2<Cre> mice was similar to the control counterparts ( Figure 1—figure supplement 3K , L , M ) . These results show that CD95-induced slow rolling is not related to the up-regulation of cell surface expression level of integrins . In inflamed tissue , inflammatory cytokines activate the expression of adhesion molecules , such as selectin and ICAM , and the synthesis of chemokines and lipid chemoattractants on the luminal surface of endothelial cells to facilitate the recruitment of leukocytes ( Ley et al . , 2007 ) . In this study , we show that in the autoperfused flow chamber assay immobilized CD95L promotes the slow rolling of neutrophils in an integrin signaling-dependent pathway . Hence , we hypothesized that in vivo activated endothelial cells present CD95L to facilitate neutrophil recruitment . In order to address this hypothesis , we crossed Cdh5<CreERT2> mice with Fasl<f/f> mice to enable inducible deletion of CD95L in endothelial cells by tamoxifen treatment ( Fasl<f/f>::Cdh5<CreERT2> ) and performed the intravital microscopy experiments as we did with Fas<f/f>::Lyz2<Cre> mice ( Figure 2A , B , Figure 2—figure supplement 1A , B ) . Interestingly , the leukocyte rolling velocity was significantly increased in mice with CD95L deficiency in endothelial cells as compared to the Fasl<f/f> mice ( Figure 2C ) . Moreover , the rolling flux fraction which shows the percentage of rolling cells was reduced in Fasl<f/f>::Cdh5<CreERT2> mice ( Figure 2D ) . These observations indicate that endothelial cell-derived CD95L is essential for leukocyte slow rolling during inflammation . In line with the results from Fas<f/f>::Lyz2<Cre> mice , the numbers of adherent cells in the inflamed venules and transmigrated cells in the cremaster muscle were also reduced in Fasl<f/f>::Cdh5<CreERT2> mice as compared to the control litter mates ( Figure 2E , F ) . 10 . 7554/eLife . 18542 . 007Figure 2 . Endothelial cells-derived CD95L is necessary for neutrophil recruitment during inflammation . ( A ) Scheme of inducible CD95L deletion in endothelial cells of Fasl<f/f>::Cdh5<CreERT2> mouse line . ( B ) Representative reflected light oblique transillumination pictures of postcapillary venules of Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice 2 hr after TNF-α application . Demarcations on each side of the venule determine the areas in which extravasated leukocytes were counted . ( C–F ) Rolling velocity of leukocytes ( C ) , rolling flux fraction ( D ) and numbers of adherent leukocytes ( E ) in inflamed cremaster muscle venules and numbers of transmigrated leukocytes in inflamed cremaster muscle ( F ) of Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice . Data are presented as mean ± SEM , n=6 . ( G ) Injection schedule of tamoxifen and thioglycollate is depicted . ( H ) Flow cytometry plot of peritoneal neutrophils at 6 hr after thioglycollate injection . ( I ) Influx of peritoneal neutrophils 6 hr after thioglycollate injection in Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice . n=11–14 . ( J ) Influx of peritoneal neutrophils 6 hr after thioglycollate injection in Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice . Data in I–J are presented as mean ± SEM and were pooled from two independent experiments , n=16–17 . Statistical significance was evaluated by two-tailed unpaired Student's t test in ( C–F , I , J ) , p<0 . 05 , p<0 . 01 , p<0 . 001 , n . s not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 00710 . 7554/eLife . 18542 . 008Figure 2—figure supplement 1 . Induced-deletion of CD95L or CD95 has no influence on ICAM and E-selectin level in endothelial cells . ( A ) Flow cytometry plot of liver endothelial cells . ( B–C ) Endothelial cells were dissociated from the liver of Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice ( B ) or Fas<f/f> and Fas<f/f>::Cdh5<CreERT2> mice ( C ) and stained with antibodies of endothelial cell markers and CD95L or CD95 . CD95L or CD95 level was analyzed by flow cytometry . ( D–G ) Endothelial cells of Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice were stained with antibodies of endothelial cell markers and subtypes of ICAM and selectin . ICAM1 ( D ) , ICAM2 ( E ) , E-selectin ( F ) and P-selectin ( G ) levels were analyzed by flow cytometry . ( H–J ) Endothelial cells of Fas<f/f> and Fas<f/f>::Cdh5<CreERT2> mice were stained with antibodies of endothelial cell markers , ICAM and selectin . ICAM1 ( H ) , E-selectin ( I ) and P-selectin ( J ) levels were analyzed by flow cytometry . Data are presented as mean ± SEM in ( B–J ) and evaluated by two-tailed unpaired Student's t test in ( B , C , G ) , p<0 . 05 , p<0 . 01 , n=4 . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 00810 . 7554/eLife . 18542 . 009Figure 2—figure supplement 2 . Characterization of Fasl<f/f>::Cdh5<CreERT2> mice . ( A , B ) Percentage of blood neutrophils , monocytes , T cells , B cells , dendritic cells and natural killer cells among CD45+ blood cells in Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice before tamoxifen induction ( A ) or 10 days after tamoxifen induction ( B ) . n=6 . ( C ) Absolute number of neutrophils in blood of Fasl<f/f> and Fasl<f/f>::Cdh5<CreERT2> mice . n=6 . Data are presented as mean ± SEM , n . s . not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 009 To further prove that endothelial cell-derived CD95L is required for myeloid cell recruitment , we used a thioglycollate-induced peritonitis model . Neutrophil extravasation into the inflamed peritoneal cavity was assessed in endothelial-CD95L deficient mice ( Figure 2G , H ) . The number of peritoneal neutrophils at 6 hr after thioglycollate injection was significantly reduced in endothelial-CD95L deficient mice as compared to control counterparts ( Figure 2I ) . Nonetheless , deletion of CD95 in the endothelial compartment ( Fas<f/f>::Cdh5<CreERT2> ) did not affect thioglycolate-induced neutrophil recruitment to the inflamed peritoneum ( Figure 2J , Figure 2—figure supplement 1C ) . Notably , the cell surface expression of adhesion molecules , E-selectin , ICAM1 and ICAM2 was similar in CD95L-deleted and non-deleted endothelial cells , with the exception of cell surface expression of P-selectin that was reduced ( Figure 2—figure supplement 1D–G ) . Similarly , deletion of CD95 in myeloid cells had no impact on ICAM1 and selectin levels in endothelial cells ( Figure 2—figure supplement 1H–J ) . Additionally , the ratio of different leukocyte populations in the blood and the absolute number of blood neutrophils were not significantly changed in Fasl<f/f>::Cdh5<CreERT2> mice before or after tamoxifen induction as compared to the control mice ( Figure 2—figure supplement 2A , B , C ) . These data demonstrate that the involvement of endothelial cell-derived CD95L in leukocyte slow rolling and transmigration is not related to the change of adhesion molecule expression level on the luminal surface of blood vessels or the homeostasis of leukocytes in the blood stream . E-selectin engagement triggers a signaling cascade which cooperates with chemokine signals to facilitate neutrophil rolling and adhesion during inflammation ( Zarbock et al . , 2007 ) . E-selectin is expressed by inflamed endothelial cells and engages PSGL-1 , CD44 and other ligands in neutrophils ( Xia et al . , 2002; Katayama et al . , 2005 ) . Upon ligand engagement , it has been reported that E-selectin activates SFKs which in turn initiate a signaling pathway involving the activation of ITAM bearing adaptors , Syk , Btk , PLCγ2 , P38 and PI3Kγ ( Yago et al . , 2010; Mueller et al . , 2010 ) . We have previously shown that CD95L triggers the recruitment of myeloid cells to inflammatory sites via SFK-Syk-PI3K pathway ( Letellier et al . , 2010 ) . To validate whether CD95 signaling can also activate Btk and PLCγ2 via Syk , we studied Btk and PLCγ2 activation upon CD95L stimulation . As Syk deficiency ( Syk-/- ) is perinatal-lethal in mice ( Turner et al . , 1995 ) , we cultured primary embryonic liver-derived macrophages . Activation of CD95 increased phosphorylation of Syk , Btk and PLCγ2 ( Figure 3A and quantified analysis in Figure 3B , C ) . CD95-mediated phosphorylation of PLCγ2 greatly decreased in macrophages isolated from Syk-/- mice as compared to wild type ( WT ) cells ( Figure 3A , C ) . The reduced basal level of p-PLCγ2 in Syk-/- cells indicates that Syk is not only involved in CD95L-mediated phosphorylation of PLCγ2 but also acts as a hub for other ligands . However , the upregulated phosphorylation of Btk was still present in Syk-/- as compared to WT cells ( Figure 3A , B ) . Activation of Btk could be explained by the previously observed interaction of the PH domain of Btk with CD95 in B-cells ( Vassilev et al . , 1999 ) . To examine this interaction we pulled down Btk from the lysates of CD95L-treated macrophages by immunoprecipitation . Binding of CD95 to Btk was detected following CD95L stimulation ( Figure 3D ) . In order to validate the functional involvement of Btk in CD95L-induced PLCγ2 phosphorylation , we used the Btk inhibitor PCI-32765 ( Ibrutinib ) 1 hr prior to CD95L stimulation of macrophages . PCI-32765 fully blocked the basal and CD95-induced phosphorylation of Btk and PLCγ2 ( Figure 3E , F , G ) , but not the phosphorylation of Syk ( Figure 3H ) , which indicates that the phosphorylation of PLCγ2 is Btk-dependent and that Btk is an essential mediator for CD95-induced PLCγ2 activation . Taken together , these data demonstrate the presence of two signaling branches downstream of CD95: CD95/SFK/Syk/Btk/PLCγ2 and CD95/SFK/Btk/PLCγ2 ( Figure 3I ) . More importantly , immobilized CD95L-induced slow rolling was abolished in PCI-32765 pretreated mice in the autoperfused flow chamber assay , implying that CD95L-induced slow rolling was Btk-dependent ( Figure 3J ) . 10 . 7554/eLife . 18542 . 010Figure 3 . CD95L stimulation induces phosphorylation of PLCγ2 via activating Syk and Btk in myeloid cells . ( A ) Macrophages cultured from WT or Syk-/- embryonic liver cells were treated with CD95L ( 40ng/ml ) . Lysates were prepared at the indicated time points and immunoblotted for the indicated proteins . ( B–C ) Quantification analysis of PLCγ2 and Btk phosphorylation level in ( A ) from three independent experiments . Data are presented as mean ± SEM , n=3 . ( D ) Macrophages cultured from bone marrow cells were treated with CD95L ( 40 ng/ml ) . Lysates were prepared at the indicated time and immunoprecipitated with anti-Btk followed by immunoblotting with CD95 and Btk antibody . ( E ) Macrophages cultured from bone marrow cells were treated with DMSO or Btk inhibitor PCI-32765 ( 1 µM ) one hour prior to CD95L stimulation ( 40 ng/ml ) . Lysates were prepared at the indicated time points and immunoblotted for the indicated proteins . ( F–H ) Quantification analysis of Btk , PLCγ2 and Syk phosphorylation level in ( E ) from three independent experiments . Data are presented as mean ± SEM , n=3 . ( I ) Scheme of CD95L stimulation-induced PLCγ2 activation . ( J ) Rolling velocity of neutrophils from DMSO or Btk inhibitor pre-treated mice in a flow chamber coated with E-selectin/ICAM1/CD95L . Data are presented as mean ± SEM , two-tailed unpaired Student's t test , p<0 . 001 , n=3 . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 010 The common final step for integrin activation has been revealed as binding of talin1 and kindlin-3 to the cytoplasmic domain of β integrin which in turn breaks the salt bridges between the cytosolic domains of integrin α and β subunits and induces integrin conformational changes ( Tadokoro et al . , 2003; Wegener et al . , 2007; Lefort et al . , 2012 ) . Recruitment of talin1 to LFA-1 is Rap1a-dependent ( Lefort et al . , 2012 ) . In order to find out whether CD95 also activates Rap1 , we performed an active Rap1 pull-down assay in CD95L-stimulated mouse bone marrow-derived neutrophils . Significant activation of Rap1 was observed in neutrophils 15 min after CD95L stimulation ( Figure 4A , B ) . Of note , CD95L-treated and control neutrophils exhibited similar levels of integrin αL , integrin αM and integrin β2 expression ( Figure 4—figure supplement A–D ) . 10 . 7554/eLife . 18542 . 011Figure 4 . CD95L stimulation induces integrin activation and recruitment of integrin to CD95 . ( A ) Bone marrow-derived murine neutrophils were treated with CD95L ( 40 ng/ml ) . Lysates were prepared at the indicated time points and GST-RalGDS-RBD peptide affinity-precipitated for Rap1 immunoblotting . ( B ) Quantification analysis of Rap1-GTP activation in ( A ) from three independent experiments . Data are presented as mean ± SEM , n=3 . ( C–D ) U937 cells were perfused through human E-selectin coated flow chamber in the presence of soluble or immobilized CD95L . The binding of KIM127 ( C ) or mAb24 ( D ) were analyzed by flow cytometry and presented as mean ± SEM , n=3 . ( E ) Bone marrow-derived murine neutrophils were treated with coated CD95L for 10 min . The binding of soluble ICAM1 was analyzed by flow cytometry and data presented as mean ± SEM , n=3 . ( F ) Representative pictures show PLA of integrin αL and CD95 in control or CD95L-treated dHL60 cells . Red , PLA; green , Phalloidin; blue , DAPI . Scale bar: 10 µm . ( G ) The number of PLA signal in each control or CD95L-treated dHL60 cell . Data are presented as violin plot , 383 control cells and 630 CD95L-treated dHL60 cells from 8 random fields were evaluated . ( H ) Ratio of PLA negative and positive cells in control or CD95L-treated dHL60 . Data are presented as stacked bar . ( I–J ) Number of PLA signal ( I ) and integrated density ( J ) of PLA signal in PLA positive cells . Data are presented as mean ± SEM , n=226–453 . ( K ) Bone marrow-derived macrophages were treated with CD95L ( 40 ng/ml ) . Lysates were prepared at the indicated time and immunoprecipitated with anti-CD11a followed by immunoblotting for CD95 and CD11a antibody . Statistical significance was evaluated by one-way ANOVA followed by Bonferroni multiple comparison post hoc test in ( B–D ) and two-tailed unpaired Student's t test in ( E , G , I , J ) , p<0 . 05 , p<0 . 01 , p<0 . 001 , n . s not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 01110 . 7554/eLife . 18542 . 012Figure 4—figure supplement 1 . CD95L treatment doesn’t influence the integrin level of neutrophils in vitro . ( A ) Flow cytometry plot of percoll isolated-bone marrow neutrophils . ( B–D ) Bone marrow-derived neutrophils were treated with CD95L ( 40 ng/ml ) and fixed at the indicated time points . Fixed neutrophils were stained with antibodies of neutrophil markers and integrin subunits and analyzed by flow cytometry . Neutrophils expression levels of integrin αL ( B ) , integrin αM ( C ) and integrin β2 ( D ) are presented as mean ± SEM , n=3 . ( E ) FACS histogram plots show KIM127 and mAb24 reporter antibodies binding on hE-seletin/CD95L stimulated U937 cells . ( F–G ) CD95L ligand-treated or non-treated U937 cells were perfused through human E-selectin coated or non-coated flow chamber . The binding of KIM127 ( E ) or mAb24 ( F ) were analyzed by flow cytometry and presented as mean ± SEM , n=3 . ( H ) Integrated density of phalloidin in control or CD95L-treated dHL60 cells is presented as mean ± SEM . ( I ) Bone marrow-derived macrophages were treated with CD95L ( 40 ng/ml ) . Lysates were prepared at the indicated time and immunoprecipitated with anti-CD11b followed by immunoblotting for CD95 and CD11b antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 012 Inside-out integrin signaling triggers conformational changes in integrin , leading to increased binding affinity ( integrin activation ) and avidity ( Abram and Lowell , 2009 ) . Different conformations of LFA-1 can be recognized by integrin epitope specific antibodies . KIM127 recognizes an epitope of β2 subunit of human LFA-1 when it is extended ( Beglova et al . , 2002 ) , whereas mab24 binds to the epitope of I-like domain in β2 subunit of human LFA-1 only if this is accessible as in the high-affinity state ( Lu et al . , 2001 ) . To examine if CD95 induces integrin conformational changes , binding of reporter antibodies was analyzed by flow cytometry in U937 cells . To this end , cells were pre-incubated with the reporter antibodies , and thereafter perfused through the flow chamber in the presence of soluble or immobilized CD95L . We observed significantly increased binding of KIM127 and mab24 in cells treated with soluble CD95L ( Figure 4C , D , Figure 4—figure supplement 1E ) . Notably , in the absence of coated E-selectin , soluble CD95L showed no effect on KIM127 and mab24 binding ( Figure 4—figure supplement 1F–G ) . These data indicate that CD95L treatment induces integrin conformational changes including its extension and full activation in an E-selectin-dependent manner . The soluble ICAM1 binding assay is a commonly used test for LFA-1 function , which is determined by affinity and avidity ( Salas et al . , 2004; Lefort et al . , 2012 ) . To further confirm that CD95L stimulation induces integrin activation , mouse bone marrow-derived neutrophils were incubated with ICAM1-Fc and the binding of ICAM1 was assessed by flow cytometry . Anti-CD11b antibody was used to block the integrin αMβ2 ( Mac-1 ) -dependent ICAM1 binding . CD95L-activated neutrophils showed significant binding of soluble ICAM1 as compared to the non-treated group ( Figure 4E ) . Taken together , these results demonstrate that CD95 signaling induces integrin activation . Compartmentalization of multi-molecular signaling complexes integrates extracellular signals and facilitates integrin activation ( Bezman and Koretzky , 2007 ) . Specifically , a signalosome consisting of Src family kinase Hck , Btk , WASp and PLCγ2 has been identified to be indispensable for fMLF-induced MAC-1 activation required for neutrophil recruitment ( Volmering et al . , 2016 ) . To test if CD95 assembles a signaling complex with LFA-1 that together with E-selectin orchestrates integrin activation , we performed proximity ligation assay of integrin αL and CD95 in control or CD95L-treated dHL60 cells . Proximity ligation assay ( PLA ) detects interaction of proteins that are 30nm or less apart from each other ( Söderberg et al . , 2006 ) . Indeed , binding of CD95 to integrin αL was detected by PLA in dHL60 cells ( Figure 4F ) . CD95L treatment significantly enhanced this binding , as shown by increased number of PLA signal of CD95-integrin clusters ( Figure 4G ) and a ratio of PLA signal positive cells ( Figure 4H ) . Among the PLA positive cells , CD95L treatment increased the number and integrated density of PLA signal as compared to the non-treated control ( Figure 4I , J ) . Interestingly , CD95L-treated cells also showed a tendency towards increased polymerization of F-actin as shown by phalloidin staining compared to the control cells ( Figure 4—figure supplement 1H ) . Moreover , immunoprecipitation of integrin αL from lysates of CD95L-treated mouse macrophages , revealed a stimulation-dependent association of CD95 with integrin αL ( Figure 4K ) . However , integrin αM did not bind to CD95 upon CD95L stimulation ( Figure 4—figure supplement 1I ) . Collectively , our data indicate that activation of CD95 leads to the assembly of multiprotein complexes including integrins and that CD95 mediates integrin activation required for rolling and adhesion of myeloid cells . In order to test the involvement of myeloid cell-specific CD95 in systemic inflammation , we used a CLP-induced animal model of sepsis . Bacterial load in blood , spleen and peritoneal cavity of Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice was assessed by the number of colony-forming unit ( CFU ) 6 hr after CLP . Importantly , Fas<f/f>::Lyz2<Cre> mice demonstrated significantly higher CFU in peritoneal lavage liquid and blood as compared to the Fas<f/f> littermate control mice ( Figure 5A , B ) . CFU in spleen homogenate was also elevated , but was not significantly higher ( Figure 5C ) . Moreover , peritoneal neutrophil infiltration in Fas<f/f>::Lyz2<Cre> mice was reduced as compared to the Fas<f/f> littermate control mice ( Figure 5D ) . These data indicate that CD95 in myeloid cells is involved in mounting an effective bacterial clearance response during systemic inflammation via recruiting neutrophils to the inflammatory sites . 10 . 7554/eLife . 18542 . 013Figure 5 . CD95 in myeloid cells is required for bacterial clearance . ( A–C ) Bacterial counts of peritoneal lavage fluid ( A ) , blood ( B ) and spleen ( C ) from Fas<f/f> or Fas<f/f>::Lyz2<Cre> mice 6 hr after CLP . Data are pooled from three independent experiments and presented as dot plot with mean ± SEM , n=16–19 . ( D ) Infiltrating peritoneal neutrophils 6 hr after CLP in Fas<f/f> or Fas<f/f>::Lyz2<Cre> mice . Data are pooled from three independent experiments and presented as mean ± SEM , n=14–18 . Significance between groups was evaluated by running a linear mixed model for the log-CFU with the random covariable of time point and the fixed covariable of gender . p<0 . 05 , n . s not significant , DOI: http://dx . doi . org/10 . 7554/eLife . 18542 . 013 In summary , inflammation induces CD95L expression in endothelial cells ( Sata and Walsh , 1998 ) . CD95 together with E-selectin orchestrate signaling events leading to integrin activation that finally result in slow rolling and adhesion of myeloid cells . CD95-induced leukocyte infiltration was first found in early studies aiming at inducing apoptosis of tumor cells in vivo ( Arai et al . , 1997; Seino et al . , 1997 ) . In these studies , transplantation of CD95L-overexpressing/CD95-negative tumor cells induced a dramatic neutrophil infiltration into the tumor xenografts . Other studies using Boyden chamber assays demonstrated that soluble CD95L induces the transmigration of human neutrophils in vitro ( Seino et al . , 1998; Ottonello et al . , 1999; Dupont and Warrens , 2007 ) . Although these findings are interesting , they did not address whether CD95-induced neutrophil recruitment was through direct CD95 activation on neutrophils or secondary to CD95-induced production of inflammatory mediators . The present study shows that CD95 is involved in induction of slow rolling and adhesion of neutrophils , and that these steps are blocked in CD95-deficient neutrophils , indicating that CD95 induces slow rolling via a direct effect on neutrophils and not via induction of inflammatory cytokines and chemokines . In this study , soluble or coated CD95L induce neutrophil slow rolling in the autoperfused flow chamber assay . However , coating of the flow chamber with CD95L alone was not sufficient to induce leukocyte tethering ( capturing ) , indicating that E-selectin is required for CD95-mediated slow rolling . It has been shown that selectin ligands PSGL-1 and CD44 are enriched in lipid rafts ( Miner et al . , 2008; Neame et al . , 1995 ) . In addition , the three SFKs of neutrophils , Fgr , Hck , Lyn , which are activated upon the engagement of selectin to its ligands ( Yago et al . , 2010 ) , also associate with cholesterol-dependent membrane rafts ( Lowell , 2004 ) . Interestingly , neutrophil slow rolling has been reported to be dependent on intact lipid microdomains to signal slow rolling on E-selectin and P-selectin ( Yago et al . , 2010 ) . The clustering of lipid microdomains is regulated by the actin cytoskeleton ( Chichili and Rodgers , 2007 ) . Ezrin/radixin/moesin ( ERM ) proteins , which link the cytoskeleton to integral membrane proteins via their FERM domains , associate with PSGL-1 and CD44 through their cytoplasmic domains ( Yonemura et al . , 1998; Urzainqui et al . , 2002 ) . Moreover , ligation of PSGL-1 to selectin recruits Syk to an atypical ITAM on ERM proteins bound to the cytoplasmic domain of PSGL-1 ( Urzainqui et al . , 2002 ) . Thus , leukocyte rolling requires the formation of multiprotein complexes at the plasma membrane . CD95 clustering is accompanied by reorganization of the actin cytoskeleton and aggregation of lipid microdomains ( Söderström et al . , 2005 ) . Accordingly , CD95 clustering in sphingolipid-rich membrane microdomains is necessary for the induction of CD95 signaling ( Grassme et al . , 2001 ) . CD95 indirectly associates with actin via direct and specific binding to ezrin FERM domains ( Lozupone et al . , 2004 ) , and the organization of the microfilaments affects the outcome of CD95 stimulation ( Parlato et al . , 2000 ) . In addition , CD44 has been reported to bind to CD95 via ezrin to block the apoptotic signal transduction ( Mielgo et al . , 2006 , 2007 ) . These findings suggest that CD95 may also be a part of a multiprotein complex encompassing selectin ligands , SFKs and cytoskeletal proteins that orchestrate leukocyte rolling . Ligand binding to external domains causes conformational changes that increase ligand affinity , and formation of integrin clustering , which in turn results in SFK autophosphorylation and Syk kinase activation in the outside-in integrin pathway ( Abram and Lowell , 2009 ) . SFK and Syk kinases can directly interact with the cytoplasmic domain of β2 , and β3 integrins ( Arias-Salgado et al . , 2003 ) . In addition , ITAM containing adaptor proteins DAP12 and FcRγ couples Syk to integrins ( Mócsai et al . , 2006 ) . The ITAM-like YXXL motif of CD95 is involved in the CD95-Lyn-Syk signaling cascade leading to myeloid cell transmigration ( Letellier et al . , 2010 ) . Interestingly , the current study identifies an association of CD95 with integrin αLβ2 in neutrophils , as assessed by PLA , and macrophages , as confirmed by direct co-immunoprecipitation . Moreover , integrin activation reporter assay and the soluble ICAM1 binding assay demonstrate that CD95-induced integrin activation is a mechanism present in both macrophages and neutrophils . Altogether , these findings strongly indicate that CD95 modulates rolling and adhesion via its participation in a multiprotein signaling complex containing selectin ligands , SFKs , integrins and cytoskeletal proteins in both neutrophils and macrophages . Alternatively , CD95 might indirectly impact myeloid cell recruitment by promoting secretion of inflammatory cytokines , as already reported in a variety of cell types ( Altemeier et al . , 2007; Dupont and Warrens , 2007; Park et al . , 2003; Wang et al . , 2010a ) . In line with this , a recent study showed that CD95 in apoptotic cells induced the production of pro-inflammatory cytokines and chemokines , which in turn attracted myeloid cells ( Cullen et al . , 2013 ) . Interestingly , the autoperfused flow chamber assay in this study shows that CD95-induced slow rolling is independent of chemokine production . The upregulation of selectins and ICAMs in endothelial cells and selectin- and ICAM-ligands in leukocytes play important roles in rolling and adhesion ( Pober and Sessa , 2007 ) . Unlike the effect of TNF-α on endothelial cells , CD95L stimulation or CD95/CD95L deletion in endothelial cells have no impact on the expression level of adhesion molecules in endothelial cells . A previous study has reported that crosslinking of CD95 with antibody rapidly triggers down-modulation of L-selectin , CD44 , LFAα and LFAβ in CD95-sensitive T cell blasts ( Kabelitz et al . , 1996 ) . In our study , in vitro or in vivo treatment with CD95L , or CD95 deletion in myeloid cells did not change cell surface expression levels of most integrins , which shows a cell type dependent effect of CD95 on the expression of adhesion molecules . These results demonstrate that CD95-induced neutrophil slow rolling is independent of regulation of adhesion molecules . Endothelial cells play a pivotal role in leukocyte recruitment by synthesizing and presenting chemokines and leukocyte adhesion molecules during inflammation ( Pober and Sessa , 2007 ) . Interestingly , CD95L has also shown to be expressed by endothelial cells ( Sata and Walsh , 1998 ) . Overexpression of CD95L by adenovirus transduction in endothelial cells markedly attenuated TNF-α-induced T cell and macrophage infiltration , and adherent mononuclear cells underwent apoptosis ( Sata and Walsh , 1998 ) . Along this line , tumor endothelial cells selectively and highly express CD95L , which serves as a barrier to prevent the infiltration of CD8 cells via induction of apoptosis in the establishment of immune tolerance ( Motz et al . , 2014 ) . Contrary to the apoptotic effect of endothelial cell-derived CD95L , the present study shows that deletion of CD95L in endothelial cells impairs neutrophil recruitment in inflamed cremaster muscle and thioglycolate-induced peritonitis . Together , these data indicate that endothelial cell-derived CD95L may serve as a chemokine to induce myeloid cell recruitment during inflammation . Endothelial cells are known to basally express CD95L ( Sata and Walsh , 1998 ) , however , as shown by our study on CD95L-induced slow rolling in the autoperfused flow chamber assay , CD95L might only induce slow rolling in an inflammatory setting , as this function requires additional presentation of E-selectin by endothelial cells . In addition , inflammatory cytokines might increase CD95L levels in endothelial cells . Along this line , IFNγ activates CD95L promoter activity in T-cells ( Kirchhoff et al . , 2002 ) . Integrin signaling plays important roles in regulating cancer 'stemness' , metastasis and drug resistance ( Seguin et al . , 2015 ) . As CD95 is now recognized as an inducer of tumor cell growth and invasion ( Martin-Villalba et al . , 2013; Peter et al . , 2015 ) , it is of great importance to study the CD95-induced integrin signaling in tumor progression . On the other hand , tumor growth and metastasis are promoted by myeloid-derived suppressor cells ( MDSC ) ( Condamine et al . , 2015 ) . Blockade of endothelial cells-derived CD95L in order to inhibit MDSCs recruitment to tumor might be used as a potential strategy for cancer therapy . Effective removal of infectious organisms is of utmost importance to attenuate the early onset of sepsis ( Bosmann and Ward , 2013 ) . It has been reported that mice with a global impairment of CD95 activity , Faslpr/lpr and Fasgld/gld mice , develop severe diarrhoea and showed impaired bacterial clearance in a bacterial-induced gut infection model ( Pearson et al . , 2013 ) . Yet , the use of a mouse with ubiquitous impairment of CD95 activity hindered clarification of the exact mechanism underlying this impairment . We now show that myeloid cells require CD95 activity to efficiently to infiltrate into the inflammatory sites and clear bacteria following CLP-induced sepsis . We observed increased bacterial load in blood and peritoneum of Fas<f/f>::Lyz2<Cre> mice as compared to control counterparts , in a CLP-induced sepsis model . Altogether these data reveal that CD95 in myeloid cells plays an important role in bacterial clearance . Altogether , this study shows a chemoattractant effect of endothelial cell-derived CD95L in induction of neutrophil slow rolling and adhesion via integrin activation . Both cancer cells and immune cells exhibit very high levels of CD95 surface expression . Therapies aimed at interfering with CD95’s activity can be used for the treatment of diseases with a major cell-extravasation component such as cancer progression and inflammation . C57BL/6N mice were purchased from Charles River Laboratories . Syk+/- mice were from Martin Turner ( The Babraham Institute , UK ) and bred as heterozygous . Fas<f/f> mice ( kind gift from K . Rajewsky , Max Delbrück Center for Molecular Medicine , Germany ) were bred with Lyz2<Cre> ( Jackson Laboratory , USA ) mice . Cdh5<CreERT2>mice ( Ralf H . Adams , University of Münster , Germany , Wang et al . , 2010b ) were bred with Fasl<f/f> or Fas<f/f> ( Karray et al . , 2004 ) mice . Animal experiments were performed in accordance with institutional guidelines of the German Cancer Research Center and were approved by the Regierungspräsidium Karlsruhe , Germany ( Permit Number: G188/13 ) . Autoperfused mouse flow chambers assay was performed as previously reported ( Chesnutt et al . , 2006 ) . Briefly , carotid artery of male , 12 weeks old , WT or Fas<f/f>::Lyz2<Cre> mice was exposed and connected to flow chamber with a PE10 tubing . The free end of the flow chamber was connected to a water-filled PE50 tubing in order to control the pressure drop in the chamber which determined the shear stress of rolling cells . Flow chambers were coated with different combinations of 30 μg/ml E-selectin , 90 μg/ml L-selectin , 20 μg/ml P-selectin , 15 μg/ml ICAM1 ( R&D systems , USA ) and 50 μg/ml CD95L . For some conditions , mice were intravenously injected with 10 μg CD95L , 40 μg anti-CD11a antibody ( M17/4 , ebioscience , USA ) , 40 μg anti-CD11b antibody ( M1/70 , ebioscience , USA ) , DMSO ( 1:100 ) or Btk inhibitor ( PCI-32765 , Sigma , 15 mg/kg ) in 100 μl saline one hour before performing the autoperfusion assay . Rolling cells in 3 random fields for each flow chamber were recorded and two flow chambers were used for each mouse . 3 to 4 mice were used for each group . The CD95L utilized in this study was a fusion protein of trimeric human CD95L-receptor binding domain fused with T4-Foldon motif from the fibritin of the bacteriophage T4 ( CD95L-T4 ) and purified from CD95L-T4 plasmid-transfected HEK293T cells ( Kleber et al . , 2008 ) . It is commercially available from IBA GmbH , Göttingen , Germany . Fas<f/f>::Lyz2<Cre±> and Fasl<f/f>::Cdh5<CreERT2±> mice were used for intravital microscopy . 2 hr before cremaster muscle exteriorization , mice received 4 µg PTx i . v . ( Sigma-Aldrich , USA ) and 500 ng TNF-α intrascrotally ( R&D systems , USA ) . Mice were anesthetized with an i . p . injection of 125 mg/kg ketamine hydrochloride ( Sanofi ) , 0 . 025 mg/kg atropine sulfate ( Fujisawa , Japan ) , and 12 . 5 mg/kg xylazine ( Tranqui Ved; Phoenix Scientific , UK ) and placed on a heating pad . The cremaster muscle was prepared as previously described ( Mueller et al . , 2010 ) . Postcapillary venules with a diameter between 20 and 40 µm were recorded using an intravital microscope ( Axioskop , SW 40/0 . 75 objective; Carl Zeiss , Inc . ) through a digital camera . Blood flow centerline velocity was measured using a dual-photodiode sensor system ( CircuSoft Instrumentation ) . Recorded images were analyzed using ImageJ and AxioVision ( Carl Zeiss , Germany ) software . Leukocyte rolling flux fraction was calculated as percentage of total leukocyte flux . Transmigrated cells were determined in an area reaching out 75 µm to each side of a vessel over a distance of 100 µm vessel length . Macrophages were cultured from femurs and tibias derived-bone marrow cells as previously described ( Letellier et al . , 2010 ) or from fetal liver cells of Syk+/- mice ( E15 ) . Neutrophils were isolated from mouse bone marrow cells over discontinuous 50%/55%/62%/81% percoll gradients . The gradients were centrifuged at 1600 g for 30 min without braking at 10°C and the interphase between 62% and 81% percoll was collected . Neutrophils were cultivated overnight in RPMI medium containing 10% fetal calf serum and 20% WEHI-3B conditioned medium . Fasl<f/f>::Cdh5<CreERT2±> and Fas<f/f>::Cdh5<CreERT2±> mice were gavaged with tamoxifen in sun flower seed oil ( 200 mg/kg , Sigma , Germany ) for 5 consecutive days and 1ml of 3% thioglycollate broth ( Fluka , Germany ) in PBS was i . p . injected 7 days after the last tamoxifen induction . In this model , neutrophil infiltration peaks at 6 hr . Peritoneal cells were collected 6 hr after thioglycollate injection and total cells were counted as previously described ( Letellier et al . , 2010 ) . Differential cell counts were accessed by flow cytometry after staining with neutrophil markers . Blood samples were stained with antibodies against leukocyte markers . Neutrophils were identified according to the profile of Forward Scatter ( FSC ) /Sider Scatter ( SSC ) , DAPI-negativity , and CD45 , CD11b , Ly6G-positivity . Neutrophils were stained with anti-CD95 ( Jo2 ) , anti-CD11b , anti-CD11a , anti-CD18 for testing the expression of CD95 and different integrins . Endothelial cells from dissociated liver cells were identified according to the profile of FSC/SSC , DAPI , CD45-negativity , and CD31-positivity . Antibodies of anti-CD95L ( MFL3 ) , anti-CD95 , anti-ICAM1 , anti-ICAM2 , anti-E-selectin , and anti-P-selectin were used for testing the expression of CD95 , CD95L and different adhesion molecules in endothelial cells . Flow cytometry data were analyzed with Flowjo software . CD95L treated ( 40 ng/ml ) or non-treated cells were washed with PBS containing phosphatase inhibitors , pelleted , and lysed on ice for 30 min with Pierce IP Lysis buffer ( Fisher Scientific , Germany ) containing vanadate , inhibitors for phosphatase and proteinase . Lysates of 500 μg protein were immunoprecipitated at 4°C for 4 hr with the anti-mouse CD11a ( M17/4 , biolegend ) , anti-mouse Btk ( Cell Signaling , USA ) , anti-mouse CD11b ( M1/70 , ebioscience , USA ) antibodies or the corresponding isotype controls . Afterward , 40 μl Dynabeads M-280 Streptavidin was added to each sample and incubated for 1 hr at 4°C with rotation . Beads were washed 5 times with 1 ml of lysis buffer . The immunoprecipitates were released by cooking the beads with 40 μl of 2x laemmli buffer at 95°C for 5 min . Immunoblotting was performed as previously described ( Letellier et al . , 2010 ) . Membranes were probed with following antibodies respectively: anti-phospho-Syk ( Tyr319 , 352 ) , anti-phospho-Btk ( Tyr223 ) , anti-phospho-PLCγ2 ( Tyr1217 ) , anti-Syk , anti-Btk , anti-PLCγ2 ( Cell Signaling , USA ) , anti-Rap1 ( Fisher Scientific ) , anti-mouse CD11a , anti-CD95 ( M20 , Santa Cruz Biotechnology , Germany ) , anti-mouse CD11b ( Novus Biologicals , USA ) . Western blots were quantified with ImageJ software and normalized to the respective loading controls . Active Rap1 Pull-Down assay was performed according to the manufacturer’s instructions ( Fisher Scientific , Germany ) . Bone marrow-derived neutrophils were stimulated with CD95L ( 40 µg/ml ) and the cell lysates were prepared . 100 µl Glutathione Resin and 20 μg of GST-RalGDS-RBD peptide were added to 500 μg lysate . GTPγS and GDP incubated lysates were used as positive and negative control respectively . After one-hour incubation at 4°C , resin beads were washed 4 times and then followed by incubation in 40 µl of 2x laemmli buffer at 95°C for 5 min . Precipitates were electrophoresed and blotted for anti-Rap1 . Soluble ICAM1 binding assay was performed as previously described ( Lefort et al . , 2012 ) . Mouse bone marrow-derived neutrophils were suspended in Hanks Balanced Salt Solution containing 1 mM CaCl2 and MgCl2 , and then the cell suspension was planted in CD95L-precoated chamber in the presence of ICAM1/FC ( 20 µg/ml , R&D systems , USA ) and PE-conjugated anti-human IgG1 ( Fc-specific; Southern Biotechnology , USA ) for 5min at 37°C . Anti-CD11b ( 10 µg/ml ) antibody was used to block the Mac-1-dependent ICAM1 binding . The binding of ICAM1 was determined by flow cytometry . Integrin conformational change upon CD95L treatment was tested by staining with reporter antibodies recognizing specific epitopes of integrin at different statuses . To test the binding , U937 cells ( 10 million/ml ) were premixed with anti-Human CD11/CD18 ( mab24 ) or anti-Human CD11/CD18 ( KIM127 ) and perfused through the human E-selectin coated flow chamber with a syringe pump ( New Era Pump Systems , USA ) at the flow rate of 3 μl/min upon the stimulation with soluble CD95L ( 60 ng/ml ) or immobilized CD95L ( 10 μg/ml for coating ) . The assembly and coating of the flow chamber was the same as described for the autoperfused mouse flow chamber assay . Cells flowed through the chamber were collected and fixed in 2% PFA . Then the fixed cells were stained with PE anti-mouse IgG and analyzed with flow cytometry . Colocalization of integrin αL to CD95 was determined by proximity ligation assay . dHL60 cells were planted in CD95L-precoated chamber for 10 min at 37°C . Following stimulation , cells were fixed with 2% PFA and stained with anti-CD11a ( EP1285Y , abcam , UK ) and anti-CD95 ( APO-1-1 , Enzo Life Science , Germany ) antibodies . After washing , the proximity ligation assay was performed by using duolink kit following the manufacturer's instruction ( Duolink In Situ Red Starter Kit Mouse/Rabbit , Sigma-Aldrich ) . Then the cells were stained with DAPI and Alexa 647-conjugated Phalloidin . Immunofluorescent signals were recorded with a TCS SP5 confocal microscope ( Leica , Germany ) . Fluorescence images were analyzed with a programmed imageJ algorithm . The number of PLA and integrated density of PLA were accessed . Sepsis was induced by CLP performed as previously described ( Rittirsch et al . , 2009 ) . Age matched littermates of Fas<f/f> and Fas<f/f>::Lyz2<Cre> mice were used . Briefly , mice were anesthetized and cecum was exposed and ligated . To induce serve sepsis , the cecum was punctured twice with an 18-gauge needle , after which a small drop of feces was extruded from each puncture site to ensure patency . Mice were sacrificed 6 hr after CLP , at which time the peritoneal cavity was lavaged with 3 ml sterile phosphate-buffered saline containing 1 mM EDTA , and blood and spleen were collected . Aliquots of peritoneal lavage , blood and spleen homogenate were serially diluted , overnight cultured on sheep blood agar plates under 37˚C , and then the number of CFUs was determined . Statistical significance between groups were evaluated by one-way ANOVA using Bonferroni multiple comparison post hoc test for multiple groups comparison or Student's t test for two groups comparison . All data were presented as mean ± standard error of the mean ( SEM ) unless otherwise indicated . Statistical significance was determined by the p-value of the statistical test and deemed as significant p<0 . 05; strongly significant p<0 . 01 and highly significant **p<0 . 001 . Statistical analysis was performed with GraphPad Prism ( Version 5 . 01 ) . Significance for bacterial clearance experiments was evaluated by running a linear mixed model for the log-CFU with the random covariable of time point and the fixed covariable of gender , using SAS ( Version 9 . 2 ) .	When tissues are damaged or infected , the body produces an inflammatory response . Neutrophils – a type of white blood cell – play an important part in this response . These cells normally circulate through the bloodstream , and are recruited to the inflamed site by chemical signals sent out by immune cells in the damaged tissue . This causes passing neutrophils to migrate through the wall of the blood vessel to gain access to the inflamed tissue . The neutrophils go through a sequence of steps before they can pass through the blood vessel wall . After initially tethering to the cells that line the blood vessel , the neutrophils experience a period of “slow rolling” across the vessel lining , before tightly adhering to one of the cells . In 2010 , researchers determined that a protein on the neutrophil’s surface , known as CD95 , helps the cell migrate through blood vessel walls . This protein interacts with a “ligand” molecule on the surface of the cells that line the blood vessel . However , it remains unclear whether CD95 and its ligand play a role in the steps that lead up to the neutrophils migrating through the blood vessel wall . Gao et al . – who include researchers involved in the 2010 study – now show that activating CD95 in neutrophils also triggers the cell’s slow rolling and adhesion . Experiments performed on mouse cells and tissues showed that the cells that line the blood vessels present the CD95 ligand on their surfaces in order to activate CD95 in the neutrophils circulating in the bloodstream . This ultimately leads to neutrophil slow rolling and adhesion . Further experiments in mice showed that this ability of CD95 to recruit neutrophils to inflamed sites was crucial for clearing bacteria in cases of sepsis , where infection causes the immune system to damage the body’s own tissues . Future studies could address whether inhibiting CD95's activity could help to treat diseases that feature uncontrolled white blood cell recruitment , including various cancers and autoimmune diseases .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2016	Endothelial cell-derived CD95 ligand serves as a chemokine in induction of neutrophil slow rolling and adhesion
The presynaptic protein complexin ( CPX ) is a critical regulator of synaptic vesicle fusion , but the mechanisms underlying its regulatory effects are not well understood . Its highly conserved central helix ( CH ) directly binds the ternary SNARE complex and is required for all known CPX functions . The adjacent accessory helix ( AH ) is not conserved despite also playing an important role in CPX function , and numerous models for its mechanism have been proposed . We examined the impact of AH mutations and chimeras on CPX function in vivo and in vitro using C . elegans . The mouse AH fully restored function when substituted into worm CPX suggesting its mechanism is evolutionarily conserved . CPX inhibitory function was impaired when helix propagation into the CH was disrupted whereas replacing the AH with a non-native helical sequence restored CPX function . We propose that the AH operates by stabilizing CH secondary structure rather than through protein or lipid interactions . Precise control of synaptic vesicle fusion at the presynaptic terminal endows a nervous system with the means to regulate functional synaptic connectivity . Although many of the molecules required for neurotransmitter exocytosis are known , the mechanisms by which this process is regulated are less well understood . The core machinery of the fusion process is composed of the neuronal SNAREs ( VAMP2 , syntaxin 1 , and SNAP-25 ) along with several SNARE-binding proteins such as synaptotagmin , Munc13 , and Munc18 ( Sudhof and Rothman , 2009; Sudhof , 2013 ) . These proteins are highly similar in sequence and function across species , reflecting the deep conservation of synaptic transmission . Complexin is another essential SNARE-binding protein , and genetic ablation of complexin profoundly impacts neurotransmitter release and its regulation in all synaptic preparations studied to date ( Reim et al . , 2001; Huntwork and Littleton , 2007; Xue et al . , 2008; Hobson et al . , 2011; Martin et al . , 2011; Lin et al . , 2013; Vaithianathan et al . , 2013 ) . Mammals possess four isoforms of complexin , and deletion of the two broadly expressed isoforms ( mCpx1/2 ) is lethal ( Reim et al . , 2001; Xue et al . , 2008 ) . This small cytoplasmic protein possesses an alpha-helical SNARE-binding domain known as the central helix ( CH ) , which is broadly conserved among metazoa ( Pabst et al . , 2000; Reim et al . , 2001; Bracher et al . , 2002; Chen et al . , 2002; Brose , 2008 ) . The CH domain of mCpx1 is flanked on its N-terminus by a stable alpha helix called the accessory helix ( AH ) ( Pabst et al . , 2000; Chen et al . , 2002 ) , and this domain has subsequently been shown to play an inhibitory role in mammalian , fly , and nematode synapses ( Xue et al . , 2007 , 2009; Yang et al . , 2010; Martin et al . , 2011; Cho et al . , 2014; Trimbuch et al . , 2014 ) . However , the primary sequence of the AH domain is poorly conserved , and its secondary structure has only been investigated in rodent complexin . A wide variety of models for AH function have been proposed including direct binding to SNAREs or other proteins ( Giraudo et al . , 2009; Lu et al . , 2010; Yang et al . , 2010; Kummel et al . , 2011; Bykhovskaia et al . , 2013; Cho et al . , 2014 ) , electrostatic interactions with membranes ( Trimbuch et al . , 2014 ) , and direct effects on the CH and its SNARE binding through secondary structure interactions ( Chen et al . , 2002 ) . Do all of these mechanisms contribute to AH domain function ? If the AH domain binds specifically to another protein , why is it so poorly conserved relative to the CH domain ? To investigate the mechanism of AH action and its conservation across phylogeny , we examined the AH domain structure and function in the Caenorhabditis elegans mCpx1/2 ortholog CPX-1 using both in vitro and in vivo approaches . While worm and mouse AH domains are two of the most divergent among published complexin sequences , the two domains could be exchanged without impairing function in vivo . Further , the recombinant worm AH formed a highly stable alpha helix in solution similar to the mouse AH . Abolishing the hydrophobic character of the worm or mouse AH domain had little effect , whereas disrupting helix stability and invasion of helical structure into the CH severely impaired complexin inhibitory function . Moreover , replacing the AH with an artificial helical sequence fully restored inhibitory function . Remarkably , this sequence was functional despite large differences in length , charge , and hydrophobicity , indicating that these properties are not critical for AH function . These experiments indicate that the principal role of the AH domain is to nucleate and propagate helical structure into the CH domain , and this function is conserved across evolution . A region of the alpha helical domain of mouse complexin binds tightly to the assembled SNARE bundle ( Figure 1A ) , positioned in the groove formed by synaptobrevin and syntaxin ( Bracher et al . , 2002; Chen et al . , 2002 ) . This so-called ‘central helix’ ( CH ) is deeply conserved across phylogeny ( 76% identity between mammals and nematodes ) , whereas the adjacent helical sequence corresponding to the mouse accessory helix ( AH ) is much more heterogeneous , with only 20% identity between mammals and nematodes based on the 18 residues N-terminal to CH ( Figure 1B ) . To compare AH function in mouse and worm , we first established whether the worm complexin protein CPX-1 possesses a stable helical region adjacent to its CH analogous to mouse complexin ( Pabst et al . , 2000; Bracher et al . , 2002; Chen et al . , 2002 ) . Computational predictions based on amino acid sequence indicated a highly stable helical region including 30 residues between P37 and G66 encompassing the AH and half of the CH domain ( Figure 1C , D ) ( Munoz and Serrano , 1997 ) . This domain was confirmed to be helical by solution-state NMR spectroscopy on recombinant full-length worm CPX-1 ( Snead et al . , 2014 ) as well as on a truncated version lacking its C-terminal domain ( Figure 1C , D ) , validating the computational predictions for this protein . The stable helical structure of the AH domain is predicted to be deeply conserved across phylogeny based on the analysis of complexin sequences in 16 diverse metazoan species from seven phyla ranging from Trychoplax to human ( Figure 1—figure supplements 1–2 ) . Thus both mouse and worm complexin possess a stable alpha helical domain N-terminal to the CH despite sharing little sequence homology , and this is likely to be a universal feature of complexin . 10 . 7554/eLife . 04553 . 003Figure 1 . The worm AH domain forms a stable helix and this structure is deeply conserved across phylogeny . ( A ) Ribbon diagram of the mammalian complexin-SNARE crystal structure ( Chen et al . , 2002 ) using PDB code 1KIL . Cytoplasmic SNARE domains from synaptobrevin ( red ) , Syntaxin ( yellow ) , and SNAP-25 ( green ) . Mouse Cpx1 ( residues 26–83 ) is divided into the accessory helix ( AH–orange ) and central helix ( CH–blue ) . ( B ) Sequence alignment of the accessory helix ( orange ) and central helix ( blue ) for C . elegans CPX-1 ( worm ) and M . musculus Cpx1 ( mouse ) . Amino acid identity indicated with gray squares in between the sequences . Helix-breaking prolines indicated in green . ( C ) Cα-Cβ shifts from a truncated worm CPX-1 peptide missing the C-terminal domain ( residues 1–77 , black circles ) are compared with Agadir predictions for the Cα shifts ( red diamonds ) . The Agadir predicted Cα shift values at residues E38 and V39 were normalized to the experimentally determined Cα-Cβ shift values at those sites ( blue bracket ) , allowing for a comparison of AH and CH shift predictions . ( D ) Predicted helical state of each residue ( Agadir ) is shown for CPX-1 . ( E ) Summary of Agadir helix prediction for the AH domain ( defined by the average helicity of 18 residues N-terminal to the CH domain ) across 16 species: Trichoplax adhaerens Ta , Mnemiopsis leidyi Ml , Caenorhabditis elegans Ce ( orange ) , Caenorhabditis briggsae Cb , Drosophila melanogaster Dm , Anopheles gambiae Ag , Loligo pealei Lp , Hirudo medicinalis Hm , Narke japonica Nj , Ciona intestinalis Ci , Danio rerio Dr , Xenopus laevis Xl , Gallus gallus Gg , Ornithorhynchus anatinus Oa , Mus musculus Mm ( pink ) , Homo sapiens Hs . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 00310 . 7554/eLife . 04553 . 004Figure 1—figure supplement 1 . Alpha helical regions of complexin across phylogeny . Alignments of the alpha helical regions of complexin across 16 species from seven phyla using Clustal Omega multiple sequence alignment . The accessory helix region ( orange ) was defined based on Agadir prediction of at least 5% helicity . The central helix domain is shaded in blue . Species are indicated on right and by a two letter code used throughout the paper ( left ) . Sequences derived from Ensembl database ( Flicek et al . , 2014 ) and aligned using Clustal Omega ( EMBL-EBI ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 00410 . 7554/eLife . 04553 . 005Figure 1—figure supplement 2 . Evolutionary conservation of helicity across >0 . 5 billion years . ( A ) Agadir helicity predictions for human ( Hs , blue ) , ctenophore ( Ml , pink ) , placazoa ( Ta , green ) , and worm ( Ce , dashed line ) complexin homologs , plotted relative to the first residue of the central helix ( position #1 ) . ( B ) . Average domain helicity from ( A ) using AH ( orange , residues −19 to 0 ) and CH ( blue , residues 1 to 25 ) domains for the four species shown in A . ( C ) Percent sequence identity relative to human mCpx1 for the AH ( orange ) and CH ( blue ) domains for the three invertebrate complexins . Note that the placazoan Trichoplax does not possess a nervous system . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 005 To investigate the functional significance of the AH domain in worm complexin , two deletion variants of CPX-1 missing either 12 ( ΔAHshort ) or 20 ( ΔAHlong ) residues of the AH were expressed in cpx-1 mutants ( Figure 2A ) . All transgenic rescue experiments reported in this study utilized a functional CPX-1::GFP fusion protein , and transgene expression was monitored by imaging neuromuscular junction ( NMJ ) fluorescence in living intact animals ( Martin et al . , 2011; Wragg et al . , 2013 ) ( Figure 2—figure supplement 1 ) . Loss of CPX-1 caused a 12-fold increase in the rate of spontaneous fusion in the absence of external calcium at cholinergic NMJs ( Hobson et al . , 2011; Martin et al . , 2011; Wragg et al . , 2013 ) ( Figure 2B , C ) . Rescue with full-length CPX-1 completely restored the basal synaptic vesicle ( SV ) fusion rate ( 3 . 3 ± 0 . 9 Hz for wild type , 2 . 7 ± 0 . 5 Hz for FL rescue ) whereas rescue with the ΔAHshort variant only partially reversed the increased rate of spontaneous fusion ( 39 . 8 ± 6 . 2 Hz for cpx-1 , 20 . 8 ± 5 . 6 Hz for ΔAHshort rescue ) . Thus CPX-1 retained a limited ability to inhibit fusion in the absence of the AH domain . In all cases , the muscle miniature EPSC amplitude was unaffected ( Figure 2D ) , consistent with the neuronal expression of CPX-1 in C . elegans ( Martin et al . , 2011 ) . cpx-1 mutants exposed to the cholinesterase inhibitor aldicarb paralyzed more rapidly than wild-type animals ( Figure 2E , F ) as described previously ( Hobson et al . , 2011; Martin et al . , 2011; Wragg et al . , 2013 ) . Consistent with the NMJ recordings , rescue with ΔAHshort or ΔAHlong only partially restored wild-type sensitivity whereas full-length CPX-1 completely rescued aldicarb sensitivity ( Figure 2F ) . The helix-breaking proline in position 37 is shared by all of the published nematode genomes , so the short AH domain is a common feature of this phylum . However , deletion of proline 37 does not impair CPX-1 inhibitory function , indicating that the short AH domain is not an essential feature of nematode complexin structure ( data not shown ) . Taken together , these structural and functional results demonstrate that , despite the lack of sequence conservation between worm and mouse AH , the worm AH domain comprises a highly stable alpha helix that plays a major role in CPX-mediated inhibition of spontaneous SV fusion . 10 . 7554/eLife . 04553 . 006Figure 2 . The worm AH contributes to CPX-1 inhibition of spontaneous vesicle fusion . ( A ) Two deletions within the worm AH domain were used: ΔAHshort ( 35–49 , red ) and ΔAHlong ( 30–50 , aqua ) . ( B ) Examples of spontaneous EPSCs in zero external Ca2+ for wild-type , cpx-1 , and transgenic animals expressing full-length CPX-1 ( FL rescue ) and the short AH deletion ( ΔAHshort rescue ) . Average spontaneous EPSC Rate ( C ) and EPSC amplitude ( D ) for the genotypes indicated in B . Data are mean ± SEM and the number of independent assays is indicated for each genotype . Using Tukey–Kramer statistics for multiple comparisons , ** denotes significantly different from wild type , # significantly different from cpx-1 but not wild type , * significantly different from both wild type and cpx-1 ( p < 0 . 01 ) . ( E ) Cartoon of aldicarb acting at the worm cholinergic neuromuscular junction . Acetylcholine ( ACh , red ) is hydrolyzed by cleft cholinesterases ( AChE , green ) . Aldicarb inhibits AChE causing an elevation in ACh and eventual paralysis depending on the level of exocytosis . ( F ) Paralysis time course on 1 mM aldicarb for wild-type ( black filled circles , n = 36 ) , cpx-1 ( pink filled circles , n = 10 ) , full-length rescue CPX-GFP ( gray open circles , n = 10 ) , ΔAHshort rescue ( red open diamonds , n = 10 ) , and ΔAHlong rescue ( aqua open diamonds , n = 10 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 00610 . 7554/eLife . 04553 . 007Figure 2—figure supplement 1 . Axonal protein abundance for CPX-1 transgenes . ( A ) Schematic of a worm depicting the region imaged for axonal expression ( pink box ) . All rescuing transgenes were quantified by imaging the C-terminal GFP tag in single axons of immobilized intact animals . ( B ) Representative confocal image of CPX-1::GFP fluorescence in a dorsal cord axon of an intact animal . Scale bar is 5 microns . ( C ) Average axonal fluorescence values for transgenic animals used in this study . Fluorescence was background-subtracted and averaged for at least 20 animals for each strain ( number indicated in the bars ) . All data were normalized to a full-length wild-type CPX-1::GFP rescue strain and plotted as mean ± SEM . The blue dotted lines represent the range of expression found to rescue aldicarb sensitivity based on prior studies using a previously described axonal expression quantification protocol ( Wragg et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 007 The similar structures of worm and mouse complexin AH domains suggest that AH function could be conserved between these highly divergent species . Indeed , several studies have indicated that mouse and fly AH domains contribute to an inhibitory activity of complexin ( Xue et al . , 2007; Giraudo et al . , 2008; Chicka and Chapman , 2009; Giraudo et al . , 2009; Kummel et al . , 2011; Cho et al . , 2014; Lai et al . , 2014; Trimbuch et al . , 2014 ) . However , the AH may operate through distinct mechanisms in distantly related species . To test for conservation of mechanism , the endogenous worm AH was replaced with the mouse AH and expressed in cpx-1 mutants . Since the length of the AH is not identical between worm and mouse , two versions of the chimeric complexin were generated: a longer version encompassing mouse residues 24–47 replaced the worm residues 26–49 ( mouse AH ) , and a shorter AH swap replaced 38–49 with mouse residues 36–47 ( mouse AHshort ) ( Figure 3A ) . Recordings from the NMJs of cpx-1 mutants expressing the mouse AH demonstrated that CPX-1 is fully functional with a mouse AH domain ( Figure 3B , C ) . Both mouse AH and mouse AHshort restored near wild-type aldicarb sensitivity , further indicating that the mouse AH is functional in the context of worm CPX-1 ( Figure 3D ) . The precise length of the AH is not conserved across species ( Figure 1—figure supplement 1 ) , and the spacing between the N-terminal domain and the CH is not likely to play a critical role in CPX-1 function since the NTD can be deleted without significantly impairing CPX-1 inhibitory function in worm ( Hobson et al . , 2011; Martin et al . , 2011 ) . 10 . 7554/eLife . 04553 . 008Figure 3 . Mouse AH functions in worm CPX-1 and neither domain requires hydrophobic residues . ( A ) Two chimeric CPX-1 constructs substituting worm AH with the mouse AH . The long form ( mouse AH ) substitutes worm residues 26–49 with mouse residues 24–47 whereas the short form ( mouse AHshort ) substitutes worm residues 38–49 with mouse residues 36–47 . Average spontaneous mEPSC rates ( B ) and amplitudes ( C ) for wild-type , cpx-1 , and transgenic rescue of cpx-1 expressing the mouse AH chimera ( mouse AH , blue ) . ( D ) Average paralysis time course in 1 mM aldicarb for wild-type ( black filled circles ) , cpx-1 ( pink filled circles ) , full-length rescue ( gray open circles ) , mouse AH rescue ( blue open diamonds ) , and mouse AHshort rescue ( orange open diamonds ) . ( E ) Worm and mouse AH domain residues shown on a helical wheel diagram with hydrophobic residues indicated in red and hydrophobic substitutions with glutamate indicated with green arrow heads . ( F ) Percent rescue of wild-type paralysis kinetics using the time to 50% paralysis ( t0 . 5 ) for the five genotypes shown in D as well as a worm CPX-1 variant with V39E , I40E , L44E substitutions ( w[h→E] ) and a mouse AHshort variant with A40E , L41E , A44E substitutions ( m[h→E] ) . Note that the residue positions are labeled based on their location in CPX-1 and mCpx1 respectively . Data are mean ± SEM with sample sizes as indicated on the bar graphs . Using Tukey–Kramer statistics for multiple comparisons , ** denotes significantly different from wild type , # significantly different from cpx-1 but not wild type , * significantly different from both wild type and cpx-1 ( p < 0 . 01 ) , n . s . is not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 00810 . 7554/eLife . 04553 . 009Figure 3—figure supplement 1 . Conservation of hydrophobic moments of complexin . ( A ) The AH hydrophobic moment was computed with angular orientation relative to the first residue of the CH . ( B ) Example of a hydrophobic moment for a particular AH ( worm ) . ( C ) Polar plot of the hydrophobic moments for 16 species ( red ) and rescuing constructs ( turquoise ) . All species besides Trychoplax cluster in a small range of orientations ( yellow ) . The moment of the worm CH is shown in green . The 16 species shown are Trichoplax adhaerens Ta , Mnemiopsis leidri Ml , Caenorhabditis elegans Ce , Caenorhabditis briggsae Cb , Drosophila melanogaster Dm , Anopheles gambiae Ag , Loligo pealei Lp , Hirudo medicinalis Hm , Narke japonica Nj , Ciona intestinalis Ci , Danio rerio Dr , Xenopus laevis Xl , Gallus gallus Gg , Ornithorhynchus anatinus Oa , Mus musculus Mm , and Homo sapiens Hs . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 00910 . 7554/eLife . 04553 . 010Figure 3—figure supplement 2 . Conservation of charge density in the AH domain . The helical region of the AH domain was defined using Agadir with a criterion of >5% predicted helicity for a continuous region N-terminal to the CH domain . The net formal charge within this stretch was divided by the number of residues to calculate a charge density and plotted as charge density/AH for 16 species . C . elegans and M . musculus ( pink ) are the species investigated in this study . The 16 species shown are Trichoplax adhaerens Ta , Mnemiopsis leidri Ml , Caenorhabditis elegans Ce , Caenorhabditis briggsae Cb , Drosophila melanogaster Dm , Anopheles gambiae Ag , Loligo pealei Lp , Hirudo medicinalis Hm , Narke japonica Nj , Ciona intestinalis Ci , Danio rerio Dr , Xenopus laevis Xl , Gallus gallus Gg , Ornithorhynchus anatinus Oa , Mus musculus Mm , and Homo sapiens Hs . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 010 We next examined the phylogenetic conservation among several AH domains assuming that shared features are more likely to indicate conservation of AH domain function . Comparing AH domain properties in 16 species across seven phyla , three conserved features were apparent: a region of hydrophobic residues along the helix ( Figure 3—figure supplement 1 ) , stable helicity ( Figure 1E ) , and high negative charge density ( Figure 3—figure supplement 2 ) . One recent model of AH function proposes that an interaction between the hydrophobic AH residues and the tSNARE complex prevents full SNARE assembly thereby inhibiting SV fusion ( Krishnakumar et al . , 2011; Kummel et al . , 2011 ) . To test this model in C . elegans , several hydrophobic residues in the AH domain ( Figure 3E ) were replaced with the charged residue , glutamate ( h→E ) and this CPX-1 variant was expressed in cpx-1 mutants . As shown in Figure 3F , the AH ( h→E ) was fully functional in the absence of all hydrophobic side chains . Moreover , the short mouse AH chimera was also functional even when its hydrophobic residues were replaced with glutamates . Similar observations have been reported at the fly NMJ using mouse Cpx1 ( Cho et al . , 2014 ) . These findings indicate that hydrophobic residues in the AH domain do not play a critical conserved role in complexin function . Previous studies on mouse complexin suggested that the helical structure of the AH domain is important for stabilizing the CH domain ( Pabst et al . , 2000; Chen et al . , 2002 ) and for the inhibitory function of mCpx1 ( Xue et al . , 2007 ) , but the reason for this requirement remains unclear . While several potential roles for the AH in mediating protein–protein interactions have been proposed , the AH may simply serve to nucleate and propagate helical structure into the CH region ( Chen et al . , 2002 ) , but this idea has never been tested . To explore this possibility , a helix-breaking proline was inserted into the AH domain ( R43P ) , and the effects on AH domain secondary structure of a recombinant truncated form of the mutant protein missing its C-terminal domain ( ΔCT ) were examined by NMR spectroscopy ( Figure 4A ) . Furthermore , because conversion from random coil to alpha helix is a highly cooperative process , helicity in the CH domain is also predicted to decrease for the R43P mutant ( Munoz and Serrano , 1997 ) . Indeed , decreased NMR carbon secondary shifts were observed throughout the AH and extending well into the CH domain , confirming decreases in both nucleation and propagation of the helical conformation ( Figure 4B ) . Circular dichroism ( CD ) spectroscopy provides another measure of overall alpha helical structure ( Greenfield and Fasman , 1969; Saxena and Wetlaufer , 1971; Rohl and Baldwin , 1997 ) . Absorption at 222 nm was monitored in recombinant ΔCT protein while titrating in 2 , 2 , 2-trifluoroethanol ( TFE ) , a co-solvent known to stabilize alpha helices in solution ( Nelson and Kallenbach , 1986; Segawa et al . , 1991; Shiraki et al . , 1995 ) . The increase in alpha helical structure with increasing concentration of TFE can be used to measure the stability and cooperativity of alpha helix formation . While some of the cooperativity arises from coordination of multiple TFE molecules ( Berkessel et al . , 2006 ) , intramolecular propagation of helical structure will also contribute . As shown in Figure 4C , the R43P variant displayed a lower propensity for helix formation with a lower cooperativity , consistent with both decreased helix nucleation and propagation , as also evident from both the computational predictions and the NMR data . In living worms , inserting a proline into the AH domain completely eliminated AH function since the CPX ( R43P ) rescue was indistinguishable from the ΔAHshort rescue in both electrophysiological ( Figure 4D , E ) and behavioral ( Figure 4F ) assays of synaptic function . Thus , inhibition of spontaneous fusion requires a helical AH domain in vivo . 10 . 7554/eLife . 04553 . 011Figure 4 . Disrupting AH helix stability impairs CPX-1 inhibitory function . ( A ) NMR derived Cα-Cβ shifts from either wild-type ( black ) or R43P ( blue ) worm CPX-1 peptide missing the C-terminal domain . R43P is indicated in red . Below , the predicted helical content using Agadir for wild-type ( black ) , and R43P complexin ( blue ) . ( B ) Average secondary chemical shift for wild-type ( black ) and R43P ( blue ) complexin either over the entire peptide ( residues 1–77 , left ) , AH domain ( 37–49 , middle ) , or CH domain ( 50–74 , right ) . The average helical content was estimated by dividing the chemical shift by 3 . 4 ( average shift of 100% helical peptide ) . The helical content was also measured by CD spectroscopy ( red arrowheads ) . ( C ) Helical content for wild type ( black ) and R43P ( blue ) complexin was measured by CD spectroscopy for increasing concentrations of 2 , 2 , 2-trifluoroethanol ( TFE ) . The resulting dose–response data was fit to a simple equilibrium binding curve with equilibrium constants and Hill coefficients indicated on the graph . Average spontaneous EPSC Rate ( D ) and EPSC amplitude ( E ) for wild-type , cpx-1 , and either the ΔAHshort or R43P rescuing transgene expressed in cpx-1 as indicated . ( F ) Sensitivity to aldicarb was quantified by the average time to 50% paralysis and then normalized to wild-type and cpx-1 mutant animals . On this scale , rescue with ΔAHshort or R43P variants of CPX-1 partially restored wild-type aldicarb sensitivity . Data are mean ± SEM and the number of independent assays is indicated for each genotype . Using Tukey–Kramer statistics for multiple comparisons , ** denotes significantly different from wild type , * significantly different from both wild type and cpx-1 ( p < 0 . 01 ) , n . s . is not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 011 These results provide support for a model in which nucleation of an alpha helix within the AH domain and propagation of this helix into the CH domain is required for CPX inhibitory function . Three predictions arise from this hypothesis: First , there should be a direct correlation between helical stability and inhibitory function . Second , disruption of helix propagation into the CH domain should impair CPX inhibition . And third , replacement of the AH with a non-native alpha helical domain should functionally substitute for the endogenous AH sequence and stabilize the CH domain . The first prediction was examined by creating a series of CPX variants predicted computationally to feature decreasing helical stability ( Figure 5A , B ) . Helical stability dropped monotonically as the residue at position 43 changed from R → V → F → P . Notably , the predictions also indicate a parallel decrease in the helicity of the CH helix , consistent with the nucleation/propagation hypothesis . This AH series was expressed in cpx-1 mutants , and rescue was quantified using aldicarb sensitivity . Supporting this hypothesis , transgenic animals expressing this series of CPX-1 variants display a level of aldicarb sensitivity that mirrors the helical stability ( Figure 5C , D ) . To test the second prediction , helical propagation was blocked by inserting a Gly–Gly ( GG ) between the AH and CH domains . Because glycines destabilize helical structure , GG insertions are expected to disrupt helical propagation ( O'Neil and DeGrado , 1990; Blaber et al . , 1993 ) as supported by the computational predictions of AH and CH helical content . Rescue of cpx-1 mutants with the GG variant demonstrated that insertion of a GG between the AH and CH was equivalent to deleting the entire AH domain ( Figure 5E ) . This disruption is not due to a shift in the orientation of AH domain hydrophobic residues ( a consequence of inserting two residues ) since inserting three glycines produced a similar loss of function ( data not shown ) . Thus helix propagation from the AH domain to the CH domain appears to play a major role in CPX-1 inhibitory function , in agreement with a similar finding at the fly NMJ ( Cho et al . , 2014 ) . 10 . 7554/eLife . 04553 . 012Figure 5 . Stability of the AH and its propagation into the CH domain are required for CPX-1 inhibitory function . ( A ) AH and CH helical content based on Agadir predictions is plotted for wild-type CPX-1 and three variants with a single substitution at residue 43 as indicated . ( B ) Average helical content of the AH domain ( orange ) , CH domain ( blue ) , or entire helical region ( gray ) for four residues in position 43: Arg , Val , Phe , and Pro . ( C ) Average paralysis time course on one millimolar aldicarb for wild-type ( black filled circles ) and cpx-1 animals ( pink filled circles ) , as well as four rescuing CPX-1 transgenes: wild type CPX-1 ( R , gray open circles ) , R43V ( V , blue diamonds ) , R43F ( F , green diamonds ) , and R43P ( P , red diamonds ) . ( D ) Normalized rescue of the t0 . 5 for paralysis for each of the four transgenic rescue strains is plotted vs the predicted helicity for either the AH domain ( orange ) , the CH domain ( blue ) , or the entire helical region ( gray ) . ( E ) Normalized rescue of aldicarb sensitivity ( t0 . 5 ) for wild-type and cpx-1 animals as well as three transgenic rescue strains: full-length wild-type CPX-1 ( FL ) , R43P , the long AH deletion ( ΔAH ) , and the GG insert in between the AH and CH domains ( GG insert ) . Data are mean ± SEM with n = 10 experiments for all strains except wild type ( n = 36 ) . *p < 0 . 01 different from wild-type , # is p < 0 . 01 different from cpx-1 but not wild type . p < 0 . 01 different from wild type and cpx-1 animals . Significance was determined by Tukey–Kramer method . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 012 As a third test of the helix nucleation/propagation hypothesis , the endogenous AH domain was replaced with an artificial alpha helix based on a Glu-Ala-Ala-Lys ( EAAK ) motif repeated seven times ( Figure 6A ) . This design was predicted to form a stable alpha-helical structure in solution ( Figure 6B ) ( Marqusee and Baldwin , 1987 ) . CD spectroscopy of TFE titrations on recombinant CPX-1 ( ΔCT ) containing the 7-turn helix revealed that this variant is more helical than wild type ( Figure 6C ) . The increased helicity values could arise because 24 residues of the CPX-1 AH were replaced with 30 residues of helical EAAK repeat , so a larger fraction of the whole polypeptide will necessarily be helical . As an independent measure of helical stability , we compared the helix-coil equilibrium constants derived from TFE titrations and found that the 7-turn polypeptide forms a more stable helix than wild-type CPX-1 ( 17 . 9% for 7-turn vs 32 . 4% for wild type ) . Thus , the 7-turn artificial alpha helix provides a stable helical substitute for the endogenous AH domain irrespective of its increased length . Additionally , the 7-turn construct displayed a lower apparent cooperativity than wild-type CPX-1 ( 2 . 8 for 7-turn vs 5 . 6 for wild type ) , but the true cooperativity of the 7-turn polypeptide was underestimated since it was quite helical even at 0% TFE ( Figure 6C ) . Surprisingly , the 7-turn helix fully rescued both aldicarb sensitivity and suppression of spontaneous fusion at cholinergic NMJs ( Figure 6D–F ) . Moreover , the AH ( 7-turn ) sequence is electrostatically neutral overall , yet it fully restored CPX inhibition . Therefore , the conserved negative charge density of the AH domain was not essential for CPX inhibition . In fact , the charge density of the rescuing transgenes used in this study generally did not correlate with their function in vivo ( Figure 6—figure supplement 1 ) . The EAAK motif also creates a strong hydrophobic moment , but the orientation of the moment is rotated approximately 90° relative to wild-type CPX-1 , indicating that AH function is not sensitive to the positioning of the hydrophobic residues ( Figure 6—figure supplement 2 ) . Together with the progressive AH destabilizing substitutions and GG insertion , these results reveal that a critical feature of AH function is the nucleation and invasion of a stable alpha helix into the CH domain . 10 . 7554/eLife . 04553 . 013Figure 6 . The AH domain can be functionally replaced by a non-native helix . ( A ) Schematic of the helix substitution strategy . A ( EAAK ) 7A sequences were substituted for residues 26–49 in the worm AH domain . ( B ) Agadir prediction for helical stability of the 7-turn helix motif compared to wild-type AH . ( C ) Helical content for the 7-turn construct was measured by CD spectroscopy , and the TFE dose–response data were fit as in Figure 4 . Average spontaneous EPSC Rate ( D ) and EPSC amplitude ( E ) for wild-type , cpx-1 , and the 7-turn rescuing transgene expressed in cpx-1 as indicated . ( F ) Sensitivity to aldicarb was quantified by monitoring the average time to 50% paralysis normalized to wild-type and cpx-1 mutant animals , and plotted for full-length wild-type CPX-1 ( FL ) , 5-turn substitution ( 5 Turn ) , and 7-turn substitution variants expressed in cpx-1 mutants . Data are mean ± SEM and the number of independent assays is indicated for each genotype . Using Tukey–Kramer statistics for multiple comparisons , ** denotes significantly different from wild type , # significantly different from cpx-1 but not wild type , * significantly different from both wild type and cpx-1 ( p < 0 . 01 ) , n . s . is not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 01310 . 7554/eLife . 04553 . 014Figure 6—figure supplement 1 . No correlation between charge density and function . Charge density of the AH domain in seven rescuing CPX-1 variants . Green indicates nearly complete rescue , gray is partial , and red is poor rescue as measured by aldicarb sensitivity . The CPX-1 variants are mouse AHshort ( SM ) , worm AH with hydrophobic to glutamates ( w[h→E] ) , mouse AH with hydrophobic to glutamates ( m[h→E] ) , 7-turn helix substitution ( 7T ) , as well as the R43P , R43F , and R43V substitutions . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 01410 . 7554/eLife . 04553 . 015Figure 6—figure supplement 2 . Hydrophobic moments of functional and nonfunctional CPX transgenes . Polar plot of hydrophobic moments for the AH domain of CPX-1 variants . The CPX-1 variants ( blue ) are mouse AHshort ( SM ) , worm AH with hydrophobic to glutamates ( w[h→E] ) , mouse AH with hydrophobic to glutamates ( m[h→E] ) , 7-turn helix substitution ( 7T ) , as well as the R43P , R43F , and R43V substitutions . The moment of the worm CH is shown in green . 7T construct designated with a pink arrow . The hydrophobic moments of 16 species from Figure 3—figure supplement 1 are included for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 04553 . 015 Why compare complexins from multiple divergent species ? Several features of the AH domain appear to be conserved across more than half a billion years of evolution based on the species examined here , so an evolutionary comparison provides insight into critical aspects of the underlying mechanisms . Beyond their primary sequence , the two most striking conserved features of the AH domain are its high negative charge density and the distribution and orientation of its hydrophobic residues ( Figure 3—figure supplements 1–2 ) . Although the deep conservation of charge density and hydrophobic moment suggests that they play a role in complexin function , the inhibition of spontaneous fusion does not require either feature in worm ( this study ) and hydrophobicity is superfluous in fly as well ( Cho et al . , 2014 ) . Perhaps other roles of complexin utilize these properties . In contrast to charge and hydrophobicity , a stable helix is both deeply conserved and required for complexin inhibitory function . How conserved is helicity vs primary protein sequence ? The SNARE-binding central helix is the defining motif of complexin homologs based on primary protein sequences ( Pabst et al . , 2000; Brose , 2008 ) . Two of the most distantly related complexin genes reported belong to the placozoan Trichoplax adherens ( Ta ) and the ctenophore Mnemiopsis leidyi ( Ml ) ( Flicek et al . , 2014 ) . Interestingly , the CH domains of these representatives of basal animal phyla , share 44% identity with the human Cpx1 CH . However , whereas the AH domain of Ta shares 40% identity with the human AH domain , there is only 10% sequence conservation between Ml and human AH domains . The predicted AH and CH structures for human and Ml both contain extended regions of stable helix propagation similar to worm , whereas Ta is predicted to have only a modest degree of stable helical structure on its own ( Figure 1—figure supplement 2 ) . This provides an evolutionary example of how primary sequence and helicity do not necessarily change in parallel in the AH domain . A speculative explanation for the noticeable increase in AH domain helical stability progressing from ctenophore ( 36% ) to worm ( 54% ) to mammal ( >80% in human ) is the higher body temperatures of warm-blooded animals compared to soil- and marine invertebrates . Helical stability is highly dependent on ambient temperature , so the higher body temperatures of mammals and many other vertebrates may necessitate more stable helical sequences ( Privalov , 1982; Scholtz and Baldwin , 1992 ) . Inhibition by complexin depends on the integrity of its AH domain in all studies where this domain has been examined ( Xue et al . , 2007; Giraudo et al . , 2008; Chicka and Chapman , 2009; Giraudo et al . , 2009; Xue et al . , 2009; Cho et al . , 2010; Yang et al . , 2010; Krishnakumar et al . , 2011; Kummel et al . , 2011; Martin et al . , 2011; Cho et al . , 2014 ) . Several models for AH-domain function have been proposed over the past few years . Based on the unusual stability of the helical structure of the AH motif noted in early studies of the complexin-SNARE complex , the AH has been speculated to nucleate and propagate helical structure into the SNARE-binding CH ( Chen et al . , 2002 ) . Several groups have proposed that the AH domain of mouse Cpx1 competes with the C-terminal region of VAMP2 for a binding site on the tSNAREs thereby preventing full SNARE assembly ( Giraudo et al . , 2009; Lu et al . , 2010 ) . The AH is also thought to mediate an intermolecular interaction between neighboring SNARE complexes , and this trans Cpx/SNARE array was proposed to underlie complexin-mediated inhibition based on in vitro assays of membrane fusion ( Krishnakumar et al . , 2011; Kummel et al . , 2011 ) . A recent study examining the trans Cpx/SNARE array model in the fly NMJ found that mutations expected to disrupt the hydrophobic interaction between the AH domain and the tSNAREs did not impair inhibition of spontaneous SV fusion in vivo ( Cho et al . , 2014 ) in agreement with the hydrophobic residue substitutions used in the present study . Based on molecular dynamics simulations , another model conjectures that the AH domain can form a tight complex with the SNARE bundle and can also stabilize a partially-assembled state by binding directly to the C-terminal region of VAMP2 ( Bykhovskaia et al . , 2013 ) . In contrast to these AH-SNARE interaction models , electrostatic repulsion between the AH domain and membranes has been suggested to be inhibitory ( Trimbuch et al . , 2014 ) . Thus a broad range of models and features have been put forward to describe AH domain function . Our results suggest that the mechanism of AH function is deeply conserved and relatively independent of the primary protein sequence , hydrophobicity , length , or net charge density , inconsistent with models that rely on specific AH-protein interactions or electrostatic AH-membrane interactions . However , the mechanistic effect of stabilizing the CH domain remains unclear . An appealing model is that a stabilized CH alpha helix interacts more efficiently with the assembling SNARE bundle thereby promoting the SNARE interaction required for complexin inhibition . Indeed , CH binding to SNAREs is reduced in the absence of the AH in mouse Cpx1 ( Xue et al . , 2007 ) . Alternatively , the CH domain could interact with an as yet unidentified synaptic protein , and this binding would then require stabilization by the AH . The models based on hydrophobic AH interactions or electrostatic contributions are not supported by the experiments described here , but these mechanisms may be more prominent in other species . Alternatively , the high negative charge and hydrophobic moment of the AH domain may be relevant for other complexin functions . Nevertheless , our data indicate that the nucleation and propagation function of the AH domain is universal for proper complexin function across species . Further exploration of the CH and its binding partners is required for a detailed mechanistic understanding of complexin inhibitory action at the synapse . Animals were maintained at 20°C on agar nematode growth media seeded with OP50 bacteria as previously described ( Brenner , 1974 ) . Strains employed in this study include: N2 Bristol and cpx-1 ( ok1552 ) . To measure aldicarb sensitivity , 20–30 young adult animals were placed on agar plates containing 1 mM aldicarb ( Watson International , China ) . Worms were scored for paralysis at ten minute intervals for 2 hr . Each genotype was coded , tested 10 times blindly , and the paralysis curves were generated by averaging paralysis time courses for each plate as described previously ( Dittman and Kaplan , 2008 ) . Percent rescue based on t0 . 5 was calculated by first interpolating the time at which 50% of the worms paralyzed for each trial , averaging the single-trial t0 . 5 values together , and then normalizing to 100% rescue for wild-type t0 . 5 and 0% rescue for cpx-1 t0 . 5 values according to:%Rstrain=100·t0 . 5[strain]−t0 . 5[cpx]t0 . 5[WT] To control for protein expression levels in the extrachromosomal arrays , animals were first immobilized using 2 , 3-butanedione monoxime ( 30 mg/ml , Alfa Aesar , Ward Hill , MA ) , mounted on 2% agarose pads , in M9 buffer ( 22 . 0 mM KH2PO4 , 42 . 3 mM Na2HPO4 , 85 . 6 mM NaCl , and 1 . 0 mM MgSO4 ) , and imaged on an inverted Olympus microscope ( IX81 ) , using a laser scanning confocal imaging system ( Olympus Fluoview FV1000 with dual confocal scan heads ) and an Olympus PlanApo 60× 1 . 42 NA objective . Rescuing complexin variants were C-terminally tagged with GFP separated by a 12 residue linker ( GGSGGSGGSAAA ) , and synaptic protein levels were estimated by measuring background-subtracted fluorescence between dorsal cord synaptic peaks . Data were analyzed with custom software in IGOR Pro ( WaveMetrics , Lake Oswego , OR; Burbea et al . , 2002; Dittman and Kaplan , 2006 ) . A fluorescent slide was imaged daily to monitor the laser stability and the dorsal cord fluorescence was normalized to the slide value . Whole-cell patch-clamp recordings were performed on dissected C . elegans as described previously ( Madison et al . , 2005; McEwen et al . , 2006 ) . Dissected worms were superfused in an extracellular solution containing 127 mM NaCl , 5 mM KCl , 26 mM NaHCO3 , 1 . 25 mM NaH2PO4 , 20 mM glucose , 1 mM CaCl2 and 4 mM MgCl2 , bubbled with 5% CO2 , 95% O2 at 20°C . Whole-cell recordings were carried out at −60 mV using an internal solution containing 105 mM CH3O3SCs , 10 mM CsCl , 15 mM CsF , 4 mM MgCl2 , 5 mM EGTA , 0 . 25 mM CaCl2 , 10 mM HEPES and 4 mM Na2ATP , adjusted to pH 7 . 2 using CsOH . Under these conditions we only observed cholinergic EPSCs . For low calcium experiments , 1 mM CaCl2 was replaced with additional MgCl2 for a total of 5 mM divalent cations . Protein expression constructs for NMR were cloned into a pET vector containing a His6 tag and SUMO cleavage site to facilitate purification . A truncated polypeptide lacking the C-terminal domain ( residues 1–77 ) was purified to simplify NMR and CD spectroscopic analysis . For NMR , BL21-DE3 E . coli cells were transformed and grown in Luria Broth ( LB ) containing 50 μg/ml Kanamycin to an optical density at 600 nm between 0 . 6 and 0 . 8 . Cells were pelleted at 6500 rpm for 15 min , washed and resuspended for 30 min in a minimal media containing 15N NH4Cl , 13C D-glucose , pelleted again , and resuspended in media containing 15N NH4Cl and 13C D-glucose prior to induction . To produce perdeuterated proteins , cells were grown directly in D2O-based minimal media containing 15N NH4Cl , 13C2H D-glucose to an optical density at 600 nm of 0 . 6–0 . 8 prior to induction . Cells were induced with 400 µg/ml isopropyl thiogalactopyranoside ( IPTG , OmniPur , Billerica , MA ) , and grown for three hours at 37°C . The cells were then pelleted , resuspended in lysis buffer ( 350 mM NaCl , 20 mM imidazole , 20 mM Tris pH 8 , 1 mM EDTA , 0 . 1 mM PMSF , 1 . 7 mM BME , and 2 mM DTT ) lysed by sonication , and pelleted at 40 , 000 rpm for 45 min . The supernatant was then bound to Ni-NTA beads and washed ( wash buffer: 350 mM NaCl , 20 mM imidazole , 20 mM Tris pH 8 , 1 . 7 mM BME , and 2 mM DTT ) . The protein was then eluted ( wash buffer with 250 mM imidazole ) , and fractions containing protein were combined and dialyzed overnight ( dialysis buffer: 20 mM Tris pH 8 , 150 mM NaCl , and 2 mM DTT ) . The dialyzed sample was incubated with a His6-tagged SUMO protease to cleave the His6 tag from CPX-1 . The cleaved sample was incubated with Ni-NTA beads again to separate the protein from His6–SUMO protease and the His6–SUMO tag . The protein-containing fractions were eluted with wash buffer ( as above ) , pooled , and dialyzed overnight into ddH2O and then lyophilized . Protein expression constructs for CD were prepared as above with the following modifications: E . coli were grown to an optical density in LB + kanamycin at 600 nm between 0 . 6 and 0 . 8 . Cells were then induced with IPTG , grown for four hours at 37°C , pelleted , and resuspended in lysis buffer ( as above ) . Following sonication , the same procedures were followed as described above . Lyophilized protein samples were resuspended to a final concentration of 50–75 µM in 100 mM NaCl , 50 mM PIPES pH 6 . 08 , and proton-nitrogen ( HSQC ) spectra as well as a standard set of heteronuclear triple resonance three-dimensional spectra were collected on a Varian Unity Inova 600 MHz ( Weill Cornell NMR Facility ) spectrometer equipped with a cryoprobe and additional 3D and 4D ( H ) N ( COCA ) NNH spectra were collected using a perdeuterated full-length construct on a Bruker Avance 800 MHz spectrometer with cryoprobe ( New York Structural Biology Center , NY ) . Previous backbone resonance assignments for full-length free wild-type complexin were confirmed and extended to achieve a completeness level of 93% . Backbone resonance assignments for free R43P complexin were obtained at a completeness level 84% . All spectra were collected at 20°C . Data were processed using NMRpipe and analyzed using NMRview ( Johnson and Blevins , 1994; Delaglio et al . , 1995 ) . Spectra were referenced indirectly to water . Protein concentrations were measured using a Bradford reagent ( Bio-Rad , Hercules , CA ) . Lyophilized protein samples were resuspended in 40 mM phosphate buffer containing 100 mM NaCl to between 1 and 2 mM . CD spectra were then taken from 250 nm to 200 nm on both Aviv 62DS and Aviv Model 410 instruments at 25°C . The 2 , 2 , 2- trifluoroethanol ( TFE ) experiments were performed by adding increasing volumes of TFE ( JT Baker ) added to the samples . Data were background subtracted , and averages of two sequential scans were computed . The process was repeated and averaged to generate the data displayed in Figures 4 , 6 . Percentage helicity from CPX-1 spectra was estimated by calculating the best fit to a linear combination of pure helix and pure random coil spectra ( Saxena and Wetlaufer , 1971 ) . The percent helicity was then plotted against the percentage of TFE used , and fits were obtained to the following equation:h=hmin+ ( hmax−hmin ) 1+ ( K/[TFE] ) nwhere h is the predicted % helicity , hmin is the minimum value of helicity observed , at 0% TFE , hmax is the maximum value of helicity , K is the dissociation constant for the equilibrium between the random coil + TFE and the alpha-helix conformation , [TFE] is the percentage of TFE , and n is the cooperativity . The hmin value was allowed to change for each protein , but the hmax value was kept constant at the wild-type protein's value . The 7-turn artificial helix construct possesses an intrinsically larger helical percentage at high [TFE] , so hmax was not held constant when fitting the TFE titration for this peptide . K and n were found using least squares minimization . Protein sequences were entered into the online helical prediction software , Agadir , http://agadir . crg . es/ using default settings ( Munoz and Serrano , 1997 ) . To calculate cross-phylogeny comparisons in Figure 1 , the full length protein was entered into Agadir , and the 18 amino acids N-terminal to the beginning of the central helix were averaged . To calculate the percent helicity and Cα shifts , only the residues corresponding to the protein constructs were entered into Agadir . Hydrophobic moments were calculated using an online tool available at http://rzlab . ucr . edu/scripts/wheel/wheel . cgi using standard interface hydrophobicity values for each residue ( Wimley and White , 1996; Wimley et al . , 1996 ) . To calculate charge density across the AH amino acids predicted to be at least 5% helical by Agadir , the number of negatively charged amino acids was subtracted from the number of positively charged amino acids and the difference was divided by the total number of residues . For all datasets in this study , statistical comparisons were made across the entire dataset using the Tukey–Kramer method for multiple comparisons with p < 0 . 01 as the significant criterion . Some of this data was then used in multiple figures . Average values for wild-type and cpx-1 mutant voltage-clamp recordings from the same dataset were shown in Figures 2–4 and Figure 6 . The aldicarb time course data for wild-type , cpx-1 , and full-length CPX-1 rescue animals is used in Figures 2–6 . Likewise , the aldicarb rescue data for ΔAHshort is displayed in Figures 2 , 4 , 5 while the R43P rescue data is used in Figures 4 , 5 . The wild-type CPX-1 ( ΔCT ) NMR chemical shift data are used in both Figure 1 and Figure 4 .	The nervous system sends information around the body in the form of electrical signals that travel through cells called neurons . These signals cannot pass across the small gaps—called synapses—that separate neighboring neurons . Instead , when electrical signals reach the synapse , chemicals called neurotransmitters are released across the gap and trigger an electrical signal in the next neuron . Neurotransmitters are stored within neurons in small envelopes of membrane known as synaptic vesicles . They are released when the vesicles fuse with the membrane that surrounds the neuron . This fusion process must be tightly controlled to ensure that information is passed between the neurons at the right time . Complexin is a small protein that controls vesicle fusion by binding to a group of proteins called the SNARE complex . It contains two structured sections called the central helix and the accessory helix , which are both important for vesicle fusion . The central helix is able to bind to the SNARE proteins , and it has the same sequence of amino acids—the building blocks of proteins—in all animals . However , the sequence of amino acids in the accessory helix varies widely across different animals and it is not clear whether it performs the same role in all of them . Radoff et al . studied complexin in the nematode worm C . elegans , and found that when its accessory helix is replaced with the amino acid sequence from the mouse one , it can still properly control vesicle fusion . Indeed , complexin can still work properly when its accessory helix is replaced with an artificial protein helix that has a similar shape . These experiments suggest that the overall structure of the accessory helix is more important than its exact sequence of amino acids . Radoff et al . propose that its role in vesicle fusion is to stabilize the structure of the central helix to allow it to bind to the SNARE proteins . The next challenge is to understand how vesicle fusion is prevented when complexin binds to the SNARE proteins .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics", "neuroscience" ]	2014	The accessory helix of complexin functions by stabilizing central helix secondary structure
As a first-line vertebrate immune defense , the polymeric immunoglobulin receptor ( pIgR ) transports polymeric IgA and IgM across epithelia to mucosal secretions , where the cleaved ectodomain ( secretory component; SC ) becomes a component of secretory antibodies , or when unliganded , binds and excludes bacteria . Here we report the 2 . 6Å crystal structure of unliganded human SC ( hSC ) and comparisons with a 1 . 7Å structure of teleost fish SC ( tSC ) , an early pIgR ancestor . The hSC structure comprises five immunoglobulin-like domains ( D1-D5 ) arranged as a triangle , with an interface between ligand-binding domains D1 and D5 . Electron paramagnetic resonance measurements confirmed the D1-D5 interface in solution and revealed that it breaks upon ligand binding . Together with binding studies of mutant and chimeric SCs , which revealed domain contributions to secretory antibody formation , these results provide detailed models for SC structure , address pIgR evolution , and demonstrate that SC uses multiple conformations to protect mammals from pathogens . The mucosa is fundamental to vertebrate survival , forming an elaborate extracellular environment , in which the immune system mediates host interactions with commensal and pathogenic agents . The human mucosa protects ~400 m2 of epithelial barriers in the gut , lungs , urogenital tract , and associated tissues such as mammary glands . Protection is conferred largely through the function of the polymeric Immunoglobulin receptor ( pIgR ) , which transports and stabilizes secretory antibodies and also functions as an innate immune factor ( Kaetzel , 2005 ) . Human pIgR is a glycosylated type I membrane protein consisting of a 620-residue ectodomain with five tandem immunoglobulin-like ( Ig-like ) domains , a 23-residue transmembrane domain , and a 103-residue intracellular domain ( Hamburger et al . , 2006 ) ( Figure 1A ) . pIgR is the oldest identifiable Fc receptor , first emerging in teleost ( bony ) fish . Throughout evolution , the number of Ig-like domains in the pIgR ectodomain increased; typically , bony fish express a two-domain variant , birds , amphibians and reptiles a four-domain variant , and mammals a five-domain variant ( D1-D5 ) ( Akula et al . , 2014 ) . Mammals , including rabbits and cows , express an alternatively-spliced variant containing D1 , D4 and D5 ( Deitcher and Mostov , 1986; Kulseth et al . , 1995 ) . 10 . 7554/eLife . 10640 . 003Figure 1 . Structure of hSC . ( A ) Schematic of mature human pIgR protein indicating Ig-domain ( D1-D5 ) boundaries . The proteolytic cut site that releases hSC from the apical membrane ( black arrow ) , the 23-residue transmembrane region ( TM ) , cytoplasmic tail , and potential N-linked glycosylation sites ( PNGS , orange arrows ) are indicated . ( B ) Schematic epithelial cell layer showing basolateral to apical transcytosis ( arrows ) of pIgR and release of free SC , SIgA , and SIgM . ( C ) Cartoon representation of the hSC structure viewed from the front face colored to highlight CDR loops ( green ) , D5 Cys468 and Cys502 ( yellow ) , PNGS ( orange ) , domain linkers ( grey ) , and hSC termini ( N-terminus: blue sphere; C-terminus: red sphere ) . ( D ) Molecular surface representation of the hSC structure shown in six orientations and colored as in ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 003 The pIgR is expressed on the basolateral surface of epithelial cells where the ectodomain binds polymeric forms of IgA and IgM produced by local plasma cells . Similar to other antibody classes , each IgA and IgM monomer contains two Fab arms and one Fc region ( a dimer of IgA heavy chain domains Cα2 and Cα3 or IgM domains Cµ2 , Cµ3 and Cµ4 ) . Unlike other antibody classes , IgA and IgM monomers can polymerize , producing dimeric IgA ( dIgA ) , pentameric IgM ( pIgM ) , and to a lesser extent , higher order polymers . Adjoining monomers are linked tail-to-tail through heavy chain C-terminal extensions ( tailpiece; tp ) and a 137-residue protein called joining ( J ) chain and are stabilized through disulfide bonds ( Hamburger et al . , 2006 ) . The transport cycle of pIgR ( Figure 1B ) starts when pIgR at the basolateral membrane binds dIgA or pIgM , and the resulting complex is transcytosed to the apical surface where proteases cleave the pIgR ectodomain ( now referred to as secretory component , SC ) , releasing it into the mucosa as a complex with dIgA or pIgM . These complexes , Secretory IgA ( SIgA; the predominant mucosal antibody ) and Secretory IgM ( SIgM ) , exclude pathogens from the epithelial barrier and promote host relationships with commensal bacteria through innate and adaptive mechanisms . SC specifically protects secretory Ig from proteolytic degradation and confers innate recognition functions upon SIgA and SIgM , allowing them to bind and exclude bacteria ( Kaetzel , 2005 ) . Although binding to polymeric Ig stimulates pIgR transcytosis ( Song et al . , 1994 ) , up to 50% of pIgR in humans trafficks to the apical surface and is released as unliganded , or free , SC ( Brandtzaeg , 1971 ) , which can exclude pathogenic bacteria and bacterial toxins through protein-protein or protein-glycan interactions ( Kaetzel , 2005 ) . In humans , SC binds the major Streptococcus pneumoniae surface protein , choline binding protein A ( CbpA ) , using protein-protein interactions , and to pathogenic bacteria such as H . pylori , E . coli and Shigella spp , using complex carbohydrates attached to one or more of its seven potential N-linked glycosylation sites ( Kaetzel , 2005 ) . SC carbohydrates have also been shown to facilitate binding to host mucus , host protein IL-8 , commensal bacteria Lactobacillus and Bifidobacteria and contribute to microbiota homeostasis of the intestinal mucosa ( Kaetzel , 2005; Mathias and Corthesy , 2011 ) . SC and SIgA interactions with pathogens and commensals are thought to be especially important for nursing infants , who ingest large quantities of maternal free SC and SIgA ( Hurley and Theil , 2011; Rogier et al . , 2014 ) . Understanding the structure ( s ) of the pIgR ectodomain ( hereafter called SC ) and how it interacts with ligands and pathogens is of interest because its critical role in immunity requires the protein to accommodate binding , transport and protection of secretory antibodies while also conferring innate protection in both free and liganded forms . High-resolution structural information for SC and the SC interactions with polymeric immunoglobulin ( pIg ) ligands is limited to a crystal structure of the human SC D1 domain , which adopts an Ig-variable ( V ) -like fold ( Hamburger et al . , 2004 ) . The structures and contributions of D2-D5 to intact SC function are largely unknown . D1 is both necessary and sufficient for binding to pIg Fcs and is also thought to interact with J-chain because pIgR transports only J-chain–containing pIgs , and isolated D1 does not bind monomeric IgA . D1 binding to pIg is partly mediated by three D1 loops that are structurally equivalent to the antigen-binding complementarity determining regions ( CDRs ) of immunoglobulin variable domains ( Hamburger et al . , 2006 ) . Binding to dIgA can be further stabilized by a disulfide bond between SC D5 and Fcα Cα2; however , this interaction is absent in some SIgA complexes and does not form in SIgM ( Almogren et al . , 2007; Hamburger et al . , 2006 ) . Here we report the first crystal structures of intact SC proteins , comparing the highly-evolved five-domain human SC ( hSC ) and a two-domain teleost fish SC ( tSC ) , a relative of the first vertebrate SC ancestor . We characterized the conformation and dynamics of free and liganded hSC in solution , and used structure-based alignments to create mutant and chimeric SCs to determine how individual domains contribute to ligand binding . These results provide a detailed model for SC structure and pIg binding mechanisms , demonstrating that mammalian SC evolved to adopt a compact , closed triangular structure , which opens upon ligand binding , whereas two-domain SC ancestors consist of tandem domains arranged in an elongated conformation . For hSC , we show that each of the five domains adopt distinct associations with each other in unliganded versus liganded forms , and that each contributes uniquely to dIgA and pIgM recognition and secretory antibody formation . The crystal structure of hSC ( Figure 1C ) was determined to 2 . 6Å resolution ( Rcryst = 20 . 1%; Rfree = 25 . 4% ) ( Supplementary file 1 ) . The final model ( 540 ordered residues of 549 total ) revealed five Ig-like domains ( D1-D5 ) arranged into a compact triangle ( three sides of ~70Å , ~70Å and ~90Å ) in which D2-D3 and D4-D5 form two of the sides , and D1 contacts both D2 and D4-D5 to form the third side ( Figure 1C ) . The domains lie in a plane such that the triangle thickness is roughly equal to that of a single domain ( ~40Å ) ( Figure 1C , D ) . The overall arrangement involves extensive interfaces between all five domains and a small solvent-accessible hole ( ~14Å diameter ) in the center . As defined in Figure 1D , the hSC front face shows all five domains . A 90° clockwise rotation reveals a side face dominated by D2 and D3; another 90° clockwise rotation reveals the back face showing all five domains , and a further 90° rotation reveals a side face comprising D4 and D5 . A fifth face is formed at the bottom of the hSC triangle ( 90° from the front and back faces ) , which includes D5 , D1 and D2 , and all domains are visible when viewed from the top . Important SC motifs , including CDRs , some residues implicated in ligand binding , and potential N-linked glycosylation sites , are largely solvent exposed and distributed on all faces of the molecule ( Figure 1C , D ) . The D1 CDR1 and CDR3 loops , which contribute to dIgA and pIgM binding ( Kaetzel , 2005 ) , are exposed exclusively on the front face , whereas the D4 CDRs are exposed only on the back face . The D2 , D3 and D5 CDRs are exposed on both faces and at least one side , and D5 residues Cys468 and Cys502 , which can disulfide bond with the dIgA residue Cys311 ( Hamburger et al . , 2006 ) , are exposed on the D4-D5 side . The D1 , D2 , D4 , and D5 domains of hSC include seven potential N-linked glycosylation sites that anchor carbohydrates involved in hSC interactions with bacterial and host lectins ( Kaetzel , 2005 ) . We observed partially ordered glycans at four sites ( D1 sites Asn65 and Asn72 , D2 site Asn168 , and D5 site Asn481 ) , which do not contact protein portions of hSC . These observations suggest that carbohydrates are unlikely to stabilize the hSC domain arrangement directly , but are optimally positioned to allow hSC to bind pIg ligands and to facilitate lectin interactions with free hSC and SIgA . At least one N-linked glycosylation site is visible from each orientation; however , six of the seven potential glycosylation sites are clustered on the back and bottom faces of the molecule ( Figure 1C , D ) . Consistent with predictions ( Mostov et al . , 1984 ) , each hSC domain adopts an Ig superfamily ( IgSF ) fold . IgSF domains share a basic folding topology consisting of two β-sheets linked through a conserved disulfide bond . The domains are classified as Variable ( V ) , Intermediate ( I ) or Constant ( C ) ( Wang , 2013 ) . D1-D4 each adopt an Ig-V-like fold , which typically contain β-strands A , B , E , D on one sheet and strands A’ , G , F , C , C’ , C” on the other . As reported for D1 , the loops connecting strands B-C , C’-C” , and F-G are similar to IgV CDRs ( Hamburger et al . , 2004 ) . In contrast to D1 , domains D2-D4 each lack an A strand that would normally pair with the β-sheet containing strands B , E , and D ( Figure 2A–D ) . The A strand-equivalent residues in these domains instead form the D1-D2 , D2-D3 and D3-D4 linkers . In addition to the canonical Ig disulfide bond connecting strands B and F , domains D1 , D3 , D4 and D5 include a second disulfide that links the C and C’ strands . Despite lacking the C-C’ disulfide , the relative orientation of the C and C’ β-strands in D2 is similar to other SC domains ( Figure 2 ) . D5 topology diverges from the canonical IgV fold because , in common with both the IgI and IgC folds , it lacks the C” strand and associated CDR2 loop ( Figure 2E ) . D5 also diverges from D2-D4 because residues following the D4-D5 linker hydrogen bond to the B strand , providing D5 with a two-residue A strand , a feature found in IgV and IgI but absent in IgC ( Wang , 2013 ) . Taken together , D5 topology resembles IgI , although it lacks conserved interactions that typically link loops equivalent to CDR1 and CDR3 ( Wang , 2013 ) . 10 . 7554/eLife . 10640 . 004Figure 2 . Individual hSC domains . ( A-E ) Cartoon representations of hSC domains shown in the same orientation . Disulfides are shown as sticks ( yellow ) , and CDR loops ( green ) , N- and C-terminal residues , and β-strands within Ig domain topology are labeled . The D4 CDR3 is likely flexible because five of its seven residues are disordered ( dashed lines ) . ( F ) Ribbon diagram showing Cα traces of aligned domains D1-D5 viewed from an orientation ~90° clockwise from ( A-E ) with CDR loops and regions with structural differences among domains indicated . See also Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 00410 . 7554/eLife . 10640 . 005Figure 2—figure supplement 1 . Comparison between human D1 and D3 CDR3 loop structures . Cartoon representations of human D3 ( A ) and D1 ( B ) with residues stabilizing the CDR3 loop positions shown as sticks and CDRs labeled . The D3 CDR3 is stabilized by the side chain of Phe251 and His310 , which forms a pH-dependent salt bridge with neighboring residue Asp312 , and by a hydrogen bond linking the Gly313 main chain oxygen and C’ residue Asn266 . D3 residue Phe251 is structurally equivalent to D1 residue Try36 , which stabilizes the position of D1 CDR3 ( Hamburger et al . , 2004 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 005 The CDR-equivalent loops in hSC D1-D5 display differences that suggest distinct functional roles . We previously reported that D1 CDR1 includes an α-helical turn formed in part by residues implicated in ligand binding ( Hamburger et al . , 2004 ) . Here we find that the CDR1 loops from all hSC domains contain a helical turn capped by a conserved lysine residue . However , the D2-D5 CDR1 loops include 310-helices instead of the CDR1 α-helix in D1 ( Figure 2 ) . The D5 CDR1 differs from the other CDR1s because it is disulfide bonded to the neighboring DE loop ( residues Cys468 in CDR1 and Cys502 in the DE loop ) ( Figure 2E ) . Notably , the D5 DE loop is 12 residues long , compared to 3-4 residues in D1-D3 and 7 residues in D4 , and extends ~10Å beyond the position occupied by DE loops in other domains ( Figure 2E , F ) . Despite the disulfide bond to the D5 CDR1 , side chain resolution was poor and the atomic B-factors were high for several residues in the DE loop , suggesting flexibility . CDR3 also exhibited divergence in conformation and length among D1-D5 , varying from two residues in D5 to ten residues in D3 , although D1 and D3 CDR3s occupy similar positions ( Figure 2 , Figure 2—figure supplement 1 ) . The two-residue CDR2s adopt similar conformations among D1-D4 . The compact , triangular arrangement of hSC domains is distinct from tandem domains in IgSF proteins , such as CD4 ( Wu et al . , 1997 ) , which are elongated with nearly co-linear domains . Instead , adjacent hSC domains are not co-linear and share distinct interfaces ( Figure 3A , Figure 3—figure supplements 1 , 2 ) . D1 and D2 are linked by a partially disordered and likely flexible linker and are related by an ~82° inter-domain angle . Their interface comprises residues in the D1 A’ , G , F , C , and C’ strands , which contact residues along the D2 A’ strand and the D2-D3 linker ( 1 , 156 Å2 total buried surface area ) ( Figure 3A , Figure 3—figure supplement 2A ) . Domains D2 and D3 ( related by ~152° ) are nearly co-linear and connected by a potentially rigid four-residue linker with two conserved prolines . Residues in the D2 A’-B loop , B strand , and the E-F loop contact residues in the D3 CDR3 and CDR1 loops , forming an interface that buries 790 Å2 total surface area ( Figure 3A , Figure 3—figure supplement 2B ) . D3 connects to D4 via a partially disordered and likely flexible seven-residue linker . The two domains are related by ~118° and share a small interface ( 596 Å2 total buried surface area ) involving residues in the D3 A’ and B strands and the D4 C-C’ loop . D4 and D5 are related by ~96° and form an extensive interface ( 1520 Å2 total buried surface area ) , in which residues in the D4 A’-B loop and B strand , CDR2 and C”-D loops , and D and E strands contact residues in the D5 G and F strands , C-C’ loop and CDR3 ( Figure 3 , Figure 3—figure supplement 2C ) , stabilizing the interface with numerous hydrogen bonds and hydrophobic interactions . 10 . 7554/eLife . 10640 . 006Figure 3 . Domain interfaces of hSC . ( A ) Cartoon representation of the hSC structure showing the front face view with interface residues colored pink and axes used to determine angles between domains shown as dashed lines ( each line is 50Å long ) . The approximate position of D1 CDR1 Pro26 and D5 CDR1 Cys468 are shown as orange circles . ( B ) Cartoon representation of the D1-D4-D5 interface with residues involved in putative hydrogen-bonding interactions shown as sticks and including three conserved D4-D5 interface residues , Trp480 , Trp523 and Trp525 , which bracket the D4 C”-D loop , and are positioned within hydrogen bonding distance of D4 Ser396 and Gly393 . See also Figure 3—figure supplements 1 , 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 00610 . 7554/eLife . 10640 . 007Figure 3—figure supplement 1 . Sequence alignment of representative SC D1-D5 domains . The secondary structures , locations of CDR loops , and hydrogen bonding ( H-bonding ) interactions based on the hSC structure are indicated above the alignment; interface residues , N-linked glycosylation sites ( PNGS ) and disulfide pairing ( numbered; green ) are indicated below . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 00710 . 7554/eLife . 10640 . 008Figure 3—figure supplement 2 . Domain interfaces . ( A ) Close-up ( front face ) view of D1-D2-D3 interface with residues involved in putative hydrogen bonding interactions shown as sticks . Interactions between D1 Arg39 and D2 Gln218 are conserved among many mammalian SC sequences ( Figure 3—figure supplement 1 ) . The D1 C-C’ loop residue Arg43 contacts D3 through a side chain-mediated salt bridge with Glu322 and uses its main chain to contact the D2 Asp125 side chain . Glu322 and Asp125 are conserved among most mammalian SCs ( the only vertebrates with an SC D2 domain ) , whereas Arg43 is not conserved ( Figure 3—figure supplement 1 ) . Inter-domain contacts involving D1 C-C’ appear necessary to stabilize this segment of D1 , as this region was disordered in the structure of isolated D1 ( Hamburger et al . , 2004 ) . ( B ) Close-up view of the D2-D3 interface . Hydrogen bonds between the D2 and D3 domains are absent . However , the close proximity of interfacing loops from each domain is likely maintained by the four-residue D2-D3 linker , which contains two conserved prolines ( shown as sticks ) . The interface is stabilized through electrostatic interactions between D2 Arg128/Arg193 and D3 Glu244 ( CDR1 ) /Glu317 ( CDR3 ) side chains , which are separated by ~4 . 5Å . ( C ) Close-up view detailing part of the D4-D5 interface . Residues near and in the D5 CDR3 involved in putative hydrogen bonding between D4 and D5 ( shown as sticks ) include Ser348 and Thr537 , and Glu400 and Tyr534; Glu400 is also within hydrogen bonding distance of D5 CDR3 His532 . Ser348 , Thr537 and Tyr534 are conserved whereas Glu400 and His532 are not ( Figure 3—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 008 An unexpected and likely functionally significant feature of the hSC structure is the large interface ( 1480 Å2 total buried surface area ) formed between domains D1-D4-D5 ( Figure 3 ) . D1 and D4 are approximately anti-parallel , with CDR loops pointing in opposite orientations and contacts formed by the C” strands and neighboring residues from both domains . D5 is almost perpendicular to D1 and D4 ( related by ~96° and ~101° , respectively ) , with its CDR loops pointing away from D1 , separating ligand-binding motifs in D1 and D5 CDR1 loops by nearly 45Å ( as measured between D1 CDR1 Pro26 and D5 CDR1 Cys468 , orange circles , Figure 3A ) . Notably , the interface includes residues from all three of the D1 CDR loops and buries CDR2 at the bottom of a depression formed by the three domains . Important contacts between D4 and D1 ( Figure 3B ) that stabilize the triangular shape of hSC include a salt bridge between D4 Glu392 and D1 Arg34 , the last residue in D1 CDR1 . Arg34 , which is critical for pIgR binding to dIgA ( Coyne et al . , 1994 ) , also hydrogen bonds to D1 CDR3 residue Asn97 , which stabilizes the position of the D1 CDR3 loop ( Hamburger et al . , 2004 ) to create a network of interactions between D4 and CDRs 1 and 3 in D1 . This interface is further stabilized through interactions with the D5 C-C’ loop and flanking strands that protrude into the D1-D4 interface where D5 residues share contacts with D4 and D1 , thereby bridging the three domains ( Figure 3B ) . The keystone in this interaction is Asn482 , whose side chain is located in the center of the C-C’ loop where it forms hydrogen bonds with the main chain oxygen of D4 Gly385 and the main chain nitrogen of D1 Val56 . Similarly , Glu53 , the residue at the center of the D1 CDR2 , is buried by and within hydrogen bonding distance of D5 Trp523 ( F strand ) , which also contacts D4 . A set of pH-dependent stabilizing salt bridges occur between D1 CDR1 residue His32 and D5 residues Glu521 and Glu545 , which are located in the D5 E-F loop and G strand , respectively ( Figure 3B ) . His32 , Arg34 , Glu392 , Asn482 , and Glu521 are conserved in mammalian SC sequences , and Arg34 and Glu392 are also conserved in reptilian , amphibian , and avian SCs ( Figure 3—figure supplement 1 ) . It is notable that conserved D1 residues His32 and Arg34 , as well as residues in D1 CDR2 , are important or necessary for binding to dIgA and pIgM ( Coyne et al . , 1994; Roe et al . , 1999 ) , yet also stabilize the D1-D4-D5 interface and are largely inaccessible , while other putative ligand-binding residues in CDR1 and CDR3 ( e . g . , Tyr24-Asn30 , Arg99-Lue101 ) remain exposed . The existence of flexible linkers , and in some cases , the absence of extensive stabilizing contacts between hSC domains , suggested that hSC might adopt a range of conformations . To evaluate the conformational flexibility of hSC in solution , we conducted double electron-electron resonance ( DEER ) spectroscopy on free and liganded forms of hSC variants containing pairs of nitroxide spin labels . A DEER experiment measures the probability distribution of inter-nitroxide distances in the 17–80Å range ( Jeschke and Polyhach , 2007 ) . The most probable distance and width of the distribution provide direct information on the structure and structural heterogeneity , respectively , and are presumed to reflect the amplitude of molecular motion in solution under physiological conditions . Compared to methods such as crystallography and electron microscopy that capture snapshots of protein structure , DEER has the advantage of providing analytical data that describe a protein’s structural heterogeneity and flexibility ( Hubbell et al . , 2013 ) . Guided by the hSC structure , we prepared spin-labeled variants , D1-67R1/D5-455R1 , D1-67R1/D5-491R1 , and D1-80R1/D5-491R1 , containing the R1 nitroxide side chain at the indicated positions ( Figure 4A , B ) , to monitor changes in spatial proximity between D1 and D5 . An intra-domain pair D5-455R1/D5-491R1 was prepared to monitor domain flexibility . All spin-labeled proteins bound dIgA and pIgM with kinetics indistinguishable from unlabeled hSC showing that spin labeling did not disrupt ligand binding ( Figure 4—figure supplement 1A ) . 10 . 7554/eLife . 10640 . 009Figure 4 . DEER spectroscopy of spin-labeled hSC . ( A ) Structure of R1 nitroxide ( unpaired electron shown in red ) attached to a cysteine residue . ( B ) Cartoon representation in two orientations showing modeling of an R1 side chain ( red sticks ) at the indicated sites on the hSC crystal structure . The distances measured are indicated as red dashed lines in the lower panel . ( C-F ) Distance distributions for the indicated mutants obtained after model-free fitting of the dipolar evolution function ( DEF ) . The vertical red lines indicate the expected interspin distance based on modeling of the R1 side chain on the hSC crystal structure . ( G ) Same as C-F but for D1-67R1/D5-491R1 complexes with dIgA ( cyan ) or pIgM ( green ) . Minor peaks at 27Å likely correspond to unliganded SC , and peaks observed with maxima at ~57Å upon dIgA binding and at ~53Å upon pIgM binding may reflect structural heterogeneity in ligand structure or arise from intermediate binding states . The relative populations of these states cannot be determined due to existence of distance probabilities outside the reliable range of detection . The gray and orange bars indicate the upper limit of reliable distance and shape of the distribution for unliganded and liganded hSC variants , respectively ( see Experimental Procedures ) . For D1-67R1/D5-491R1 complexes , distances above 70 Å are beyond detection limits ( dashed traces ) . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 00910 . 7554/eLife . 10640 . 010Figure 4—figure supplement 1 . SPR data , time domain data and CW EPR spectra and control DEER measurements . ( A ) SPR measurements showing similar sensorgrams for unmodified and spin-labeled hSC proteins binding to dIgA and pIgM . hSC or D1-67R1/D5-491RI spin-labeled hSC were injected over dIgA or pIgM immobilized surfaces in a two-fold dilution series starting at a highest concentrations of 1 . 02 µM ( dIgA ) and 0 . 26 µM ( pIgM ) . Uncertainty of the binding mechanism precluded our ability to accurately model kinetics for these data . ( B ) Time domain of the dipolar evolution function ( DEF ) ( shown in black ) for the indicated mutants with model-free fit shown in red . ( C ) Same as B but for D1-67R1/D5-491R1 complexes with dIgA or pIgM . ( D ) Time domain data ( left ) and corresponding distance distribution ( right ) obtained after model-free fitting of the DEF ( red ) for control hSC variants D5-455R1/D5-491R1 before ( black ) and after incubation with dIgA ( cyan ) and pIgM ( green ) ligands . The vertical red line on the distance distribution indicates the distance measured between 455R1 and 491R1 on the hSC crystal structure ( Figure 4B ) . ( E ) CW spectra recorded for three indicated hSC variants labeled with the R1 sidechain ( left ) and for the D1-67R1/D5-491R1 variant before and after its incubation with dIgA or pIgM ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 010 DEER time domain data and resulting distance distributions for all R1 pairs in the unliganded protein demonstrated most probable distances within 1Å of the expected distances based on the hSC crystal structure ( Figure 4C–F , Figure 4—figure supplement 1B ) , indicating that the crystal structure represents the predominant solution structure of hSC at the D1-D5 interface . The intra-domain distance distribution for D5-455R1/D5-491R1 showed one dominant peak at 25Å with a narrow width at half-height of 3 . 5Å , characteristic of R1 pairs in a rigid protein structure ( Lerch et al . , 2014 ) . By contrast , distance distributions for the inter-domain pairs were multimodal and broader , suggesting contributions from protein flexibility . The flexibility could involve local motions of secondary structure elements containing the spin labels and/or subtle inter-domain motions between D1 and D5 , but the relatively small amplitude of these motions ( <10Å ) indicated that D1 and D5 are in contact . To investigate whether ligand binding induces structural changes in the hSC D1-D5 interface , and to obtain structural data describing SIgA and SIgM complexes , we measured the interspin distance between D1-67R1 and D5-491R1 in the presence of dIgA and pIgM . As shown in Figure 4G and Figure 4—figure supplement 1C , ligand binding induced a dramatic separation of D1 and D5 , leading to a broad distance distribution between the R1 residues , extending beyond 70Å . The position and widths of the distance distribution for distances beyond 70Å are not well-determined , but the data are consistent with movements of more than ~42Å between D1-D5 . Despite this remarkable domain separation , the intra-domain distance distribution of D5-455R1/D5-491R1 remained unchanged ( Figure 4—figure supplement 1D ) , and the continuous wavelength ( CW ) EPR spectra did not change for any of the sites ( Figure 4—figure supplement 1E ) , indicating that the local structure around R1 was unperturbed . The broad distance distributions we observed for hSC in the hSC:dIgA and hSC:pIgM complexes suggest that hSC binding progresses through intermediate states and/or that liganded hSC is highly flexible . These observations further suggest that SIgA and SIgM structures are richly heterogeneous . To compare SC structures across evolution and to facilitate the design of chimeric proteins for characterizing SC-ligand interactions , we solved the 1 . 7Å crystal structure of teleost fish SC ( tSC ) ( Rcryst = 18 . 5%; Rfree = 21 . 5% ) ( Supplementary file 1 ) . Teleost fish express the oldest recognizable pIgR protein ( Akula et al . , 2014 ) , which transports polymeric versions of IgM and IgT/IgZ ( a teleost Ig specialized in mucosal immunity ) to the mucosa ( Sunyer , 2013 ) . The fish genome does not encode a J-chain , thus fish pIgs are thought to be structurally divergent from their human counterparts ( Flajnik , 2010 ) . In addition , tSC contains just two extracellular domains , tD1 and tD2 . tD1 and tD2 share homology with mammalian D1 and D5 , respectively , but CDR loop residues implicated in mammalian SC interactions with pIg are not conserved ( Feng et al . , 2009; Hamuro et al . , 2007; Rombout et al . , 2008 ) . The structure of tSC revealed two tandem Ig-like domains related by ~90° ( Figure 5A ) . Like hSC domains D1 and D3-D5 , tSC domains are IgSF folds that include the canonical B to F strand disulfide bond and a second disulfide bond linking the C and C’ strands . tD1 and tD2 share a similar topology ( Figure 5—figure supplement 1A , B ) , but only 72 of 97 Cα atoms could be aligned ( rmsd = 0 . 87Å ) . Aligning fish and human SC domains resulted in rmsds of 0 . 77Å for tD1 and D1 ( 83 of 100 Cα atoms ) and 1 . 08Å for tD2 and D5 ( 64 of 97 Cα atoms ) ( Figure 5B–D ) . The largest structural differences between tSC and hSC domains are in regions between the C’ and D strands , where tD1 residues at positions similar to the C” strand are disordered and the tD2 C” strand hydrogen bonds to the D strand on the A-B-E-D face rather than to the C’ strand on the A’-G-F-C face ( Figure 5 , Figure 5—figure supplement 1A , B ) , similar to other IgSF proteins such as IREMs ( immune receptors expressed on myeloid cells ) ( Marquez et al . , 2007 ) . These differences are coupled with distinct features in tSC CDR2 loops and neighboring residues , which include a patch of negative electrostatic potential on tD2 formed from six negatively charged residues in the CDR2 , C” , and D strands . The tSC CDR1 and CDR3 loops also adopt distinct conformations . In contrast to the human D1 CDR1 α-helical turn , both tSC CDR1 loops include a 310 helical turn and are structurally similar to human D5 CDR1 . The tD1 CDR1 is solvent exposed and highly conserved among representative fish SC sequences , while the tD2 CDR1 is partly in contact with tD1 and contains fewer conserved residues ( Figure 5A , Figure 5—figure supplement 1C , D ) . The residues in the CDR3 loops are largely conserved among fish ( Figure 5—figure supplement 1C ) , but differ between tSC domains; the tD1 CDR3 is elongated , extending toward CDR1 and containing a short helical turn , whereas the tD2 CDR3 is shorter , similar to the D5 CDR3 ( Figure 5A–C ) . tD2 has a 4-residue DE loop , similar to counterpart loops in human D1-D4 and in contrast to the extended loop and disulfide link to CDR1 found in human D5 . The interface between the two tSC domains is formed by residues in the tD1 A’-B and E-F loops and residues in tD2 CDR1 and DE loop , burying a total surface area of 591Å2 . Ten of the 17 residues in the interface are well conserved among representative fish species , and five inter-domain hydrogen bonds may limit inter-domain flexibility ( Figure 5—figure supplement 1C , D ) . Collectively , structural differences between the tD1-tD2 and their human D1 and D5 counterparts may represent adaptations that are advantageous for binding to species-specific ligands . 10 . 7554/eLife . 10640 . 011Figure 5 . Structure of tSC . ( A ) Cartoon representation of the tSC structure with disulfides shown as yellow sticks , CDR loops colored cyan , N- and C-termini indicated ( blue and red spheres ) and Ig domain topology labeled . ( B ) Ribbon diagram showing Cα traces of tD1 and human D1 ( hD1 ) following alignment . tD1 is deep salmon with cyan CDRs , and hD1 is light purple with green CDRs . ( C ) Ribbon diagram showing aligned Cα traces of tD2 and human D5 ( hD5 ) . tD2 is pink with cyan CDRs , and hD5 is dark blue with green CDRs . ( D ) Structure-based sequence alignments of tD1-tD2 , tD1-hD1 , and tD2-hD5 with corresponding secondary structure and CDR boundaries shown . See also Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 01110 . 7554/eLife . 10640 . 012Figure 5—figure supplement 1 . tSC domain interfaces and sequence conservation . ( A-B ) Topology diagrams of tSC domain 1 ( A ) and domain 2 ( B ) . β-strands are shown as arrows with residue numbers indicated . Disulfide bonds between strands are indicated by gold lines . ( C ) Sequence alignment of teleost fish SC showing the secondary structure from the tSC crystal structure ( O . mykiss SC; top ) along with the positions of CDR loops ( cyan numbers ) and tD1-tD2 interface residues ( blue ovals ) . ( D ) Cartoon representation of tSC D1-D2 interface with putative hydrogen bonding interactions shown as sticks . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 012 Differences between the fish and human SC structures , together with data demonstrating that hSC undergoes a conformational change upon pIg binding , motivated us to investigate how each domain contributes to ligand binding . Our approach involved evaluating the ability of short hSC variants and human-fish SC chimeric proteins to bind human dIgA ( Figure 6 ) and pIgM ( Figure 6—figure supplement 1 ) . Unless otherwise noted , hSC D5 residues Cys468 and Cys502 were mutated to alanine to prevent covalent binding to dIgA . All hSC variants and human-fish chimeric proteins were monomeric and monodisperse as verified by size exclusion chromatography ( SEC ) and/or SEC with in-line multi-angle light scattering ( MALS ) ( data not shown ) . Although the stoichiometry of hSC to pIg in SIg complexes is reportedly 1:1 , binding is thought to occur through a multi-step mechanism ( Hamburger et al . , 2006 ) . Consistent with this model and the demonstration of a ligand binding-induced conformational change in hSC , the binding profiles of our SC variants failed to fit single state kinetic models . Because uncertainties in the binding mechanism precluded accurate selection of a kinetic model , we qualitatively monitored differences in the SPR binding profiles to identify changes in kinetics among protein variants . 10 . 7554/eLife . 10640 . 013Figure 6 . hSC and chimeric SC binding to dIgA . ( A-C ) Sensorgrams showing hSC , D1 , and D1-D3 binding to dIgA . ( D ) Cartoon representations and associated sequences of CDR1 and DE loop residues substituted in chimeric proteins following structural alignments and modeling . ( E-G ) Sensorgrams for interactions of chimeric SC proteins and dIgA . The time at which complete dissociation occurs is indicated ( astrisk ) . ( H ) Equilibrium binding response versus the log of concentration for the sensorgrams in ( A , E-G ) and replicate experiments ( not shown ) . Average KD values for two replicates for each SC variant were: hSC ( 63+/-4nM ) , hSC tCDR1 ( 82+/-1nM ) , hSC tDE ( 65+/-3nM ) , hSC tCDR1/tDE ( 140+/-7nM ) . ( I-J ) Sensorgrams showing hSC D1-D3-D4-D5 and D1-D4-D5 binding to dIgA . See also Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 01310 . 7554/eLife . 10640 . 014Figure 6—figure supplement 1 . hSC and chimeric SC binding to pIgM . ( A-F ) Sensorgrams showing the responses of hSC D1-D5 , D1 , D1-D3 and chimeric SCs binding to pIgM . ( G ) Overlay of sensorgrams showing normalized response of D1-D5 and chimeric SC binding to pIgM . ( H-I ) Sensorgrams showing the response of D1-D3-D4-D5 and D1-D4-D5 binding to pIgM . All results were consistent among independent replicate experiments ( not shown ) . Highest concentrations for hSC D1-D5 , D1-D3 , D1-D3-D4-D5 D1-D4-D5 , tCDR1 , tDE , tCDR1/tDE were 256nM . Highest concentration for D1 was 16nM . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 014 hSC D1-D5 exhibited concentration-dependent binding to dIgA with relatively slow association and dissociation phases ( Figure 6A ) . In contrast , and consistent with previously published SPR experiments ( Hamburger et al . , 2004 ) , D1 association and dissociation were rapid ( Figure 6B ) , demonstrating that D2-D5 contribute to the dIgA binding mechanism even in the absence of covalent interactions mediated by D5 Cys468/Cys502 . To investigate potential contributions from other hSC domains , we tested binding of D1-D3 , D2-D3 , D4-D5 , and D2-D5 . The D1-D3 variant exhibited rapid association similar to D1 and an initially rapid dissociation that subsequently slowed ( Figure 6C ) , suggesting that D2-D3 either indirectly altered D1 binding to dIgA or contributed additional specific interactions . D2-D3 gave inconsistent results among replicate experiments , with weak binding detected in two experiments and no binding detected in a third experiment ( data not shown ) . We did not detect binding of D4-D5 and D2-D5 even at concentrations above 32 µM ( data not shown ) , supporting reports that D1 is required for binding of SC to dIgA ( Brandtzaeg , 2013 ) . Differences in binding of hSC D1-D3 versus D1-D5 to dIgA suggested that domains D4 and/or D5 contribute to non-covalent interactions between hSC and dIgA . Hypothesizing that residues near D5 Cys 468/Cys502 are involved , we engineered a series of chimeric proteins ( hSC-tCDR1 , hSC-tDE and hSC-tCDR1/tDE ) , in which the human D5 domain contained the CDR1 loop and/or DE loop from counterpart regions of tD2 exhibiting divergent sequence and structure ( Figures 5 , 6D ) . We characterized chimeric protein binding to immobilized dIgA by SPR ( Figure 6E–G ) and calculated equilibrium binding affinities ( KD ) from maximal binding response values ( Figure 6H ) . Binding of hSC ( KD~63+/-4nM ) and the hSC-tDE chimera ( KD~65+/-3nM ) were indistinguishable . By contrast , the hSC-tCDR1 chimera exhibited subtle changes during dissociation from dIgA , and the affinity ( KD~82+/-1nM ) was slightly reduced compared with the hSC D1-D5 affinity . The affinity for dIgA was further reduced ( KD~140+/-7nM ) , and the dissociation phase was shorter when the binding of a double chimera ( hSC-tCDR1/tDE ) was evaluated ( asterisk , Figure 6G ) . Taken together , these data suggested that D5 mediates direct non-covalent interactions between hSC and dIgA . Since the double chimera exhibited a greater reduction in binding than the hSC-tCDR1 chimeric protein , the D5 DE loop may play a role in stabilizing the position of D5 CDR1 , which could be enhanced in the presence of the disulfide that normally links the two motifs ( absent in our binding studies ) . The hSC-tCDR1/tDE did not mimic the binding kinetics of D1 for dIgA , suggesting that additional , as yet unidentified , interfaces contributed to binding affinity and/or that the binding mechanism of D1 was indirectly altered by the presence of D2-D5 , perhaps because isolated D1 would not require a conformational change to maximally expose surface area involved in binding . Alternatively , fish motifs found in the chimeric proteins could confer positive binding interactions for dIgA . Having shown that hSC binds dIgA non-covalently using its D1 and D5 domains and undergoes a conformational change that separates these domains upon SIgA formation ( Figure 4G , 6E–H ) , we characterized binding of shorter hSC proteins , in which the distance between D1 and D5 was constrained . For these studies , we created two hSC proteins , one comprising domains D1-D3-D4-D5 ( analogous to avian , reptilian , and amphibian SC , which lack a domain homologous to D2 ) and another comprising D1-D4-D5 ( analogous to a naturally-occurring mammalian splice variant ) ( Akula et al . , 2014; Deitcher et al . , 1986 ) . Both variants exhibited robust , concentration-dependent binding to dIgA; however , dissociation occurred rapidly , with all of the injected protein dissociating ~200 s into the experiment compared to ~400 s for hSC ( Figure 6A , I–J ) , similar to the binding kinetics of the D1–dIgA interaction ( Figure 6B ) . These results suggest that D2 , and possibly D3 , contribute to binding by providing direct interactions with ligand and/or by promoting interactions between D5 and dIgA . The latter possibility is attractive , given that chimeric SC with mutations in D5 exhibited similar binding profiles that were characterized by rapid dissociation ( Figure 6G–H ) . We also evaluated the interactions of the hSC variants and chimeras with pIgM and found that D1-D5 , D1 , D1-D3 , D1-D3-D4-D5 , and D1-D4-D5 exhibited concentration-dependent binding to pIgM ( Figure 6—figure supplement 1A–C , H , I ) . Although qualitative differences in binding among these variants was subtle , the association phase of D1 required a lower starting concentration ( 16nM ) to reach equilibrium compared with the starting concentration required for D1-D5 ( 256nM ) , suggesting differences in binding ( Figure 6—figure supplement 1A , B ) . Similar to dIgA binding studies , we detected inconsistent responses for D2-D3 and did not detect binding of D4-D5 and D2-D5 ( data not shown ) . In contrast to the dIgA binding studies , we could not detect differences in pIgM binding between hSC D1-D5 and the hSC-tCDR1 , hSC-tDE or hSC-tCDR1/tDE chimeras , suggesting that the D5 CDR1 and DE loop do not play a prominent role in the hSC-pIgM interaction ( Figure 6—figure supplement 1D–G ) . Since the discovery of SC ( Tomasi et al . , 1965 ) and later identification of its membrane-bound form ( Mostov et al . , 1980 ) , the pIgR has been established as a central component of the vertebrate immune response , transporting and stabilizing secretory antibodies , excluding pathogenic bacteria , providing immune protection at epithelial barriers such as the lungs , gut , urogenital tract , and conferring protection to offspring through breast milk ( Kaetzel , 2005 ) . Despite its importance , relatively little was known of pIgR/SC structure . Models ranged from schematic representations of elongated tandemly-arranged domains ( Monteiro and Van De Winkel , 2003 ) to solution scattering-based , computational models of J-shaped molecules , in which D1 makes contacts only with D2 ( via the D1-D2 linker ) , and D2-D3 are folded back toward D4-D5 , leaving all D1 binding motifs free to interact with ligand ( Almogren et al . , 2009; Bonner et al . , 2007 ) . Here we present atomic resolution structures that detail how the SC structure has changed over the course of vertebrate evolution , showing that bony fish utilize the two domains of ancestral SC to form an open , elongated structure whereas mammals utilize five domains to form a closed , triangular structure that opens upon ligand binding . These differences are accompanied by domain-specific structural variations and together suggest that SC has evolved into a structurally-plastic molecule that protects mammals by specifically modulating its conformation for its known functions as free SC , SIgA , and SIgM . The mammalian SC might be described as the host immune system’s hand , whose thumb ( D1 ) and four fingers ( D2-D5 ) can form a fist or open to grasp a polymeric antibody in order to fight potential invaders . By contrast , the more primitive two-domain fish SC could be described as a single finger , already in an open conformation for ligand binding and lacking the dexterity of a hand . The expansion of SC from two domains to five parallels increasing organism complexity as well as changes in antibody isotypes and species-specific organization of polymeric Igs ( Akula et al . , 2014; Flajnik , 2010 ) . While parallel evolutionary changes in receptors and their ligands are common , it was unexpected to find that the addition of D2 , D3 and D4 domains in mammals was associated with a closed conformation that occluded putative ligand-binding motifs . Burial of ligand-binding motifs hints that the closed conformation provides mammalian SC with advantages that are separate from its binding to polymeric antibodies . For unliganded SC released into the mucosa , the closed conformation could protect SC from proteases by burying susceptible inter-domain linkers and loops , and also promote interactions with commensal and/or pathogenic bacteria . Supporting the notion that the closed conformation is important for unliganded hSC function , our structure suggests that the D1-D5 interface would be stabilized in acidic mucosa because protonation of D1 His32 that occurs at acidic pH would facilitate electrostatic interactions with conserved acidic residues in D5 ( Figure 3B ) . The hSC structure also provides insight into hSC interactions with the human pathogen Streptococcus pneumoniae . Unliganded hSC and SIgA use residues in D3 and D4 to bind to the major S . pneumoniae adhesion protein CbpA , and a peptide corresponding to hSC D4 residues 349–375 inhibited S . pneumoniae adherence to epithelial cells ( Kaetzel , 2005 ) . The hSC structure shows that residues 349–375 occupy solvent-exposed regions of D4 CDR1 and the D3-contacting regions of the C-C’ loop , rationalizing why both D3 and D4 are required for the interaction with CbpA ( Figure 7 ) . In addition , since SIgA binds CbpA , this suggests that these regions of hSC remain exposed upon binding to dIgA . 10 . 7554/eLife . 10640 . 015Figure 7 . Model for pIgR transcytosis , ligand binding and release of free SC and SIgA . Schematic model depicting a mammalian epithelial barrier with membrane-bound SC ( pIgR ) shown on the basolateral side in its closed , unliganded conformation . The pIgR D1-D5 binding to dIgA is depicted as a three-step mechanism whereas pIgR D1-D4-D5 splice variant binding to dIgA is depicted as a one-step mechanism . pIgR binding to pIgM is not shown but occurs via a mechanism related to dIgA binding . The boxed region ( pIgR D1-D5 and dIgA ) is enlarged and associated atomic resolution models shown ( inset ) . The hSC crystal structure is colored as in Figure 1 . A model for dIgA based on the Fcα crystal structure ( pdb code 1OW0 ) is shown with the Fabs , tailpiece ( TP ) and J chain shown schematically , approximate distances between known SC binding sites labeled ( C311 and F443 ) , and an arrow indicating possible bending at the dimer interface . Following transcytosis , free SC and SIgA ( and SIgM ) are released into the mucosa . Unliganded mucosal SC is depicted as a schematic and as the hSC structure ( same colors as Figure 1C ) with residues 349–375 , the putative D4 S . pneumonia CbpA binding site , colored magenta . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 015 Our data support an accepted model for mammalian pIgR binding to dIgA , in which initial non-covalent binding of SC D1 is followed by covalent binding of SC D5 ( Hamburger et al . , 2006 ) , but further suggest that additional steps involving all five domains bridge these two events . Of particular significance is the change in the positions of D1 and D5 upon binding to both dIgA and pIgM , as identified in DEER experiments . The large magnitude of this change ( more than 40Å ) , along with evidence for flexible inter-domain linkers , implies that the D1-D5 interface is broken during ligand binding and suggests that the conformational change involves other domains and exposes additional ligand binding motifs on SC ( e . g . , by enhancing accessibility to D1 binding motifs ) . Our SPR data demonstrated that during this process , all five domains contribute to binding kinetics and that D5 binds dIgA independent of covalent bond formation . These observations suggest that non-covalent interactions between D5 CDR1 and DE loops can stabilize the complex during the disulfide exchange reaction and also suggest that both D1 and D5 contact dIgA directly in endogenous SIgA that lack covalent interactions with SC ( Almogren et al . , 2007; Lindh and Bjork , 1976 ) and in the R1-labeled SC-dIgA complexes used for DEER measurements . In the case of SC binding to IgM , which does not involve covalent bond formation ( Hamburger et al . , 2006 ) , we found that binding also involves a conformational change , which could facilitate ligand binding at secondary sites on SC D1-D5 , although such sites do not appear to involve the D5 CDR and DE loops because chimeric SC and hSC exhibited similar binding profiles . The observation that a conformational change occurs even when binding is dominated by D1 supports the hypotheses that the closed SC conformation occludes D1 binding motifs and that increasing D1 accessibility is an important part of the ligand binding mechanism . The demonstration that hSC D1-D3-D4-D5 and D1-D4-D5 bound dIgA with kinetics similar to D1 rather than to D1-D5 ( Figure 6 ) also sheds light on specific roles for pIgR D2-D4 during pIgR-dIgA recognition . These results suggest that the gain of the D2 domain in mammalian pIgR resulted in enhanced binding to dIgA , perhaps by providing direct contacts and/or by facilitating interactions between D5 and ligand , and further suggest that binding of mammalian splice variants to pIg ligands are mediated primarily through interactions with D1 . This possibility is supported by earlier work demonstrating that D2 and D3 were required for covalent binding of D5 to dIgA in other mammals ( Crottet and Corthesy , 1999; Solari et al . , 1985 ) , an observation we confirmed with an hSC D1-D4-D5 variant ( data not shown ) . When unliganded , short pIgR/SC variants are likely to adopt conformations that differ from the closed conformation of unliganded hSC . The tSC structure reveals an elongated conformation for a two-domain variant , and analysis of the hSC structure indicates that steric constraints would prevent mammalian D1-D4-D5 splice variants from adopting a closed conformation with a D1-D4-D5 interface equivalent to that of hSC . This implies that D2-D3 play a role in stabilizing the closed hSC conformation . The presence of D3 may be sufficient to stabilize a closed conformation because analysis of the hSC structure suggests that D1-D3-D4-D5 variants could adopt a closed conformation with a D1-D4-D5 interface equivalent to that of hSC . While the absence of a closed conformation might enhance ligand accessibility to D1 binding motifs , it is unclear if this would be advantageous because D1-D4-D5 dissociates from dIgA rapidly compared to D1-D5 ( Figure 6I ) . Mammalian SC splice variants transport dIgA in vivo ( Kuhn et al . , 1983 ) ; however , the utility of these variants may be impaired by altered interactions with ligands and/or in any context in which the closed state is advantageous , whether for ligand binding and/or free SC functions . Structural characterization of hSC and SIg complexes using DEER allowed us to capture hSC in both a compact unliganded state and heterogeneous liganded states , information that we could not obtain using crystallography or single particle electron microscopy ( unpublished results ) . While it is notable that SIgA and SIgM appear heterogeneous in structure , predicting how hSC domains are arranged in these complexes is challenging . hSC could bind a single Fc , or bind across the dIgA dimer and contact two Fcs ( or more in the case of pIgM ) . Putative hSC D1 interaction sites on dIgA include Fcα residues 377 , 402–411 , 413-414 , 440–443 , and D5 forms a covalent bond with Fcα residue C311 ( Woof and Russell , 2011 ) . Interaction sites are maximally separated by ~50Å when measured between Phe443 and Cys311 in monomeric Fcα ( Herr et al . , 2003 ) , and likely separated by more than 50Å when measuring between adjacent Fcs in the dIgA dimer ( Figure 7 ) . Although dIgA contains four copies of all Fc residues implicated in binding , SC binds dIgA with 1:1 stoichiometry ( Kaetzel , 2005 ) , suggesting either that dimeric IgA formation induces conformational differences in Fcα that alter the binding site environments and/or that a 1:1 stoichiometry is enforced by hSC interactions with J-chain . Solution scattering models suggested that hSC binds along one side of dIgA in an extended conformation that contacts both Fcs ( Bonner et al . , 2009 ) . These results , taken together with our DEER measurements showing that the D1 and D5 CDR1s could be separated by more than 85Å upon hSC binding to dIgA in a SIgA molecule ( 45Å between D1 and D5 CDR1s in the closed conformation plus >40Å change in positions of D1 and D5 upon ligand binding ) , favor a model in which hSC contacts both Fcs in dIgA . Our results suggest a revised model for pIgR/SC function and ligand binding in mammals ( Figure 7 ) . In this model , unliganded pIgR at the basolateral membrane remains in the closed state until encountering ligand . The closed state , characterized by partially-buried ligand binding motifs including D1 CDR2 and CDR1 residues His32 and Arg34 ( Coyne et al . , 1994; Roe et al . , 1999 ) , binds to a single Fcα in dIgA ( making additional , potential contacts with J-chain ) using exposed motifs in D1 CDR1 and CDR3 , facilitating an initial recognition event that disrupts the D1-D4-D5 interface and induces a conformational change . The conformational change frees D5 ( and perhaps other domains ) to form interactions with the second Fc on dIgA ( or another Fc on pIgM ) and likely exposes D1 residues such as His32 and Arg34 , permitting the ligand to interact with regions previously buried by D4-D5 . This incremental binding mechanism , particularly the secondary binding at a previously occluded site , might be necessary or advantageous for binding to a ligand in which binding sites are separated by a long distance ( e . g . , between adjacent Fcs in dIgA or pIgM ) and/or to ensure binding at the correct site ( s ) if multiple binding sites are exposed on a ligand . Following transcytosis to the apical membrane , the pIgR-pIg complexes are released into the mucosa as SIgA or SIgM , and pIg-associated SC retains an open conformation . In contrast , unliganded pIgR released at the apical membrane as free SC would maintain a stable , closed conformation that promotes SC’s innate immune function to recognize and exclude bacteria . In summary , the high-resolution insights on SC structure and mechanism presented here provide a new framework on which to build understanding of mucosal immune evolution and function . Our model for mammalian SC mechanism provides an explanation for how mammalian SC structure evolved to maintain innate functions of unliganded SC while also adapting to bind mammalian pIgs , which are large , complicated molecules . Together with current and forthcoming biological data , these models facilitate insight into normal immune function and disease states while also providing a molecular basis on which to engineer new SC and SIg-based therapeutics . The gene fragment encoding the human pIgR signal peptide and residues 1–547 was amplified from cDNA encoding full-length human pIgR ( a gift from Roland Strong , Fred Hutchinson Cancer Research Center ) and sub-cloned into the pTT5 expression vector ( NRC Biotechnology Research Institute ) along with a 5’ Kozak sequence and a C-terminal StrepII affinity tag ( residues Trp-Ser-His-Pro-Gln-Phe-Glu-Lys ) . This construct was modified using site-directed mutagenesis to create hSC domain variants , chimeric SC , and mutants used for spin-labeling . In addition to mutating human D5 CDR1 loop residues in the hSC-tCDR1 and hSC-tCDR1/tDE chimeras , we also mutated the flanking residue , Phe466 , to a Tyr to preserve stabilizing interactions found in the corresponding fish motifs . To improve purification efficiency , the StrepII tag was replaced with a 6x-His tag for hSC variants produced for spin-labeling and for hSC D1-D3-D4-D5 . Residue numbers and sequences of expression constructs are in Supplementary file 2 . The gene encoding the O . mykiss gene ( GenBank: ADB81776 . 1 ) was codon-optimized for expression in human cells , synthesized ( Blue Heron Biotechnology; Bothell , WA ) and cloned into the pTT5 expression vector ( NRC Biotechnology Research Institute ) with a 5’ Kozak sequence . The construct was modified by site directed mutagenesis to insert a C-terminal 6x-His tag and stop codon following residue Ser213 ( mature sequence numbering ) . SC proteins and variants were expressed in HEK293-6E cells ( NRC Biotechnology Research Institute ) transiently transfected with plasmid DNA using 25 kDa linear polyethylenimine ( Polysciences; Warrington , PA ) . Six days following transfection , cell culture supernatants were subjected to affinity chromatography , either StrepTrap or HisTrap ( GE Healthcare Bio-Sciences; Pittsburgh , PA ) , and further purified by SEC using a Superdex 200 column ( GE Healthcare ) . Purifications were conducted in TBS ( 20 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl ) supplemented with either 2 . 5 mM desthiobiotin or 250 mM imidazole for StrepTrap or HisTrap elution . Human pIgA was provided by J . Vaerman ( Catholic University of Louvain , Brussels , Belgium ) ( Vaerman et al . , 1995 ) and further purified by SEC using a Superose 6 column ( GE Healthcare ) to isolate dIgA . Human pIgM was purchased from Sigma-Aldrich and further purified by SEC , as described for pIgA , to isolate pentameric IgM . Crystals of Strep-tagged hSC ( hSC residues 1–547 plus a Gly-Ser linker ) ( space group P21; a = 61 . 26 Å , b = 242 . 43 Å , c = 63 . 05 Å; β = 114 . 89°; two molecules per asymmetric unit ) were grown by combining 5-7mg/ml protein in TBS with 0 . 2 M ammonium sulfate , 0 . 1 M BisTris pH 5 . 5 , 31% Polyethylene glycol ( PEG ) 3350 using a 1:1 ratio . Crystals grew in sitting drop vapor diffusion at 25°C , and were subsequently soaked in cryoprotectant containing the mother liquor supplemented with 30% glycerol . Iodine derivative crystals of tSC ( residues 1–213 ) were grown by combining 8 . 2 mg/ml tSC with 75 mM MES monohydrate pH 6 . 5 , 9% w/v PEG 20 , 000 , and 7 . 5% glycerol in a 1:1 ratio and iteratively soaking resulting crystals in mother liquor supplemented with 10% glycerol/250 mM NaI , 20% glycerol/500 mM NaI , and 30% glycerol/1000 mM NaI for 5 , 10 , and 60 min respectively . Isomorphous , native crystals ( space group P43212 ( No 96 . ) ; a = 54 . 94 Å , b = 54 . 94 Å , c = 187 . 26 Å; one molecule per asymmetric unit ) were obtained by combining 8 . 2 mg/ml tSC with 75 mM MES monohydrate pH 6 . 5 , 9% w/v PEG 20 , 000 , and 25 mM phenol in a 1:1 ratio and were cryoprotected in mother liquor supplemented with 30% glycerol . Diffraction data from cryopreserved crystals were collected at the Stanford Synchrotron Radiation Laboratory on beamline 12–2 using a PILATUS 6M PAD detector . Data processing was done using Autoxds ( A . Gonzoles and Y . Tsai , SSRL ) , XDS ( Kabsch , 2010 ) , Pointless , Scala and Truncate as implemented in CCP4i ( Collaborative Computational Project , 1994 ) . The crystal structure of hSC was determined using molecular replacement ( MR ) . Search models for each of the five domains were generated by modifying the hSC D1 crystal structure ( pdb code 1XED ) ( Hamburger et al . , 2004 ) using Sculptor ( Bunkoczi and Read , 2011 ) and performing searches using Phaser ( McCoy et al . , 2007 ) as implemented in the AutoMR wizard in Phenix ( Adams et al . , 2010 ) . Successful search strategies resulted in placement of eight of ten possible domains in the asymmetric unit , although only D1 was placed in the same position in both hSC copies . Correct domain assignments were determined by inspection of simulated annealing composite omit maps generated in Phenix ( Adams et al . , 2010 ) and manual building of alternative sequences for D2-D4 using Coot ( Emsley and Cowtan , 2004 ) . D5 , the most divergent domain , was manually built into resulting difference maps using Coot ( Emsley and Cowtan , 2004 ) . The crystallographic asymmetric unit ( ASU ) contained two structurally-similar copies of SC ( room mean square deviation , rmsd , of 0 . 29Å for all common Cα atoms ) . The final hSC model included residues 2–113 , 116–333 , 335–426 , 432–549 ( Chain A ) , residues 2–256 , 260–330 , 335–426 , 432–549 ( Chain B ) , and N-acetyl glucosamine ( NAG ) residues attached to Asn65 , Asn72 , Asn168 , and Asn481 , an α-1 , 6 fucose on the Asn72 NAG , and two SO42- ions . One SO42- is bound to D2 at a position between the A’ and B strands; the other is at the interface between D1 and D4 , where it contacts the C’ and CDR3 from D1 and CDR2 , C” and the C”-D loop in D4 . Regions with unresolved main chain density that were not modeled were restricted to the D1-D2 and D3-D4 linkers and the D4 CDR3 loop . The model , including riding hydrogens and water molecules , was refined using Phenix Refine ( Afonine et al . , 2012 ) using non-crystallographic symmetry ( NCS ) restraints and individual B-factors and validated using MolProbity ( Chen et al . , 2010 ) . NCS differences are largely restricted to the linkers between domains and flexible loops . For example , the D1-D2 linker residues 111–117 , which form crystal contacts with D2 from a neighboring molecule , follows a different path in each NCS copy . This difference causes a 2–3Å shift of Cα atoms in the neighboring D2 C-C’ loop as well different side chain conformations of several loop residues . The structure of tSC was determined by Single Wavelength Anomalous Dispersion ( SAD ) using iodide ions . Derivative data were collected on SSRL 12–2 operating at 7KeV ( 1 . 77Å ) . Iodide sites were located and phasing was completed using the Shelx CDE pipeline ( Sheldrick , 2010 ) as implemented in CCP4 and a preliminary model was built using Phenix Autobuild ( Terwilliger et al . , 2008 ) . A 1 . 7Å resolution native data set , with the same Rfree-flagged reflections as the Iodide derivative , was used to complete model building and refinement using Coot ( Emsley and Cowtan , 2004 ) and Phenix refine ( Afonine et al . , 2012 ) . The structure was validated using MolProbity ( Chen et al . , 2010 ) . The final model included residues tSC 1–52 , 60–213 plus four residues of the C-terminal linker/affinity tag . Structures were analyzed and figures were generated using the Pymol Molecular Graphics System ( Schrodinger LLC ) . Individual domains and NCS copies of hSC were aligned using CEalign in the Pymol Molecular Graphics System ( Schrodinger LLC ) . The hSC domain boundaries ( not including linkers ) were defined as follows: hD1 residues 2–110 , hD2 residues 118–218 , hD3 residues 225–329 , hD4 residues 339–440 and hD5 residues 446–545 and included sulfate ions and glycans . The tSC domains ( not including linker ) were defined as tD1 residues 1–52/60–111 and tD2 114–210 . Domain interfaces were analyzed using PISA , 'Protein interfaces , surfaces and assemblies' service at the European Bioinformatics Institute . ( http://www . ebi . ac . uk/pdbe/prot_int/pistart . html ) ( Krissinel and Henrick , 2007 ) . These analyses utilized modified pdb files in which SC domains were renamed as individual chains that included linkers and were defined as follows: hD1 residues 2–113 , hD2 residues 116–222 , hD3 residues 223–331 , hD4 residues 335–442 , hD5 residues 443–549 , tD1 1–52 , 60–112 and tD2 113–210 and included NAG ligands . Alignments used to generate figures and rmsd calculations for tSC utilized the following residues: tD1 2–6 , 7–35 , 61–67 , 71–91 , 102–111 and tD2 114–118 , 120–148 , 168–174 , 177–197 , 201–210; tD1 1–23 , 29–36 , 41–50 , 61–93 , 103–111 and hD1 3–25 , 33–40 , 45–54 , 63–95 , 102–110; tD2 114–115 , 119–149 , 178–198 , 201–210 , and hD5 446–447 , 450–480 , 508–528 , 536–545 . Inter-domain angles were calculated by determining the angle between the long axes of adjacent domains that had been approximated by ellipsoids calculated from the coordinates using the program Dom_angle ( Su et al . , 1998 ) . The conformations of CDR loops in the hSC structure do not appear to be heavily influenced by crystal packing because the D1 CDRs are similar in conformation to CDRs in the isolated D1 crystal structure ( Hamburger et al . , 2004 ) ( ~0 . 5–1 . 0 Å rmsd after aligning D1 Cα atoms ) and , crystal contacts involving the CDR1 loops in domains D1–D4 are limited to a single side chain interaction . Several side chains of the D5 CDR1 contact a symmetry-related molecule in the crystal; however , these contacts are unlikely to alter the overall D5 CDR1 conformation because the main chain adopts the same conformation as the D2-D4 CDR1 loops . Crystal contacts with the D2 and D3 CDR2 loops involve extensive interactions that might influence the position of side chains . However , the overall conformations of the short , two-residue CDR2 loop in all SC domains appears more likely constrained by its flanking C’-C” strands than by crystal contacts , and the D2 and D3 CDR2 loops adopt the same conformation as the CDR2s in other domains , which are not influenced by crystal contacts . NCBI Genbank accession numbers for SC sequences in alignment figures are: DOG ( NP_001274081 ) , COW ( NP_776568 ) , MOUSE ( NP_035212 ) , RAT ( NP_036855 ) , POSSUM ( AAD41688 ) BOAR ( NP_999324 ) , CHICKEN ( AAQ14493 . 1 ) , FROG ( ABK62772 ) , ANOLE ( XP_008113873 ) , Oncorhynchus mykiss ( ADB81776 . 1 ) , Salmo salar ( ACX44838 . 1 ) , Epinephelus coioides ( ACV91878 . 1 ) , Scophthalmus maximus ( AGN54539 . 1 ) , Danio rerio ( NP_001289179 . 1 ) , Paralichthys olivaceus ( ADK91435 . 1 ) , Cyprinus carpio ( ADB97624 . 1 ) , Takifugu rubripes ( NP_001266944 . 1 ) , Gadus morhua ( AIR74929 . 1 ) . Sequence alignments were completed using ClustalOmega ( Sievers et al . , 2011 ) and corresponding figures were made using Espript 3 ( http://espript . ibcp . fr ) ( Gouet et al . , 1999; Gouet et al . , 2003; Robert and Gouet , 2014 ) . SEC-purified cysteine-substitution variants of hSC for spin-labeling ( T67C/V455C , T67C/Q491C , A80C/V491C , and V455C/Q491C; all including C468A and C502A substitutions ) were diluted by 75% in TBS ( 20 mM Tris-HCl pH 7 . 4 and 150 mM NaCl ) , supplemented with 1 . 5 mM dithiothreitol ( DTT ) to a final concentration of 0 . 5 mM DTT , and incubated at 4°C for 20–60 min . DTT was removed using BioSpin P-6 Columns ( Bio-Rad; Hercules , CA ) in which the protein collection tube contained the required volume of 200 mM 2 , 2 , 5 , 5-tetramethyl-pyrroline-1-oxyl methanethiosulfonate ( HO-225 ) in acetonitrile to produce a five-fold molar excess relative to free cysteine ( 0 . 5–1 µl ) . The HO-225 reagent , which reacts with cysteine to generate the R1 side chain ( Berliner et al . , 1982 ) , was the generous gift of Kalman Hideg ( University of Pecs , Hungary ) . The protein was incubated with HO-225 for 4 hr at 25°C and overnight at 4°C , and then purified by SEC using a Superdex 200 column ( GE Healthcare ) to remove excess HO-225 and isolate monodisperse protein . hSC proteins containing R1 were exchanged into TBS supplemented with 20% glycerol and concentrated to 168–220 µM ( 10–13 . 5 mg/ml ) . SIgA and SIgM complexes were formed by mixing either dIgA or pIgM with D1-67R1/D5-491R1 in a 1:1 molar ratio and exchanged into TBS buffer containing 90% D20 ( Sigma-Aldrich; St . Louis , MO ) and 20% d-8 glycerol ( Sigma-Aldrich ) prior to data collection . The final hSC concentration in the samples with dIgA and pIgM was around ~40 μM . For continuous wave ( CW ) EPR , samples of 5 μl were loaded into 0 . 64 mm ( inner dimension; i . d . ) x 0 . 84 mm ( outer dimension; o . d . ) glass capillaries . Spectra were recorded at room temperature on a Varian E-109 spectrometer fitted with a two-loop one-gap resonator ( Hubbell et al . , 1987 ) at 2 mW incident microwave power and 1 G field modulation amplitude at 100 kHz . For DEER , samples of 20 μl were loaded into glass capillaries ( 1 . 4 i . d . × 1 . 7 o . d . ; VitroCom Inc . , NJ ) and flash-frozen in liquid nitrogen . Four-pulse DEER data at 80K were obtained on a Bruker ELEXSYS 580 operated at Q-band as previously described ( Lopez et al . , 2013 ) with some modifications . A standard four-pulse DEER sequence [ ( π/2 ) νo - τ1 - ( π ) νo - T - ( π ) νp - τ2 - ( π ) νo - τ1 – echo] was employed where νo and νp are the observe and the pump frequencies , respectively . A 36 ns π–pump pulse was set at the maximum of the absorption spectrum and the observer π/2 ( 16 ns ) and π ( 32 ns ) pulses were positioned 50 MHz ( 17 . 8 Gauss ) upfield at the maximum of the center-field absorption line . The delay time τ1 was 200 ns; τ2 varied from 2 . 5 to 6 . 0 μs depending on the sample with a constant step size of 16ns . In samples with deuterated buffer , electron spin echo envelope modulation due to the deuterium nuclei was averaged by adding traces at 8 different τ1 values , starting at 200ns and incrementing by 16ns . Distance distributions were obtained from the raw dipolar evolution data using the program LongDistance available at: http://www . biochemistry . ucla . edu/biochem/Faculty/Hubbell/ . The upper limit of reliable distance ( r ) and width determination ( σ ) for each mutant in nanometers was calculated according to ( Jeschke , 2012 ) : rmax , r≈5tmax/2μs3 rmax , σ≈4tmax/2μs3 , where tmax is the maximum time domain recorded for each sample . Surface plasmon resonance ( SPR ) binding studies were performed using a Biacore T200 instrument ( GE Healthcare ) . Human dIgA or pIgM was immobilized on three of four flow cells of CM5 biosensor chips ( GE Healthcare ) using primary amine coupling ( Biacore manual ) ; the remaining flow cell was mock-coupled and used as a reference surface . Ligand densities used for all experiments except those involving D1-D3-D4-D5 were 164 , 451 , and 765 response units ( RUs ) for the dIgA surfaces , and 569 , 1577 , and 4221 RU for pIgM surfaces . Ligand densities used to evaluate D1-D3-D4-D5 binding were 137 , 483 , and 790 RUs for the dIgA surfaces , and 517 , 1779 , and 3076 RUs for pIgM surfaces . In experiments evaluating D1-D3-D4-D5 binding , D1-D5 and D1 were included as controls to ensure that proteins bound to different sensor chips reproducibly . Sensorgrams shown in figures were generated from the highest density surfaces . A two-fold dilution series in HBS-EP+ buffer ( 10 mM HEPES pH 7 . 4 , 150 mM NaCl , 3 mM EDTA , 0 . 05% ( v/v ) P20 ) was used to test binding of the following analytes ( where the highest concentrations of analyte used on dIgA and pIgM surfaces , respectively , are indicated in parentheses following the analyte name: D1-D5 ( 1 . 02 µM , 0 . 26 µM ) , D1-D5 C468A/C502A ( 1 . 02 µM , 0 . 26 µM ) , D1 ( 1 . 16 µM , 0 . 016 µM ) , D1-D3 ( 1 . 02 µM , 0 . 26 µM ) , D1-D3-D4-D5 ( 1 . 02 µM , 0 . 26 µM ) , D1-D4-D5 ( 1 . 02 µM , 0 . 26 µM ) , D1-D4-D5 C468A/C502A ( 1 . 02 µM , 0 . 26 µM ) , D2-D3 ( 65 . 54 µM , 65 . 54 µM ) , D4-D5 C468A/C502A ( 65 . 54 µM , 65 . 54 µM ) , D2-D5 ( 32 . 8 µM , 32 . 8 µM ) , CDR1 chimera ( 1 . 02 µM , 0 . 26 µM ) , DE chimera ( 1 . 02 µM , 0 . 26 µM ) , CDR1 DE chimera ( 1 . 02 µM , 0 . 26 µM ) . Additional experiments testing hSC D1-D5 ( 1 . 02 µM , 0 . 26 µM ) and D1-67R1/D5-491R1 ( 1 . 02 µM , 0 . 26 µM ) binding to dIgA and pIgM utilized identical parameters except that the HBS-EP+ buffer was supplemented with 20% glycerol to mimic conditions used in DEER experiments . Covalent binding of D1-D4-D5 was tested on a single flow cell of a CM5 biosensor chip with human dIgA surface density of 456 RU . D1-D4-D5 ( C468/C502 intact ) was manually injected at concentrations up to 4 µM and the total response ( RU ) was recorded at pre-injection baseline , binding , stability , and following regeneration . The baseline response returned to pre-injection levels following regeneration , indicating that no D1-D4-D5 was covalently bound to the surface . In contrast , the same experiment utilizing 1 . 02 µM hSC with intact cysteines ( C468/C502 ) resulted in an ~5RU increase in baseline following each injection/regeneration cycle , indicative of covalent binding to the dIgA-coupled surface . In all experiments the flow rate was 50 µl/min , and surfaces were regenerated using 2 . 5 M MgCl2 . All data were collected at 25°C and processed using T200 Evaluation software ( GE Heathcare ) . Equilibrium binding affinity values were calculated using the T200 Evaluation software ( GE Healthcare ) and the 1:1 Langmuir binding model . The reported affinity values were generated by averaging KD values obtained from two independent experiments; the corresponding standard deviation was calculated using Excel ( Microsoft Corporation ) . Response data and equilibrium binding models were exported and re-plotted using Prism ( GraphPad Software ) . Crystallographic atomic coordinates and structure factors have been deposited in the Protein Data Bank ( http://www . rcsb . org ) with codes 5D4K ( hSC ) and 5F1S ( tSC ) .	A sticky substance called mucus lines our airways and gut , where it acts as a physical barrier to prevent bacteria and other microbes from entering the body . Mucus also contains proteins called antibodies that can bind to and neutralize molecules from microbes ( known as antigens ) . The primary antibody found in mucus is called Immunoglobulin A . This antibody is produced by immune cells within the body and must pass through the “epithelial” cells that line the airway or gut to reach the layer of mucus . These epithelial cells have a receptor protein called the polymeric immunoglobulin receptor ( plgR ) that binds to Immunoglobulin A molecules , transports them across the cell , and then releases them into the mucus layer . The pIgR also releases Immunoglobulin A into breast milk , which protects nursing infants until their own immune system has developed . When released into the mucus layer , the Immunoglobulin A antibodies remain attached to a portion of pIgR known as the secretory component . This part of the receptor serves to stabilize and protect the antibodies from being degraded and helps the antibodies to bind to other host and bacterial proteins . Researchers have noted that the secretory component can be released into the mucus even when it is not attached to an antibody . These “free” secretory components have been shown to help prevent bacteria and the toxins they produce from entering the body . Despite the importance of secretory component in immune responses , the three-dimensional structure of the secretory component and how it interacts with antibodies and bacteria remained unknown . Here , Stadtmueller et al . use a technique called X-ray crystallography to determine a three-dimensional model of the free form of a secretory component from humans , and compare it to an ancestral secretory component protein found in fish . Further experiments on the human protein revealed how the structure of the secretory component changes when antibodies bind to it . Stadtmueller et al . propose a model for how both forms of the secretory component can protect the body from microbes and other external agents . The next challenge is to develop a three-dimensional model of the secretory component when it is bound to Immunoglobulin A . DOI: http://dx . doi . org/10 . 7554/eLife . 10640 . 002	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics", "immunology", "and", "inflammation" ]	2016	The structure and dynamics of secretory component and its interactions with polymeric immunoglobulins
How metabolism is reprogrammed during neuronal differentiation is unknown . We found that the loss of hexokinase ( HK2 ) and lactate dehydrogenase ( LDHA ) expression , together with a switch in pyruvate kinase gene splicing from PKM2 to PKM1 , marks the transition from aerobic glycolysis in neural progenitor cells ( NPC ) to neuronal oxidative phosphorylation . The protein levels of c-MYC and N-MYC , transcriptional activators of the HK2 and LDHA genes , decrease dramatically . Constitutive expression of HK2 and LDHA during differentiation leads to neuronal cell death , indicating that the shut-off aerobic glycolysis is essential for neuronal survival . The metabolic regulators PGC-1α and ERRγ increase significantly upon neuronal differentiation to sustain the transcription of metabolic and mitochondrial genes , whose levels are unchanged compared to NPCs , revealing distinct transcriptional regulation of metabolic genes in the proliferation and post-mitotic differentiation states . Mitochondrial mass increases proportionally with neuronal mass growth , indicating an unknown mechanism linking mitochondrial biogenesis to cell size . Neurons rely on oxidative phosphorylation to meet energy demands . Malfunctions of mitochondrial oxidative phosphorylation ( OXPHOS ) lead to a wide range of neurological disorders , and are frequently observed in neurodegenerative diseases ( Lin and Beal , 2006; Schon and Przedborski , 2011; Koopman et al . , 2013 ) . For example , chronic exposure of the brain to the lipophilic pesticide rotenone causes dopaminergic neuron degeneration ( Betarbet et al . , 2000 ) . Hereditary mutations in OXPHOS genes cause Leigh syndrome , a severe early childhood neurodegeneration ( Finsterer , 2008 ) . Little is known about the molecular basis of neurons’ preference for oxidative phosphorylation , or neuronal metabolic responses to the energy crisis and their implications for disease progression . It is imperative to understand how neuronal metabolism , defined by the metabolic enzymes in pathways , such as glycolysis , TCA cycle and mitochondrial OXPHOS , is set up during development , is maintained during adult life and is altered in neurological disorders . Glycolysis and TCA cycle are major pathways providing metabolic precursors for biosynthesis and energy production . The activities and metabolic flux of these pathways are delicately tuned to ensure optimal resource distribution , conforming to cellular function . The balance between glycolysis and mitochondrial oxidative phosphorylation is essential for stem cell function ( Shyh-Chang et al . , 2013; Teslaa and Teitell , 2015 ) . For instance , deletion of lactate dehydrogenase A ( LDHA ) impairs maintenance and proliferation of hematopoietic stem and progenitor cells ( Wang et al . , 2014 ) . Most cancer cells use aerobic glycolysis to generate energy , a phenomenon called the Warburg effect , in which a large fraction of glycolytic pyruvate is converted into lactate ( Vander Heiden et al . , 2009 ) . Moreover , under low oxygen condition , the transcription of glycolytic genes is upregulated by hypoxia-inducible factor ( HIF ) , and increased glycolytic ATP produced by enhanced glycolytic flux can partially supply cellular energy demands ( Iyer et al . , 1998; Seagroves et al . , 2001; Semenza , 2012 ) . A pioneering study using embryonic Xenopus retina revealed that neural progenitor cells ( NPCs ) are less reliant on oxidative phosphorylation for ATP production than are non-dividing differentiated neurons , and the transition from glycolysis to oxidative phosphorylation is tightly coupled to neuronal differentiation , though the exact molecular basis underlying the transition is unknown ( Agathocleous et al . , 2012 ) . Studies in cardiomyocytes provide an example of how a metabolic transition is regulated during development ( Leone and Kelly , 2011 ) . Around the postnatal stage , cardiomyocytes exit from the cell cycle and gradually enter a maturation process; mitochondrial oxidative activity increases concurrently with elevated expression of mitochondrial genes . The key transcription factors involved are PPARα and its coactivator PGC-1α , which control a broad range of metabolic and mitochondrial genes . PGC-1α may also play a key role in neuronal metabolism , as PGC-1α knockout mice show obvious neurodegenerative pathology ( Lin et al . , 2004 ) . Neuronal differentiation from human NPCs derived from embryonic stem cells or induced pluripotent stem cells ( iPSCs ) is able to recapitulate the in vivo developmental process and has been successfully used to model a variety of neurological diseases ( Qiang et al . , 2013 ) . We used this neuronal differentiation model to explore neuronal metabolic differentiation . The disappearance of HK2 and LDHA , together with a PKM2 splicing shift to PKM1 , marks the transition from aerobic glycolysis in NPCs to oxidative phosphorylation in neurons . The protein levels of c-MYC and N-MYC , which are transcriptional activators of HK2 , LDHA and PKM splicing , decrease dramatically . Constitutive expression of HK2 and LDHA results in neuronal cell death , indicating that turning off aerobic glycolysis is essential for neuronal differentiation . The metabolic regulators PGC-1α and ERRγ increase significantly upon differentiation; and their up-regulation is required for maintaining the expression of TCA and mitochondrial respiratory complex genes , which , surprisingly , are largely unchanged compared to NPCs , revealing distinct transcriptional regulation of metabolic genes in the proliferation and post-mitotic differentiation states . Mitochondrial mass increases proportionally with neuronal mass growth , indicating an unknown mechanism linking neuronal mitochondrial biogenesis to cell size . In addition , OGDH , a key enzyme in the TCA cycle , has a novel and conserved neuronal splicing shift , resulting in the loss of a calcium binding motif . NPCs were derived from iPSCs reprogrammed from the human BJ male fibroblast line . The protocol for NPC establishment and neuronal differentiation is outlined in Figure 1—figure supplement 1 . To obtain NPC lines of high purity , colonies containing neural rosettes were manually selected and picked as described in Materials and methods and Figure 1—figure supplement 2 . The identity and purity of NPCs were examined by anti-Sox2 and Nestin staining ( Figure 1A ) . Only high-quality NPC lines containing more than 90% Sox2 and Nestin double-positive cells were used for experiments . After 3 weeks of differentiation , a majority ( ~85% ) of cells expressed the neuronal marker MAP2 ( Figure 1B ) . Although rare at 3 weeks , glial cells emerged and proliferated after 4–5 weeks; therefore , 3-week neuronal cultures were used to represent a population of developing neurons . Consistent with a previous study ( Johnson et al . , 2007 ) , neurons after ~4 weeks could be induced to fire multiple action potentials ( Figure 1—figure supplement 3 ) . 10 . 7554/eLife . 13374 . 003Figure 1 . Gene expression of glycolysis and TCA during neuronal differentiation . ( A ) Human BJ iPSC-derived NPCs showed homogeneous expression of the NPC markers Nestin and Sox2 . ( B ) The left panel shows NPC-derived neurons after 3 weeks of differentiation; right panel shows staining of MAP2 , a neuronal marker . ( C ) The top 200 up- and down-regulated genes during neuronal differentiation are depicted by a heatmap . Red and green intensities indicate fold increases and decreases , respectively , in gene expression ( expressed as log2 ) . ( D ) GO term analysis of genes up-regulated during neuronal differentiation . The top eight GO term biological process categories obtained are ranked by p-value . ( E ) The FPKM values of known neuron-specific genes are shown as log10-fold change . ( F ) The fold changes of FPKM values of proliferating NPC markers are shown . Bars represent mean ± SD of four RNA-seq replicates for NPCs and neurons differentiated at 1 and 3 weeks . ( G , H , I and J ) Relative expression levels of the key metabolic genes in glycolysis , tricarboxylic acid cycle ( TCA ) , pyruvate dehydrogenase ( PDH ) complex and pentose phosphate pathway . Bars show the mean of FPKM values of differentiated neurons at 1 and 3 weeks relative to FPKM values of NPCs . Error bars represent SD of four RNA-seq replicates at each time point . Abbreviations , dihydroxyacetone phosphate ( DHAP ) ; 1 , 3-bisphosphoglyceric acid ( 1 , 3 BPG ) ; 3-phosphoglyceric acid ( 3PG ) ; phosphoenolpyruvic acid ( PEP ) , α-ketoglutarate ( α-KG ) ; oxaloacetate ( OAA ) . ( Figure 1—sourcer data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 00310 . 7554/eLife . 13374 . 004Figure 1—source data 1 . FPKM values of glycolysis , TCA , PDH and pentose phosphate pathway in NPCs and differentiated neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 00410 . 7554/eLife . 13374 . 005Figure 1—figure supplement 1 . The outline of the protocol used to differentiate neurons from iPSCs ( upper panel ) . Representative pictures of fibroblasts , iPSCs , NPCs and neurons ( lower panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 00510 . 7554/eLife . 13374 . 006Figure 1—figure supplement 2 . Colonies containing neural rosettes , and type 4 colony produces NPCs of the best quality . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 00610 . 7554/eLife . 13374 . 007Figure 1—figure supplement 3 . Electrophysiological study of BJ 5-week neurons . Representative results were shown . ( a ) Patched neuron filled with rhodamine; ( b ) Evoked voltage dependent sodium and potassium currents recorded in voltage-clamp ( −70 mV ) . ( c ) Evoked action potential . ( d ) Spontaneous burst of action potentials . ( e ) Zoom of 'd' . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 00710 . 7554/eLife . 13374 . 008Figure 1—figure supplement 4 . FPKM values of paralogous genes in glycolysis . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 00810 . 7554/eLife . 13374 . 009Figure 1—figure supplement 5 . Metabolites quantified by gas chromatography mass spectrometry ( GC-MS ) . NPCs at early passage ( P3 ) and 3-week neurons were incubated in fresh medium for 12 hr , and metabolites in cells were extracted and analyzed by GC-MS . The relative amount of metabolites extracted from neurons versus NPCs was obtained after normalization to protein content . Bars represent mean ± SD of three biological replicates for NPCs and 3-week neurons . ( Figure 1—figure supplement 5—sourcer data 1 ) DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 00910 . 7554/eLife . 13374 . 010Figure 1—figure supplement 5—source data 1 . Metabolites quantified by GC-MS . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01010 . 7554/eLife . 13374 . 011Figure 1—figure supplement 6 . Oxygen consumption rate ( OCR ) analysis by Seahorse extracellular flux analysis . FCCP ( F ) is a chemical uncoupler of electron transport and oxidative phosphorylation; Rotenone and Antimycin A ( R&A ) are complex I and III inhibitors . Error bars represent SD . None mitochondrial OCR has been subtracted . ( Figure 1—figure supplement 6—source data 1 ) . In this experiment , NPCs had to be grown at high density to avoid differentiation and ensure optimal proliferation . The large amount of lactate secreted by high-density NPCs under these conditions decrease the pH value of the medium , which lacks sodium bicarbonate , significantly , from 7 . 4 to 7 . 0 , during incubation and measurement . The NPC ECAR readings were out of the linear measurement range and not shown here . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01110 . 7554/eLife . 13374 . 012Figure 1—figure supplement 6—source data 1 . OCR measurement by Seahorse extracellular flux analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01210 . 7554/eLife . 13374 . 013Figure 1—figure supplement 7 . The glucose and lactate concentrations in the medium growing NPCs and 3-week neurons were quantified by YSI 2950 metabolite analyzer . The lactate production/ glucose consumption ratio was calculated as the amount of lactate divided by used glucose . Bars represent mean ± SD . n= 3 . In theory , if one glucose molecule is completely converted into lactate ( no entry into mitochondrial TCA cycle ) , the ratio of lactate/glucose would be 2 . ( Figure 1—figure supplement 7—source data 1 ) DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01310 . 7554/eLife . 13374 . 014Figure 1—figure supplement 7—source data 1 . The ratio of lactate production/glucose consumption . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 014 To profile the transcriptomes of NPCs and 1- and 3-week-old differentiated neurons , RNA-seq experiments were performed with two NPC lines established from independent BJ iPSC clones . For each time point , two experimental duplicates were used for each NPC line and their differentiated neurons . The sequencing reads were mapped to a human reference genome by STAR and assembled with Cufflinks ( Dobin et al . , 2013; Trapnell et al . , 2012 ) . The median sequencing read yield for each sample was ~35 million 100-base reads , of which ~85% were mapped . The FPKM ( Fragments Per Kilobase of transcript per Million mapped reads ) value calculated by the Cufflink algorithm was used to represent the gene expression level ( Trapnell et al . , 2010 ) . The top 200 genes up- and down-regulated during neuronal differentiation were clustered and are presented as a heatmap ( Figure 1C ) . Most of the down-regulated genes were cell cycle related , whereas up-regulated genes were enriched in neuronal function pathways , including synapse formation , transmission of nerve impulses , gated channel activity and neuronal projections ( Figure 1D ) . Neuron-specific genes such as NEUROD2 , MYT1L , MAPT ( Tau ) , SYN1 ( synapsin ) , CHD5 ( Egan et al . , 2013 ) were significantly increased ( Figure 1E ) , whereas the markers associated with proliferating NPCs , LIN28A/B and FOXM1 ( Cimadamore et al . , 2013; Karsten et al . , 2003 ) , were considerably decreased ( Figure 1F ) . These data indicate that our neuronal differentiation model is reliable . Through the 10 enzymatic steps of glycolysis , glucose is metabolized to pyruvate . Subsequently , glycolytic pyruvate enters into the mitochondrial TCA cycle to generate NADH and FADH2 , which are utilized by mitochondrial OXPHOS complexes to generate ATP . For most of the glycolytic reactions , there are multiple paralogous enzymes , but according to our RNA-seq data , for a majority of the reactions only a single enzyme was predominantly expressed in NPCs or neurons ( Figure 1—figure supplement 4 ) . As shown in Figure 1G , expression of the majority of glycolysis genes was decreased in neurons , particularly enzymes acting at the steps after fructose 1 , 6 phosphate formation , i . e . , ALDOA to LDHA/B , which declined to half to one-third of their level in NPCs . An exception was ENO2 , a known neuron-specific gene , which increased by ~two fold . GLUT1/3 and HK2 , encoding glucose transporters and hexokinase , respectively , dropped ~ten fold . Notably , HK1 , the other hexokinase , and PFKM , a phosphofructokinase , functioning at the first two irreversible steps of glycolysis , exhibited minor increases . In contrast to a general decrease in glycolysis genes , TCA genes remained at the same level during differentiation; an exception was isocitrate dehydrogenase , IDH2 , the mitochondrial isoform of IDH , which dropped nearly 50% ( Figure 1I ) . The mitochondrial pyruvate dehydrogenase complex ( PDC ) converts pyruvate to acetyl-CoA for TCA cycle entry . Expression of the genes encoding the subunits of PDC itself , including PDHA1 , PDHB , PDHX , DLAT , and DLD , showed no significant changes ( Figure 1H ) . The enzymatic activity of PDC is regulated by phosphorylation , being inhibited by a family of protein kinases ( PDKs ) and activated by a family of protein phosphatases ( PDPs ) . Except for PDK2 , RNA levels of the other three PDKs declined significantly during neuronal differentiation; in contrast , PDP1 RNA levels increased ~2 . 5 fold at 3 weeks of differentiation ( Figure 1H ) . The overall changes in PDC kinase and phosphatase levels would favor increased PDC complex activity in neurons , and indeed PDH Ser300 phosphorylation , a major inhibitory site , decreased considerably during neuronal differentiation consistent with an increase in PDC activity ( see Figure 3A ) . The pentose phosphate pathway generates NADPH and ribose-5-phosphate . NADPH is used in reductive biosynthesis reactions; ribose-5-phosphate is a precursor for nucleotide synthesis . Overall , the genes at the beginning steps involved in NADPH production showed unchanged expression levels during differentiation , but the levels of the ones on the branch required for nucleotide synthesis decreased , consistent with neuronal post-mitotic status ( Figure 1J ) . To further define the metabolic profiles of glycolysis and TCA pathways , we measured representative glycolytic and TCA metabolites by gas chromatography mass spectrometry ( GC-MS ) ( Figure 1—figure supplement 5 ) . Consistent with the overall decreased gene expression in glycolysis and glucose transporters , the levels of 3-phosphoglyerate ( 3PG ) and pyruvate in neurons dropped to ~52% and 32% of those in NPCs . The levels of TCA intermediates in neurons were also lower than those in NPCs: citrate was ~32% of that in NPCs , α-ketoglutarate was ~17% , fumarate was ~30% , and malate was ~22%; succinate was relatively higher , at ~62% , which may be due to the decreased protein level of SDHB in complex II , which catalyzes the conversion from succinate to fumarate ( see Figure 5C ) . It should be noted that absolute metabolite levels do not directly reflect true metabolite flux , i . e . , the observed reduction in TCA cycle intermediates does not necessarily mean that carbon flux through the TCA cycle is decreased . To directly estimate mitochondrial oxidative activity , the oxygen consumption rate ( OCR ) was measured in 3-week differentiated neurons and in early passage ( P3 ) NPCs using the Seahorse extracellular flux analyzer ( Figure 1—figure supplement 6 ) . The basal and maximum OCRs of NPCs were ~82 and 130 ( pmoles/min/10 µg ) , whereas those in 3-week differentiated neurons were ~98 and 183 , significantly higher than in NPCs , consistent with increased TCA cycle flux . The lactate production/glucose consumption ratio was measured to estimate the fraction of pyruvate that was converted into lactate and secreted or else transported into mitochondria . As shown in Figure 1—figure supplement 7 , the ratio was ~1 . 61 lactate/glucose for NPCs , while in neurons , it was ~0 . 35 , consistent with a significant decrease in aerobic glycolysis in neurons . Besides expression level , mRNA splicing also influences the activities of several metabolic enzymes . Pyruvate kinase ( PKM ) is such an example . Alternative RNA splicing generates two isoforms from the PKM gene , PKM1 and PKM2 , which have their own exclusive exons , exon 9 for PKM1 and exon 10 for PKM2 . PKM1 is normally found in tissues where oxidative phosphorylation is preferentially used to generate ATP , whereas PKM2 is expressed in proliferating cells with high anabolic activity ( Imamura and Tanaka , 1972 ) . PKM1 is constitutively active , while , PKM2 is subjected to allosteric regulation and can adopt a low or high activity state ( Anastasiou et al . , 2012 ) . It is still unclear how PKM1/2 isoforms control cell metabolism in proliferation and non-proliferation states . Replacing PKM2 with PKM1 in cancer cells reduces lactate production and increases oxygen consumption ( Christofk et al . , 2008 ) , indicating that PKM2 may favor high flux glycolysis . A recent study reveals that PKM1 expression does not affect upstream glycolytic intermediates but significantly reduces nucleotide biosynthesis ( Lunt et al . , 2015 ) . To find out whether changes in alternative mRNA splicing occurred in glycolysis , TCA and mitochondrial respiratory complex genes during neuronal differentiation , the distributions of exonic RNA-seq reads were manually compared between NPCs and neurons using Integrative Genomics Viewer ( Robinson et al . , 2011 ) . Two genes were found to exhibit such mRNA splicing shifts ( Figure 2A ) . One was PKM in glycolysis; the other was OGDH ( α-ketoglutarate dehydrogenase ) in the TCA cycle . NPCs mainly expressed PKM2 , whereas neurons expressed PKM1 , as confirmed by RT-PCR with isoform-specific primers ( Figure 2B ) . In contrast , primary human astrocytes only expressed PKM2 . This result was consistent with a recent finding that mouse neurons exclusively expressed PKM1 , whereas astrocytes only expressed PKM2 ( Zhang et al . , 2014 ) . Our results established that the splicing shift from PKM2 to PKM1 occurred during neuronal differentiation from NPCs . PKM splicing or the PKM2/PKM1 ratio is known to be controlled by the expression levels of nuclear ribonucleoprotein ( hnRNP ) proteins , hnRNPI , hnRNPA1 and hnRNPA2 , and in cancer cells , these proteins are expressed at high levels and bind repressively to the sequences flanking exon 9 , favoring the exon 10 addition thus generating more PKM2 ( David et al . , 2010; Clower et al . , 2010 ) . Consistent with this model , during neuronal differentiation , hnRNPI , hnRNPA1 and hnRNPA2 RNA levels decreased to nearly 40% compared to NPCs ( Figure 2C ) . 10 . 7554/eLife . 13374 . 015Figure 2 . Neuron-specific splicing of PKM and OGDH . ( A ) The upper panel shows the RNA-seq reads obtained from NPCs and neurons mapped to the PKM and OGDH chromosome locus using Integrative Genomics Viewer . The lower schematic diagram depicts the organization of exons near the cell type-specific splicing site . Red box and line represent exon splicing unique to NPCs; blue ones represent that unique to neurons . PKM2 and OGDH1 are unique to NPCs; PKM1 and OGDHneu are predominantly for neurons . ( B ) Validation of PKM , OGDH splicing by PCR . Primers were designed to amplify the unique splicing region and common region of PKM1/2 and OGDH1/neu mRNA . PCR was carried out with cDNA prepared from NPCs , neurons at 3 week and primary human astrocytes . ( C ) The fold changes of FPKM values of hnRNPI , hnRNPA1 and hnRNPA2 are shown . Bars represent mean ± SD of four RNA-seq replicates for NPCs and neurons differentiated at 1 and 3 week . ( D ) The RNA-seq reads obtained from purified mouse astrocytes and neurons mapped to the PKM and OGDH chromosome locus using Integrative Genomics Viewer . The original RNA-seq data were from Zhang et al . ( 2014 ) . ( E ) Alignment of amino acid sequences encoded by the alternative exons of human and mouse OGDH1/neu . OGDH1 contains a calcium-binding motif underlined in red , absent in OGDHneu . ( Figure 2—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01510 . 7554/eLife . 13374 . 016Figure 2—source data 1 . FPKM values of hnRNPI , hnRNPA1 and hnRNPA2 . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 016 OGDH encodes the E1 subunit of the α-ketoglutarate dehydrogenase complex , catalyzing the conversion of α-ketoglutarate to succinyl-CoA in the TCA cycle . Mitochondrial calcium increases can further boost its activity ( Denton , 2009 ) . OGDH in NPCs ( OGDH1 ) uses exon 4 , whereas the isoform in neurons , here called OGDHneu , uses two adjacent exons ( Figure 2A ) . The alternative splicing of OGDH was confirmed by RT-PCR with isoform-specific primers ( Figure 2B ) . Primary human astrocytes also only expressed OGDH1 ( Figure 2B ) . Different from the more exclusive PKM splicing , there was still considerable expression of OGDH1 in neurons . OGDHneu splicing is conserved in mouse neurons , whereas mouse astrocytes , like human astrocytes , only expressed OGDH1 ( Figure 2D ) . As a result of this splice switch , a stretch of 34 amino acids in OGDH1 , containing a Ca2+-binding motif ( Armstrong et al . , 2014 ) , is replaced by 45 amino acids specific to OGDHneu ( Figure 2E ) . In a recent study , Denton et al . ( 2016 ) also detected this splice form of OGDH in the whole brain tissue and pancreatic islets , calling it LS1 , and , importantly , they confirmed that LS1 isoform is insensitive to Ca2+ . Note , that we did not find the OGDHneu isoform in RNA-seq data from cardiomyocytes , indicating that , unlike PKM1 , OGDHneu is not specific for cells that preferentially use oxidative phosphorylation . Calcium is used in a variety of neuronal functions . We suspect that OGDH neuron-specific splicing may avoid a prolonged activation of the OGDH complex as a result of frequent intracellular calcium increases . Next , we surveyed the protein levels of representative glycolysis and TCA enzymes by immunoblotting ( Figure 3A ) . Equal amounts of protein extracts from NPCs and neurons at days 3 , 7 , and 21 of differentiation ( D3 , D7 , D21 , respectively ) were resolved by SDS-PAGE . The protein levels of the enzymes examined were largely consistent with their RNA levels . HK2 and LDHA mRNAs were about 15% and 25% of those in NPCs , but strikingly , both proteins almost completely disappeared upon differentiation ( Figure 3A ) . To further confirm HK2 and LDHA expression patterns , immunostaining was done on NPCs and 3-week differentiated neurons . Consistent with the immunoblotting data , HK2 and LDHA could readily be detected in NPCs but not in neurons ( Figure 3B ) . HK2 catalyzes the ATP-dependent phosphorylation of glucose to yield glucose-6-phosphate , an irreversible step in glycolysis dictating the amount of glucose entry . LDHA catalyzes the conversion pyruvate into lactate at the last step of extended glycolysis , diverting pyruvate from the mitochondrial TCA cycle and recycling NAD+ to sustain high flux glycolysis . The drastic decreases in HK2 and LDHA , the two key enzymes supporting aerobic glycolysis ( DeBerardinis and Thompson , 2012; Dang , 2012 ) , are consistent with neurons not having high flux aerobic glycolysis and exhibiting low lactate production . For the TCA enzymes , the levels of citrate synthase ( CS ) , isocitrate dehydrogenase ( IDH2 ) and succinyl-CoA ligase β subunit ( SUCLA2 ) were constant during differentiation ( Figure 3A ) , even though IDH2 showed ~50% decrease at the RNA level . Based on the RNA and protein data , we conclude that there are no significant changes in TCA gene expression in neurons compared to NPCs . To explore if the loss of HK2 and LDHA expression during neuronal differentiation is conserved , we checked HK2 and LDHA expression during differentiation of mouse neuroprogenitor cells derived from embryonic stem cell . As shown in Figure 3—figure supplement 1 , both HK2 and LDHA mRNA levels dropped , falling to ~8 . 5% and ~11% of the level in NPCs , respectively , and , consistently , their protein levels were also greatly decreased . In contrast , as in human cells , sustained levels of HK1 mRNA and protein , were observed during mouse neuronal differentiation . Similar results were observed in primary mouse embryonic neurons at E18 . Thus , it appears that the disappearance of LDHA is , mechanistically , a major switch for downregulating aerobic glycolysis as NPCs differentiate , allowing a transition into a neuronal mitochondrial oxidative phosphorylation state . 10 . 7554/eLife . 13374 . 017Figure 3 . Characterization of HK2 and LDHA in NPCs , differentiated neurons and neurons directly converted from fibroblasts . ( A ) Immunoblotting analysis of the representative metabolic genes in glycolysis , tricarboxylic acid cycle ( TCA ) pathways . 20 µg of protein lysate from NPCs and from neurons differentiated for 3 , 7 and 21 days ( D3 , D7 , D21 ) were loaded . ( B ) Immunostaining analysis of HK2 and LDHA in NPCs and 3-week neurons . ( C ) Effects of HK2 or LDHA knockdown on NPC proliferation . NPCs at early passage ( P2 ) were seeded in 24-well plates one day before infection . The NPC number was determined at 5 days after infection with lenti-shRNA virus against HK2 or LDHA . Two effective shRNA lentiviral vectors targeting different regions of HK2 or LDHA were used . Scramble shRNA vector was used as control . Error bars represent ± SD , n= 3 . The knockdown efficiency was confirmed by immunoblotting . ( D , E ) Immunostaining analysis of LDHA in neurons directly converted from fibroblasts . Two protocols were applied: one was to knockdown PTB1 , a single RNA-binding protein , and the other was to overexpress proneuronal transcription factors Ngn2 and Ascl1 . Tuj1 ( ß-III tubulin ) and Tau were stained as early and mature neuronal markers , respectively . The above experiments were repeated at least three times . ( F ) ChIP analysis of HK2 and LDHA promoters using anti-cMYC or N-MYC antibodies and rabbit IgG as control . Chromatin were prepared from NPCs . The enrichment values are shown as percentage normalized to input . N . C . stands for non-specific control . Bars are mean ± SD , n= 3 . Immunoblotting and real time PCR analysis of HK2 and LDHA expression in NPCs with inducible c-MYC . mRNA expression levels of HK2 and LDHA relative to those from non-induction control were calculated after normalization to β-actin . Bars are mean ± SD , n= 3 . ( G ) Immunoblotting analysis of c-MYC , N-MYC and Max in NPCs and 3-week neurons . ( Figure 3—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01710 . 7554/eLife . 13374 . 018Figure 3—source data 1 . Knockdown effect on NPC proliferation and Myc control of HK2 and LDHA in NPCs . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01810 . 7554/eLife . 13374 . 019Figure 3—figure supplement 1 . Immunoblotting analysis of mouse HK1 , HK2 and LDHA in mouse NPCs derived from embryonic stem cell ( ES-E14TG2a ) , 2-week differentiated neurons and embryonic neurons at E18 . The relative mRNA expression levels of HK1 , HK2 and LDHA in neurons were calculated against the levels in NPCs . Bars are mean ± SD , n= 3 . ( Figure 3—figure supplement 1—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 01910 . 7554/eLife . 13374 . 020Figure 3—figure supplement 1—source data 1 . RT-PCR analysis of HK1 , HK2 and LDHA during mouse neuronal differentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 02010 . 7554/eLife . 13374 . 021Figure 3—figure supplement 2 . Immunoblotting and real time PCR analysis of HK2 expression in neurons with inducible c-MYC . mRNA expression levels of HK2 relative to those from non-induction control were calculated after normalization to β-actin . Bars are mean ± SD , n= 3 . ( Figure 3—figure supplement 2—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 02110 . 7554/eLife . 13374 . 022Figure 3—figure supplement 2—source data 1 . Activation of HK2 by ectopic c-Myc expression in neuron . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 022 Because loss of LDHA considerably impairs hematopoietic stem and progenitor cells during hematopoiesis ( Wang et al . , 2014 ) , we examined whether HK2 and LDHA are required for human NPC propagation by performing knockdown experiments . BJ NPC cells at passage 3 were infected with lenti-shRNAs targeting HK2 or LDHA . For each gene , two shRNA constructs against different mRNA region were used; scramble shRNA was used as control . Depletion of HK2 or LDHA significantly slowed the propagation of NPCs by 2 . 6~3 . 5 fold ( Figure 3C ) . These results indicate that , as found in cancer cell lines ( Fantin et al . , 2006; Wolf et al . , 2011 ) , aerobic glycolysis is required for NPC proliferation . The inhibitory effect of HK2 depletion on NPC proliferation is consistent with a recent finding that ablation of HK2 diminishes aerobic glycolysis and disrupts cerebellar granule neuronal progenitor development ( Gershon et al . , 2013 ) . MYC , a key transcription factor associated with cell cycle progression , directly activates HK2 and LDHA transcription , and the promoter regions of HK2 and LDHA contain MYC binding sites ( 5'-CACGTG-3' ) , also known as E box ( Kim et al . , 2004 ) . MYC also promotes a high PKM2/PKM1 ratio by upregulating heterogeneous nuclear ribonucleoproteins ( hnRNP ) that regulate PKM alternative splicing ( David et al . , 2010 ) . c-MYC and N-MYC redundantly promote neural progenitor cell proliferation in the developing brain , with the brains of double knockout mice exhibiting profoundly impaired growth ( Wey and Knoepfler , 2010 ) . As shown in Figure 3F , chromatin immunopreciptation with anti-c-MYC and N-MYC antibodies in human NPCs showed that both MYCs could bind to the HK2 and LDHA promoters in the region containing the E boxes as reported ( Kim et al . , 2004 ) . Consistently , we found that overexpression of c-MYC using a doxycycline-inducible lentivirus upregulated HK2 and LDHA expression in NPCs ( Figure 3F ) , indicating that MYC controls HK2 and LDHA expression in these cells similarly to other types of cells previously studied . Both c-MYC and N-MYC protein levels decreased significantly , by ~70% and ~85% respectively , once NPCs differentiated into neurons ( Figure 3G ) . Dox-induced expression of c-Myc is sufficient to re-activate HK2 expression in neurons ( Figure 3—figure supplement 2 ) . Therefore , the decreased expression of HK2 and LDHA during neuronal differentiation was most likely attributable predominantly to the observed MYC decrease occurring as NPCs exited from the cell cycle and entered the differentiation process . Functional induced neurons ( iN ) can be generated directly from fibroblasts by forced expression of neural specific transcription factors , a process also called transdifferentiation ( Vierbuchen et al . , 2010 ) . However , it is unknown whether the metabolism of directly converted neurons is the same as that of NPC-differentiated neurons . To test if LDHA was also turned off in neuronal transdifferentiation , two reprogramming protocols were applied . One was to knock down a single RNA-binding protein PTB ( Xue et al . , 2013 ) , and the other was to overexpress the proneuronal transcription factors Ngn2 and Ascl1 ( Ladewig et al . , 2012 ) . Both approaches generated Tuj1-positive neurons after 3 weeks of transduction . The second protocol using Ngn2 and Ascl1 overexpression generated more iNs , and the Tuj1-positive neurons did not have an LDHA immunofluorescence signal ( Figure 3D ) . In contrast , with PTB knockdown , a much higher rate ( ~80% ) of colocalization of LDHA and Tuj1 was observed ( Figure 3D ) . Furthermore , while Ngn2/Ascl1-derived iNs expressed Tau , a mature neuronal marker , no Tau expression was detected in iNs from PTB knockdown , suggesting that PTB KD iNs were not as mature as Ngn2/Ascl1-derived iNs ( Figure 3E ) . These data established that LDHA was also downregulated in neurons directly converted from fibroblasts , indicating a metabolic resetting during transdifferentiation . To determine if shutoff of aerobic glycolysis is critical for neuronal differentiation , we constitutively co-expressed HK2 and LDHA ( Flag-tagged ) in NPCs using a genome-integrating transposon-based vector , and examined the effects on neuronal differentiation . Their expression was confirmed by immunoblotting ( Figure 4A ) , and immunostaining with anti-FLAG antibody revealed that ~80% of NPCs have discernible exogenous LDHA expression ( Figure 4B ) . Three-week neuronal cultures differentiated from NPCs constitutively expressing HK2 and LDHA showed a significant fraction of GFAP-positive glial cells , ~40% of total cells , compared to only ~4% GFAP-positive cells in neuronal cultures from control vector-transduced NPCs . Consistently , real-time PCR analysis of GFAP abundance in total mRNA also showed a nearly eight-fold increase ( Figure 4C ) . The neuronal cultures from NPCs constitutively expressing HK2 and LDHA had ~40% condensed nuclei , an indicator of apoptotic cells , which was much higher than control , ~6% , and such increased cell death was not observed in NPCs ( Figure 4D ) . The LDHA signal was clearly detected in GFAP-positive cells with anti-LDHA antibody ( Figure 4E ) or anti-FLAG antibody staining ( Figure 4H ) , but there was no colocalization with MAP2 , a neuronal marker . Anti-LDHA and FLAG staining also showed punctate patterns ( Figure 4E and H ) , which were associated with punctate MAP2 staining and condensed nuclei ( Figure 4F and G ) , indicating that dead cells arose from neurons with exogenous LDHA expression . The results suggest that neurons cannot tolerate constitutive HK2 and LDHA expression and die during differentiation . Neurons detected in the culture differentiated from NPCs constitutively expressing HK2 and LDHA may have been derived from the fraction of NPCs not expressing HK2 and LDHA . Currently , we do not understand why glial cells increase in the population . 10 . 7554/eLife . 13374 . 023Figure 4 . Shutoff of aerobic glycolysis is critical for neuronal differentiation . ( A ) Immunoblotting analysis of HK2 and LDHA in NPCs and 3-week neurons constitutively expressing HK2 and LDHA ( Flag-tagged ) . ( B ) Immunostaining of NPCs with anti-FLAG antibody ( green ) , and nuclear staining was done with Hoechst ( red ) . The percentage of Flag-positive cells were quantified , and 100 cells were counted for each group . ( C ) Immunostaining analysis of MAP2 and GFAP in 3-week neurons . The percentage of GFAP and MAP2 cells were quantified , and 100 cells were counted for each group , and three times of neuronal differentiation were included . Bars are mean ± SD , n= 3 . The GFAP mRNA abundance in the RNA extracted from neuronal culture was quantified by real-time PCR and normalized to β-actin , and presented as a fold increase compared to neurons differentiated from control NPCs . Bars are mean ± SD , n = 3 . ( D ) Nuclear staining with Hoechst in NPCs and 3-week neurons . The percentages of condensed nuclear were quantified , and 50 cells were counted for each group . Bars are mean ± SD , n= 3 . ( E ) Immunostaining analysis of LDHA , MAP2 and GFAP in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA . ( F , G ) Colocalization of irregular puntated staining of LDHA ( green ) with MAP2 ( red , in F ) or condensed nuclear stained with Hoechst ( red , in G ) in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA . ( H ) Immunostaining analysis of anti-FLAG and GFAP in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA . ( I ) Immunoblotting analysis of AMPK T172 phosphorylation in the cell lysate extracted from day 4 and day 21 neuronal culture differentiated from NPCs expressing HK2 and LDHA . The AMPK T172 phosphorylation was quantified after normalized to non-phosphoAMPK , and presented as fold increase . ( J ) Lactate in medium from day-4 neuronal culture differentiated from NPC expressing HK2 and LDHA were quantified and normalized by protein content , and presented as percentage compared to those from control neurons , bars are mean ± SD , n= 3 . ( K ) Immunostaining analysis of LDHA , MAP2 and GFAP in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA . The neuronal differentiation medium used contained 5 mM sodium pyruvate , ten fold of the standard concentration ( 0 . 5 mM ) . The above experiments were repeated at least three times . ( Figure 4—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 02310 . 7554/eLife . 13374 . 024Figure 4—source data 1 . Constitutive expression of HK2 and LDHA is detrimental for neuronal differentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 024 As aerobic glycolysis diverts pyruvate from mitochondrial oxidative phosphorylation for energy production by converting it into secreted lactate , a likely reason for the neuronal death is mitochondrial ATP production deficiency . As neuronal cultures differentiated from NPCs constitutively expressing HK2 and LDHA had a large fraction of GFAP-positive glial cells , it is challenging to make a meaningful comparison of lactate secretion or ATP content at the later stages of differentiation . Therefore , we chose to measure AMPK T172 phosphorylation , a cellular energy sensor , during the early stage of neuronal differentiation . Indeed , AMPK pT172 was clearly increased at day 4 of neuronal differentiation compared to control , but the phosphorylation returned to a level comparable to control at day 21 , when extensive cell death had already occurred ( Figure 4I ) . Day-4 neuronal cultures constitutively expressing HK2 and LDHA showed ~four fold increase of lactate secretion compared to control cultures ( Figure 4J ) , indicating increased aerobic glycolysis . These results imply that turning off aerobic glycolysis is critical to maintain the normal neuronal ATP level . If the neuronal death were due to excessive conversion of pyruvate to lactate resulting in reduced pyruvate entry into mitochondria , increasing the level of pyruvate in neuronal differentiation medium might be able to rescue the neuronal death . This was indeed the case; as shown in Figure 4K , coexpression of LDHA and MAP2 was readily detected in cells grown in differentiation medium containing 5 mM sodium pyruvate , ten-fold higher than the standard concentration ( 0 . 5 mM ) , and there was no irregular punctate staining of LDHA as seen with standard neuronal differentiation medium ( Figure 4E ) . However , GFAP-positive glial cells were still abundant ( Figure 4K ) . A transcriptional circuit , including PGC-1 family coactivators , estrogen-related receptors ( ERRs ) and nuclear respiratory factor ( NRF ) , is responsible for the transcription of metabolic and mitochondrial genes in multiple tissues , such as heart and skeletal muscle ( Scarpulla et al . , 2012 ) . PGC-1α is induced in the neonatal mouse heart concurrent with the increase of mitochondrial energy-producing capacity ( Lehman et al . , 2000 ) . However , the control of mitochondrial biogenesis during neuronal differentiation is still not well understood . To examine if there were any changes in mitochondrial mass during neuronal differentiation , we first measured mtDNA copy number , a commonly used surrogate of mitochondrial mass , calculated by normalizing to nuclear genomes . As shown in Figure 5A , after the first week , the neuronal mtDNA copy number doubled compared to NPCs , and at 3 weeks was increased to ~four fold . To roughly estimate the extent of cell growth , we measured the protein mass of NPCs and 3-week differentiated neurons . Consistent with neurons being larger cells than NPCs , we found that 3-week differentiated neurons contained ~four-fold higher protein mass than NPCs ( Figure 5B ) , similar to the increase in mtDNA copy number . When the amounts of mitochondrial respiratory complexes were examined by immunoblotting with equal protein loading of cell lysates , no significant difference was found between NPCs and neurons at 1 and 3 weeks , except for SDHB of complex II , which was consistently ~50% lower in neurons ( Figure 5C ) . The same was true for other mitochondrial markers , such as TFAM , Hsp60 and ATP5O ( Figure 5D ) . It appears that although mitochondrial mass increases on a per cell basis , mitochondrial density remains largely unchanged during neuronal differentiation . 10 . 7554/eLife . 13374 . 025Figure 5 . Increased neuronal PGC-1α /ERRγ maintain the transcription of metabolic and mitochondrial genes during neuronal differentiation . ( A ) mtDNA copy number was measured during neuronal differentiation . Real-time PCR was done with NPCs and neurons differentiated at 1 , 3 , and 7 weeks . Bars represent mean ± SD n= 3 . ( B ) Measurement of protein mass content of NPCs and neurons at 3 weeks normalized as in per million cells . Bars represent mean ± SD , n =3 . ( C ) Immunoblotting analysis of the representative component of each mitochondrial respiratory complex . 10 µg protein lysate from NPCs and neurons at 1 and 3 weeks were loaded in SDS-PAGE gel . ( D ) Expression changes of main transcription factors involved in the transcription of metabolic and mitochondrial genes . Bars show the mean of FPKM values of differentiated neurons at 1 and 3 weeks relative to those of NPCs . Error bars represent SD of four RNA-seq replicates at each time point . Immunoblotting analysis confirmed the upregulation of PGC-1α and ERRγ . ( E ) Relative expression levels of genes encoding the mitochondrial respiratory complexes . Bars show the mean of FPKM values of differentiated neuron at 1 and 3 weeks relative to those of NPCs . Error bars represent SD of four RNA-seq replicates at each time point . ( F ) Effects of PGC-1α or ERRγ knockdown on the gene expression of glycolysis , tricarboxylic acid cycle ( TCA ) and mitochondrial respiratory complexes . Neurons differentiated at 3 week were infected with lenti-shRNA virus against PGC-1α or ERRγ , and the total RNA was extracted 5 days after infection . Two effective shRNA lentiviral vectors targeting different regions of PGC-1α or ERRγ were used . Scramble shRNA vector was used as control . In real-time PCR experiments , the relative mRNA expression levels in neurons depleted of PGC-1α or ERRγ to scramble control were calculated after normalization to β-actin . Bars are mean ± SD , n= 3 . Similar results were obtained for both shRNA knockdown constructs . ( G ) A hypothetical model depicting distinct transcriptional regulation of metabolic genes in the proliferation and post-mitotic differentiation states . ( Figure 5—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 02510 . 7554/eLife . 13374 . 026Figure 5—source data 1 . PGC-1α and ERRγ maintain the metabolic gene expression during neuronal differentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 02610 . 7554/eLife . 13374 . 027Figure 5—figure supplement 1 . The fold changes of FPKM values of UCP2 are shown . Bars represent mean ± SD of four RNA-seq replicates for NPCs and neurons differentiated at 1 and 3 weeks . ( Figure 5—figure supplement 1—source data 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 02710 . 7554/eLife . 13374 . 028Figure 5—figure supplement 1—source data 1 . UCP2 expression during neuronal differentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 028 We went on to analyze the expression profiles of transcription factors involved in mitochondrial biogenesis: PGC-1α levels were significantly increased , ~three fold increase at 1 week and ~four fold at 3 week; ERRγ levels were also upregulated , ~2 . 8 fold at 1 week and ~3 . 4 at 3 week , and similar changes were observed in their protein levels ( Figure 5D ) . Surprisingly , despite the significant increases in PGC-1α and ERRγ , the expression of the majority of the genes encoding mitochondrial respiratory complexes was either slightly decreased or unchanged in neurons ( Figure 5E ) . At first glance , this was confusing , because overexpression of PGC-1α in cardiac cells triggers the upregulation of hundreds of genes in glycolysis , TCA and mitochondrial respiratory complexes ( Rowe et al . , 2010 ) . To examine whether the increased PGC-1α and ERRγ affected TCA and mitochondrial OXPHOS gene expression , knockdown experiments were performed . Three-week neurons were infected with lenti-shRNAs targeting PGC-1α or ERRγ . For each gene , two effective shRNA constructs against different mRNA regions were used , and scramble shRNA was used as control . We found that PGC-1α was required to maintain the levels of ENO1 and PKM in glycolysis; CS , ACO2 , OGDH and SUCLA2 in TCA cycle; TFAM and TFBM1 in mtDNA replication and transcription , and NRF1 ( nuclear respiratory factor-1 ) , a key transcription factor that activates nuclear genes encoding respiratory complex subunits ( Figure 5F ) . PGC-1α not only stimulates the induction of NRF1 , but also binds to and coactivates the transcriptional function of NRF1 in muscle cells ( Wu et al . , 1999 ) . Consistently , almost all the representative genes for each mitochondrial respiratory complex exhibited decreased expression in PGC-1α knockdown neurons ( Figure 5F ) . In contrast to the broad effect of PGC-1α , ERRγ was mainly required for maintaining expression of NRF1 and mitochondrial respiratory complexes ( Figure 5F ) , which is consistent with a recent finding that in neurons the genomic binding sites for ERRγ have extensive overlap with NRFs ( Pei et al . , 2015 ) . Therefore , the increased expression of PGC-1α/ERRγ during neuronal differentiation appears to maintain , rather than increase as assumed , the transcription of metabolic and mitochondrial genes as in NPCs . This observation also implies that there is distinct transcriptional control of metabolic and mitochondrial OXPHOS genes in proliferating versus post-mitotic differentiated cells as discussed below . A series of transcriptional and protein level changes during neuronal differentiation , as summarized in Figure 6 , define the neuronal energy preference for oxidative phosphorylation . First , downregulation of LDHA is a key switch for turning off aerobic glycolysis . Lactate dehydrogenase , catalyzing the conversion between pyruvate and lactate , is a tetramer , and the LDHA and LDHB genes encode the two common subunits A and B respectively , which can assemble into five isozymes - A4 , A3B1 , A2B2 , A1B3 , and B4 . The A4 isozyme kinetically favors the conversion from pyruvate to lactate , while the B4 isozyme prefer converting lactate to pyruvate ( Cahn et al . , 1962 ) . The critical role of LDHA in aerobic glycolysis has been proven in cancer cells , and LDHA depletion greatly decreases aerobic glycolysis and increases mitochondrial OXPHOS activity ( Fantin et al . , 2006 ) . The absence of LDHA in neurons is consistent with previous immunohistochemistry studies in human brain showing that neurons are exclusively stained with anti-LDHB antibody while astrocytes are stained by both anti-LDHA and LDHB antibodies ( Bittar et al . , 1996 ) . Second , the shift of pyruvate kinase from PKM2 to PKM1 by alternative mRNA splicing is likely to be important . A current hypothesis is that PKM2 expressed in proliferating cells is in a less-active state , thus resulting in accumulation of upstream metabolites that can be used for biosynthetic pathways ( Christofk et al . , 2008; Eigenbrodt and Glossmann , 1980 ) , although this model is challenged by a recent finding that PKM1 expression does not decrease upstream glycolytic intermediates but significantly reduces nucleotide biosynthesis ( Lunt et al . , 2015 ) . Third , the decreased expression of HK2 , and the GLUT1 and GLUT3 glucose transporters reduce glucose entry , terminating high flux glycolysis . HK1 is still expressed in neurons , and this will be needed to provide a low level flux though the glycolytic pathway in order to generate enough pyruvate to feed the TCA cycle . Fourth , the decrease in expression of PDKs and the increase in PDPs may result in more active pyruvate dehydrogenase complexes ( PDC ) , as demonstrated by decreased phosphorylation at Ser 300 of PDH E1 alpha protein ( PDHA1 ) , promoting pyruvate entry into the TCA cycle . Consistent with decreased levels of glycolytic genes , 3-phosphoglyceric acid ( 3PG ) and pyruvate levels decline in neurons compared to NPCs . Surprisingly , even though neurons rely on oxidative phosphorylation to generate energy; the expression of TCA and mitochondrial genes is not increased . Thus , it appears that a need to avoid aerobic glycolysis is the major reason underlying neuronal reliance on oxidative phosphorylation . 10 . 7554/eLife . 13374 . 029Figure 6 . A model depicting transcriptional changes of metabolic genes underlying the switch from aerobic glycolysis in NPCs to oxidative phosphorylation in neurons . The genes with decreased expression are dimmed . The width of the arrows indicates increased and decreased pyruvate and lactate utilization at different steps in NPCs and neurons . Abbreviations: glucose 6-phosphate ( G6P ) ; fructose 1 , 6-bisphosphate ( FBP ) ; glycerol 3-phosphate ( G3P ) ; dihydroxyacetone phosphate ( DHAP ) ; 1 , 3-bisphosphoglyceric acid ( BPG ) ; 3-phosphoglyceric acid ( 3PG ) ; phosphoenolpyruvic acid ( PEP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13374 . 029 In addition to the conventional regulators of glycolysis , TCA and oxidative phosphorylation , the critical role of UCP2 ( uncoupling protein 2 ) in promoting aerobic glycolysis and inhibiting oxidative phosphorylation has been established from multiple experimental models ( Pecqueur et al . , 2008; Samudio et al . , 2008; Bouillaud , 2009; Zhang et al . , 2011 ) . Distinct from UCP1 , which uncouples ATP synthesis from the proton gradient by transporting protons into the mitochondrial matrix , UCP2 exports malate and oxaloacetate from mitochondria into the cytosol thus limiting the entry of pyruvate into the TCA cycle ( Vozza et al . , 2014 ) . As demonstrated in human embryonic stem cells and hematopoietic stem cells , UCP2 is critical for maintaining stem cell glycolytic metabolism , and its level decreases during differentiation ( Zhang et al . , 2011; Yu et al . , 2013 ) . We found that during neuronal differentiation , UCP2 expression levels dropped significantly , and in 3-week neurons the UCP2 level was only 20% of that in NPCs ( Figure 5—figure supplement 1 ) , indicating that similar mitochondrial metabolism remodeling occurs in neurons . It should be noted that the expression levels of metabolic enzymes alone do not completely reflect their biochemical/physiological activities , which could be regulated at multiple levels , such as post-translational modifications , which have not been explored in the current study . Moreover , mitochondrial structure ( cristae organization and size ) , dynamics ( fusion and fission ) and calcium concentration in the mitochondrial matrix are all involved in mitochondrial energy metabolism ( Mishra and Chan , 2014 ) . Possible changes in these processes need to be investigated to further understand how mitochondrial metabolism is reprogrammed during neuronal differentiation . In addition to transcriptional regulation , protein degradation could also be used to downregulate key metabolic enzymes . For instance , Pfkfb3 , the enzyme generating fructose-2 , 6-bisphosphate , a potent activator of phosphofructokinase , is constantly degraded by proteasome through anaphase-promoting complex/cyclosome ( APC/C ) -Cdh1 to suppress neuronal glycolysis ( Herrero-Mendez , et al . , 2009 ) . We suspect there might be additional mechanisms of this sort at the protein level accounting for the extremely low levels of HK2 and LDHA protein in neurons . Even under conditions of energy shortage due to mitochondrial deficiency , neurons cannot turn on aerobic glycolysis genes , such as HK2 and LDHA . In an iPSC-based disease model of maternally inherited Leign syndrome ( MILS ) , an early childhood neurodegenerative disease due to ATP synthase defect , we found that MILS neurons but not their NPCs show severe energy deficiency; MILS NPCs generate more lactate than healthy control NPCs . MILS neurons also have very low expression of HK2 and LDHA , and do not exhibit a significant increase in mitochondrial density ( Zheng et al . , 2016 ) . It appears that neurons cannot use aerobic glycolysis or mitochondrial biogenesis to compensate for energy shortage . To explore the importance of turning off aerobic glycolysis , we attempted to reactivate aerobic glycolysis by constitutive expression of HK2 and LDHA during neuronal differentiation . We found that neurons cannot survive the sustained high levels of HK2 and LDHA that were tolerated by NPCs . Glucose transporter levels dramatically decrease in neurons , which limits glucose uptake and the production of glycolytic pyruvate . Reactivating the conversion of pyruvate to lactate by expression of LDHA would decrease the amount of pyruvate available for mitochondrial oxidation . Indeed , increasing pyruvate in the neuronal differentiation medium prevented the neuronal death observed upon expressing HK2/LDHA , indicating that glycolytic pyruvate deficiency is a major cause of death . Therefore , turning off aerobic glycolysis allows the efficient use of pyruvate for energy production . Many lines of evidence support the conclusion that lactate secreted by glial cells is a critical energy source for neurons in vivo ( Pellerin and Magistretti , 2012 ) , and blocking neuronal oxidative utilization of lactate affects neuronal survival and even memory formation ( Suzuki et al . , 2011; Lee et al . , 2012 ) . Obviously , downregulation of LDHA , an enzyme catalyzing the conversion of pyruvate to lactate , would favor the reverse reaction , which is catalyzed by a tetramer composed of LDHB to generate pyruvate from exogenous lactate . To our surprise , a significant fraction of GFAP-positive glial cells was detected at early times in neuronal cultures differentiated from NPCs constitutively expressing HK2 and LDHA . This phenotype could not be reversed by extra pyruvate in the medium . Interestingly , it has been reported that exposure of NPCs to hypoxia , which boosts aerobic glycolysis , leads to more glial cells during neuronal differentiation ( Xie et al . , 2014 ) . We confirmed this observation in wild-type NPCs ( data not shown ) . Aerobic glycolysis is tightly associated with cellular redox status ( Ochocki and Simon , 2013 ) . Interestingly , Sirt1 , a histone deacetylase and sensor of NADH/NAD+ , has been shown to direct the differentiation into the astroglial lineage at the expense of the neuronal lineage ( Prozorovski et al . , 2008 ) . We surmise that enhanced glycolysis may generate a cellular redox status that shifts the lineage choice toward glial cells during differentiation of NPCs . Initially , we were surprised to find that , despite increased neuronal PGC-1α and ERRγ expression , the majority of the genes encoding mitochondrial respiratory complexes and TCA enzymes were not increased . Our knockdown studies revealed that depleting PGC-1α in neurons led to a significant decrease in a wide spectrum of genes in glycolysis , TCA pathway and mitochondrial respiratory complexes . Thus , the increased expression of PGC-1α and ERRγ in neurons is essential to maintain the expression level of these metabolic genes . Although NPCs do not rely exclusively on oxidative phosphorylation to generate energy , mitochondria are used for generation of biosynthetic precursors in these proliferating cells , and mitochondria themselves have to be duplicated for daughter cells . Cell cycle transcription factors , such as MYC and E2F , have been shown to promote metabolic and mitochondrial gene expression . c-MYC-null fibroblasts have diminished mitochondrial mass ( Li et al . , 2005 ) . c-MYC activation in the myocardium of adult mice induces mitochondrial biogenesis and glycolysis but reduces PGC-1α level ( Ahuja et al . , 2010 ) . E2F also upregulates mitochondrial genes , a function conserved from flies to mammals ( Ambrus et al . , 2013 ) . It appears that , during neuronal differentiation , the transcriptional control of mitochondrial genes and metabolic genes shifts from a cell-cycle mode to a post-mitotic neuronal program: in proliferating NPCs , MYC and E2Fs activate the transcription of metabolic genes , while in differentiated neurons , PGC-1α and ERRγ are responsible ( Figure 5G ) . This hypothetical model does not exclude the possibility that low levels of PGC-1α and ERRγ take part in metabolic and mitochondrial gene transcription in NPCs together with cell cycle related transcription factors . During neuronal differentiation , mitochondrial mass increases proportionally with neuronal mass growth , indicating an unknown mechanism linking mitochondrial biogenesis to cell size . Interestingly in this regard , in a high-throughput screen to identify small molecules interfering with mitochondrial abundance , hundreds of compounds were found to be capable of changing the cellular mitochondrial content; the majority of them also change cell size accordingly ( Kitami et al . , 2012 ) , indicating a fundamental relationship between cell size and mitochondrial number . Our finding illustrates an example of this relationship in a normal developmental context . Cells were fixed in cold 4% paraformaldehyde in PBS for 10 min . NPCs and neurons were permeabilized at room temperature for 15 min in 0 . 2% TritonX-100 in PBS . Samples were blocked in 5% BSA with 0 . 1% Tween 20 for 30 min at room temperature . The following primary antibodies and dilutions were used: goat anti-Sox2 ( Santa Cruz Biotechnology , Dallas , TX ) , 1:200; mouse anti-Nestin ( EMD Millipore , Temecula , California ) , 1:200; rabbit anti-βIII-tubulin ( Covance , San Diego , CA ) , 1:200; mouse anti-βIII-tubulin ( Covance ) , 1:200; rabbit anti-GFAP ( Dako , Carpinteria , CA ) 1:200; mouse anti-MAP2AB ( Sigma-Aldrich , St . Louis , MO ) , 1:200; rabbit anti-LDHA ( Cell signaling , Danvers , MA ) , 1:200 , and rabbit anti-HK2 ( Cell signaling ) , 1:200 . Secondary antibodies were Alexa donkey 488 and 568 anti-mouse , rabbit and goat ( Invitrogen , Carlsbad , CA ) , used at 1:1000 . Nuclear staining was done with Hoechst ( Invitrogen ) . Cell lysates were prepared with lysis buffer containing 20 mM Tris ( pH 7 . 5 ) , 150 mM NaCl , 1 mM EDTA , 1 mM EGTA , 1% Triton X-100 , 2 . 5 mM sodium pyrophosphate , 1 mM β-glycerophosphate , 1 mM Na3VO4 , 1 μg/ml leupeptin . 1 mM PMSF was added immediately prior to use . The protein concentration was measured by DC protein assay ( Bio-Rad , Irvine , CA ) . Quick neuron nuclear extract preparation: the cells were rinsed with PBS once , and plates placed on ice and 1 ml of ice cold Buffer A ( 25 mM Hepes pH 7 . 0 , 25 mM KCl , 0 . 05 mM EDTA , 5 mM MgCl2 , 10% glycerol , 0 . 1% NP-40 , 1 mM DTT ) added . The plate was scraped and cells were transferred into an Eppendorf tube , which was centrifuged and rinsed once with Buffer A ( no NP-40 ) . The pellet was resuspended in PBS , and 2x SDS PAGE sample buffer added , prior to boiling for 10 min at 95°C . The following primary antibodies and dilutions were used: antibodies against the major glycolysis enzymes were from Cell Signaling sold as glycolysis antibody sampler kits ( #8337&12866 ) ; all were used at 1:1000; Rabbit anti-TFAM ( Cell Signaling ) used at 1:1000; Rabbit anti-CS , IDH2 , PGC-1α and ATP5O ( Abcam , Cambridge , United Kingdom ) used at 1:1000 , and mouse anti-Sucla2 and goat anti-Hsp60 ( Santa Cruz Biotechnology ) used at 1:1000 . Immunoblotting results were analyzed by Odyssey Imager ( Licor , Lincoln , NE ) scanning . The establishment of neural progenitor cells from iPSCs and neuron differentiation were performed as previously described ( Brennand et al . , 2011 ) . Human embryonic stem cell ( hESC ) and iPSC lines were mainly maintained on Matrigel using mTeSR1 . For embryoid body formation , hESC and iPSC lines were cultured on a mitotically inactive mouse embryonic fibroblast feeder layer in hESC medium , DMEM/F12 supplemented with 20% knockout serum replacement , 1 mM L-glutamine , 0 . 1 mM non-essential amino acids , β-mercaptoethanol and 10 ng /ml FGF2 . Neural differentiation was induced as follows: hESCs grown on inactivated mEFs were fed N2/B27 medium without retinoic acid for 2 days , and then colonies were lifted with collagenase treatment for 1 hr at 37°C . The cell clumps were then transferred to ultra-low attachment plates . After growth in suspension for 1 week in N2/B27 medium , aggregates formed embryoid bodies , which were then transferred onto polyornithine ( PORN ) /laminin-coated plates and developed into neural rosettes in N2/B27 medium . After another week , colonies , showing mature neural rosettes with NPCs migrating out from the colony border , were picked under a dissecting microscope , digested with accutase for 10 min at 37°C and then cultured on PORN/laminin-coated plates in N2/B27 medium supplemented with FGF2 . This step is critical for the purity of NPCs; only colonies ( type 4 ) showing sufficient maturity as described in Figure 1—figure supplement 2 were picked . For neuronal differentiation , NPCs were dissociated with accutase and plated in neural differentiation media , 500 ml DMEM/F12 GlutaMAX , 1x N2 , 1X B27+RA , 20 ng/ml BDNF ( Peprotech , Rocky Hill , NJ ) , 20 ng/ml GDNF ( Peprotech ) , 200 nM ascorbic acid ( Sigma ) , 1 mM dibutyrl-cyclicAMP ( Sigma ) onto PORN/laminin-coated plates . For one well of a 6-well plate , 200 , 000 cells/well were seeded; for one well of a 12-well plate , 80 , 000 cells were seeded . Neurons can be maintained for 3 months in a 5% CO2 37°C incubator . For Ngn2/Ascl1-based conversion , postnatal human fibroblasts were transduced with lentiviral particles for EtO and XTP-Ngn2:2A:Ascl1 and selected with G418 ( 200 μg/ml; Life Technologies ) and puromycin ( 1 μg/ml; Sigma Aldrich ) . For knockdown PTB1-based conversion , cells were transduced with pLVTHM carrying a shRNA against PTB1 ( GCGTGAAGATCCTGTTCAATACTCGAGTATTGAACAGGATCTTCACGC ) and selected with puromycin . To initiate conversion , Ngn2/Ascl1 or shPTB1 fibroblasts were pooled at high densities and after 24 hr the medium was changed to neuron conversion medium based on DMEM:F12/Neurobasal ( 1:1 ) for 3 weeks . The iN conversion medium contains the following supplements: N2 supplement , B27 supplement ( both 1x; Gibco ) , doxycycline ( 2 μg/ml , Sigma Aldrich ) , laminin ( 1μg/ml , Life Technologies , Carlsbad , CA ) , dibutyryl cyclic-AMP ( 500 μg/ml , Sigma Aldrich ) , human recombinant Noggin ( 150 ng/ml; Preprotech ) , LDN-193189 ( 5 μM; Cayman Chemical Co , Ann Arbor , MI ) and A83-1 ( 5 μM; Stemgent , Cambridge , MA ) , CHIR99021 ( 3 μM , LC Laboratories , Woburn , MA ) , Forskolin ( 5 μM , LC Laboratories ) and SB-431542 ( 10 μM; Cayman Chemicals Co . ) . Medium was changed every third day up to 3 weeks . The protocol was adapted from the previous method ( Ladewig et al . , 2012 ) . Total RNA was isolated using RNeasy kit ( QIAGEN , Hilden , Germany ) . 500 ng of total RNA from each sample was used for cDNA synthesis by MMLV reverse transcriptase; and quantitative real-time polymerase chain reaction ( PCR ) was performed with SYBR Green Master Mix on ABI 7000 cycler ( Applied Biosystems , Foster City , CA ) and normalization to β-actin . Primer sequences were listed in Supplementary file 1 . mtDNA copy number was estimated by comparing SYBR Green real time PCR amplification of a mitochondrial DNA amplicon , tRNA Leu ( UUR ) , with a nuclear DNA amplicon ( β2-microglobulin ) from DNA isolated using a Qiagen genomic DNA kit according to the protocol ( Venegas et al . , 2011 ) . The primer pair for mtDNA tRNA Leu: 'CACCCAAGAACAGGGTTTGT' and TGGCCATGGGTATGTTGTTA'; for β2-microglobulin: 'TGCTGTCTCCATGTTTGATGTATCT' and TCTCTGCTCCCCACCTCTAAGT' . The ChIP assays were performed using ChIP-IT high sensitivity kit ( Active Motif , Carlsbad , CA ) . The polyclonal antibodies against c-MYC , N-MYC , and anti-HA used were purchased from Santa Cruz Biotechnology , Active Motif and Abcam . The real-time PCR primers were designed according to the work by Kim et al . ( 2004 ) and ( 2007 ) . NPCs ( at passage 3 ) and 1- and 3-week differentiated neurons were harvested; total RNA was isolated using RNeasy kit with in-column DNase digestion ( QIAGEN ) . RNA-seq experiments were performed with two NPC lines established from independent iPSC clones . For each time point , two experimental duplicates were used for each independent NPC line and its differentiated neurons . The quality of RNA was quantified by RNA integrity number ( RIN ) by Bioanalyzer ( Agilent , La Jolla , CA ) ; only samples with RIN greater than 8 . 5 were used for library preparation . Stranded mRNA-seq libraries were prepared from poly ( A ) RNA after oligo ( dT ) selection . RNA-seq reactions were done on an Illumina HiSeq 2500 system ( Illumina , San Diego , CA ) and the sequenced reads were mapped to an annotated human genome ( version GRch37/hg19 ) using STAR ( Dobin et al . , 2013 ) . The value of FPKM ( Fragments Per Kilobase of transcript per Million mapped reads ) calculated by Cufflink algorithm was used to represent the gene expression level ( Trapnell et al . , 2010 ) . The RNAseq data ( GSE75719 ) was submitted to NCBI GEO database . Lentiviral plasmids containing shRNA ( Mission shRNA , Sigma ) against human LDHA ( TRCN0000026538; TRCN0000164922 ) ; HK2 ( TRCN0000195582; TRCN0000232926 ) ; PGC-1α ( TRCN0000001169; TRCN0000001166 ) , ERRγ ( TRCN0000033645; TRCN0000033647 ) were transfected into 293T cells using Lipofectamine 2000 ( Invitrogen ) together with the third generation packaging plasmids pMD2 . G , pRRE and pRSV/REV . Cells were cultured in DMEM containing 10% fetal bovine serum , 100 U/mL penicillin and 100 μg/mL streptomycin . Lentivirus was concentrated from filtered culture media ( 0 . 45 μm filters ) by ultracentrifugation at 25 , 000 rpm for 90 min . After two days of infection , NPCs or neurons were selected with 200 ng/ml puromycin ( Sigma ) . The inducible c-MYC ( T58A , a mutation stabilizing c-MYC ) lentivirus expression vectors , FUdeltaGW-rtTA ( #19780 ) and FU-tet-o-hc-MYC ( #19775 ) were developed by Maherali et al . ( 2008 ) and obtained from Addgene . Doxycycline ( Sigma ) was used at 2 μg/ml . HK2 and LDHA ( Flag-tagged ) were cloned into home-made piggyBac transposon-based vector . HK2 and LDHA are spaced by a self-splicing 2A peptide and their expression is controlled by CMV promoter . NPCs were transfected with high-efficiency Amaxa Nuclearfector technology ( Lonza , Basel , Switzerland ) . Neurons were grown in a 6-well plate . After growth in fresh medium for 12 hr , cells were washed quickly 3 times with cold PBS , and 0 . 45 ml cold methanol ( 50% v/v in water with 20 µM L-norvaline as internal standard ) was added to each well . Culture plates were transferred to dry ice for 30 min . After thawing on ice , the methanol extract was transferred to a microcentrifuge tube . Chloroform ( 0 . 225 ml ) was added and the tubes were vortexed and centrifuged at 10 , 000 g for 5 min at 4°C . The upper layer was dried in a centrifugal evaporator and derivatized with 30 µl O-isobutylhydroxylamine hydrochloride ( 20 mg/ml in pyridine , TCI ) for 20 min at 80°C , followed by 30 µl N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide ( Sigma ) for 60 min at 80°C . After cooling , the derivatization mixture was transferred to an autosampler vial for analysis . GC-MS analysis was performed in the Cancer Metabolism Core at the Sanford-Burnham Medical Research Institute ( La Jolla , California ) . More details including the parameters of machine settings can be found in the publication from the center ( Scott et al . , 2011 ) . The OCR of NPCs and neurons grown in a laminin-coated Seahorse plate was measured using a Seahorse extracellular Flux Analyzer ( Agilent Technologies Inc , La Jolla , CA ) , following the manufacturer’s instructions . After the measurement , cells were lysed in 60~100 µl lysis buffer with two 'freeze and thaw' cycles on dry ice . Protein concentrations were determined by DC protein assay ( Bio-Rad ) . The OCR values were normalized by protein mass . For measurement of lactate levels in medium , medium from cultures of iPSCs , NPCs and neurons was freshly changed and collected after 12 hr; cells were then frozen on the plate and lysed by two freeze-and-thaw cycles in dry ice . Medium lactate was measured by Lactate Assay kit ( BioVision , Milpitas , CA ) and normalized by total protein content . Comparisons were done using Student's t-test . Statistical analyses were performed using GraphPad Prism .	Structures called mitochondria act like the batteries of cells , and use several different metabolic processes to release energy . For example , neurons rely on a metabolic process called oxidative phosphorylation , while neural progenitor cells ( which develop , or differentiate , into neurons ) use a process called aerobic glycolysis instead . Little is known about why neurons prefer to use oxidative phosphorylation to provide them with energy , and it is also not clear why problems that affect this process are often seen in neurological disorders and neurodegenerative diseases . Zheng , Boyer et al . have now used human neural progenitor cells to explore the metabolic changes that occur as these cells develop into neurons . It appears that the loss of two metabolic enzymes , called hexokinase and lactate dehydrogenase , marks the transition from aerobic glycolysis to oxidative phosphorylation . In addition , the instructions to produce an enzyme called pyruvate kinase are altered or “alternatively spliced” when progenitor cells differentiate , which in turn changes the structure of the enzyme . The levels of the proteins that activate and regulate the production of these three metabolic enzymes also decrease dramatically during this transition . Further experiments showed that neurons that produce hexokinase and lactate dehydrogenase while they differentiate die , which means that neurons must shut off aerobic glycolysis in order to survive . The amounts of two proteins that regulate metabolism ( called PGC-1α and ERRγ ) increase significantly when a neuron differentiates . This sustains a constant level of activity for several metabolic and mitochondrial genes as neural progenitor cells differentiate to form neurons . Zheng , Boyer et al . also found that neurons build more mitochondria as they grow; this suggests that an unknown mechanism exists that links the creation of mitochondria to the size of the neuron . Zheng , Boyer et al . have mainly focused on how much of each metabolic enzyme is produced inside cells , but these levels may not completely reflect the actual level of enzyme activity . The next steps are therefore to investigate whether any other processes or modifications play a part in regulating the enzymes . Further investigation is also needed to determine the effects of changes in mitochondrial structure that occur as a neuron develops from a neural progenitor cell .	[ "Abstract", "Introduction", "Result", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology" ]	2016	Metabolic reprogramming during neuronal differentiation from aerobic glycolysis to neuronal oxidative phosphorylation
Medial and lateral hypothalamic loci are known to suppress and enhance appetite , respectively , but the dynamics and functional significance of their interaction have yet to be explored . Here we report that , in larval zebrafish , primarily serotonergic neurons of the ventromedial caudal hypothalamus ( cH ) become increasingly active during food deprivation , whereas activity in the lateral hypothalamus ( LH ) is reduced . Exposure to food sensory and consummatory cues reverses the activity patterns of these two nuclei , consistent with their representation of opposing internal hunger states . Baseline activity is restored as food-deprived animals return to satiety via voracious feeding . The antagonistic relationship and functional importance of cH and LH activity patterns were confirmed by targeted stimulation and ablation of cH neurons . Collectively , the data allow us to propose a model in which these hypothalamic nuclei regulate different phases of hunger and satiety and coordinate energy balance via antagonistic control of distinct behavioral outputs . The regulated intake of food based on caloric needs is a fundamental homeostatically controlled process that is essential for health and survival . The hypothalamus is a highly conserved central convergence point for the neural and biochemical pathways that underlie this regulatory mechanism . Early studies demonstrated by way of electrical stimulation or lesions that specific hypothalamic regions play important roles in the regulation of appetite . For example , while stimulation of ventromedial hypothalamic loci in rodents and cats reduced feeding , activation of more lateral hypothalamic loci increased both hunting behavior and food intake ( Anand and Brobeck , 1951; Brobeck et al . , 1956; Delgado and Anand , 1952; Krasne , 1962 ) . Conversely , lateral hypothalamic lesions were found to reduce feeding to the point of starvation , whereas medial hypothalamic lesions resulted in overeating ( Anand and Brobeck , 1951; Hoebel , 1965; Teitelbaum and Epstein , 1962 ) . Thus , the lateral and medial hypothalamic regions came to be regarded as ‘hunger’ and ‘satiety’ centers , respectively . Recent experiments employing optical and electrophysiological methods have lent support to these early studies . For example , GABAergic neurons in the lateral hypothalamus were observed to be activated during feeding and essential for enhanced food intake during hunger ( Jennings et al . , 2015; Stuber and Wise , 2016 ) . However , these experiments have examined only subsets of hypothalamic neurons; their activity patterns and function within the context of the entire network remain unknown . This limited view hampers our understanding of the dynamical interactions between the ensemble of brain circuits thought to be important for the initiation , maintenance and termination of food consumption ( Sternson and Eiselt , 2017 ) . Here , we leverage the small and optically accessible larval zebrafish to identify modulatory regions central to the control of appetite and to shed light on their specific roles and dynamical activity patterns in relation to behavior . Using pERK-based brain-wide activity mapping ( Randlett et al . , 2015 ) , we first identified neuronal populations that display differential neural activity under conditions that would yield hunger and satiety . We show that lateral and medial hypothalamic regions have anti-correlated activity patterns during food deprivation , and voracious or steady state feeding . Next , through a combination of calcium imaging , optogenetics and ablation analysis , we show that serotonergic neurons in the caudal periventricular zone of the medial hypothalamus ( cH ) are state-dependent regulators of feeding behavior , most likely via their modulation of lateral hypothalamic activity . These results allow us to propose a model where mutually antagonistic brain states regulate energy balance by encoding distinct signals for different facets of appetite control . Larval zebrafish hunt prey such as paramecia through a sequence of motor actions that has been considered a hardwired reflex response to external prey stimuli ( Bianco et al . , 2011; Semmelhack et al . , 2015; Trivedi and Bollmann , 2013 ) . Only recently has evidence emerged that this behavior is flexibly modulated by satiation state ( Filosa et al . , 2016; Jordi et al . , 2015; Jordi et al . , 2018 ) and that larvae at 7 days post-fertilization ( dpf ) display enhanced hunting and enhanced food intake after a period of food deprivation . A robust readout of food intake in larval zebrafish was obtained both by the ingestion of fluorescently-labeled paramecia and by behavioral analysis , using protocols adapted for this study ( Johnson et al . , 2019; Jordi et al . , 2015; Jordi et al . , 2018; Shimada et al . , 2012 ) . A 2 hr period of food deprivation robustly enhances subsequent food intake ( Figure 1a ) . Up to 15 min after the presentation of prey , food-deprived animals display a strong upregulation of hunting and prey intake relative to fish that have continuous access to food ( referred to as fed fish; Figure 1a ) , on the basis of fluorescent food ingestion ( left panel , Figure 1a ) and hunting bouts ( right panel , Figure 1a ) . We refer to this behavior as ‘voracious feeding’ . Finally , as the fish consume food , their rate of food intake declines to that of continuously fed fish ( Figure 1a ) . These behaviors likely represent internal states that are commonly referred to as hunger and satiety , and reflect the animal’s underlying caloric or metabolic needs . As a first step toward understanding the homeostatic control of feeding in this simple vertebrate system , we employed whole-brain neuronal activity mapping via phosphorylated ERK visualization in post-fixed animals ( MAP-mapping; Randlett et al . , 2015 ) . Whole brain confocal image datasets of phospho-ERK expression were gathered from animals sacrificed after 15 min of voracious feeding that followed a 2 hr period of food deprivation . For comparison , image sets were also gathered from animals that had been fed continuously ( fed fish ) . The image volumes were registered to a standardized brain atlas . A difference map ( Figure 1b ) reveals significant specific differences in neural activity when comparing voracious feeding with continuous feeding ( Figure 1b–d , Video 1 , Supplementary files 1–2 ) . Since both experimental groups experienced the same sensory stimuli ( i . e . exposure to paramecia ) prior to sacrifice , differences in brain activity should primarily reflect the animals’ internal states , which could include manifestations of an altered sensitivity to food cues , activity related to hunting and prey capture , or the motivational history resulting from food deprivation . Indeed , multiple sensorimotor loci related to hunting showed enhanced activity during feeding that followed the food-deprived condition , consistent with the increased feeding behavior observed in food-deprived animals . These loci included the retinal Arborization Fields ( AFs; optic tectum and AF7 ) , pretectum , as well as downstream hindbrain loci , such as reticulospinal and oculomotor neurons , all of which are known to be engaged during prey capture behavior ( Bianco and Engert , 2015; Muto et al . , 2017; Semmelhack et al . , 2015 ) . In addition , enhanced activity was observed in the cerebellum , inferior olive , vagal sensory and motor neurons , area postrema and locus coeruleus , areas that have been implicated in feeding regulation and behavior ( Ahima and Antwi , 2008; Ammar et al . , 2001; Dockray , 2009; Zhu and Wang , 2008 ) . We focused our attention on brain areas likely to be involved in motivational states related to feeding . These included an area of particularly strong differential activity in the lateral region of the intermediate hypothalamus ( Lateral Hypothalamus , LH; Figure 1b–d ) , which has recently been identified as part of the feeding pathway in larval zebrafish ( Muto et al . , 2017 ) and whose mammalian analog has been strongly implicated in appetite control ( Sternson and Eiselt , 2017 ) . However , the zebrafish LH , unlike its mammalian counterpart , does not harbor melanin-concentrating hormone ( MCH ) -positive , orexin ( hypocretin ) -positive neurons , or other major feeding-related peptides ( Figure 1—figure supplements 1 and 2 ) . We therefore characterized the expression of multiple appetite-related neuromodulators ( AgRP , MSH , CART , NPY , MCH , Orexin ) and found that they are instead expressed in nearby areas of the hypothalamus ( Figure 1—figure supplement 1 ) . The zebrafish LH region does however contain glutamatergic and GABAergic cell types ( Figure 1—figure supplement 2 ) ; these non-peptidergic LH cell types have been shown in rodents to be important for the regulation of feeding ( Jennings et al . , 2015; Stuber and Wise , 2016 ) . Among areas that showed relatively decreased neural activity upon feeding food-deprived animals , the most significant was the adjacent caudal hypothalamus ( cH ) , which contains monoaminergic neurons -- mainly serotonergic and dopaminergic cells , with a small fraction of histaminergic cells ( Chen et al . , 2016; Kaslin and Panula , 2001; Lillesaar , 2011 ) . Indeed , in all of nine independent MAP-mapping experiments , activity was reduced in the cH and increased in the LH within 15 min of food presentation ( Figure 1c ) . The evident inverse relationship between LH and cH neural activity is supported by independent component analysis ( Randlett et al . , 2015 ) , which was applied to feeding-related MAP-mapping data ( Figure 1e , Figure 1—figure supplement 3 ) . Multiple components were uncovered in which cH and LH activities were strongly anti-correlated . These results led us to hypothesize that the lateral and caudal hypothalamic regions form a functionally interconnected network with opposing activity patterns . To probe these neural activity changes at higher resolution , we performed anti-pERK antibody staining on isolated brains and examined the hypothalamus in time course experiments spanning a period of food deprivation and subsequent feeding ( Figure 1f–h , Figure 2 ) . We quantified the mean anti-pERK fluorescence within a region-of-interest ( ROI; Figure 1g ) as well as the number of active cells or cell clusters ( Figure 1h; Figure 1—figure supplement 4 ) . These two metrics were employed because the high density of pERK-positive cells in the cH of food-deprived animals made high-throughput quantitation of active cells unreliable , whereas use of this metric in areas of sparse activity ( e . g . mLH and lLH ) yielded better differential sensitivity than ROI averaging . Using these respective metrics , we observed that mean fluorescence in the cH was significantly increased in food-deprived fish , while the number of active neurons in the medial and lateral lobes of the LH ( mLH and lLH , respectively ) was relatively low ( Figure 1f–h ) . Within the cH , enhanced pERK activity during food deprivation was most prevalent in serotonergic neurons , but also present in a smaller proportion of dopaminergic neurons ( Figure 1—figure supplement 5 , Videos 2 and 3 ) . During the period of voracious feeding that followed food deprivation , the pERK-reported activity of cH neurons fell dramatically to a level significantly below that observed in continuously fed fish ( Figure 1f–h ) . This characteristically low cH activity level coincided with a large increase in LH activity , measured by either mean anti-pERK fluorescence or by measurement of the number of individually active neurons , that lasted throughout the period of voracious feeding . Thereafter , as feeding continued at a more moderate pace , and the rate of food ingestion declined , LH neuronal activity likewise declined ( especially for lLH neurons; Figure 1h ) . Reciprocally , cH activity slowly increased back towards baseline levels . After 30 min of feeding , neural activity in both the cH and LH had mostly converged to the baseline level observed for continuously fed fish , consistent with the time course of hunting behavior reduction ( Figure 1a , right panel ) . Thus these cH and LH populations displayed anti-correlated activity over time frames that spanned a progression of distinct behaviors associated with food deprivation , voracious feeding and a gradual return to apparent satiety ( Figure 1i ) . To more closely align the activity patterns of cH and LH neuronal populations with feeding behavior , we examined these areas after a 30 min ( i . e . short ) or 2–4 hr ( i . e . long ) period of food deprivation , with or without a subsequent period of feeding ( Figure 2 , Figure 2—figure supplement 1 ) . Following food removal , cH activity increased , with an especially large anti-pERK average fluorescence intensity increase after 2 hrs of food deprivation ( Figure 2a–b ) . In contrast to the cH , food removal quickly reduced the frequency of active mLH and lLH neurons ( Figure 2a , c ) . Despite the reduction in LH active cell count over food deprivation , there were no obvious changes in mean LH anti-pERK fluorescence over the course of food deprivation ( Figure 2b ) . This is because there are few active LH cells in continuously fed and food-deprived fish , thus their overall contribution to the fluorescence average of the mLH and lLH regions of interest is small . Notably , the addition of prey ( paramecia ) rapidly reversed the food deprivation- induced patterns of cH and LH neural activity , with an amplitude of change that was correlated with the length of food deprivation ( Figure 2a–c , Figure 2—figure supplement 1d–e ) . Fish that had been food-deprived for longer periods ( 2 hr or 4 hr ) displayed a greater increase in the number of active LH neurons compared to feeding animals that had been food-deprived for only 30 min ( Figure 2a–c; Figure 2—figure supplement 1d–e ) . Likewise , the reduction in cH activity after food presentation was greater when it followed a longer period of prior deprivation ( Figure 2a–b; Figure 2—figure supplement 1d ) . In general , the presence of highly active neurons in the LH was correlated with higher food consumption ( as measured by gut fluorescence , Figure 2—figure supplement 1a–e ) . We next set out to characterize acute effects of food sensory cues on both the cH and LH , and also to analyze in more detail the apparent negative activity relationship between these two nuclei . Such analyses require higher temporal resolution than afforded by anti-pERK staining analysis , thus we switched to in vivo calcium imaging of the cH and LH in live animals ( Figure 3 ) . To that end , two transgenic Gal4 drivers , Tg ( 116A:Gal4 ) and Tg ( 76A:Gal4 ) , were combined to express GCaMP6s ( Tg ( UAS:GCaMP6s ) ) in neuronal subsets of both the cH and LH ( Figure 3—figure supplements 1–2 ) . The 116A:Gal4 transgene drives expression mainly in serotonergic neurons of the cH ( 88 . 9 ± 0 . 8% 5-HT positive ) and paraventricular organ ( PVO; Figure 3—figure supplement 1 ) , whereas 76A:Gal4 drives expression in a large proportion of LH cells ( Figure 3—figure supplement 2; Muto et al . , 2017 ) . Using these transgenic animals , we examined calcium dynamics in the cH and LH regions in tethered animals during the controlled presentation of prey stimuli ( Figure 3a ) . In these experiments , live paramecia were released in a puff of water in the vicinity of the immobilized fish , which can neither hunt nor ingest prey . Consistent with the results of anti-pERK analysis of post-fixed brains ( Figures 1 and 2 ) , activity in the mLH and lLH regions was increased and cH activity quickly reduced , in fact within seconds of paramecia release ( Figure 3b , d ) . Neurons in all three hypothalamic loci also responded to water flow alone , but these responses were significantly less than those elicited by paramecia ( Figure 3b , d , e ) . These prey-induced changes in activity were particularly striking for the mLH region , which displayed both a strongly enhanced calcium spike frequency and spike amplitude upon the introduction of prey . Thus , prey sensory cues , even in the absence of hunting or prey ingestion , strongly and differentially regulate neuronal activity in the caudal and lateral hypothalamus . The activities of cH and LH neurons also appeared remarkably anti-correlated; both spontaneous and prey-induced fluctuations in one population were accompanied by corresponding opposing activity changes in the other ( Figure 3b–c ) . This observation was supported by cross-correlation analysis between cH , mLH and lLH voxels ( Figure 3f ) , which revealed high correlation within the same hypothalamic region ( red color ) , and anti-correlation between cH and LH regions ( blue color ) . Further , lLH voxels showed more spatial heterogeneity than mLH voxels ( Figure 3f ) , though a small cluster of cells at the most-anterior part of the lLH was most consistently anti-correlated with cH activity ( Fish C and D , black arrowheads ) . When ranked according to their degrees of anti-correlation with voxels from other lobes , the cH and lLH displayed the greatest anti-correlation ( Figure 3g ) . Overall , these results indicate that cH and LH neurons display generally anti-correlated activities over short timescales , in addition to the anti-correlation observed over longer epochs reflecting motivational states imposed by food deprivation and feeding . In addition to these studies over short timescales , we also analyzed live imaging traces that spanned extended time periods ( up to 2 hr ) of food deprivation ( Figure 3—figure supplement 3a ) . This long-term imaging resulted in some confounding modulation of baseline fluorescence over these timescales ( Figure 3—figure supplement 3a , particularly lLH trace ) , that do not necessarily reflect changes in neural firing ( Berridge , 1998; Verkhratsky , 2005 ) and may well be related to modified internal states caused by tethering and immobilization . Nonetheless , we observed significantly higher calcium spike frequencies and amplitudes in the cH as compared to LH regions over the course of food deprivation ( Figure 3—figure supplement 3a , c–d ) , activity patterns that were the opposite of those observed for these regions when prey was presented ( Figure 3b , e ) . For example , the calcium spike amplitude and frequency of the cH region were many-fold greater than those observed in the mLH region during food deprivation ( Figure 3—figure supplement 3d ) , whereas after prey presentation , these relative activities were reversed , with the mLH displaying significantly greater spike amplitude and frequency than the cH ( Figure 3b , e ) . Likewise , lLH calcium spike frequency is significantly lower than the cH during food deprivation , but increases significantly after prey presentation ( Figure 3—figure supplement 3d , Figure 3e ) . Thus , the cH is more active over food deprivation , and the LH under conditions where food is present . We next sought to characterize the responses of hypothalamic regions to prey ingestion , as opposed to the mere detection of prey . To distinguish between the consequences of sensory and consummatory inputs , we compared neural activities in food-deprived fish exposed to paramecia or artemia . Artemia are live prey commonly fed to adult zebrafish and are actively hunted by fish at all stages , including larvae ( Figure 4a , Video 4 ) . Thus , artemia provide sensory inputs that elicit hunting behavior in larval animals . They are however too large to be swallowed and consumed by larvae . Hence , the comparison between these two types of prey dissociates neural activity triggered by prey detection and hunting from that of food ingestion . Prey ingestion can only occur in freely behaving animals and thus we needed to return to pERK- based activity mapping in post-fixed animals for our analysis . We found that artemia exposure caused significant increases in both mLH and lLH activity , whereas little change was detected in cH neurons ( Figure 4a–c ) . Exposure to paramecia on the other hand triggered an even larger response in both LH lobes and led , as expected , to a significant reduction in cH activity . In order to quantify the relative changes in the mLH and lLH lobes , we compared the artemia-induced activity change ( θA ) to the paramecia-induced activity change ( θP ) for each lobe . The average mLH anti-pERK fluorescence only displayed a marginally greater artemia-induced increase ( θA/θP = 41% ) than the lLH region ( θA/θP = 38%; Figure 4c , top panel ) . However , when the frequency of active neurons was compared , the mLH displayed a much larger response ( θA/θP = 32% ) to artemia than the lLH ( θA/θP = 15% ) ( Figure 4c , bottom panel ) . Taken together with our calcium imaging results ( Figure 3 ) , these observations indicate that while all three hypothalamic regions ( cH , mLH and lLH ) are modulated by prey sensory cues , they respond more strongly to prey ingestion . Among these regions , the mLH appears to be the most highly tuned to prey detection in the absence of prey ingestion ( Figure 4d ) . The observed anti-correlated patterns of caudal and lateral hypothalamus neural activity in both our calcium imaging and pERK-based activity data suggest they might interact via mutual inhibition . For example , during food deprivation , rising cH activity ( and the absence of food ) could restrain LH activity , while a subsequent experience of prey detection and ingestion might trigger LH activity that inhibits cH activity . This reduction in cH activity may , in turn , relieve suppression of LH activity , a neural ‘switch’ that could drive voracious feeding behavior . As an initial test of this hypothesis , we determined whether optogenetic excitation of cH neurons would be sufficient to inhibit LH neural activity . We used the Tg ( y333:Gal4 ) line ( Marquart et al . , 2015 ) to drive expression of a red-shifted channelrhodopsin ( Tg ( UAS:ReaChR-RFP ) ) ( Dunn et al . , 2016; Lin et al . , 2013 ) in cH neurons ( see Figure 5—figure supplement 1 regarding choice of Tg ( y333:Gal4 ) ) . The Tg ( y333:Gal4 ) line drives ReaChR expression in a large fraction of cH serotonergic neurons ( 57 . 4 ± 2 . 1%; Figure 5—figure supplement 1 ) , as well as a smaller fraction of dopaminergic cells ( 23 . 9 ± 2 . 2%; up to 30% overlap observed , Figure 5—figure supplement 2 ) . Tg ( HuC:GCaMP6s ) was co-expressed to monitor spontaneous LH neuron calcium activity . These tethered transgenic fish were subjected to targeted laser ( 633 nm ) illumination of the cH region to locally activate the ReaChR channel . We showed that ReaChR activation in the cH was sufficient to induce cH neural activity ( Figure 5a , c ) . In contrast , ReaChR activation significantly reduced spontaneous lLH calcium spike activity within a 90 s period that followed laser illumination ( Figure 5b , d ) , whereas no significant decrease was observed in mLH activity ( Figure 5b , d ) . Illumination of a control preoptic area region , where Tg ( y333:Gal4 ) -driven ReaChR is not expressed , did not affect lLH activity , though we did observe a small increase in mLH activity ( Figure 5e ) . This effect might be visually induced or driven by light-sensitive opsins known to be expressed in the preoptic area ( Fernandes et al . , 2012 ) . Since no such increase was observed when the cH itself was optogenetically activated , it is plausible that an inhibitory effect of cH stimulation on the mLH is masked by an opposing light response sensitivity . In sum , optogenetic stimulation of cH neural activity is sufficient to inhibit lLH neural activity , consistent with the notion that cH and LH regions interact to modulate the animal’s motivational state in response to food deprivation and feeding . The opposing patterns of cH and LH activity suggest they might encode opposing functions in the motivation and control of feeding behavior . Increased cH activity during food deprivation might encode a motivated state that leads to enhanced prey detection , enhanced hunting behavior and increased prey ingestion following food presentation . In contrast , the incremental increase in cH activity during feeding ( Figure 1g ) might progressively inhibit lLH activity ( Figure 5 ) and thus inhibit prey ingestion ( Muto et al . , 2017 ) . To test these expectations , we used optogenetic ReaChR activation to increase cH neuron activity during food deprivation or during voracious feeding . We reasoned that since after a short period of food deprivation ( ≤30 minutes ) , cH activity is relatively low ( Figure 2a , b ) , optogenetic cH neuron activation in such animals would mimic a longer food deprivation and yield subsequent voracious feeding . In contrast , animals that are already feeding voraciously will have very low cH activity ( Figures 1f–g and 2a–b ) ; cH activation in these animals might thus reduce voracious feeding by mimicking the ‘satiated’ state ( Figure 1f , g ) . Accordingly , animals expressing ReaChR in cH neurons ( Tg ( y333:Gal4;UAS:ReaChR-RFP ) ) were exposed to 630 nm illumination and assessed for ingestion of fluorescently labeled paramecia ( Figure 6 ) . Such animals exhibited enhanced cH activity following illumination ( Figure 6; Figure 6—figure supplement 1 ) . As expected , animals that had been illuminated during a short period of food deprivation subsequently consumed significantly more paramecia than control fish , which were similarly food-deprived and illuminated , but lacked the ReaChR transgene ( Figure 6a ) . In contrast , fish that had been illuminated at the end of a two-hour food deprivation period displayed a high level of prey ingestion irrespective of whether the ReaChR channel was present . Thus , the high level of cH activity produced by two hours long food deprivation could not be augmented by optogenetic activation . On the other hand , when cH activity was optogenetically excited during voracious feeding ( where cH activity would normally be very low ) , prey ingestion was reduced ( Figure 6b ) . We presume that increased cH activity inhibits lLH activity ( Figure 5 ) , which in turn is associated with satiation and lack of feeding ( Figure 1f , g ) . Indeed , inhibition of LH signaling has been shown to reduce prey capture success in comparable studies ( Muto et al . , 2017 ) . Finally , we asked what would happen if cH activity was reduced by partial ablation of serotonergic cells . Chemical-genetic ablation was performed via expression of a transgenic bacterial nitroreductase ( Tg ( UAS:nfsb-mCherry ) ) ( Curado et al . , 2008; Davison et al . , 2007; Pisharath and Parsons , 2009 ) that was driven in cH serotonergic neurons by Tg ( 116A:Gal4 ) ( Figure 3—figure supplement 1 ) . Tg ( 116A:Gal4; UAS:nfsb-mCherry ) -positive animals displayed a loss of nfsb-mCherry-expressing neurons after treatment with the chemical MTZ ( Figure 6—figure supplement 2 ) . These animals were compared to MTZ-treated sibling control animals lacking the Tg ( UAS:nfsb-mCherry ) transgene ( Figure 6c ) . Fish with ablated cH serotonergic neurons displayed greater food ingestion than control animals irrespective of whether the animals had been food-deprived or continuously fed ( Figure 6c ) . Animals that had been continuously fed displayed greater prey ingestion . They thus appear to display a defect in cH-mediated inhibition of feeding ( Figure 6b ) that could underlie satiety . Animals that had been food-deprived displayed greater than normal ( relative to non-ablated control animals ) voracious feeding ( Figure 6c ) . Taken together , these results are consistent with the notion that cH activity regulates hunting and prey ingestion , at least partially via inhibition of hunting and prey ingestion behaviors . We show that the medial hypothalamic zone , especially the caudal hypothalamus ( cH ) , is strongly activated by food deprivation and silent during voracious feeding , and that these changes in activity occur on a timescale of seconds to minutes . Here , we focused mainly on the cH serotonergic neurons , although many medially localized neurons show similar activity patterns . In contrast , the lateral hypothalamus ( LH ) , which contains GABAergic and glutamatergic neurons , can be inhibited by the cH ( Figure 5 ) and is weakly active in the absence of food; conversely it is most strongly active during voracious feeding when cH serotonergic neurons are silent . Interestingly , fish that display satiated feeding behavior exhibit intermediate activity levels in the two hypothalamic regions ( Figure 1 ) . Thus , "hunger" in the larval zebrafish is encoded by two alternative and distinct states of activity in opposing brain regions , depending on whether food is absent or present , with the restoration of energy homeostasis ( i . e . satiety ) paralleled by a return to an intermediate state of balanced activity . While generally anti-correlated , the cH and LH also appear to be differentially modulated both by internal energy states and external factors such as prey . In the absence of food , LH neural activity decreases rapidly ( Figure 2 ) , suggesting a requirement of external food cues to drive LH activity , though some modest rate of spontaneous activity is still observed ( Figure 5 , Figure 3—figure supplement 3 ) . In contrast , the slower timescale of increasing cH activity during food deprivation ( Figure 2 , Figure 3—figure supplement 3 ) may reflect a rising caloric deficit . Notably , many of the cH neurons are cerebrospinal fluid-contacting and thus have access to circulatory nutrient and hormone information ( Lillesaar , 2011; Pérez et al . , 2013 ) . When prey is presented to a food-deprived animal , a rapid state change occurs as LH neural activity is strongly increased and cH activity rapidly diminishes ( Figures 1–4 ) . Importantly , the silence of cH neurons and strength of LH activity were correlated with the extent of prior food deprivation ( Figure 2 ) , suggesting a role for these nuclei in regulating food intake based on energy needs . The quick timescale of these changes in activity suggests that they do not reflect an alleviation of caloric deficit ( i . e . a change in hunger state ) , which would take a significantly longer time to occur . Further , the striking anti-correlation between the cH and LH is consistent with their mutual inhibition , and suggests that the acute reduction in cH activity allows for rapid LH excitation upon the presentation of prey cues . We supported this notion by showing that optogenetic stimulation of a subset of cH neurons could inhibit lLH activity ( Figure 5 ) . However , the mechanisms for cH and LH mutual interactions are still unknown . It is possible that the cH may act via nearby inhibitory GABAergic neurons , and/or exert its effects through direct secretion of monoamines into the ventricles or perineuronal space . The fast ( seconds ) anti-correlation between cH and LH calcium activity ( Figure 3 ) , suggests the presence of direct inhibitory connections . The LH , which was previously characterized in Muto et al . ( 2017 ) , similarly does not appear to send direct projections to the cH , but could potentially interact via intermediary neurons in the medial/periventricular regions of the hypothalamus . Ingestive behavior has been proposed to comprise a series of sequential phases: 1 ) the initiation phase , triggered by energy deficit , in which the animal begins to forage; 2 ) the procurement phase , triggered by the presence of food sensory cues , in which the animal seeks and pursues food; and 3 ) the consummatory phase , which usually involves more stereotyped motor programs ( Berthoud , 2002; Watts , 2000 ) . An animal’s energy status is sensed internally and may influence the initiation , procurement and consummatory stages of ingestive behavior . Thus , a hungry animal will be more alert to food cues , seek food more persistently and also eat more voraciously . In mammals , LH neurons are responsive to both external food sensory cues and consummatory cues ( Jennings et al . , 2015 ) . Here , we show that the LH lobes in zebrafish also respond to both types of food cues . In the ‘sensory’ stage , the mLH and lLH are already activated , which may reflect an enhanced sensitivity to food cues during hunger . In contrast , cH activity transiently falls ( as shown by calcium imaging in Figure 3 ) but remains overall high . Notably , cH inhibition and LH activation during the sensory stage is not as strong as post-food consumption ( Figure 4 ) , which induces massive and opposing changes in the activity of both domains . Since LH and cH activity are modulated within minutes of food consumption , they are unlikely to reflect satiety signals , and rather might play a role in further driving voracious food consumption , at least until the activity of both populations returns to baseline . While it is unclear which consummatory cues modulate LH and cH activity , based on live imaging results from Muto et al . ( 2017 ) , the greatest enhancement of LH activity was observed almost immediately ( milliseconds to seconds ) after paramecia consumption . Thus , the cue is likely a fast pregastric signal ( taste/tactile/swallowing ) , rather than postgastric absorption or hormone secretion . Finally , our data raise the possibility of functional compartmentalization within the LH . Especially in terms of cellular pERK activity , the lLH is more weakly activated by food sensory cues compared to the mLH , suggesting that the lLH , similar to the cH , may be more sensitive to consummatory cues than sensory food cues alone . These results are also consistent with a generally stronger anti-correlation of lLH and cH activity ( compared to mLH ) , as observed in our calcium imaging and optogenetic experiments . Further molecular , cellular , and functional dissection of the individual LH lobes will allow for a better understanding of their behavioral roles . Finally , we test the hypothesis that the cH and LH form mutually antagonistic functional units that dominate different phases of hunger and drive appropriate behavioral responses during each phase ( Figure 6 ) . In particular , we show that the activation state of the cH is a crucial regulator of satiation state-dependent food intake . Artificial cH activation in satiated fish prior to feeding is sufficient to drive subsequent voracious feeding . Based on observed cH dynamics , we propose that the degree of cH inhibition during voracious feeding is proportional to the degree of cH activation prior to feeding . This could be mediated by the release of serotonin/other neuromodulators over the course of food deprivation , which may be capable of sensitizing the LH even in the absence of food cues . In this way , zebrafish are able to retain a ‘memory’ of their hunger state , which is released once food is presented . This motif might help ensure that the animal eventually returns to a stable equilibrium , that is , satiety . We furthermore show that the acute effect of cH activation during feeding is suppression of food intake , whereas cH ablation enhances food intake , which is again consistent with mammalian studies of medial hypothalamic areas . At first glance , the observation that the cH acutely suppresses food intake is inconsistent with the idea that it is most active during food deprivation . However , the critical difference here is the presence or absence of food . Once food is presented to a hungry fish , high activity in the cH may simply suppress LH activity , and hence elevate the initial threshold for food intake . The seemingly paradoxical roles of the cH during hunger may also make sense when considering that , in the absence of food , consummatory behavior would in fact be counterproductive . Thus , during food deprivation , the cH may play complementary roles such as the sensitization of the LH and/or other feeding-related circuits ( as discussed above ) , or drive alternative behavioral programs , like foraging or energy-conserving measures ( see Appendix 1 - Conceptual Circuit Model for a more in-depth discussion ) . Given that cH neurons are also activated by aversive stimuli ( Randlett et al . , 2015; Wee et al . , 2019 ) , they might generally encode a negative valence state , of which being hungry in the absence of food is an example . The silence of these neurons in a hungry fish where food is present may then imply a positive valence state , a notion that is in ready agreement with human subjective experience . Similar features of hunger-related ( i . e . AgRP ) neurons have also been described in mammals ( Betley et al . , 2015; Chen et al . , 2015; Dietrich et al . , 2015; Mandelblat-Cerf et al . , 2015 ) . Although the cH does not have an exact mammalian homolog , its functions have been proposed to be adopted by other modulatory populations , such as the serotonergic raphe nucleus in mammals ( Gaspar and Lillesaar , 2012; Lillesaar , 2011 ) . While shown to be a potent appetite suppressant , serotonin is also released during food deprivation , and can enhance food-seeking behavior ( Elipot et al . , 2013; Kantak et al . , 1978; Pollock and Rowland , 1981; Voigt and Fink , 2015 ) . Thus , our results revealing opposing cH activity patterns during hunger could reflect similarly complex roles of serotonin in zebrafish , potentially explaining some of its paradoxical effects on food intake and weight control in mammals ( Harvey and Bouwer , 2000 ) . The cH and PVO also express dopaminergic ( intermingled with 5-HT ) and a much smaller fraction of histaminergic neurons , which appear to be densely interconnected ( Chen et al . , 2016; Kaslin and Panula , 2001 ) . We note that our data , while confirming a role of serotonergic neurons , does not rule out an involvement of these other neuromodulators in appetite control , particularly dopamine . Further , we do not rule out the involvement of other circuits in appetite control; in fact , there are likely numerous players involved . For example , the PVO appears to be modulated by food cues and food deprivation , is anti-correlated with LH activity , and labeled by our transgenic lines ( albeit more sparsely ) , suggesting it may complement the role of the cH . Our conclusions are also limited by the available tools and methodologies -- since different transgenic lines were utilized for stimulation and ablation , we cannot be certain that we are manipulating the same population of neurons , though both share mutual overlap with serotonergic cells . Also , due to the lack of complete transgene specificity , there is a possibility that our manipulations may affect non-specific targets such as the olfactory bulb . The strong LH activation by the presentation of food after food deprivation suggests that this region is involved in the induction of voracious feeding . This notion is supported by Muto et al . ( 2017 ) who recently demonstrated that inhibition of the LH impairs prey capture , a behavior that is clearly related to voracious feeding . Furthermore , electrical stimulation of the homologous region ( lateral recess nuclei ) in adult cichlids and bluegills ( Demski , 1973; Demski and Knigge , 1971 ) can elicit feeding behavior , which is consistent with our hypothesis . Interestingly , while stimulating parts of this region induced food intake , the activation of other parts induced behaviors such as the ‘snapping of gravel’ , which are reminiscent of food search or procurement . In mammals , electrical or optogenetic stimulation of LH neurons triggers voracious feeding , again consistent with our findings that the LH is highly activated during the voracious feeding phase in hungry fish ( Delgado and Anand , 1952 ) . In particular , GABAergic neurons that do not co-express MCH or Orexin have been shown to be responsive to food cues and are sufficient to stimulate food intake in mammals ( Jennings et al . , 2015 ) . Whether the GABAergic and glutamatergic neurons of the zebrafish LH co-express other neuromodulators , as has been recently discovered in mammals ( Mickelsen et al . , 2019 ) remains to be explored . Overall , these data suggest that the zebrafish LH may play an important role in driving food intake during hunger , despite some differences in peptidergic expression from the mammalian LH . Certainly , since cues such as water flow and optogenetic stimulation light are sufficient to modulate cH and/or LH neurons , these hypothalamic loci may be also involved in other sensorimotor behaviors beyond appetite regulation . In conclusion , we have shown here how anatomically-segregated hypothalamic nuclei might interact to control energy homeostasis . We argue that the medial-lateral logic of hypothalamic function that is well established in mammalian systems may be conserved even in non-mammalian vertebrates , though their activity patterns might possibly be more complex than originally believed . Our data suggest diverse roles of neuromodulators such as serotonin in regulating behavioral responses during hunger , which complement mammalian observations . Finally , we propose that investigating large-scale network dynamics can reveal an additional layer of insight into the principles underlying homeostatic behavior , which might be overlooked when studies are restricted to the observation and perturbation of smaller subpopulations . Larvae and adults were raised in facility water and maintained on a 14:10 hr light:dark cycle at 28°C . All protocols and procedures involving zebrafish were approved by the Harvard University/Faculty of Arts and Sciences Standing Committee on the Use of Animals in Research and Teaching ( IACUC ) . WIK wildtype larvae and mit1fa-/- ( nacre ) larvae in the AB background , raised at a density of ~40 fish per 10 cm petri dish , were used for behavioral and MAP-mapping experiments . Transgenic lines Tg ( UAS-E1b:NTR-mCherry ) ( Davison et al . , 2007 ) ( referred to as UAS:nfsb-mCherry ) , Tg ( UAS:GCaMP6s ) ( Muto and Kawakami , 2011; Muto et al . , 2017 ) Tg ( HuC:GCaMP6s ) ( Kim et al . , 2017 ) , Tg ( Vglut2a:dsRed ) ( Miyasaka et al . , 2009 ) , Tg ( Gad1b:loxP-dsRed-loxP-GFP and Tg ( Gad1b:GFP ) ( Satou et al . , 2013 ) , Tg ( TH2:GCaMP5 ) ( McPherson et al . , 2016 ) , Tg ( ETvmat2:GFP ) ( referred to as VMAT:GFP ) ( Wen et al . , 2008 ) , Tg ( HCRT:RFP ) ( Liu et al . , 2015 ) have all been previously described and characterized . Tg ( pGal4FF:116A ) ( referred to as 116A:Gal4 ) was isolated from a gene trap screen by the Kawakami group ( Kawakami et al . , 2010 ) , Tg ( pGal4FF:76A ) was recently published by the same group ( Muto et al . , 2017 ) . Tg ( y333:Gal4 ) from a different enhancer trap screen was used to drive expression in the cH in cases where 116A:Gal4-driven expression was sparse ( Marquart et al . , 2015 ) . Tg ( UAS:ReaChR-RFP ) was generated by Chao-Tsung Yang ( Ahrens lab , Janelia Research Campus ) using Tol2 transgenesis . The same optogenetic channel was previously validated in zebrafish in Dunn et al . ( 2016 ) . More details on the MAP-mapping procedure can be found in Randlett et al . ( 2015 ) . 5–6 dpf , mit1fa-/- ( nacre ) larvae in the AB background were fed an excess of paramecia once daily . On the day of the experiment ( at 7 dpf ) , the larvae were distributed randomly into two treatment groups: 1 ) Food-deprived , where larvae were transferred into a clean petri dish of facility water , taking care to rinse out all remaining paramecia or 2 ) Fed , where after washing and transferring they were fed again with an excess of paramecia . After two hours , larvae in both groups were fed with paramecia . After 15 min , larvae were quickly funneled through a fine-mesh sieve , and the sieve was then immediately dropped into ice-cold 4% paraformaldehyde ( PFA ) in PBS ( PH 7 . 2–7 . 4 ) . Fish were then immunostained with procedures as reported below ( see Immunostaining methods ) . The rabbit anti-pERK antibody ( Cell Signaling , #4370 ) and mouse anti-ERK ( p44/42 MAPK ( Erk1/2 ) ( L34F12 ) ( Cell Signaling , #4696 ) were used at a 1:500 dilution . Secondary antibodies conjugated with alexa-fluorophores ( Life Technologies ) were diluted 1:500 . For imaging , fish were mounted dorsal-up in 2% ( w/v ) low melting agarose in PBS ( Invitrogen ) and imaged at ~0 . 8/0 . 8/2 μm voxel size ( x/y/z ) using an upright confocal microscope ( Olympus FV1000 ) , using a 20 × 1 . 0 NA water dipping objective . All fish to be analyzed in a MAP-Mapping experiment were mounted together on a single imaging dish , and imaged in a single run , alternating between treatment groups . ICA analysis was performed exactly as reported in Randlett et al . ( 2015 ) . The central brain ( not including eyes , ganglia , or olfactory epithelia ) from each fish was downsampled into 4 . 7 um3 sized voxels to generate a pERK level vector for each fish . Fish in which any of the voxels was not imaged ( due to incomplete coverage ) were excluded from the analysis . Fish were normalized for overall brightness by dividing by the 10th percentile intensity value , and voxels normalized by subtracting the mean value across fish . The fish-by-voxel array was then analyzed for spatially independent components using FastICA ( http://research . ics . aalto . fi/ica/fastica/ , Version 2 . 5 ) , treating each fish as a signal and each voxel as sample , using the symmetric approach , ‘pow3’ nonlinearity , retaining the first 30 principal components and calculating 30 independent components . Independent component ( IC ) maps are displayed as the z-score values of the IC signals . Since ICA analysis requires a substantial sample size , the original analysis reported in Randlett et al . ( 2015 ) included 820 fish exposed to various treatments , including fish sampled at different points of the day and night , and fish given various noxious or food stimuli , additional fish stimulated with electric shocks , light flashes , moving gratings , heat , mustard oil , melatonin , clonidine , nicotine , cocaine , ethanol and d-amphetamine . Here , to focus the analysis on more naturalistic feeding conditions , we restricted the dataset to n = 300 fish that were either food-deprived ( 2 hr ) , or presented with food in food-deprived or fed conditions . 24 hr after fixation ( 4% paraformaldehyde ( PFA ) in PBS ) , fish were washed in PBS + 0 . 25% Triton ( PBT ) , incubated in 150 mM Tris-HCl at pH 9 for 15 min at 70°C ( antigen retrieval ) , washed in PBT , permeabilized in 0 . 05% Trypsin-EDTA for 45 min on ice , washed in PBT , blocked in blocking solution ( 10% Goat Serum , 0 . 3% Triton in Balanced Salt Solution or 2% BSA in PBS , 0 . 3% Triton ) for at least an hour and then incubated in primary and secondary antibodies for up to 3 days at 4°C diluted in blocking solution . In-between primary and secondary antibodies , fish were washed in PBT and blocked for an hour . If necessary , pigmented embryos were bleached for 5 min after fixation with a 5%KOH/3%H2O2 solution . The protocol was similar for dissected brains , except that the brains were dissected in PBS after 24 hr of fixation , and the permeabilization step in Trypsin-EDTA and occasionally Tris-HCL antigen retrieval were omitted . Dissected brains were mounted ventral up on slides in 70% glycerol prior to imaging . Confocal images of dissected brains were obtained using either a Zeiss LSM 700 or Olympus FV1000 . Paramecia cultures ( ~1–2 500 ml bottles ) were harvested , spun down gently ( <3000 rpm ) and concentrated , and subsequently incubated with lipid dye ( DiD’ solid , D-7757 , Thermo Fisher Scientific , dissolved in ethanol ) for >2 hr ( 5 µl of 2 . 5 mg/ml working solution per 1 ml of concentrated paramecia ) on a rotator with mild agitation . They were then spun down gently ( <3000 rpm ) , rinsed and reconstituted in deionized water . An equal amount ( 100 µl , ~500 paramecia ) was pipetted into each 10 cm dish of larvae . This method was adapted from Shimada et al . ( 2012 ) . After the experiment , larvae were fixed and mounted on their sides on glass slides or placed in wells of a 96 well plate . They were then imaged using the AxioZoom V16 ( Zeiss ) and analyzed using custom Fiji ( Schindelin et al . , 2012 ) software . In cases where the identity of larvae needed to be maintained , for example , to correlate food intake with brain activity , larvae were imaged and subsequently stained individually in 96 well plates . This led to more variable staining which affects analysis of mean fluorescence . Larvae were always distributed randomly into experimental groups . Brains within each dataset were usually registered onto a selected reference image from the same dataset using the same CMTK registration software used in MAP-mapping . Further analysis was then performed using custom Fiji and MATLAB software . For confocal calcium imaging of the cH and LH simultaneously in the presence of food , Tg ( 76A:Gal4;116A:Gal4; UAS:GCaMP6s ) triple transgenic fish were embedded in 1 . 8% agarose , with their eyes/nostrils released . GCaMP activity from a single z-plane ( where the cH and LH neurons could be seen ) was imaged using a confocal microscope ( Olympus FV1000 ) at one fps . After a 5 min habituation period and a 5 min baseline period , a dense drop of water , followed by paramecia ( 5 min later ) was pipetted into the dish . Due to paramecia phototaxis , most of the paramecia moved into close vicinity of the fish’s head under the laser , allowing for strong visual/olfactory exposure to paramecia . After image registration ( TurboReg Fiji Plugin , Thévenaz et al . , 1998 ) , and downsampling ( Fiji/MATLAB ) , manually-segmented ROIs were selected and total fluorescence within the ROI was calculated . Cross-correlation and other analyses were performed using custom MATLAB software . For long-term 2P imaging of the cH and LH simultaneously in the absence of food ( Figure 3—figure supplement 3 ) , Tg ( 76A:Gal4;116A:Gal4; UAS:GCaMP6s ) triple transgenic fish were embedded in 1 . 8% agarose . GCaMP activity from either multiple slices ( 3 z-planes spanning a ~ 20 µm volume of the intermediate hypothalamus using an electrically-tunable liquid lens ( Edmund Optics , 83–922 ) , 237 ms per z-plane ) or a single z-plane where the cH and LH neurons ( 1 . 5 fps ) could be seen was imaged using custom 2P microscopes . After image registration ( Fiji/MATLAB ) , manually segmented ROIs were selected and total fluorescence within the ROI was calculated . Calcium spike detection and other analyses were performed using custom MATLAB software . Baseline detrending was performed on ‘raw’ Δf/f traces by fitting a quadratic polynomial and subtracting it from the trace . Calculations on calcium spike frequency and amplitude were subsequently performed using baseline-detrended calcium traces . Optogenetic stimulation and calcium imaging was performed on a confocal microscope ( Zeiss LSM 880 ) using a 633 nm laser for ReaChR activation , and a 488 nm laser for calcium imaging . Tg ( y333:Gal4;UAS:ReaChR-RFP; HuCGCaMP6s ) triple-transgenic fish were used to record LH activity after ReaChR activation . As Tg ( HuC:GCaMP6 ) does not label the cH , in some cases we used fish that also had Tg ( UAS:GCaMP6s ) co-expressed in the cH , allowing for monitoring of cH activity directly . The ReaChR activation spectrum is wide and 488 nm laser power at sufficiently high intensities is sufficient to activate ReaChR . Since Tg ( y333:Gal4;UASGCaMP6s ) is expressed strongly in the cH , weak 488 nm laser power can be used to monitor cH activity after ReaChR activation of cH . On the other hand , Tg ( HuC:GCaMP6s ) expression in the LH is considerably weaker than Tg ( UAS:GCaMP6s ) expression driven by Tg ( y333:Gal4 ) , and recording LH activity requires high laser power . Thus , during LH recording trials , we could not simultaneously image the cH . Fed fish were embedded in 1 . 8–2% agarose , with tails , mouth and eyes freed , 15–20 min before imaging in the absence of food . For baseline recording , spontaneous activities in cH or LH were recorded . ReachR activation was then induced in one side of cH periodically for 10–15 s , and ensuing activity in one or both sides of LH or cH was recorded continuously during intervals ( of 120–180 s ) between stimuli . Larvae expressing Tg ( 116A:Gal4;UAS:nfsb-mCherry ) , or their non-transgenic siblings were incubated in 2 . 5 mM Metronidazole ( Sigma-Aldrich , M3761 ) from 4-6 dpf/5–7 dpf . MTZ was subsequently washed out , and food intake was measured at 7 or 8 dpf . For these experiments , the MTZ-treated non-transgenic siblings were used as the control group . Each control or ablated group was food-deprived or fed for 2 hr , and labeled food was added to quantify food intake . In the case of fed fish , unlabeled food was very gently washed out 15 mins before the experiment and the food-deprived fish were also agitated slightly to simulate a short washout . Optogenetic stimulation was done by placing a square LED panel ( 630 nm , 0 . 12 mW/mm2 driven at full current , Soda Vision , Singapore ) directly on top of petri dishes containing ReaChR positive or negative fish , for 10 min continuously before or during feeding . We had attempted other methods of stimulating the fish ( e . g . pulsed LED stimulation ) but found that it was disruptive to behavior . 7 dpf larval fish were food-deprived for 2 hr , acclimatized in 24 well plates for 30 min , and then fed either an excess of hatched artemia or paramecia . Raw videos of hunting behavior were then recorded for 10 min at 30 fps using a high-resolution monochrome camera ( Basler acA4924 ) and custom Python-based acquisition software . We developed a system ( Johnson et al . , 2019 ) in which a high-speed infrared camera moves on motorized rails to automatically track a zebrafish larvae in a large pool ( 300 × 300×4 mm ) . A single fish is recruited to the arena center with motion cues delivered from a projector to initiate each trial . Paramecia are dispersed throughout the middle of the pool . For analysis 60 Hz image frames are centered and aligned . In every frame , the tail was skeletonized and the gaze angle of each eye is calculated . The eyes can each move from around zero degrees ( parallel to body-axis ) to 40 degrees ( converged for hunting ) . Each bout was then represented as a point in 220-dimensional posture space by accumulating 22 posture measurements ( 20 tail tangent angles to encode tail shape , and two eye gaze angles ) across 10 image frames ( ~167 ms ) from the beginning of each bout . All bouts were then mapped to a 2-D space with t-distributed stochastic neighbor embedding ( t-SNE ) , Four major hunting bout types can be identified from this embedding . Hunts begin with the ‘j-turn’ , and fish follow and advance toward prey objects with ‘pursuit’ bouts . Hunts end with an ‘abort’ or a ‘strike’ . When the fish is not actively involved in a hunt , it explores the arena with ‘exploratory’ bouts . Fractions of hunting bouts were then compared between fed and food-deprived fish in 3 min time bins over 45 min . All error bars show mean ± SEM over fish . Significance was reported as follows: *p<0 . 05 . Significance was determined using the non-parametric Wilcoxon signed-rank test for paired data and the Wilcoxon rank-sum test for independent samples . One-tailed tests were performed in cases where there was a prior prediction regarding the direction of change . A one-or two-way ANOVA ( Tukey-Kramer correction , MATLAB statistical toolbox ) was used in cases where multiple comparisons were involved . Analysis code used in this manuscript is available at https://github . com/carolinewee/ROIbasedpERKanalysis ( Wee , 2019a; copy archived at https://github . com/elifesciences-publications/ROIbasedpERKanalysis ) , https://github . com/carolinewee/gutfluorescence ( Wee , 2019b; copy archived at https://github . com/elifesciences-publications/gutfluorescence ) and https://github . com/carolinewee/CellularpERKanalysis ( Wee , 2019c; copy archived at https://github . com/elifesciences-publications/CellularpERKanalysis ) .	How soon after a meal do you start feeling hungry again ? The answer depends on a complex set of processes within the brain that regulate appetite . A key player in these processes is the hypothalamus , a small structure at the base of the brain . The hypothalamus consists of many different subregions , some of which are responsible for increasing or decreasing hunger . Wee , Song et al . now show how two of these subregions interact to regulate appetite and feeding , by studying them in hungry zebrafish larvae . The brains of zebrafish have many features in common with the brains of mammals , but they are smaller and transparent , which makes them easier to study . Wee , Song et al . show that as larvae become hungry , an area called the caudal hypothalamus increases its activity . But when the larvae find food and start feeding , activity in this area falls sharply . It then remains low while the hungry larvae eat as much as possible . Eventually the larvae become full and start eating more slowly . As they do so , the activity of the caudal hypothalamus goes back to normal levels . While this is happening , activity in a different area called the lateral hypothalamus shows the opposite pattern . It has low activity in hungry larvae , which increases when food becomes available and feeding begins . When the larvae finally reduce their rate of feeding , the activity in the lateral hypothalamus drops back down . The authors posit that by inhibiting each other’s activity , the caudal and lateral hypothalamus work together to ensure that animals search for food when necessary , but switch to feeding behavior when food becomes available . Serotonin – which is produced by the caudal hypothalamus – and drugs that act like it have been proposed to suppress appetite , but they have varied and complex effects on food intake and weight gain . By showing that activity in the caudal hypothalamus changes depending on whether food is present , the current findings may provide insights into this complexity . More generally , they show that mapping the circuits that regulate appetite and feeding in simple organisms could help us understand the same processes in humans .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2019	A bidirectional network for appetite control in larval zebrafish
Synaptic transmission between midbrain dopamine neurons and target neurons in the striatum is essential for the selection and reinforcement of movements . Recent evidence indicates that nigrostriatal dopamine neurons inhibit striatal projection neurons by releasing a neurotransmitter that activates GABAA receptors . Here , we demonstrate that this phenomenon extends to mesolimbic afferents , and confirm that the released neurotransmitter is GABA . However , the GABA synthetic enzymes GAD65 and GAD67 are not detected in midbrain dopamine neurons . Instead , these cells express the membrane GABA transporters mGAT1 ( Slc6a1 ) and mGAT4 ( Slc6a11 ) and inhibition of these transporters prevents GABA co-release . These findings therefore indicate that GABA co-release is a general feature of midbrain dopaminergic neurons that relies on GABA uptake from the extracellular milieu as opposed to de novo synthesis . This atypical mechanism may confer dopaminergic neurons the flexibility to differentially control GABAergic transmission in a target-dependent manner across their extensive axonal arbors . Dopamine ( DA ) -releasing neurons in the mammalian midbrain play an important role in fundamental behaviors , including motivation , reinforcement learning and motor control , and their dysfunction is associated with a wide range of neuropsychiatric disorders ( Wise , 2004; Costa , 2007; Schultz , 2007; Morris et al . , 2009; Redgrave et al . , 2010 ) . The major target of midbrain DA neurons is the striatum , a large subcortical structure implicated in the selection and reinforcement of motor actions . DA neurons located in the substantia nigra pars compacta ( SNc ) project mainly to the dorsal striatum ( also known as the caudate and putamen , CPu ) , whereas those in the ventral tegmental area ( VTA ) innervate the ventral striatum ( or nucleus accumbens , NAc ) , forming the nigrostriatal and mesolimbic pathways , respectively . The striatum controls motor behavior through two parallel output streams with opposing effects; the so-called ‘direct’ and ‘indirect’ pathways . Each pathway arises from distinct groups of GABAergic striatal projection neurons ( SPNs ) that differ , amongst other things , in their response to DA ( Gerfen and Surmeier , 2011; Tritsch and Sabatini , 2012 ) . By providing DA to the striatum , SNc and VTA neurons are believed to play a pivotal role in balancing the activity of direct- and indirect-pathway SPNs . However , our understanding of the cellular and molecular mechanisms employed by DA neurons to modulate striatal function remains incomplete . Although the activation of metabotropic receptors following the release of DA from SNc/VTA neurons is undoubtedly central to their function , DA neurons also co-release several other transmitters that shape striatal output ( Hnasko et al . , 2010; Stuber et al . , 2010; Tecuapetla et al . , 2010; Tritsch et al . , 2012 ) . In particular , we recently showed that DA neurons in SNc potently inhibit action potential firing in SPNs by releasing a transmitter that activates GABAA receptors ( Tritsch et al . , 2012 ) . Release of this neurotransmitter requires activity of the vesicular monoamine transporter Slc18a2 ( VMAT2 ) , which can be replaced by exogenous expression of the vesicular GABA transporter Slc32a1 ( also known as VGAT or VIAAT ) . In addition , VMAT2 expression in SPNs lacking VGAT is sufficient to sustain GABAergic transmission . Collectively , these findings suggest that SNc neurons co-release GABA using VMAT2 for vesicular loading . However , this study raises several important questions . First , does GABAergic transmission generalize to DA neurons in the VTA , or is it limited to nigrostriatal afferents ? Second , what is the identity of the transmitter released by DA neurons ? GABA exists as a zwitterion at neutral pH and lacks the characteristic molecular structure of VMAT2 substrates , which typically feature an aromatic ring and a positive charge ( Yelin and Schuldiner , 1995 ) . GABAA receptors can be activated by several naturally-occurring agonists and allosteric modulators , including β-alanine , taurine , imidazole-4-acetic acid and neurosteroids ( Johnston , 1996; Belelli and Lambert , 2005 ) . Although none of these molecules constitute ideal candidates for VMAT2-dependent vesicular transport based on their structure , charge , or both , they nonetheless raise the possibility that DA neurons liberate a transmitter other than GABA to inhibit SPNs . Third , do all DA neurons contribute to GABAergic signaling , or is it reserved to a subpopulation of cells , similar to glutamate co-release from DA neurons ( Hnasko and Edwards , 2011 ) . Finally , how is inhibitory synaptic transmission sustained in DA neurons ? Addressing this point will help identify molecules required for GABAergic transmission by DA neurons and will permit the development of genetic tools to determine the relative contribution of SPN inhibition by DA neurons in vivo under normal and pathological conditions . In this study , we address these questions by examining the cellular and molecular mechanisms that underlie the rapid GABAA receptor-mediated inhibition of SPNs upon stimulation of DA axons . Our results provide strong evidence that the inhibitory transmitter released by midbrain DA neurons is GABA , and suggest that GABA co-release is a general feature of all midbrain DA neurons . Moreover , we reveal that DA neurons rely on GABA uptake through the plasma membrane—but not on de novo GABA synthesis—to sustain GABAergic transmission . We recently reported that activation of SNc axons in dorsal striatum evokes monosynaptic , GABAA receptor-mediated inhibitory postsynaptic currents ( IPSCs ) in SPNs ( Tritsch et al . , 2012; Figure 1—figure supplement 1 ) . To determine whether this GABAergic signaling extends to DA neurons projecting to ventral striatum , we expressed channelrhodopsin 2 ( ChR2 ) in the VTA using one of two approaches ( ‘Materials and methods’ ) . We either injected Slc6a3-ires-Cre mice ( Backman et al . , 2006 ) stereotaxically with an adeno-associated virus expressing Cre recombinase-dependent ChR2 ( Figure 1A ) or drove ChR2 expression genetically by crossing Slc6a3-ires-Cre mice to transgenic mice ( Madisen et al . , 2012 ) containing a conditional allele of ChR2 in the Rosa26 locus ( Figure 1D ) . Mice also harbored Drd2-Egfp or Drd1a-tdTomato bacterial artificial chromosome ( BAC ) transgenes to permit distinction between direct- and indirect-pathway SPNs , respectively ( Gong et al . , 2003; Ade et al . , 2011 ) . We performed whole-cell voltage-clamp recordings from SPNs in sagittal brain slices of NAc in the presence of inhibitors of ionotropic glutamate receptors ( NBQX and R-CPP ) and metabotropic GABAB receptors ( CGP55845 ) to prevent excitatory synaptic transmission by dopaminergic axons , as well as modulatory effects of GABAB receptors , respectively . Under our recording conditions , optogenetic stimulation of VTA axons with brief ( 1 ms ) flashes of blue light reliably evoked inward IPSCs in direct- and indirect-pathway SPNs that were blocked by the GABAA receptor antagonists picrotoxin ( n = 12; Figure 1B , E ) and SR95531 ( n = 4 , not shown ) . With the exception of synaptic latency , optogenetically-evoked IPSCs ( oIPSCs ) exhibited similar properties in both experimental systems and were consequently pooled for analysis . They averaged 353 ± 60 pA in peak amplitude ( range = 68–905 pA; n = 16 ) and their kinetics were similar to those observed in dorsal striatum ( Tritsch et al . , 2012; Figure 1—figure supplement 1 ) , with a 10–90% rise time of 2 . 4 ± 0 . 2 ms and a decay time constant of 46 . 8 ± 7 . 2 ms . The synaptic latency of oIPSCs in NAc did not differ from that observed in dorsal striatum under similar experimental conditions ( Figure 1C , F ) , indicating that the connection is monosynaptic . Synaptic stimulation of midbrain DA neurons therefore directly engages GABAA receptors on SPNs in both dorsal and ventral striatum . 10 . 7554/eLife . 01936 . 003Figure 1 . Stimulation of VTA axons evokes GABAergic currents in nucleus accumbens SPNs . ChR2 was expressed in VTA DA neurons virally ( A–C ) or genetically ( D–F ) . ( A ) Coronal cross section of a Slc6a3-ires-Cre mouse ventral midbrain immunolabeled for TH ( red ) showing viral transduction of ChR2-EYFP in the VTA ( green ) . ( B ) Representative oIPSC recorded from an indirect-pathway SPN in NAc before ( black trace ) and after ( gray trace ) bath application of the GABAA receptor antagonist picrotoxin ( 100 μM ) . Blue line depicts a 1-ms full field flash of 473 nm laser light ( 5 mW·mm−2 ) . All recordings in this and subsequent figures were performed at −70 mV using a high Cl−internal solution in the presence of NBQX ( 10 μM ) , R-CPP ( 10 μM ) , and CGP55845 ( 2–5 μM ) in the perfusate . ( C ) Latency from flash onset to oIPSC onset in NAc and dorsal striatum ( Caudate/Putamen , CPu ) SPNs . For analysis of oIPSCs in CPu , ChR2 was expressed in SNc in a separate cohort of mice . White circles depict individual recordings , red circles are mean ± SEM . ( D–F ) As in ( A–C ) for recordings in Slc6a3-ires-Cre;Ai32 mice . Note that the synaptic latency in these mice is significantly longer than in virally transduced Slc6a3-ires-Cre mice in both CPu and NAc ( p<0 . 001 , Mann–Whitney test ) , presumably because of lower ChR2 expression in Ai32 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 00310 . 7554/eLife . 01936 . 004Figure 1—figure supplement 1 . Properties of DA neuron oIPSCs in dorsal striatum . ChR2 was expressed in SNc DA neurons virally . ( A ) or genetically ( B ) . ( A ) Overlay of fifteen consecutive light-evoked ( 1ms , 473 nm , 5 mW·mm−2; blue line ) IPSCs recorded under control conditions at −70 mV from a SPN in the dorsal striatum of a Slc6a3-ires-Cre mouse previously injected in the SNc with an AAV encoding Cre-dependent ChR2-mCherry . The first trace obtained after break-in is in black and subsequent oIPSCs are shown in progressively lighter shades of gray . The magenta trace depicts the average waveform of the first five oIPSCs . Recordings were performed with NBQX ( 10 μM ) , R-CPP ( 10 μM ) , and CGP55845 ( 2 μM ) in the perfusate . ( B ) As in ( A ) for a SPN recorded under control conditions in the dorsal striatum of a Slc6a3-ires-Cre;Ai32 mouse . ( C ) Plot of the amplitude of consecutive oIPSCs shown in ( A; magenta ) and ( B; green ) over time . ( D ) As in ( C ) for oIPSCs recorded in dorsal striatum SPNs from AAV-infected mice ( n = 17; magenta ) , Ai32 mice ( n = 8; green ) , or both ( n = 25 , black ) . Amplitude is normalized to the first oIPSC after break-in . Note how the amplitude of oIPSCs progressively decreases with each stimulus under control conditions , regardless of the method used for expressing ChR2 . This decrement in oIPSC amplitude is specific to dopaminergic synapses , as oIPSCs recorded from the collaterals of iSPNs in Adora2a-Cre expressing ChR2-mCherry in the dorsal striatum remained maintained their amplitude for the duration of the recording ( n = 8; gray ) . Data represent mean ± SEM . p<0 . 001 vs black trace; Sidak's multiple comparison test . ( E ) Mean latency from flash onset to oIPSC onset in individual dorsal striatum SPNs recordings from AAV-infected ( magenta ) and Ai32 ( green ) mice . The difference in synaptic delay are likely due to differences in expression levels of ChR2 . Mean ( ±SEM ) shown in red . p<0 . 001 , Mann–Whitney test . ( F ) Mean standard deviation ( SD ) of the synaptic latency of oIPSCs in individual cells indicates little temporal jitter in both experimental models , in agreement with monosynaptic transmission . Mean ( ±SEM ) shown in red . n . s . , no significant difference between means ( p=0 . 8; Mann–Whitney test ) . ( G ) Time course of oIPSC amplitude rundown starting at the time of drug application ( or the 5th stimulus for control recordings in ACSF ) . The number of SPN recordings in each condition are: ACSF , 25; picrotoxin , 19; SR95531 , 11; SKF 89976A , 9; SKF 89976A+SNAP-5114 , 8 . Changes in the decay kinetics of oIPSCs shown in Figure 2 were quantified 3–4 min after initiating the perfusion of SKF89976A , just as the effects of the GABAA receptor antagonists begin to significantly inhibit oIPSCs . Note that acute application of GAT antagonists does not immediately affect the amplitude of oIPSCs in SPNs . Data represent mean ± SEM . p<0 . 01 , picrotoxin or SR95531 vs black trace; Tukey's multiple comparison test . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 004 The chemical identity of a synaptic neurotransmitter is traditionally established by determining whether a synapse fulfills several necessary criteria . They include the presence of ( 1 ) presynaptic synthetic enzymes , ( 2 ) presynaptic vesicular transporters , ( 3 ) postsynaptic receptors , and ( 4 ) a biochemical mechanism for inactivation . In the case of GABAergic signaling at dopaminergic synapses , conditions 2 and 3 are respectively satisfied by the presence of VMAT2 in DA neurons and GABAA receptors in SPNs . For many central nervous system transmitters , condition 4 is mediated by diffusion and reuptake through plasma membrane transporters . Of the four genes identified in mice that encode high affinity membrane GABA transporters , Slc6a1 is expressed predominantly throughout the brain ( Borden , 1996; Lein et al . , 2007 ) . Its gene product , mGAT1 , distributes mainly to presynaptic terminals of GABAergic neurons and serves to shape the amplitude and time course of IPSCs as well as regulate extrasynaptic levels of GABA ( Isaacson et al . , 1993; Jensen et al . , 2003; Overstreet and Westbrook , 2003; Conti et al . , 2004; Chiu et al . , 2005; Bragina et al . , 2008; Kirmse et al . , 2008; Cepeda et al . , 2013 ) . Importantly , mGAT1 is highly selective for GABA and does not transport structurally similar GABAA receptor agonists such as β-alanine , taurine , and muscimol ( Johnston et al . , 1978; Tamura et al . , 1995 ) . Thus , we reasoned that if dopaminergic IPSCs are modulated by inhibition of mGAT1 , it would provide compelling evidence that GABA is the neurotransmitter released . To test this , we recorded dopaminergic oIPSCs from SPNs in dorsal striatum ( Figure 2 ) . Stimulation intensity was calibrated to evoke sub-maximal IPSCs ( approximately 20% of IPSC peak amplitude at maximum intensity; mean: 232 ± 40 pA , n = 19 ) in order to minimize spillover of GABA at dopaminergic synapses ( Figure 2—figure supplement 1 ) . We compared the amplitude and kinetics of IPSCs recorded before ( baseline ) and during the first 3–4 min following bath application of control saline ( ACSF; n = 9; Figure 2A ) or saline containing SKF 89976A , a selective mGAT1 antagonist ( n = 10; Figure 2B ) . In both conditions , the IPSC amplitude progressively declined to ∼70% of baseline ( p<0 . 01 each , Wilcoxon signed-rank test; Figure 2C ) , whereas the synaptic latency and 10–90% rise time remained unchanged relative to baseline ( p>0 . 1 , Wilcoxon signed-rank test ) . The decrease in amplitude is consistent with previous reports of activity-dependent rundown of GABA and DA release from midbrain DA neurons in vitro ( Schmitz et al . , 2003; Tritsch et al . , 2012; Ishikawa et al . , 2013 ) and did not differ between conditions ( p=0 . 8 , Mann–Whitney test; Figure 2C , Figure 1—figure supplement 1G ) , indicating that acute inhibition of mGAT1 does not interfere with neurotransmitter release or postsynaptic GABAA receptors . By contrast , mGAT1 antagonism significantly affected decay kinetics: whereas oIPSC decay time constants remained stable in control recordings ( baseline: 33 . 6 ± 5 . 8 ms; 3–4 min after ACSF wash-in: 32 . 1 ± 5 . 8 ms; p=0 . 6 , Wilcoxon signed-rank test ) , they were prolonged by a factor of 3 in the presence of SKF 89976A ( baseline: 38 . 6 ± 6 . 0 ms; 3–4 min after SKF 89976A wash-in: 115 . 9 ± 26 . 3 ms; p=0 . 002 , Wilcoxon signed-rank test; Figure 2D ) . These results indicate that mGAT1 activity normally shortens the duration of dopaminergic IPSCs , likely by clearing released neurotransmitter from the extracellular space . 10 . 7554/eLife . 01936 . 005Figure 2 . The decay kinetics of dopaminergic IPSCs are shaped by membrane GABA transporters . ( A ) Two representative oIPSCs recorded in SPNs using sub-maximal ChR2 stimulation ( 1ms; 0 . 3–2 mW·mm−2; blue line ) before ( baseline , left black trace ) and 3–4 min after ( middle gray trace ) bath application of control saline ( ACSF ) . Right , overlay of peak-normalized oIPSCs showing identical decay kinetics . ( B ) As in ( A ) for oIPSCs recorded in SKF 89976A ( 10 μM , in green ) . ( C ) Histogram of mean ( ±SEM ) peak oIPSC amplitude normalized to baseline for SPNs perfused in ACSF ( gray ) and SKF 89976A ( green ) . Number of recordings indicated in parentheses . ( D ) Plot of individual oIPSC decay time constants before and after bath application of ACSF and SKF 89976A . Mean ( ±SEM ) shown in red . p=0 . 002 vs baseline , Wilcoxon signed-rank test . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 00510 . 7554/eLife . 01936 . 006Figure 2—figure supplement 1 . oIPSC decay time constant increases with stimulus strength . ( A ) The decay time constant of oIPSCs evoked with maximal light intensity ( 5 mW·mm−2 ) does not correlate with oIPSC peak amplitude ( n = 25 SPNs; R2 = 0 . 01 , linear regression; red line ) . ( B–C ) In a subset of SPN recordings ( n = 6 ) , the amplitude ( B ) and decay time constant ( C ) of oIPSCs were measured at different light power densities ( blue: 5 mW·mm−2; gray: 0 . 3 mW·mm−2 ) . Maximal stimulation strength increased the mean amplitude and mean decay time constant of oIPSCs ( red circles; p<0 . 04 for both , Wilcoxon signed-rank test ) , suggesting that the ability of GATs to limit pooling and/or spillover of GABA at dopaminergic synapses is compromised at high light stimulation intensities . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 00610 . 7554/eLife . 01936 . 007Figure 2—figure supplement 2 . mGAT1 controls ambient levels of GABA in the striatum . ( A ) Continuous whole-cell voltage-clamp recording ( Vhold = −70 mV ) from a direct pathway SPN ( dSPN ) upon bath application of the GABAA receptor antagonist picrotoxin ( 100 μM , black bar ) in the presence of NBQX ( 10 μM ) , R-CPP ( 10 μM ) , and CGP55845 ( 5 μM ) . Downward deflections represent spontaneous IPSCs . The presence of a standing inward current mediated by extracellular GABA is revealed upon application of picrotoxin , which evokes a shift in baseline holding current . ( B ) As in ( A ) for an indirect pathway SPN ( iSPN ) upon bath application of SKF 89976A ( 10 μM ) and picrotoxin ( indicated by green and black bars , respectively ) . Green box indicates the 3–4 min window following the onset of SKF 89976A perfusion during which oIPSCs , eIPSCs , sIPSCs and SKF 89976A-evoked currents are quantified . ( C ) Histogram of picrotoxin-evoked shift in holding current recorded in dSPNs under control conditions ( ACSF , gray ) or upon acute inhibition of mGAT1 ( SKF 89976A , green ) . Number of recordings is indicated in parentheses . p<0 . 01 , Mann–Whitney test . ( D ) As in ( C ) for iSPNs . Note that no differences in tonic GABAA-receptor-mediated currents were observed between iSPNs and dSPNs , either in ACSF or SKF 89976A . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 007 In the striatum , mGAT1 controls ambient levels of GABA that evoke a tonic GABAA receptor-mediated conductance in SPNs ( Ade et al . , 2008; Kirmse et al . , 2008; Santhakumar et al . , 2010; Cepeda et al . , 2013 ) . Consistent with this , acute pharmacological inhibition of mGAT1 was accompanied by a significant increase in holding current caused by a threefold increase in tonic GABA current in both direct- and indirect-pathway SPNs ( Figure 2—figure supplement 2 ) . To exclude the possibility that the increase in oIPSC decay time constant stems from changes in recording conditions during bath application of SKF 89976A , we also monitored spontaneous ( s ) IPSCs as well as IPSCs evoked electrically within the striatum ( eIPSCs; Figure 3 ) . Unlike dopaminergic oIPSCs , neither the amplitude nor the kinetics of eIPSCs were significantly affected after acute mGAT1 inhibition ( baseline amplitude: 304 ± 70 pA; 3–4 min after SKF 89976A wash-in: 339 ± 80 pA . Baseline 10–90% rise time: 1 . 3 ± 0 . 4 ms; 3–4 min after SKF 89976A wash-in: 1 . 3 ± 0 . 4 ms . Baseline decay time constant: 14 . 8 ± 2 . 6 ms; 3–4 min after SKF 89976A wash-in: 19 . 6 ± 4 . 9 ms; n = 9; p>0 . 1 for all , Wilcoxon signed-rank test; Figure 3A , B ) , despite a notable increase in holding current during the first few minutes following bath application of SKF 89976A ( baseline: −82 ± 16 pA; SKF 89976A: −143 ± 21 pA; n = 9; p<0 . 001 , Wilcoxon signed-rank test ) . Moreover , although the amplitude of sIPSCs decreased slightly in SKF 89976A relative to baseline ( 97 ± 9 vs 83 ± 7 pA , n = 14; p=0 . 01 , Wilcoxon signed-rank test ) , the frequency and kinetics of sIPSCs remained unchanged ( baseline frequency: 2 . 7 ± 0 . 5 Hz; 3–4 min after SKF 89976A wash-in: 2 . 5 ± 0 . 5 Hz . Baseline 10–90% rise time: 0 . 9 ± 0 . 1 ms; 3–4 min after SKF 89976A wash-in: 0 . 8 ± 0 . 1 ms . Baseline decay time constant: 6 . 8 ± 0 . 4 ms; 3–4 min after SKF 89976A wash-in: 6 . 4 ± 0 . 5 ms; p>0 . 05 for all , Wilcoxon signed-rank test; Figure 3C , D ) . These results are consistent with previous reports in the striatum and hippocampus ( Isaacson et al . , 1993; Kirmse et al . , 2008 ) . Thus , our results indicate that prolongation of light-evoked IPSCs by SKF 89976A is specific to dopaminergic synapses and not secondary to increased GABAergic tone . Together , these data reveal that the duration of dopaminergic IPSCs is critically dependent on mGAT1 function and , therefore , strongly support that GABA is the transmitter co-released by DA neurons . 10 . 7554/eLife . 01936 . 008Figure 3 . The decay kinetics of electrically-evoked and spontaneous IPSCs are insensitive to mGAT1 inhibition . ( A ) Representative electrically-evoked IPSC before ( baseline , black trace ) and 3–4 min after ( green trace ) bath application of SKF 89976A ( 10 μM ) . Inset , overlay of peak-normalized eIPSCs . ( B ) Plot of individual eIPSC decay time constants before ( black ) and after ( green ) SKF 89976A application . Mean ( ±SEM ) shown in red . ( C and D ) As in ( A and B ) for spontaneous IPSCs . Note that sIPSCs have much faster kinetics compared to eIPSCs . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 008 Previous reports indicate that up to 10% of midbrain DA neurons in rat contain detectable levels of mRNA for the 65 kDa isoform of glutamic acid decarboxylase ( GAD65 ) ( Gonzalez-Hernandez et al . , 2001; 2004 ) . Together with GAD67 , these enzymes constitute the major biosynthetic pathway for GABA in the central nervous system ( Soghomonian and Martin , 1998 ) . To determine the fraction of midbrain DA neurons capable of synthesizing ( and by extension releasing ) GABA in mice , we performed double fluorescence in situ hybridization for Slc18a2 ( Vmat2 ) and Gad1 or Gad2 ( which encode GAD67 and GAD65 , respectively ) . We focused our analyses to regions highlighted in Figure 4A , which consist of the SNc and lateral VTA . In agreement with previous findings ( Gonzalez-Hernandez et al . , 2001 , 2004 ) , we did not detect significant overlap between Vmat2 and Gad1 , as only 4 out of 958 Vmat2+ neurons ( 0 . 4% ) showed weak signal for Gad1 in SNc and lateral VTA ( Figure 4A , B ) . To our surprise , we were also unable to detect Gad2 mRNA in these regions ( Figure 4D , E ) : only 7 out of 1200 Vmat2+ cells ( 0 . 6% ) were weakly Gad2+ , despite the presence of numerous brightly-labeled GABAergic neurons in neighboring substantia nigra pars reticulata ( SNr ) and within the SNc and lateral VTA . We confirmed our ability to detect co-labeling by examining the histaminergic tuberomamillary nucleus ( TMN ) and dopaminergic A13 cell group ( Figure 4C , F ) , both of which express GADs ( Lin et al . , 1990; Esclapez et al . , 1993 ) . 10 . 7554/eLife . 01936 . 009Figure 4 . Midbrain DA neurons do not express Gad1 or Gad2 . ( A ) Two-color in situ hybridization of Slc18a2 ( Vmat2; top , red ) and Gad1 ( middle , green ) demonstrates the absence of co-labeled DA neurons ( bottom ) in a coronal section through lateral VTA and SNc ( dashed outline ) . Nuclei are stained blue . SNr , substantia nigra pars reticulata; D , dorsal; V , ventral; M , medial; L , lateral . ( B ) Representative high magnification confocal image of Slc18a2 ( red ) and Gad1 ( green ) expression in SNc . ( C ) Double fluorescence in situ hybridization for Slc18a2 and Gad1 exhibits considerable overlap in the tuberomamillary nucleus ( TMN ) . ( D and E ) As in ( A and B ) for Slc18a2 ( Vmat2 ) and Gad2 expression . ( F ) Slc18a2 and Gad2 expression co-localize in the A13 dopaminergic cell group . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 009 Levels of mRNA for GAD65 and GAD67 in DA neurons may be below the detection threshold , yet high enough to sustain GABA synthesis . To address this concern , we attempted to directly visualize GAD protein in DA neurons by immunofluorescence using antibodies directed against GAD65/67 and tyrosine hydroxylase ( TH ) to label catecholaminergic neurons . However , the high density of GABAergic axons converging into the substantia nigra , combined with the low concentration of GADs in cell bodies compared to axon terminals prevented us from clearly identifying DA neurons that expressed GADs ( not shown ) . We instead imaged coronal brain sections from knock-in mice ( Tamamaki et al . , 2003; Taniguchi et al . , 2011 ) expressing either EGFP or Cre recombinase under transcriptional control of the endogenous promoter for Gad1 or Gad2 ( Figure 5 ) . Cre expression was visualized using a sensitive fluorescent reporter allele ( Madisen et al . , 2010 ) and TH expression was revealed via immunofluorescence . These mice offer the advantage of having bright somatic labeling including , in the case of Gad2-ires-Cre mice , of cells with very little transcriptional activity . However , we were unable to detect any DA neuron expressing fluorescent reporter protein driven by either Gad1 ( 0 out of 519 TH+ neurons ) or Gad2 ( 0 out of 526 TH+ neurons ) promoters ( Figure 5A–D ) . Collectively , these results indicate that midbrain DA neurons in the SNc and lateral VTA of mice do not express the GABA synthetic enzymes GAD65 and GAD67 . 10 . 7554/eLife . 01936 . 010Figure 5 . DA neurons in SNc/VTA do not express GAD65 or GAD67 . ( A ) Low magnification epifluorescence image of a coronal section through the ventral midbrain showing the absence of overlap between tyrosine hydroxylase ( TH ) immunofluorescence ( red ) and endogenous EGFP ( green ) in the SNc and lateral VTA of Gad1-Egfp knock-in mice . Blue , nuclear stain . D , dorsal; V , ventral; M , medial; L , lateral . ( B ) Representative high magnification confocal image of SNc in Gad1-Egfp mice showing mutually exclusive expression of EFGP and TH , confirming that GAD67 is not expressed in SNc DA neurons . ( C and D ) As in ( A and B ) for coronal brain sections from Gad2-ires-Cre knock-in mice expressing a fluorescence Cre reporter allele ( Ai14 ) . TH immunolabeling and Cre reporter fluorescence respectively depicted in red and green for consistency . The absence of overlap indicates that GAD65 is not expressed in DA neurons of the SNc and lateral VTA . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 010 Mammalian cells possess alternative means of synthesizing GABA ( Seiler , 1980; Caron et al . , 1987; Petroff , 2002 ) . GABA transaminase is an enzyme expressed at high levels in GABAergic neurons , including in the ventral midbrain that can reversibly convert succinate semialdehyde into GABA , and vice versa ( Bessman et al . , 1953; Roberts and Bregoff , 1953; Medina-Kauwe et al . , 1994 ) . Because succinate semialdehyde is readily oxidized to succinate to sustain energy production as part of the Krebs cycle , GABA transaminase is thought to participate in the degradation of GABA in most cells , not its synthesis ( Balazs et al . , 1970 ) . Nevertheless , we considered the possibility that GABA transaminase might contribute GABA in DA neurons . Double fluorescence in situ hybridization for Vmat2 and Abat ( the gene encoding GABA transaminase ) revealed strong co-labeling in 701 out of 738 DA neurons ( 95% ) within the SNc and lateral VTA ( Figure 6A , B ) , indicating that DA neurons express genes involved in GABA metabolism . To test whether DA neurons employ GABA transaminase to synthesize GABA , we incubated slices of striatum in control ACSF or vigabatrin ( VGT ) —an irreversible inhibitor of GABA transaminase—for at least thirty minutes prior to recording oIPSCs in SPNs . In agreement with prior work ( Overstreet and Westbrook , 2001 ) , this manipulation significantly elevated extracellular GABA levels ( picrotoxin-sensitive tonic current in SPNs in ACSF: 24 ± 3 pA , n = 11; in VGT: 115 ± 23 pA , n = 12; p<0 . 001 , Dunn's Multiple Comparison Test ) , confirming the effectiveness of the drug . To minimize differences in light-evoked responses within experiments , we obtained maximal oIPSCs using strong ChR2 stimulation and limited comparisons to SPNs located in similar areas of dorsal striatum in adjacent slices . Under these conditions , we did not detect significant differences in oIPSC amplitude between both groups ( ACSF: 1 . 1 ± 0 . 3 nA; VGT: 1 . 4 ± 0 . 2 nA; n = 14; p=0 . 3 , Wilcoxon signed-rank test; Figure 6C , D ) , indicating that GABA transaminase function is not required for GABAergic signaling by DA neurons . 10 . 7554/eLife . 01936 . 011Figure 6 . GABA transaminase is not required for GABA release by DA neurons . ( A ) Double fluorescence in situ hybridization for Slc18a2 ( Vmat2; red ) and Abat ( green ) reveals that midbrain DA neurons overwhelmingly express GABA transaminase . Blue , nuclear stain . D , dorsal; V , ventral; M , medial; L , lateral . ( B ) High magnification confocal image of Slc18a2 ( red ) and Abat ( green ) mRNA distribution in SNc . ( C ) Light-evoked IPSCs recorded from SPNs upon strong ChR2 stimulation ( 1 ms , 5 mW·mm−2; blue line ) after prolonged incubation in ACSF ( black ) or vigabatrin ( VGT , 100 μM; magenta ) . ( D ) Plot of mean peak oIPSC amplitude in SPNs recorded in similar regions of dorsal striatum in adjacent slices ( depicted by gray lines ) incubated in either ACSF ( black ) or VGT ( magenta ) . Mean ( ±SEM ) values for each group shown in red . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 011 Our data indicate that inhibitory synaptic transmission from DA neurons does not depend on synthesis of GABA by either GADs or GABA transaminase . In the central nervous system , several synapses rely on neurotransmitter reuptake across the plasma membrane as opposed to de novo synthesis to sustain synaptic transmission ( Torres et al . , 2003; Edwards , 2007 ) . It is therefore conceivable that DA neurons inhibit SPNs by releasing GABA they acquire from the extracellular environment . mGAT1 and mGAT4 ( encoded by Slc6a11 ) are the two major plasma membrane GABA transporters expressed in the mouse midbrain ( Borden , 1996; Lein et al . , 2007 ) . We therefore performed double fluorescence in situ hybridization for Vmat2 and mGat1 or mGat4 to establish whether either isoform is present in DA neurons ( Figure 7 ) . Surprisingly , we found considerable expression of mGat1 in 667 out of 748 Vmat2+ neurons ( 89% ) , particularly within the SNc . mGat4 localized predominantly to glial cells , consistent with previous reports ( Borden , 1996 ) , although faint labeling above background was also detected in a substantial fraction ( 80% ) of DA neurons in SNc and lateral VTA ( 568 out of 713 Vmat2+ neurons ) . 10 . 7554/eLife . 01936 . 012Figure 7 . Midbrain DA neurons express plasma membrane GABA transporters . ( A ) Two color in situ hybridization for Slc18a2 ( Vmat2; top , red ) and mGat1 ( middle , green ) shows considerable overlap in SNc and lateral VTA ( bottom ) . Nuclei are stained blue . D , dorsal; V , ventral; M , medial; L , lateral . ( B ) Representative high magnification confocal image of Slc18a2 ( red ) and mGat1 ( green ) in SNc confirms that DA neurons express mRNA for mGAT1 . ( C ) Same as ( A ) for Slc18a2 ( top , red ) and mGat4 ( middle , green ) . Note that mGat4 is most strongly expressed in star-shaped glial cells . ( D ) Confocal image through SNc reveals strong expression of mGat4 mRNA in glial cells and weak labeling in DA neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 012 The expression of mRNA for mGAT1 and mGAT4 in DA neurons raises the possibility that DA neurons acquire GABA not through de novo synthesis , but rather through plasma membrane uptake . Being almost exclusively composed of GABAergic neurons , the striatum represents a rich source of extracellular GABA for dopaminergic terminals ( Ade et al . , 2008; Kirmse et al . , 2008; Janssen et al . , 2009; Santhakumar et al . , 2010; Cepeda et al . , 2013 ) . Inhibition of mGAT1 for a few minutes does not affect the release of GABA from SNc axons ( Figure 2C ) , indicating either that this pharmacological manipulation incompletely blocks GABA reuptake , or that it is too short to deplete presynaptic GABA levels . To determine whether uptake of ambient GABA is necessary for sustaining GABAergic transmission by DA neurons , we therefore inhibited both mGAT1 and mGAT4 in striatal slices using a cocktail of SKF 89976A and SNAP-5114 ( a mGAT4 antagonist ) for at least 30 min prior to examining oIPSCs in SPNs in the continued presence of GAT antagonists . Whereas oIPSCs obtained under control conditions averaged 1 . 3 ± 0 . 2 nA in amplitude ( n = 15 ) , oIPSCs recorded in adjacent slices with GABA transport chronically blocked were significantly smaller ( 0 . 20 ± 0 . 07 nA , n = 16; p<0 . 001 vs ACSF , Dunn's Multiple Comparison Test; Figure 8A , B ) . Importantly , prolonged exposure to SKF 89976A and SNAP-5114 did not prevent transmitter release non-specifically , as light-evoked dopamine release from SNc axons as well as GABA release from striatopallidal terminals were maintained under these conditions ( Figure 8—figure supplement 1 ) . Moreover , the small picrotoxin-sensitive oIPSCs that remained exhibited extremely long decay time constants ( >1 . 5 s ) , consistent with persistent GABAergic signaling in the absence of plasma membrane reuptake . 10 . 7554/eLife . 01936 . 013Figure 8 . Sustained GABAergic signaling from DA neurons requires GAT function . ( A ) Dopaminergic IPSCs evoked by strong ChR2 stimulation ( 1 ms , 5 mW·mm−2; blue line ) in slices incubated for at least 30 min in control ACSF ( left , black ) , muscimol ( 0 . 1 μM; right , gray ) or a cocktail of mGAT1 and mGAT4 antagonists ( 10 μM SKF 89976A + 50 μM SNAP-5114 , respectively; middle , green ) . Recordings were performed in the continued presence of each drug , in addition to the GABAB receptor antagonist CGP55845 ( 2–5 μM ) and the glutamate receptor blockers NBQX ( 10 μM ) and R-CPP ( 10 μM ) . Inset , oIPSC in GAT antagonists shown on longer time scale to illustrate slow kinetics . ( B ) Plot of peak oIPSC amplitudes recorded from individual SPNs in adjacent slices after prolonged incubation in ACSF ( black ) , mGAT1 and mGAT4 inhibitors ( green ) and muscimol ( gray ) . Mean ( ±SEM ) indicated in red . p<0 . 01 for indicated comparisons , Dunn's Multiple Comparison Test . ( C ) Histogram of tonic GABA current ( IGABA ) measured in SPNs as the reduction in holding current evoked by bath application of the GABAA receptor antagonist picrotoxin ( 100 μM ) under control conditions ( black ) , or after prolonged incubation in mGAT1 and mGAT4 blockers ( 10 μM SKF 89976A + 50 μM SNAP-5114; green ) , muscimol ( 0 . 1 μM; gray ) or vigabatrin ( 100 μM; magenta ) . p<0 . 01 vs ACSF , Dunn's Multiple Comparison Test . Number of recordings indicated in parentheses . ( D ) Plot of consecutive oIPSC amplitudes ( normalized to the first light-evoked response ) over time under control conditions ( ACSF; black; n = 25 ) and after prolonged incubation in a cocktail of GAT antagonists ( 10 μM SKF 89976A + 50 μM SNAP-5114; green; n = 11 ) . p<0 . 001 vs ACSF , Sidak's multiple comparison test . ( E ) As in ( D ) for slices supplied with exogenous GABA ( 100 μM ) for 5–10 min before obtaining oIPSCs from SPNs in dorsal striatum ( blue; n = 10 ) , and slices supplied with exogenous GABA after prolonged inhibition of GATs with 10 μM SKF 89976A + 50 μM SNAP-5114 ( magenta; n = 12 ) . *p<0 . 05 vs ACSF , #p<0 . 005 vs GABA , Tukey's multiple comparison test . Control traces in ( D ) and ( E ) are the same as in Figure 1—figure supplement 1D . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 01310 . 7554/eLife . 01936 . 014Figure 8—figure supplement 1 . Chronic GAT block does not non-specifically affect synaptic transmission at GABAergic and dopaminergic synapses . ( A ) Example traces of light-evoked ( 1 ms , 5 mW·mm−2; blue line ) IPSCs recorded from neurons in the external segment of the globus pallidus ( GPe; Vhold = −70 mV ) in Adora2a-Cre;Ai32 mice , which express ChR2-EYFP in indirect-pathway SPNs . GPe boundaries were identified using the strong EYFP fluorescence of iSPN axonal terminals . Slices were incubated for at least 30 min in ACSF ( left , black ) or in a cocktail of 10 μM SKF 89976A and 50 μM SNAP-5114 ( right , green ) . Pre-incubations and recordings were performed in the continued presence of each drug , in addition to CGP55845 ( 5 μM ) , NBQX ( 10 μM ) and R-CPP ( 10 μM ) . ( B ) Amplitude normalized traces from ( A ) shown on an expanded time scale to reveal the slow decay kinetics of IPSCs in the presence of GAT antagonists . ( C ) The peak amplitude of striatopallidal oIPSCs did not significantly differ ( p=0 . 7; Mann–Whitney test ) between slices bathed in ACSF ( black; n = 11 GPe neurons ) and slices chronically incubated in GAT blockers ( green; n = 16 GPe neurons ) , indicating that this pharmacological manipulation does not inhibit synaptic transmission at ‘classical’ GABAergic synapses . Mean ( ±SEM ) shown in red . ( D ) Representative traces of light-evoked ( 1 ms , 5 mW·mm−2; blue lines ) DA release in the dorsal striatum of Slc6a3-ires-Cre;Ai32 measured by carbon-fiber amperometry . Pharmacological conditions are identical to those in ( A–C ) . Stimulation artifacts are blanked for clarity . ( E ) Peak extracellular DA concentration measured in dorsal striatum slices incubated in ACSF ( black; n = 9 slices ) or in GAT antagonists ( green; n = 14 slices ) . Population means ( ±SEM in red ) did not differ significantly ( p=0 . 09; Mann–Whitney test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 01410 . 7554/eLife . 01936 . 015Figure 8—figure supplement 2 . oIPSC rundown is activity dependent . ( A ) The peak amplitude of the first oIPSC recorded in dorsal striatum SPNs within the first 1 . 5 hr after slicing ( black ) was comparable to that recorded during the next 1 . 5 hr ( gray ) , indicating that oIPSCs do not rundown with time in the absence of stimulation . Data in this and subsequent panels represent mean ± SEM . The number of SPNs recorded is indicated in parentheses . ( B ) The time course of synaptic transmission rundown by consecutive light stimuli was similar in the two groups of slices , confirming that recording conditions in the first and second halves of recording sessions are comparable , and that synaptic rundown of DA neuron oIPSCs is dependent on activity . ( C–D ) As in ( A–B ) for slices chronically incubated for 0 . 5–1 . 0 hr ( dark green ) or 1 . 1–2 . 0 hr ( light green ) in SKF 89976A ( 10 μM ) + SNAP-5114 ( 50 μM ) . The effect of GAT inhibition on oIPSC amplitude does not vary with pre-incubation duration ( C ) , indicating that our pharmacological manipulation depletes synaptic GABA levels within 30 min . Importantly , stimulation-evoked rundown in GAT blockers is accelerated ( contrast panels B and D ) , consistent with activity dependent depletion of vesicular GABA when cytoplasmic GABA levels are not replenished by GATs . DOI: http://dx . doi . org/10 . 7554/eLife . 01936 . 015 Chronic inhibition of GABA reuptake was also accompanied by considerable elevation of ambient GABA in the striatum , as evidenced by a 7 . 5-fold increase in tonic GABAA receptor-mediated current in SPNs ( picrotoxin-evoked tonic current in ACSF: 24 ± 3 pA , n = 11; in GAT antagonists: 182 ± 32 pA , n = 11; p<0 . 001 , Dunn's Multiple Comparison Test; Figure 8C ) . This increase in extracellular GABA might account for the reduced amplitude of oIPSCs by promoting the desensitization of GABAA receptors , by shunting synaptic currents , or both . To examine these possibilities , we included an additional experimental condition in which slices were pre-incubated in muscimol before evoking dopaminergic oIPSCs ( Figure 8A , B ) . Muscimol is a potent synaptic and extrasynaptic GABAA receptor agonist that promotes GABAA receptor desensitization to the same extent as GABA ( Mortensen et al . , 2010 ) , but displays minimal affinity for plasma membrane GABA transporters ( Johnston et al . , 1978 ) . The concentration of muscimol was titrated to evoke tonic GABAA receptor-mediated currents similar those that developed upon persistent GAT inhibition ( picrotoxin-sensitive tonic current: 212 ± 46 pA , n = 10; p<0 . 001 vs ACSF; p>0 . 05 vs GAT antagonists , Dunn's Multiple Comparison Test; Figure 8C ) . Strikingly , this manipulation did not reduce the amplitude of dopaminergic oIPSCs ( 1 . 3 ± 0 . 3 nA , n = 12; p>0 . 05 vs ACSF , p<0 . 01 vs GAT antagonists , Dunn's Multiple Comparison Test; Figure 8A , B ) , indicating that the reduction in oIPSC amplitude observed following GAT inhibition is not secondary to elevation of ambient GABAergic tone . In agreement with this conclusion , the amplitude of oIPSCs was not diminished by prolonged treatment with VGT either ( Figure 6C , D ) , despite significant increases in extracellular GABA ( picrotoxin-sensitive tonic current in VGT: 115 ± 23 pA , n = 12; p<0 . 01 vs ACSF , p>0 . 05 vs GAT antagonists or muscimol , Dunn's Multiple Comparison Test; Figure 8C ) . These results collectively indicate that GAT function is necessary for GABAergic transmission from SNc axons . If membrane GABA reuptake is required for supplying GABA for vesicular exocytosis , inhibiting GATs should prevent DA neurons from maintaining GABA release with repeated stimulation . Indeed , chronic inhibition of GATs precipitated the rundown of GABAergic transmission from DA neurons in an activity-dependent manner ( Figure 8D , Figure 8—figure supplement 2 ) . A second prediction is that elevating extracellular GABA should attenuate synaptic rundown . To test this , we locally applied GABA ( 100 μM ) to the dorsal striatum for 5–10 min prior to recording oIPSCs from SPNs . Following this manipulation , the amplitude of oIPSCs was maintained for several minutes after break-in ( Figure 8E ) . Importantly , this increase in oIPSC amplitude by exogenous GABA was prevented in slices incubated in GAT antagonists ( Figure 8E ) . These data therefore provide strong evidence that membrane GABA transporters play an important role in sustaining GABAergic transmission from DA neurons by supplying the neurotransmitter GABA to presynaptic terminals for vesicular loading and exocytosis . The chemical identity of a synaptic transmitter is typically established only after several independent observations converge on a plausible candidate . In this case , the transmitter released by midbrain DA neurons ( 1 ) functions as a GABAA receptor agonist , ( 2 ) serves as a substrate for VGAT ( Tritsch et al . , 2012 ) , and ( 3 ) is a substrate for mGAT1 . Although VGAT can transport GABA , glycine , and β-alanine into synaptic vesicles ( Gasnier , 2000; Wojcik et al . , 2006; Juge et al . , 2013 ) , glycine does not function as a GABAA receptor agonist , and the low affinity of β-alanine for GABAA receptors would give rise to IPSCs with faster kinetics than the ones we observed ( Jones et al . , 1998a ) . In addition , mGAT1 is highly selective for GABA ( Borden , 1996 ) , and applying GABA exogenously helped sustain IPSCs from DA neurons in a GAT-dependent fashion . Although we did not detect GABA synthetic enzymes in DA neurons in our experiments , SNc and VTA neurons display a GABAergic phenotype , as they contain mRNA for GABA transaminase and the membrane GABA transporters mGAT1 and mGAT4 . Together , these results strongly suggest that the neurotransmitter released by midbrain DA neurons is GABA . In agreement with this conclusion , a recent electron microscopic study detected GABA in close association with VMAT2-containing synaptic vesicles in dopaminergic terminals within striatum ( Stensrud et al . , 2013 ) . Moreover , other dopaminergic neurons in the central nervous system have been reported to contain and/or release GABA , including amacrine cells in the retina ( Hirasawa et al . , 2012 ) , periglomerular cells in the olfactory bulb ( Maher and Westbrook , 2008; Borisovska et al . , 2013; Liu et al . , 2013 ) , as well as dopaminergic cell groups throughout the brain ( Bjorklund and Dunnett , 2007; Stamatakis et al . , 2013 ) , suggesting that GABA co-release may be a general property of dopaminergic cells . Synaptic transmission requires constant refilling of new and recycled vesicles , which is contingent upon the availability of transmitter in the cytosol . Most GABAergic neurons sustain inhibitory transmission by maintaining high cytosolic concentrations of GABA using de novo synthesis from glutamate by GAD67 and to a lesser extent GAD65 ( Asada et al . , 1997; Tian et al . , 1999; Petroff , 2002 ) . Given that all cells contain glutamate , an amino acid necessary for the production of proteins , the expression of either GAD67 or GAD65 in conjunction with a vesicular GABA transporter is sufficient to mediate vesicular release of GABA . Previous reports have estimated that up to 10% of DA neurons in the SNc and VTA of rats express mRNA for GAD65 ( Gonzalez-Hernandez et al . , 2001 , 2004 ) , suggesting that GABA release may be limited to a subpopulation of DA neurons . We were unable to replicate this observation using two separate approaches , indicating that DA neurons in the SNc and lateral VTA of mice do not rely on de novo synthesis of GABA and have instead adopted other means of obtaining GABA . Most classical transmitters are transported back into the presynaptic terminal after vesicular release using plasma membrane transporters . This process is important not only for limiting the duration of synaptic currents and preventing extrasynaptic spillover , but also for replenishing presynaptic transmitter levels ( Edwards , 2007; Conti et al . , 2011 ) . In fact , several synapses , including monoaminergic ( Giros et al . , 1996; Bengel et al . , 1998; Jones et al . , 1998b; Xu et al . , 2000 ) , glycinergic ( Gomeza et al . , 2003a , 2003b; Rousseau et al . , 2008; Apostolides and Trussell , 2013 ) as well as some GABAergic terminals ( Mathews and Diamond , 2003; Bak et al . , 2006; Fricke et al . , 2007; Brown and Mathews , 2010; Wang et al . , 2013 ) rely heavily on membrane transporter function to maintain cytosolic transmitter pools available for vesicular loading . Our results show that DA neurons contain mRNA for mGAT1 and mGAT4 , indicating that DA neurons might acquire GABA through membrane transport . Indeed , we find that prolonged inhibition of mGAT1 and mGAT4 inhibited GABAergic transmission by DA neurons and precipitated the rundown of oIPSCs , suggesting that DA neurons critically depend on extracellular GABA uptake to maintain cytoplasmic GABA levels and fill synaptic vesicles . By contrast , application of exogenous GABA helped sustain inhibitory transmission from DA neurons . By virtue of the fact that the vast majority of DA neurons in the SNc and lateral VTA contain mRNA for mGATs , our results suggest that GABA co-release is , unlike glutamate co-transmission ( Hnasko and Edwards , 2011 ) , a common feature of these cells in the adult nervous system . The reliance on plasma membrane uptake may partially explain why release of both DA and GABA from these neurons is prone to rundown upon repeated stimulation in brain slices ( Schmitz et al . , 2003; Tritsch et al . , 2012; Ishikawa et al . , 2013 ) , whereas release of glutamate from dopaminergic neurons or GABA from SPNs ( which presumably depend on neurotransmitter synthesis to sustain release ) do not suffer from the same shortcoming ( Tritsch et al . , 2012; Figure 1—figure supplement 1D ) . Nevertheless , these findings provide evidence that the expression of GABA synthetic enzymes is not required to sustain GABAergic transmission , and identify a GABAergic synapse that instead relies entirely on recycling extracellular GABA . Interestingly , a similar mechanism underlies exocytic release of serotonin from dopaminergic neurons ( Zhou et al . , 2005 ) as well as from glutamatergic thalamocortical neurons during development ( Lebrand et al . , 1996 ) . Although DA neurons in SNc and VTA differ in the inputs they receive and the signals they convey ( Matsumoto and Hikosaka , 2009; Lammel et al . , 2011 , 2012; Watabe-Uchida et al . , 2012; Roeper , 2013 ) , we find that both cell populations are capable of co-releasing GABA with DA , pointing to a fundamental property of dopaminergic signaling . Morphological studies of SPNs at the electron microscopic level have revealed that the majority of dopaminergic synapses terminate on the neck of spines that receive excitatory inputs from cortex and thalamus ( Wickens and Arbuthnott , 2005 ) . Although the actions of DA are not believed to be spatially localized ( Arbuthnott and Wickens , 2007 ) , this unique synaptic arrangement is indicative of an additional , point-to-point mode of action . An intriguing possibility is that GABA co-release may function to hyperpolarize spines or shunt excitatory cortical and thalamic inputs by activating GABAA receptors . Phasic activation of DA neurons may thereby dampen ongoing cortical and thalamic drive onto SPNs to limit DA receptor-mediated plastic changes to synaptic inputs most strongly activated by salient or rewarding stimuli . In addition , the mechanism adopted by DA neurons to obtain GABA may confer these cells the flexibility to dynamically and locally control GABAergic transmission across their extensive axonal arbors . For instance , VTA neurons projecting to NAc and medial prefrontal cortex may only co-release GABA in the former , where extracellular GABA levels are high compared to cortex ( Drasbek and Jensen , 2006; Weitlauf and Woodward , 2008 ) . Alternatively , phasic and chronic changes in mGAT function or extracellular GABA levels resulting from synaptic activity , drugs or disease may alter the amplitude and kinetics of GABAergic currents arising from midbrain dopamine neurons . A greater understanding of the relative effects of DA and GABA on the activity of striatal circuits will help reveal how DA neurons contribute to behavior in health and disease . All experimental manipulations were performed in accordance with protocols approved by the Harvard Standing Committee on Animal Care following guidelines described in the US National Institutes of Health Guide for the Care and Use of Laboratory Animals . All mice were group-housed and maintained on a 12-hr light cycle with ad libitum access to food and water . Slc6a3-ires-Cre knock-in mice expressing Cre recombinase in DA neurons ( Backman et al . , 2006 ) were obtained from Jackson Labs ( Bar Harbor , ME; stock # 006660 ) . These mice were crossed with Drd2-Egfp ( GENSAT , founder line S118 ) or Drd1a-tdTomato ( stock # 016204; Jackson Labs ) BAC transgenic mice to permit distinction between direct- and indirect-pathway SPNs ( Gong et al . , 2003; Ade et al . , 2011 ) . To genetically target indirect pathway SPNs , Adora2a-Cre BAC transgenic mice ( GENSAT , founder line KG139 ) were used . For stereotaxic viral injections , postnatal day 18–25 mice were anesthetized with isoflurane , placed in a small animal stereotaxic frame ( David Kopf Instruments , Tujunga , CA ) and injected with 1 μl of an adeno-associated virus ( ∼1012 genome copies per ml; UNC Vector Core Facility , Chapel Hill , NC ) encoding Cre-dependent ChR2 ( AAV2/8 . EF1α . DIO . hChR2 ( H134R ) -mCherry or AAV2/8 . EF1α . DIO . hChR2 ( H134R ) -EYFP ) , as described previously ( Tritsch et al . , 2012 ) . Injection coordinates were 0 . 8 mm anterior from Lambda , 1 . 3 mm lateral and 4 . 4 mm below pia for SNc , and 0 . 8 mm anterior from Lambda , 0 . 6 mm lateral and 4 . 4 mm below pia for VTA . Alternatively , ChR2 ( H134R ) -EYFP was expressed genetically in Cre-containing cells by crossing Slc6a3-ires-Cre;Drd2-Egfp or Slc6a3-ires-Cre;Drd1a-tdTomato mice with Ai32 mice ( stock # 012569; Jackson Labs ) ( Madisen et al . , 2012 ) . The knock-in lines used in Figure 5 include Gad1-Egfp ( Tamamaki et al . , 2003 ) and Gad2-ires-Cre mice ( Taniguchi et al . , 2011 ) . The latter were crossed with mice bearing a Cre-dependent tdTomato reporter transgene ( Ai14; stock # 007914; Jackson Labs ) to reveal the distribution of Cre+ cells ( Madisen et al . , 2010 ) . Mice were maintained on a C57BL/6 background . Only mice heterozygous for all transgenes were used for experiments . Mice were anesthetized with isoflurane and transcardially perfused with phosphate buffered saline followed by 4% ( wt/vol ) paraformaldehyde in 0 . 1 M sodium phosphate buffer . 50-micrometer coronal brain sections were subject to immunohistochemical staining for tyrosine hydroxylase ( AB152; 1:2000; Millipore , Billerica , MA ) , as described previously ( Tritsch et al . , 2012 ) . Endogenous tdTomato and EGFP fluorescence were not immuno-enhanced . Double fluorescence in situ hybridization was performed using a tyramide signal amplification method according to the manufacturer's instructions ( NEL753001KT; PerkinElmer , Waltham , MA ) , as previously described ( Kwon et al . , 2012 ) . Briefly , brains from 4-week old mice were dissected and immediately frozen in liquid nitrogen . They were then cut in 25-μm-thick sections with a cryostat ( Leica , Buffalo Grove , IL ) , postfixed in 4% PFA , acetylated in 1% triethanolamine and 0 . 25% acetic anhydride , dehydrated serially in ethanol , bleached in 3% hydrogen peroxide diluted in methanol , prehybridized , and hybridized at 65°C using the following anti-sense probes provided by the Allen Institute for Brain Science ( riboprobe ID specified in parentheses ) : Gad1 ( RP_040324_01_F01 ) , Gad2 ( RP_071018_02_B07 ) , Vmat2 ( RP_071218_02_B08 ) , mouse Gat1 ( RP_071204_04_H01 ) , mouse Gat4 ( RP_050428_04_D04 ) , and Abat ( RP_050301_02_C02 ) . For in vitro transcription , each cDNA was cloned out from a mouse brain cDNA library using the primer sets suggested in the in situ hybridization data portal from the Allen Institute for Brain Science . Two fluorescein- or digoxigenin-labeled riboprobes generated by an in vitro transcription method ( Promega , Madison , WI ) were hybridized simultaneously and stained by fluorescein or Cy3 chromogens , respectively . After staining , sections were mounted with Prolong Gold antifade reagent with DAPI nuclear stain ( Life Technologies , Grand Island , NY ) . Acute brain slices and whole-cell voltage-clamp recordings from identified SPNs were obtained using standard methods , as described previously ( Tritsch et al . , 2012 ) . Briefly , mice ( 49–178 days old; median = 107 days ) were anesthetized and perfused with ice-cold artificial cerebrospinal fluid ( ACSF ) containing ( in mM ) 125 NaCl , 2 . 5 KCl , 25 NaHCO3 , 2 CaCl2 , 1 MgCl2 , 1 . 25 NaH2PO4 and 11 glucose ( 295 mOsm·kg−1 ) . Parasagittal slices of striatum ( 275-μm thick ) were subsequently obtained in cold choline-based cutting solution ( in mM: 110 choline chloride , 25 NaHCO3 , 2 . 5 KCl , 7 MgCl2 , 0 . 5 CaCl2 , 1 . 25 NaH2PO4 , 25 glucose , 11 . 6 ascorbic acid , and 3 . 1 pyruvic acid ) . Following 15 min recovery in ACSF at 34°C , slices were kept at room temperature ( 20–22°C ) until use . All solutions were constantly bubbled with 95% O2/5% CO2 . Whole-cell voltage-clamp recordings were established from direct- and indirect-pathway SPNs in dorsal or ventral striatum resting 18–75 μm below the slice surface ( median: 40 μm ) in ACSF warmed to 32–34°C . Direct-pathway SPNs were identified as tdTomato+ cells in Drd1a-tdTomato mice or EGFP–cells in Drd2-EGFP mice , and indirect-pathway SPNs as tdTomato−or EGFP+ cells in Drd1a-tdTomato and Drd2-EGFP mice , respectively . Patch pipettes ( 2–4 MΩ ) were filled with ( in mM ) 125 CsCl , 10 TEA-Cl , 10 HEPES , 0 . 1 Cs-EGTA , 3 . 3 QX-314 ( Cl−salt ) , 4 Mg-ATP , 0 . 3 Na-GTP , 8 Na2-Phosphocreatine ( pH 7 . 3 adjusted with CsOH; 295 mOsm·kg−1 ) . The recording perfusate always contained NBQX ( 10 μM ) and R-CPP ( 10 μM ) to block AMPA and NMDA receptor-mediated inward currents , respectively , as well as CGP55845 ( 2–5 μM ) to prevent GABAB receptor-evoked pre- and post-synaptic modulation . For some experiments ( Figures 6 and 8 , Figure 8—figure supplements 1 and 2 ) , slices were incubated for at least 30 min ( and up to two hours ) in ACSF containing CGP55845 ( 2–5 μM ) in addition to vigabatrin ( 100 μM ) , muscimol ( 0 . 1 μM ) , or SKF 89976A ( 10 μM ) + SNAP-5114 ( 50 μM ) to allow for the depletion of cytosolic GABA levels and synaptic vesicles containing GABA , and to control for the effects of chronically elevated GABAergic signaling . In these cases , drugs continued to be present in the recording chamber for the duration of the experiment . For Figure 8E , GABA ( 100 μM , in ACSF containing either 5 μM CGP55845 or 5 μM CGP55845 + 10 μM SKF 89976A + 50 μM SNAP-5114 ) was locally applied to the slice via a gravity-fed 250-μm-wide flow pipe for 5–10 min prior to recording . For all voltage-clamp experiments , errors due to the voltage drop across the series resistance ( <20 MΩ ) were left uncompensated . Membrane potentials were corrected for a ∼5-mV liquid junction potential . Under these conditions , GABAA receptor-mediated currents appeared inward when SPNs were held at negative membrane potentials ( Vhold = −70 mV ) . To activate ChR2-expressing fibers , light from a 473-nm laser ( Optoengine , Midvale , UT ) was focused on the back aperture of the microscope objective to produce wide-field illumination of the recorded cell . Brief pulses of light ( 1-ms duration; 5 mW·mm−2 under the objective for maximal stimulation , 0 . 3–2 mW·mm−2 , for sub-maximal stimulation ) were delivered at the recording site at 30 s intervals under control of the acquisition software . Epifluorescence illumination was used sparingly to minimize ChR2 activation prior to recording . eIPSCs were evoked using constant-current pulses ( 0 . 1 ms , 7–90 μA ) delivered every 15 s through a bipolar tungsten stimulating electrodes positioned within striatum , 100–200 μm away from the recorded cell . Constant-potential amperometry was performed as before ( Tritsch et al . , 2012 ) using commercial glass-encased carbon-fiber microelectrodes ( Carbostar-1 , Kation Scientific , Minneapolis , MN ) placed within dorsal striatum slices and held at 600 mV . All recordings were obtained within 4 hr of slicing . All pharmacological agents were obtained from Tocris ( Minneapolis , MN ) . Brain sections processed for in situ hybridization or immunofluorescence were imaged with an Olympus VS110 slide-scanning microscope . High-resolution images of regions of interest were subsequently acquired with a Zeiss LSM 510 META confocal microscope ( Harvard NeuroDiscovery Center ) . Individual imaging planes were overlaid and quantified for colocalization in ImageJ ( NIH ) and thresholded for display in Photoshop ( Adobe ) . Confocal images in figures represent maximum intensity projections of 3-µm confocal stacks . Membrane currents were amplified and low-pass filtered at 3 kHz using a Multiclamp 700B amplifier ( Molecular Devices , Sunnyvale , CA ) , digitized at 10 kHz and acquired using National Instruments acquisition boards and a custom version of ScanImage ( Pologruto et al . , 2003; available upon request or from https://openwiki . janelia . org/wiki/display/ephus/ScanImage ) written in MATLAB ( Mathworks ) . Electrophysiology data were analyzed offline using Igor Pro ( Wavemetrics , Natick , MA ) . In figures , voltage-clamp traces represent the averaged waveform of 3–5 consecutive acquisitions . Averaged waveforms were used to obtain current latency , peak amplitude , 10–90% rise time and decay time constant . The latter was estimated by measuring the time elapsed from the peak of the IPSC to 1/e ( 36 . 8% ) of the peak amplitude . Detection threshold for sIPSCs was set at 40 pA to facilitate event detection in the presence of large and noisy tonic GABA currents . For pharmacological analyses in Figures 2 and 3 , the peak amplitude of IPSCs measured 3–4 min following the onset of drug perfusion were averaged , normalized to baseline averages obtained immediately prior to drug application and compared statistically to values obtained at corresponding times in control preparations bathed in ACSF . Data ( reported in text and figures as mean ± SEM ) were compared using Prism 6 ( GraphPad , La Jolla , CA ) with the following non-parametric statistical tests ( as indicated in the text ) : Mann–Whitney for group comparisons , Wilcoxon signed-rank for paired comparisons , and Kruskal–Wallis analysis of variance ( ANOVA ) followed by Dunn's Multiple Comparison Test for multiple group comparisons . In experiments characterizing the time course of synaptic transmission rundown , two-way ANOVAs were used followed by the Sidak's and Tukey's multiple comparison tests for comparisons between two and more conditions , respectively . p values smaller than 0 . 05 were considered statistically significant . N values represent the number of recorded cells . For most experiments , a single cell was recorded from in each slice , with each animal contributing fewer than five recordings to individual data sets .	The electrical signals that are fired along neurons cannot be transmitted across the small gaps , called synapses that are found between neurons . Instead , the neuron sending the signal releases chemicals called neurotransmitters into the synapse . These neurotransmitters bind to receptor proteins on the surface of the second neuron and control how it fires . A neurotransmitter called dopamine plays a key role in the circuits of the brain that control how we learn certain tasks involving movement . In particular , two populations of neurons from the midbrain that release dopamine target the striatum , an area of the brain that is responsible for motor control . These neurons also release other neurotransmitters , but the identity of these other chemicals is not known , and the details of the interaction between the neurons and the striatum are poorly understood . Previous research showed that some of the midbrain neurons activate receptors that normally respond to a neurotransmitter called gamma-aminobutyric acid ( GABA ) . However , several different chemicals can trigger this receptor . Using a range of techniques , Tritsch et al . now confirm that dopamine neurons release GABA alongside dopamine , and that this applies to both sets of the dopamine-producing neurons that feed into the striatum . Some neurons can manufacture GABA from amino acids found in their internal fluid . However , Tritsch et al . could not detect the enzymes needed for this reaction in dopamine-producing neurons . Instead , these neurons contain proteins that can transport GABA across the cell membrane , which suggests that the neurons collect GABA from the extracellular fluid that surrounds them .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2014	Midbrain dopamine neurons sustain inhibitory transmission using plasma membrane uptake of GABA, not synthesis
Botryllus schlosseri is a colonial urochordate that follows the chordate plan of development following sexual reproduction , but invokes a stem cell-mediated budding program during subsequent rounds of asexual reproduction . As urochordates are considered to be the closest living invertebrate relatives of vertebrates , they are ideal subjects for whole genome sequence analyses . Using a novel method for high-throughput sequencing of eukaryotic genomes , we sequenced and assembled 580 Mbp of the B . schlosseri genome . The genome assembly is comprised of nearly 14 , 000 intron-containing predicted genes , and 13 , 500 intron-less predicted genes , 40% of which could be confidently parceled into 13 ( of 16 haploid ) chromosomes . A comparison of homologous genes between B . schlosseri and other diverse taxonomic groups revealed genomic events underlying the evolution of vertebrates and lymphoid-mediated immunity . The B . schlosseri genome is a community resource for studying alternative modes of reproduction , natural transplantation reactions , and stem cell-mediated regeneration . In 1866 , Russian embryologist Alexander Kowalevsky wrote to Charles Darwin about the extensive developmental and morphological similarities between ascidian larvae and vertebrates , leading Darwin to hypothesize that ascidians ( belonging to urochordates or tunicates ) might be crucial to understanding the origin of the vertebrate phylum ( Darwin , 1874 ) . Indeed , tunicates are the closest extant relatives of vertebrates ( Delsuc et al . , 2006 ) , and represent an investigative model for evolutionary events leading to adaptive immunity ( Sabbadin , 1962; Scofield et al . , 1982 ) and vertebrate-specific organ/tissue complexity ( Dehal et al . , 2002; Jeffery et al . , 2004; Abitua et al . , 2012 ) . The colonial tunicate species , Botryllus schlosseri , represents an important model organism for studying unique aspects of a pre-vertebrate colonial lifestyle , such as self recognition ( Sabbadin , 1962; Scofield et al . , 1982 ) , vasculature and blood development ( Schlumpberger et al . , 1984; Gasparini et al . , 2008; Tiozzo et al . , 2008 ) , apoptosis ( Lauzon et al . , 1993; Cima et al . , 2009 ) , and alternative reproduction pathways ( Sabbadin et al . , 1975; Manni and Burighel , 2006; Voskoboynik et al . , 2007; Lemaire , 2011 ) , including stem cell-mediated regeneration of complete individuals within a colony unit ( Laird et al . , 2005; Voskoboynik et al . , 2008; Rinkevich et al . , 2013 ) . Botryllus schlosseri is an invasive colonial urochordate , living in large communities consisting of multiple colonies organized into expansive mats that coat a variety of marine surfaces , such as rocks , molluscs , multicellular algae , and ship hulls ( Stoner et al . , 2002 ) . Communities develop among compatible colonies , governed by a genetically encoded histocompatibility system ( Sabbadin , 1962; Scofield et al . , 1982 ) . The progeny of each colony usually represents a clone of the vascularly connected , asexually reproducing individuals ( zooids ) derived from a single planktonic larva ( Manni and Burighel , 2006; Figure 1A–D ) . Compatible colonies fuse their blood vessels to generate a chimera , while incompatible colonies reject one another , maintaining individuality ( Sabbadin , 1962; Scofield et al . , 1982; Voskoboynik et al . , 2013 ) . Following the fusion of blood vessels between colonies , the circulating stem cells of one partner colony can compete and replace the germline and/or the soma of the other partner ( Stoner and Weissman , 1996; Stoner et al . , 1999; Laird et al . , 2005; Voskoboynik et al . , 2008; Rinkevich et al . , 2013 ) , a phenomenon analogous to allogeneic transplantation . 10 . 7554/eLife . 00569 . 003Figure 1 . Botryllus schlosseri anatomy , life cycle , and phylogeny . B . schlosseri reproduces both through sexual and asexual ( budding ) pathways , giving rise to virtually identical adult body plans . Upon settlement , the tadpole phase of the B . schlosseri lifecycle ( A ) will metamorphose into a founder individual ( oozooid ) ( B ) , which through asexual budding , generates a colony . The colony includes three overlapping generations: an adult zooid , a primary bud , and a secondary bud , all of which are connected via a vascular network ( bv ) embedded within a gelatinous matrix ( termed tunic ) . The common vasculature terminates in finger-like protrusions ( termed ampullae; B–D ) . Bud development commences in stage A ( C ) . Through budding , B . schlosseri generates its entire body , including digestive ( ds ) and respiratory ( brs ) systems , a simple tube-like heart ( h ) , an endostyle ( en ) that harbors a stem cell niche , a primitive neural complex , and siphons used for feeding , waste , and releasing larvae ( B–D ) . Each week , successive buds grow large ( D ) and complete replication of all zooids in the colony , ultimately replacing the previous generation’s zooids , which die through a massive apoptosis . ( E ) A phylogenomic tree produced from analysis of 521 nuclear genes ( 40 , 798 aligned amino acids ) from 15 species , including B . schlosseri . Scale bar-1 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 00310 . 7554/eLife . 00569 . 004Figure 1—figure supplement 1 . Mitogenomic analysis of tunicates and deuterostomes . Based on the 13 mitochondrially-encoded proteins . The tree was inferred by PhyloBayes under a GTR+G+CAT model . Support values at nodes represents Bayesian Posterior Probability ( PP ) and are reported only when >0 . 5 and <0 . 95 . Nodes with PP < 0 . 5 were collapsed . The tree was rooted with the non-deuterostome Drosophila and Aplysia species . The main deuterostome lineages are represented in different colours . Abbreviations for tunicate orders: Stolido: Stolidobranchia; Phlebo: Phlebobranchia; Aplouso: Aplousobranchia . Colonial tunicates are indicated by an asterisk and include Botryllus schlosseri , all Aplousobranchia ascidians , and the thaliacean Doliolum nationalis . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 004 Tunicates are classified as chordates because their planktonic larva stage ( Figure 1A ) shares structural characteristics with all chordates: a notochord , dorsal neural tube , segmented musculature , and gill slits ( Darwin , 1874; Dehal et al . , 2002 ) . Larvae settle in response to light and metamorphose into sessile individuals ( Figure 1B ) , which lose most of their chordate phenotypes ( Darwin , 1874; Dehal et al . , 2002 ) . Tunicates reproduce either sexually ( solitary tunicates; Dehal et al . , 2002; Lemaire , 2011 ) , or sexually and asexually ( colonial tunicates; Manni and Burighel , 2006; Lemaire , 2011 ) . These two reproductive modes give rise to nearly identical complex adult body plans , including digestive and respiratory systems , a simple tube-like heart , siphons , an endostyle , a neural complex , ovary and testis ( Manni and Burighel , 2006; Figure 1A–D ) . The ability to reproduce asexually renders colonial tunicates robust survivors , capable of rapid proliferation and whole body regeneration . These unique features of colonial tunicates coupled with their key evolutionary position and long history of scientific study prompted us to sequence the B . schlosseri genome . The B . schlosseri genome was previously estimated to be 725 Mb based on flow cytometry analysis ( De Tomaso et al . , 1998 ) , and metaphase spreads suggested that it is organized into 16 chromosomes ( Colombera , 1963 ) . To accurately assemble this relatively large genome , we developed a novel method to accurately sequence many large fragments in parallel . This long read sequencing approach ( LRseq ) effectively increases the read length of a next generation sequencer by 50-fold , while decreasing the error rate by orders of magnitude ( Figure 2; ‘Materials and methods’ under ‘Genome sequencing and assembly’ ) . Our approach began with genomic DNA sheared to 6–8 kb fragments . Limiting dilution was used to create aliquots of a few hundred to a few thousand DNA molecules . Each aliquot was amplified with PCR , fragmented ( 600–800 bp ) , barcoded , and sequenced by Illumina HiSeq 2000 ( Figure 2 ) . The Velvet assembler ( Zerbino , 2010 ) was used to assemble short paired-end reads from each barcode ( i . e . , well ) separately , thus simplifying the assembly problem and creating effective read lengths corresponding to the original large fragment sizes ( Figure 2B; Supplementary file 2A , Supplementary file 2B ) . Limiting the number of DNA molecules per well greatly reduces or eliminates chances of having a repeated or duplicate sequence within a defined partition . Furthermore , since each well was over-sequenced , the error rate is reduced by the coverage and is substantially improved from the intrinsic error rate of the sequencer ( Supplementary file 2C ) . This procedure is amenable to automation in multiwell plates , and we obtained data from twelve 96-well plates ( Supplementary file 2A , Supplementary file 2B ) . We validated this method on human genomic DNA , for which an independent reference is available ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 00569 . 005Figure 2 . A novel short read genome sequencing and assembly method for complex , repeat-rich genomes . ( A ) Genomic DNA is sheared into 6–8 kb fragments , partitioned into twelve 96-well plates , further fragmented to 600–800 bp , barcoded and sequenced separately for each well ( Illumina HiSeq 2000 2x100bp ) , and assembled by Velvet . ( B ) Size distribution of contigs assembled from a representative library preparation ( BL5 ) . ( C ) Limiting the number of amplifiable molecules per well ( barcode ) to the level that almost 100% of all amplifiable molecules are present as single copies ( <1000 gDNA molecules ) greatly reduces the chance of having a repeated or homologous sequence within a well . Thus , sample complexity is significantly reduced , which reduces ambiguity in the reconstruction of a consensus sequence . As an example , two different predicted repeat-containing genes ( g2001 , 1189bp; and g2002 , 688bp ) were assembled from two different wells ( 005 and 145 respectively ) . Although they contain highly homologous repeats ( represented as a Dot Matrix plot , ( D ) these repetitive genes were resolved and reconstructed properly in the final assembly . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 00510 . 7554/eLife . 00569 . 006Figure 2—figure supplement 1 . Validation of LRseq approach on human genomic DNA . Genomic DNA from HapMap NA7019 was prepared for LRseq . These figures show LRseq assembly statistics , obtained by mapping sequenced reads to human genome reference 36 . These data were also used to estimate the concentration of amplifiable molecules in B . schlosseri 356a DNA samples prepared by an identical protocol . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 00610 . 7554/eLife . 00569 . 007Figure 2—figure supplement 2 . Clonality confirmation of the genome of clone Sc6a-b and clone 356a . ( A ) Sc6a-b clone , a long lived ( 7 years old when sampled ) , highly regenerative colony was chosen to be sequenced . Sc6a-b subclones were starved for 48 hr prior to sampling , and 400 individuals ( zooids ) were sampled for sequencing . Subclones of this colony are still alive and maintained in our mariculture facility . ( B ) A few zooids were taken from every sample set and tested via AFLP’s genotyping analysis , confirming that all zooids belong to one genotype . ( C and D ) . Sc6a-b microsatellite loci were homozygous ( 2 loci ) and heterozygous ( 1 loci ) confirming one genotype . ( E and F ) 356a clone was a highly regenerative long lived colony . 150 individuals were sampled and their gDNA was sequenced . Microsatellite loci were homozygous ( E and F ) , confirming one genotype . Scale bar-1 mmDOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 00710 . 7554/eLife . 00569 . 008Figure 2—figure supplement 3 . Statistics for 356a assembly . ( A ) Contig length distribution . ( B ) Distribution of coverage of 356a assembled Celera contigs by Velvet assembled fragments . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 00810 . 7554/eLife . 00569 . 009Figure 2—figure supplement 4 . Interspersed and tandem repeats distribution in the B . schlosseri genome . ( A ) RepeatScout ( version 1 . 0 . 5; Price et al . , 2005 ) was used to identify interspersed repeat elements de novo using a k-mer length of 14 . All identified repeats were subsequently filtered for tandem repeat and low complexity content , using RepeatScout . Genome-wide interspersed repeats were catalogued using RepeatMasker ( version open-4 . 0; Smit et al . , 1996-2010 ) . The distribution of large interspersed repeats families ( ≥1kb ) ordered by copy number is presented . ( B ) To identify both perfect ( 100% sequence identity ) and degenerate genomic tandem repeats , we used XSTREAM ( Newman and Cooper , 2007 ) , with a minimum repeat length of 20 bp , minimum word match of 0 . 8 , and otherwise default parameters . 3 , 183 , 988 tandem repeats were identified , period range: 1–6525 bp , copy number range: 2 . 7–1096xDOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 00910 . 7554/eLife . 00569 . 010Figure 2—figure supplement 5 . Coverage of 4 fosmids by the B . schlosseri assembly . Fosmid sequences ( red lines; gi; ac numbers are shown , number=bp ) , were compared with B . schlosseri contigs using blast ( e-value < e−10 ) . Best alignments between contigs >500bp ( black lines ) are shown . Repetitive regions are marked ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 01010 . 7554/eLife . 00569 . 011Figure 2—figure supplement 6 . Validation of putative B . schlosseri genes . We experimentally validated 145 B . schlosseri predicted genes . Genes were validated by observing expression in B . schlosseri cDNAs and gDNA via PCR and qPCR assays and resequencing them on Sanger . ( A ) cDNA PCR product of several early erythroid and HSC putative genes identified in B . schlosseri tissues ( endostyle , blood or zooid ) . Names of the putative genes and the tissues that were tested in this experiment are indicated on the gel image . ( B ) qPCR expression in B . schlosseri blood of six putative immunity genes . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 011 Genomic DNA ( gDNA ) was extracted from tissue from two long-lived B . schlosseri colonies ( Sc6a-b and 356a ) raised in our mariculture facility ( ‘Materials and methods’ under ‘Animals and genomic DNA sample collection’ ) . Microsatellite heterogeneity confirmed clonality ( Figure 2—figure supplement 2 ) . Each colony was sequenced and assembled separately . We first attempted conventional sequencing and assembly from colony Sc6a-b DNA using Roche 454 Titanium ( Branford , USA ) and Illumina GAII ( San Diego , USA ) sequences ( Supplementary file 2C , ‘Materials and methods’ under ‘Genome sequencing and assembly’ ) . This Sc6a-b assembly achieved an average N50 of 1 kb , yielding short contigs that were insufficient for whole genome assembly ( Supplementary file 2D ) . By contrast , when we applied LRseq to the 356a clone , we obtained a 566 Mbp assembly with a dramatically improved N50 of 7kb ( Supplementary file 2D; Figure 2—figure supplement 3 ) . This approach not only simplified the assembly of a complex eukaryotic genome , but also reduced the confounding impact of repetitive DNA on contig assembly ( Figure 2C–D; Figure 2—figure supplement 3 ) . We sought to determine the chromosomal organization of the B . schlosseri genome . Using embryos from a wild B . schlosseri colony from Monterey Bay , we loaded a dilute solution of dispersed metaphase chromosomes into a microfluidic device as previously described ( Fan et al . , 2011 ) . The isolated metaphase chromosome mixtures from 21 individual wells were amplified , barcoded , and sequenced separately ( ‘Materials and methods’ under ‘Chromosome sequencing , assignment and assembly’; Fan et al . , 2011; Xu et al . , 2011 ) . Using the 21 chromosome mixtures , containing between 1 and 4 chromosomes each , 356a genomic contigs larger than 7 kb were aligned to the chromosome reads using BWA . Then , scaffolds were assigned to chromosome clusters by iterative K-means clustering on the correlation matrix between each scaffold ( Figure 3; ‘Materials and methods’ under ‘Chromosome sequencing , assignment and assembly’ ) . Assuming that B . schlosseri carries 16x2 chromosomes ( Colombera , 1963 ) , this approach clearly resolves 13 chromosomes with a mean chromosome meta-scaffold size of 16 , 234 kb and a mean N50 of 38 kb ( Figure 3; Figure 3—figure supplement 1; Supplementary file 2D ) . Finally , we attempted to improve our genomic assembly by incorporating the additional 21 chromosome assemblies into a hybrid assembly ( ‘Materials and methods’ under ‘Chromosome sequencing , assignment and assembly’; Figure 3—figure supplement 2; Figure 3—figure supplement 1 ) . An overall improvement in N50 was achieved , yielding a final 580 Mbp draft assembly ( Supplementary file 2D ) . 10 . 7554/eLife . 00569 . 012Figure 3 . Clustering and assignment of B . schlosseri chromosomes . ( A ) We isolated and sequenced 21 metaphase chromosome mixtures using a microfluidic device . Each chromosome mixtures was amplified , barcoded and sequenced separately ( IlluminaHiSeq ) . Genomic contigs larger than 7 kb were aligned to the chromosome reads using BWA . Subsequently , assignment of scaffolds to chromosome cluster was performed using iterative K-means clustering on the correlation matrix between each scaffold . In addition , to find the number of clusters/chromosomes we performed k-means clustering iteratively across different cluster numbers . This plot demonstrates that increasing beyond 13 clusters does little to reduce the error; therefore 13 chromosomes were successfully resolved . ( B ) To estimate the configuration after the clustering step , 17 out of the 21 wells were deduced to contain information that is used in the clustering process . The average number of normalized reads counts from each metaphase chromosome mixture ( well ) that align to each scaffold in a cluster group was calculated and plotted . Each peak represented can be inferred to denote the presence of a specific chromosome in the well . Examples of four representative wells are presented , metaphase chromosome mixtures contained between 1–4 chromosomes ( see also Figure 3—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 01210 . 7554/eLife . 00569 . 013Figure 3—figure supplement 1 . Distribution of B . schlosseri chromosome groups across different wells . We isolated and sequenced metaphase diluted chromosome mixtures using a microfluidic device . Each chromosome mixture was amplified , barcoded and sequenced separately ( IlluminaHiSeq ) . The average number of normalized reads counts from each diluted chromosome mixture ( well ) that align to each scaffold in a cluster group was calculated and plotted . Each peak represents the presence of a specific chromosome in the well . In the 17 wells presented above , chromosome mixtures contained between 1–4 chromosomes . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 01310 . 7554/eLife . 00569 . 014Figure 3—figure supplement 2 . Pipeline for the assignment of chromosome scaffolds and the 356a–chromosomes hybrid assembly process . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 01410 . 7554/eLife . 00569 . 015Figure 3—figure supplement 3 . 356a-Chromosome hybrid assembly of B . schlosseri . Reads from each of the individual chromosome sample preparations were subsequently assembled using Velvet . The resulting chromosome level contigs were then merged with the 356a assembly to create a 356a-chromosome hybrid assembly . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 01510 . 7554/eLife . 00569 . 016Figure 3—figure supplement 4 . The fraction of B . schlosseri predicted intron-less genes ( blue ) and genes with introns ( red ) in the different chromosomes . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 016 Repetitive elements can confound traditional genome assembly methods ( Salzberg et al . , 2012 ) , and are often removed to avoid assembly errors ( e . g . , Dehal et al . , 2002; Putnam et al . , 2007; Shinzato et al . , 2011; Supplementary file 2H ) . However , because LRSeq was designed to explicitly resolve long read sequences even in the presence of repeats , we further evaluated LRSeq performance by enumerating two major repeat classes in the assembly , interspersed repeats and tandem repeats . We used RepeatScout for de novo identification of interspersed elements ( Price et al . , 2005 ) , and RepeatMasker ( Smit et al . , 1996–2010 ) for analysis of genome-wide repeat demographics . We identified 6601 interspersed repeat families , each present in at least three copies , that together cover ∼65% of the B . schlosseri genome assembly ( Supplementary file 2E ) . We also identified 1400 large repeat families , defined as interspersed repeats with genomic alignments of at least 1 kb . Notably , large interspersed repeats are found in a median of four chromosomes ( of 13 chromosome assignments ) , and >10% of large interspersed repeat families occur in over 100 copies ( Supplementary file 2E; Figure 2—figure supplement 4A ) . Despite considerable repetitive content , we observed a strong concordance between genomic contigs and Sanger fosmid sequences , supporting the effectiveness of the LRseq approach ( e . g . , see Figure 2—figure supplement 5 ) . As a further validation , we interrogated our former sc6ab 380 Mb assembly for the same interspersed repeat elements , with the expectation of recovering less repeats . Indeed , only 52 . 27% of sc6ab base pairs were masked using the same repeat library . These results validate the repeat families and support their widespread presence in the B . schlosseri genome . Finally , we analyzed the assembly for perfect ( 100% sequence identity ) and degenerate tandem repeat content using XSTREAM ( Newman and Cooper , 2007 ) . In all , ∼3 . 2 million tandem repeats were identified , with periods ranging from 1–6525 bp and copy numbers ranging from 2–1096x ( Figure 2—figure supplement 4B ) . By comparison , the human genome was assembled to a very high standard using conventional Sanger technology and later Illumina technology , and was found to contain over 50% repeats ( de Konning et al . , 2011 ) . The considerable repeat content and diversity in the B . schlosseri genome indicates that LRseq may have general utility for resolving repeat architectures in diverse eukaryotic genomes . We further validated the assembly by comparison to a variety of independently generated B . schlosseri sequence data . All B . schlosseri genes ( n = 66 ) , fosmid sequences ( n = 11 ) and most of the 98 , 611 expressed sequence tags ( ESTs ) available from NCBI aligned with the B . schlosseri draft assembly ( Supplementary file 2F , Figure 2—figure supplement 5; ‘Materials and methods’ under ‘Evaluation of 365a-chromosomes hybrid assembly’ ) . Moreover , nearly all of the independently sequenced and assembled Roche 454 Sc6a-b contigs ( 93% ) were successfully mapped to the assembly ( Supplementary files 2F; ‘Materials and methods’ under ‘Evaluation of 365a-chromosomes hybrid assembly’ ) . Taken together , these data represent independent validation of the quality and integrity of the B . schlosseri draft assembly which compares favorably to , and in some cases exceeds , existing wild type genomes with respect to ungapped chromosome contig N50 , chromosome assignments , and repeat sequence integration ( e . g . , see Supplementary file 2G ) . Next , to identify protein-encoding genes , we generated RNA-Seq data ( 88 Gb; Supplementary file 2C ) from 19 different colonies to guide the gene prediction program Augustus ( Stanke et al . , 2008 ) . In total , 38 , 730 putative protein-coding loci were identified , all of which have at least 30% transcript support ( ‘Materials and methods’ under ‘RNA sequencing’ , ‘Gene prediction’ , ‘Gene annotation’; Supplementary file 2I ) . Among these predicted genes , 27 , 463 include a start and stop codon , 13 , 910 genes have at least one intron , and 13 , 553 are intron-less ( Supplementary file 2H ) . Moreover , for each of the B . schlosseri chromosomes 55% of genes have at least one intron while ∼45% are intron-less ( Figure 3—figure supplement 4 ) . In addition , the mean B . schlosseri gene length is predicted to be 3 . 6 kbp with a mean exon length of 170 bp ( Supplementary file 2H ) . We tested a set of 145 genes by PCR and Sanger-sequencing , and were able to confirm 144 of them ( 99 . 3% ) , further validating the genome assembly ( Figure 2—figure supplement 6 , ‘Materials and methods’ under ‘Evaluation of genes’ ) . Using these predicted genes , we investigated the evolutionary position of B . schlosseri . Phylogenomic analysis of 425 conserved homologous genes across 15 diverse species , and mitogenomic analysis of 65 species both support the phylogenic position of tunicates within Chordata ( Delsuc et al . , 2006; Figure 1E; Figure 1—figure supplement 1 , Supplementary file 1; ‘Materials and methods’ under ‘Mitochondrial phylogeny’ , ‘Phylogenomic analyses’ ) , and provide strong evidence that colonial and solitary tunicates represent the closest living relative of vertebrates . We investigated the B . schlosseri genome for molecular events underlying the emergence and early diversification of vertebrates . Protein-encoding genes in B . schlosseri were compared to a diverse sampling of 18 well-annotated genomes from other species , and for each genome , we assessed the presence or absence of significant homology to human or mouse proteins ( Figure 4—source data 1A; ‘Materials and methods’ under ‘Evolution analysis’ ) . All proteomes were combined into a single data set ( of constant size ) for blast analysis . As such , differences in the number of genes per genome would not have impacted our results . An e-value cutoff of e−10 was selected to strike a balance between statistical significance and the detection of remote homology ( ‘Materials and methods’ under ‘Evolution analysis’ ) . Among the analyzed species , we found that 77% of human genes could be traced back to protochordates with at least some homology ( e-value ≤ e−10 ) , around 10% less than chicken ( 85% ) and frog ( 86% ) genomes , indicating that the common ancestral genome of tunicates and vertebrates had homology to at least 77% of the human gene repertoire . This list includes about 660 genes present in the common ancestor , but absent in non-chordate species ( Figure 4—source data 1B ) . Among the genes found in B . schlosseri ( either alone or in combination with other protochordate species ) and vertebrates ( Figure 4—source data 1B , Figure 4—source data 1C ) , we found genes that are critical to the development and function of the vertebrate heart ( e . g . , ALPK3 , TNNT2; Hosoda et al . , 2001; Frey et al . , 2012 ) , and eye ( gamma and beta crystallins; Sun et al . , 2011 ) , and the ability to hear ( GJB2/3/6 CLDN; Rabionet et al . , 2000; Wilcox et al . , 2001; Figure 4 , Figure 4—source data 1C ) . Mutations in these genes are implicated in a variety of human diseases and disorders , including heart diseases ( Frey et al . , 2012 ) , cataracts ( Sun et al . , 2011 ) , deafness ( Rabionet et al . , 2000; Wilcox et al . , 2001 ) , and nemaline myopathy ( Johnston et al . , 2000; Figure 4 , Figure 4—source data 1C ) . In addition , B . schlosseri was the only protochordate in our analysis with proteins homologous to pregnancy-specific glycoproteins ( PSGs ) . PSGs are the major placental polypeptides , and complications in pregnancies and spontaneous abortions have been associated with abnormally low levels of PSGs in the maternal blood ( Camolotto et al . , 2010 ) . Analogous to mammalian pregnancies , a common blood supply among kin is established and tolerated in B . schlosseri chimeras ( Voskoboynik et al . , 2009 ) . Thus , by studying PSG-like proteins in B . schlosseri , new insights might be gained into maternal and fetal medicine . 10 . 7554/eLife . 00569 . 017Figure 4 . Innovations underlying the emergence and early diversification of vertebrates . Protein-encoding genes in B . schlosseri were compared to a diverse sampling of 18 well-annotated genomes from other species , and for each genome , the presence or absence of homology to human or mouse proteins was assessed ( all vs all blastp e-value threshold of 1e−10; Figure 4—source data 1A ) . Our data indicate that homologs of ∼660 human/mouse genes were present in the common ancestor of tunicates and vertebrates , but not non-chordate species Figure 4—source data 1B ) . Among them are genes associated with the development , function , and pathology of vertebrate features , including heart , eye , hearing , immunity , pregnancy and cancer ( Figure 4—source data 1C ) . Gray box = no homology; Yellow box = homology . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 01710 . 7554/eLife . 00569 . 018Figure 4—source data 1 . Vertebrates evolution . ( A ) Innovations that underline the emergence and early diversification of vertebrates . We compared protein-encoding genes in B . schlosseri to a diverse sampling of 18 well-annotated genomes from other species . All protein sequences were compared by blastp against all other protein sequences . Based on this data set a list was generated of genes known from human and mouse and their existence ( 1 ) or absence ( 0 ) in the tested species ( e-value < e−10 ) . ( B ) The 660 putative genes present in protochordates , human and mouse , but absent in non-chordate species . This list was generated from Figure 4—source data 1A . Per every species , or species group we filtered for genes that were present in the tested species/species group and in human or mouse , but were absent in non-chordate species . ( C ) Innovations that underline the emergence and early diversification of vertebrates . This table is based on data gathered in Figure 4—source data 1B and is focused on the genes that are present in B . schlosseri and vertebrates ( either alone or in combination with other protochordate species ) but are absent in non-chordate species . A ToppGene analysis is presented of these sets of genes which summarized their molecular functions , biological processes , human and mouse phenotypes , and pathways they are involved in , gene families , drugs and human diseases . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 018 Numerous genes predicted to have evolved in a common ancestor of B . schlosseri and vertebrates are essential to the immune system and hematopoiesis ( Figure 4 , Figure 4—source data 1B , Figure 4—source data 1C ) . Six genes unique to B . schlosseri and vertebrates ( ZBTB1 , MEFV , DSG3 , NQO1 , NQO2 and BHLHE40 ) are associated with increased leukocyte and hematopoietic cell numbers ( Figure 4—source data 1C; Chen et al . , 2009 ) . In our analysis , these genes are absent in cephalochordates and solitary urochordates , which all lack a defined vascular system ( Moller and Philpott , 2005; Lemaire , 2011 ) . In contrast , the heart in each individual zooid in a B . schlosseri colony beats synchronously with the hearts of other zooids in the colony , driving a bidirectional blood flow throughout an interconnected vasculature ( Video 1 ) . Moreover , this blood system carries at least ten morphologically different cell types ( Schlumpberger et al . , 1984; Ballarin et al . , 2008 ) . Because of the anatomy of B . schlosseri , coupled with its hematopoietic-related gene repertoire , we hypothesize that colonial ascidians may have retained and elaborated many components of the ancestral hematopoietic program , much of which has been lost in extant solitary urochordates and cephalochordates . 10 . 7554/eLife . 00569 . 019Video 1 . B . schlosseri blood circulation . ( A ) Time-lapse acquisition of blood flow in the blood vessels ( bv ) and ampullae of a chimeric B . schlosseri colony , generated from a fusion between a mother and its offspring ( fused ) . ( B ) Ampullae contract , buds develop , and a colony gets ready to replace the old generation . ( C ) Old generation zooids are getting resorbed ( res . z ) and replaced by the new generation ( buds ) . ( D ) A heart beating and pumping blood in the primary bud of a different colony . ( E ) Blood flow through a common blood vessel between two allogeneic/compatible colonies , creating a natural chimera . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 019 We next attempted to identify potential precursors of human hematopoietic populations in B . schlosseri and 17 other diverse species , including fungal , plant , and mammalian species . We analyzed gene expression microarray data from 26 different human blood cell populations , and additional non-blood human tissue samples . We identified a set of twenty signature genes that were highly expressed in each of the 26 hematopoietic populations ( Benita et al . , 2010; Seita et al . , 2012; ‘Materials and methods’ under ‘Evolution analysis’ ) . For each blood-related gene set , we identified homologous gene sequences in B . schlosseri and 17 other species ( Figure 5 , Supplementary file 3 ) . Among B . schlosseri homologs , we found high enrichment for gene sets predominantly expressed in human hematopoietic stem cells ( HSCs; i . e . , 14 of 20 cord blood HSC genes ) , myeloid populations ( i . e . , 14 of 20 early erythroid CD71+ genes ) , and early but not mature lymphoid populations . Consistent with previous studies ( Bartl et al . , 1994; Laird et al . , 2000; Dishaw and Litman , 2009; Guo et al . , 2009; Flajnik and Kasahara , 2010; Bajoghli et al . , 2011; Hirano et al . , 2011 ) this analysis indicates that the evolution of adaptive immunity progressed rapidly beginning with jawless vertebrates , with much of the genetic repertoire in place by the emergence of jawed vertebrates ( Figure 5 ) . However , homologs of human genes with specific expression in HSC and blood progenitor populations , including T and B progenitor cells , appear early in metazoan evolution ( Figure 5; Supplementary file 3 ) . 10 . 7554/eLife . 00569 . 020Figure 5 . Analysis of blood and immune cell type-specific genes across evolution reveals evidence for hematopoietic precursors in B . schlosseri . We analyzed gene expression microarray data from 26 different human blood cell populations , organized into four cell lineages ( HSC; Lymphoid Progenitors; Myeloid and Lymphoid Lineage ) , and identified a set of twenty signature genes with highly enriched expression profiles for each population ( Supplementary file 3 ) . For each blood-related gene set , we identified homologous gene sequences in B . schlosseri and 17 other species; the fraction of genes ( out of 20 ) found for each species is displayed as a heat map . Within each major lineage , cell populations are sorted in decreasing order by a conservation index , calculated as the average number of genes found across the 18 species ( indicated by a blue bar graph ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00569 . 020 Unlike solitary tunicates ( e . g . , Ciona ) , B . schlosseri has a defined vasculature with circulating blood cells ( including cells with lymphocyte-like and macrophage-like morphology; Schlumpberger et al . , 1984; Ballarin et al . , 2008; Video 1 ) . As such , we further investigated by PCR and re-sequencing the expression of all 28 B . schlosseri genes with homology to human HSCs ( n = 14 ) and early erythroid CD71+ blood cell ( n = 14 ) gene sets . Strikingly , we found evidence for expression of 13 HSC homologs in the B . schlosseri endostyle stem cell niche ( Voskoboynik et al . , 2008 ) , and 7 in the vasculature . We also confirmed expression of all 14 early erythroid CD71+ genes in the vasculature and endostyle ( Supplementary file 3 ) . Thus , our analysis not only identified B . schlosseri genes that may define evolutionary precursor cells of human hematopoietic lineages , but also indicates that the evolution of hematopoiesis proceeded from stem cells to myeloid populations to lymphoid populations , leading to the eventual emergence , absent in B . schlosseri , of T/B-cell based adaptive immunity in vertebrates ( Figure 5; Supplementary file 3 ) . Not surprisingly , the B . schlosseri genome lacks significant homology to most genes known to play an important role in the vertebrate adaptive immune system . For instance , no evidence for the following immune-related genes could be found: ( i ) assembled major histocompatibility genes , ( ii ) genes with homology to RAG1/RAG2 , which are involved in immunoglobulin and T-cell receptor rearrangements , ( iii ) terminal deoxynucleotidyl transferase , which adds nucleotides to the rearranging VDJ elements to create receptor diversity , ( iv ) V region subgenic elements encoding T cell and immunoglobulin antigen receptor domains , or ( v ) VLR like immune receptor elements found in lampreys ( Weigert et al . , 1970; Davis et al . , 1984; Oettinger et al . , 1990; Fagan and Weissman , 1998; Laird et al . , 2000; Muramatsu et al . , 2000; Pancer et al . , 2004; Rogozin et al . , 2007; Dishaw and Litman , 2009; Flajnik and Kasahara , 2010; Hirano et al . , 2011 ) . We identified a large fraction ( ∼45%; Supplementary file 2H; Figure 3—figure supplement 4 ) of intron-less genes in the B . schlosseri draft genome , including retroviral genes such as Gag , Poli , Env and LTRs , which are used by viruses to insert their genetic sequences into the host genomes . As adaptive immunity genes like RAG1/RAG2 are intron-less and first appear in jawed vertebrates , it has been suggested that they may have originated via horizontal infections of primitive retroviral like agents , and/or gene transfer ( Bartl et al . , 1994 ) . In addition , the B . schlosseri genome encodes homologues of Foxn1 , the thymus epithelial gene mutated in the immunodeficient nude mouse ( nu/nu ) , a marker of the thymopoietic microenvironment in vertebrates ( Nehls et al . , 1996; Bajoghli et al . , 2011 ) . These data indicate that at least some genetic circuitry relevant for vertebrate adaptive immunity was already in place in the common ancestor of the protochordate B . schlosseri and vertebrates . It leaves open the question of whether Ig or TCR genes , and the MHC proteins that capture and present intracellular peptides to T cells expressing these TCR proteins , existed in antecedents to B . schlosseri but were lost or somehow introduced after the line from colonial tunicates to the organisms that have an adaptive immune system . As omnis DNA e DNA , this question is perhaps the most puzzling of our findings . In conclusion , using a novel method for deciphering eukaryotic genomes , we assembled and analyzed the B . schlosseri genome , the first colonial tunicate to be sequenced . One of the great challenges in evolutionary biology is to understand how differences in DNA sequences between species underlie distinct phenotypes . The B . schlosseri genome provides an important new resource for unraveling the genes and regulatory logic that led to the emergence of vertebrates and lymphoid-mediated immunity . Moreover , the many important features encoded by the B . schlosseri genome will facilitate new insights into complex vasculature , chimerism among kin , whole-body stem cell-mediated regeneration , and a colonial lifestyle . Mature reproductive colonies of Botryllus schlosseri ( Pallas ) were collected from Santa Cruz and Monterey marinas , California . Hatched larvae were settled raised , and crossed in our mariculture facility as described ( Boyd et al . , 1986; De Tomaso et al . , 1998 ) . Long lived , highly regenerative colonies Sc6a-b and 356a , were chosen to be sequenced . Sc6a-b subclones were starved for 48 hr prior to sampling ( to minimize DNA contamination ) , and 400 individuals ( zooids ) were sampled for gDNA sequencing . Subclones of this colony are still alive and maintained in our mariculture facility ( Figure 2—figure supplement 2A ) . 150 individuals from colony 356a were sampled and their gDNA was sequenced . To confirm that all zooids belonged to one genotype , a few zooids were taken from every sample set and screened for polymorphism via amplified fragment length polymorphism ( AFLP ) analysis as described in ( Voskoboynik et al . , 2008 ) and microsatellite loci analysis as described in ( Stoner et al . , 2002 ) , confirming one genotype for Sc6a-b and 356a colonies ( Figure 2—figure supplement 2B–F ) . Tissue samples were dissected and flash-frozen in liquid nitrogen . Genomic DNA samples were extracted using a modified version of the Hoss and Paabo protocol ( Hoss and Paabo , 1993 ) as described ( De Tomaso et al . , 1998 ) . The B . schlosseri genome sequence assembly was performed using two independent methods . Colony Sc6a-b genome sequence data was obtained using single read and paired-end protocols on the Roche ( Roche , Branford , CT ) 454 GS-FLX and Illumina Genome Analyzer II ( GAII; Illumina , San Diego , CA ) instruments ( Supplementary file 2C ) . Sc6a-b gDNA was fragmented , libraries prepared , and sequencing conducted according to the manufacturer’s protocols . The 454 platform generated a total of 3086 Mb sequence data , the Illumina platform generated 3597 Mbp sequence data . The 6683 Mbp of sequence data obtained corresponds to ∼11-fold coverage of the B . schlosseri genome ( estimated size of 600 Mbp , 454kmer estimation; Supplementary file 2C ) . The 454 shotgun and Illumina GAII paired-end reads were assembled de novo using Newbler v2 . 5 ( Roche ) with default settings , heterozygote mode . 380 Mbp comprised of 518 , 856 contigs were assembled with N50 of 1160 bp ( Sc6a-b draft assembly; Supplementary file 2D ) . Colony 356a We developed a novel method to obtain a sequence in order to assemble larger contigs and reduce assembly complexity . Colony 356a gDNA was sheared using HydroShear ( speed setting 16; 20 cycles ) into random fragments of 6–10 kb . Sheared gDNA was run on a 0 . 8% agarose gel , the 6–8 kb band was cut and the DNA extracted using Qiagen gel purification kit . Fragmented DNA was repaired using NEB end repair module ( E6050S ) to produce blunt ends . Blunt end DNA was purified on a Qiagen column . After purification standard double stranded adapters from the Roche 454 kit were ligated with NEB Quick Ligase , following Roche 454 Titanium protocols . 454 adaptor mix 27 , 145 , a mix of two sequences was used: Primer A1: 5′-CCATCTCATCCCTGCGTGTCTCCGACTCAG-3′; 3′-TCTCCGACTCAG-5′ Primer B: 5″-/5BioTEG/CCTATCCCCTGTGTGCCTTGGCAGTCTCAG-3′; 3′-TGGCAGTCTCAG-5′ Amplification primer mix: Ti forward: 5′-CCATCTCATCCCTGCGTGTC-3′; Ti reverse: 5′-CCTATCCCCTGTGTGCCTTG-3′ These adapters serve as priming sites for the downstream amplification of the long fragment library . Following DNA purification ( Agencourt AMPureXP bead purification ) and fill-in reaction , second size selection was performed to remove adapter dimers and narrow down DNA size range ( 6–8 kb ) . Qiagen gel purification kit was used to purify DNA . PCR amplification for long range amplification was performed as follows: initial denaturation at 94°C for 30 s , followed by 23 cycles of ( 94°C for 15 s , 65°C for 7 min ) , followed by a final extension 65°C for 7 min . Concentration of amplifiable molecules carrying both amplification adapters was estimated by comparing B . schlosseri samples to a human standard sample prepared using an identical protocol . qPCR with nonspecific intercalating dye ( EvaGreen , Biotinum ) was used to calculate concentrations . Human standard was prepared following an identical protocol from genomic DNA derived from HapMap sample NA7019 . The amount of amplifiable DNA was obtained by mapping short reads to a human genome reference 36 and measuring the fraction of the genome that was covered ( Figure 2—figure supplement 1 ) . Mapping was done using Novoalign with default settings , counting the amount of bases covered by regions of more than 1500 bp with at least 2x coverage of properly mapping paired end reads . We then aliquoted the resulting library of B . schlosseri gDNA with amplification adapters into wells of two 96-well plates such that , on average , each well contained a predefined amount of amplifiable molecules ( estimated number of 200–2000 molecule per well; 1–6 Mb of total amplifiable sequence ) . Randomly sampled molecules in each well were amplified using NEB LongAmp master mix in the presence of 400 nM of primers: LA-V2-LEFT 5′-CCATCTCATCCCTGCGTGTCTCCG-3′; LA-V2-RIGHT 5′-CCTATCCCCTGTGTGCCTTGGCAGT-3′ ) complementary to previously ligated adapters following the two-step protocol . The resulting library of amplicons was purified using Zymo ZR-96 DNA Clean & Concentrator-5 kit . Purified DNA was eluted into two 96-well plates according to the manufacturer’s protocol . DNA was fragmented and tagged using Nextera DNA sample prep kit . Following the standard protocol , samples were incubated for 5 min at 55°C in the provided high molecular weight buffer ( Nextera DNA sample prep kit , Epicentre ) . Fragmented DNA was purified using Zymo ZR-96 DNA Clean & Concentrator-5 kit and was converted into Illumina compatible sequencing library using a custom protocol . Four oligos described in the Nextera DNA sample prep kit ( Epicentre and Illumina ) were added to the every well containing purified fragmented DNA in concentrations recommended by Epicentre . Adaptor 1: 5′AATGATACGGCGACCACCGAGATCTACACGCCTCCCTCGCGCCATCAG-3′; Primer 1: 5′-AATGATACGGCGACCACCGA-3′; Primer 2: 5′-CAAGCAGAAGACGGCATACGA-3′; Adaptor 2: 5′-CAAGCAGAAGACGGCATACGAGAT-[BAR CODE]-CGGTCTGCCTTGCCAGCCCGCTCAG-3′ Adapter 2 carried a 7 bp barcode sequence unique for each well of two 96-well plates ( Supplementary file 2I ) . Amplification was done using NEB Phusion GC master mix ( 2x ) following the recommended Nextera limited cycle PCR protocol designed to incorporate barcoded adapters: 72°C for 3 min , 95°C for 40 s , followed by nine cycles of 62°C for 30 s , 72°C for 3 min . The resulting 192 Illumina libraries were pooled together and purified using Qiagen Quiavac96 DNA purification kit . Size selection was performed by running a 2% agarose gel and excising the 400–900 bp band . This gel band was purified ( Qiagen ) , quantitated using Agilent Bioanalyzer 2100 High-Sensitivity chip and sequenced on Illumina HiSeq 2000 sequencer following manufacturer recommended protocols . After sequencing , multiplexed libraries reads were de-multiplexed and separated into independent files ( according to barcodes ) . Reads were then screened to remove reads with low overall quality and reads containing Nextera adapters ( 5′-AGATGTGTATAAGAGACAG-3′ ) resulting from imperfect size selection . We have sequenced a total of eight libraries ( Supplementary files 2A , B ) . The resulting pool of reads was assembled using Velvet ( Zerbino , 2010 ) to reconstruct 6–8 kb original fragments using Velvet optimizer mode ( Figure 2 , Supplementary file 2A , B ) . Resulting contigs from Velvet were treated as input reads for downstream assembly with Celera Assembler ( Myers et al . , 2000 ) version 6 . 1 , to produce 356a draft assembly ( Figure 2—figure supplement 3 , Supplementary file 2D ) . Embryos were isolated from a wild B . schlosseri colony from Monterey Bay . Metaphase chromosomes were isolated as previously described ( Shoguchi et al . , 2004 ) . B . schlosseri metaphase chromosome suspension was partitioned into wells in the microfluidic device as previously described ( Fan et al . , 2011; Xu et al . , 2011 ) . The contents of each microfluidic well were amplified individually and prepared for sequencing . Each well contained between 1–4 metaphase chromosomes ( Figure 3 , Figure 3—figure supplement 1 , Figure 3—figure Supplement 2 ) . 21 wells were made into libraries and sequenced using Illumina Hiseq ( 2 x 100 ) . All of the 66 B . schlosseri genes , 11 fosmid sequences and most of the 98 , 611 expressed sequence tags ( ESTs ) available from NCBI aligned with the B . schlosseri final draft assembly ( Supplementary file 2F; Figure 2—figure supplement 5 ) . Nearly all of the independently sequenced and assembled Roche 454 Sc6a-b contigs ( 93% ) were successfully mapped to the final assembly ( Supplementary file 2F ) . The Sanger sequenced NCBI genes and ESTs , and the 454 and Illumina GAII sequenced Sc6a-b genome ( ∼11x fold coverage , Supplementary file 2C ) , provide validation and an independent test to the quality and integrity of the final assembly . RNA was isolated from 19 different individuals ( developmental stages A–D; different ages ) . To minimize DNA contamination , colonies were starved for 48 hr before sampling . Total RNA was extracted following the manufacturer’s instructions ( Ambion; Purelink RNA mini kit ) and purified using the Purelink DNase kit ( Invitrogen ) . cDNA libraries for Illumina HiSeq and MiSeq were prepared ( Ovation RNA-Seq v1 system , Nugen; NEBnext DNA Master Mix for Illumina ( New England Biolabs ) and standard Illumina adapters and primers from IDT . RNA-Seq ( 2x100 bp; Illumina HiSeq ) was performed . Each genotype was sequenced separately . In total , ∼88 Gb of raw transcriptome sequence data were generated for the 19 colonies ( Supplementary file 2C ) . To investigate expression of genes in particular tissues , we also generated tissue-specific RNA-Seq libraries using the Illumina GAII ( single-end 36 bp reads ) and Illumina MiSeq ( 100-bp paired-end ) . For Illumina GAII RNAseq total RNA was isolated following Invitrogen’s recommendations . PolyA+ mRNA were isolated using oligoT Dynal beads . Reverse transcription and library construction protocols were provided by Illumina . In total , ∼4 Gb of raw transcriptome sequence data were generated for the tissue specific ( Supplementary file 2C ) . Using Cufflinks ( Trapnell et al . , 2010 ) with default parameters , B . schlosseri cDNA reads were aligned to the draft assembly and a reference-guided transcript assembly was produced . To predict genes , we used the program Augustus v2 . 5 . 5 ( Stanke et al . , 2008 ) . The reference-guided transcripts assembly was aligned to the draft genome assembly and a ‘Hints’ gff file was generated to guide gene prediction . Augustus was run ( using human HMM and parameters ) , and from 121 , 094 contigs , a total of 38 , 730 genes with a minimum of 30% transcript support were predicted ( Supplementary file 2H ) . All B . schlosseri candidate protein-coding genes were compared to human and mouse proteomes ( UniProtKB/Swiss-Prot; see Sequence data in ‘Materials and methods ’ under ‘Phylogenomic analyses’ ) using blastp and an e-value threshold of e−10 ( see Sequence Data , ‘Materials and methods’ under ‘Evolution analysis’ ) . In addition , all B . schlosseri candidate protein-coding genes were compared to the NCBI non-redundant protein database ( NR ) using blastx and an e-value threshold of e−10 . For every predicted gene 2 annotations were assigned ( 1 ) . The best hit ( smallest e-value ) from NR and ( 2 ) . if available , the best hit from the UniProt mouse/human blastp results . RNA from B . schlosseri endostyles , vasculature ( ampullae ) and blood cells were isolated using an Ambion Purelink RNA minikit and cDNA prepared using Protoscript AMV LongAmp Taq RT-PCR kit ( NEB ) . cDNA was amplified using GE Illustra Hot Start RTG beads and amplified using primers designed for the tested genes . For the amplification and Sanger sequencing of the putative genes from specific blood groups , specific primers were used . Per every gene , expression was tested using cDNA prepared from endostyle , vasculature and blood . PCR was performed on the MJ Research PTC-200 thermal cycler as follows: Initial denaturation at 95°C for 4 min , followed by 34 cycles of 95°C for 1 min , 59ofor 1 min , 72°C for 1 min , followed by a final extension 72°C for 20 min . Amplified products were run on an E-Gel EX 1% agarose gel ( Invitrogen ) to validate size , then sent to MCLAB ( 384 Oyster Pt Rd . S . San Francisco , CA ) for sequencing . We tested a set of 145 predicted genes by PCR and Sanger-sequencing , and were able to confirm 144 of them ( 99 . 3% ) , further validating the genome assembly ( Figure 2—figure supplement 6 ) . The sequence of the B . schlosseri mitochondrial genome has been submitted to the European Nucleotide Archive under accession number HF548551 . An integrated genome and transcriptome browser of B . schlosseri has been developed and is available at: http://genepyramid . stanford . edu/botryllusgenome/ .	The tunicates are an evolutionary group that includes species such as sea squirts and sea tulips . Their name comes from the structure known as a ‘tunic’ that surrounds their sac-like bodies . As marine filter feeders , tunicates obtain nutrients by straining food particles from water , and they can live either alone or in colonies depending on the species . Charles Darwin suggested that tunicates may be the key to understanding the evolution of vertebrates , and indeed today they are regarded as the closest living relatives of this group . Colonial tunicates can reproduce either sexually , or asexually by budding . Compatible colonies have the ability to recognize one another and to fuse their blood vessels to form a single organism , whereas incompatible colonies reject one another and remain separate . This recognition process bears some resemblance to the rejection of foreign organ transplants in mammals . Here , Voskoboynik and co-workers present the first genome sequence of a colonial tunicate , Botryllus schlosseri . They used a novel sequencing approach that significantly increased the length of a DNA molecule that can be determined by next-generation sequencing , and allowed large DNA repeat regions to be easily resolved . In total , they sequenced 580 million base pairs of DNA , which they estimate contains roughly 27 , 000 genes . By comparing the B . schlosseri genome with those of a number of vertebrates , Voskoboynik et al . identified multiple B . schlosseri genes that also participate in the development and functioning of the vertebrate eye , heart , and auditory system , as well as others that may have contributed to the evolution of the immune system and of blood cells . The genome of B . schlosseri thus provides an important new tool for studying the genetic basis of the evolution of vertebrates .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "genetics", "and", "genomics" ]	2013	The genome sequence of the colonial chordate, Botryllus schlosseri
How mammalian tissues maintain their architecture and tissue-specificity is poorly understood . Previously , we documented both the indispensable role of the extracellular matrix ( ECM ) protein , laminin-111 ( LN1 ) , in the formation of normal breast acini , and the phenotypic reversion of cancer cells to acini-like structures in 3-dimensional ( 3D ) gels with inhibitors of oncogenic pathways . Here , we asked how laminin ( LN ) proteins integrate the signaling pathways necessary for morphogenesis . We report a surprising reciprocal circuitry comprising positive players: laminin-5 ( LN5 ) , nitric oxide ( NO ) , p53 , HOXD10 and three microRNAs ( miRNAs ) — that are involved in the formation of mammary acini in 3D . Significantly , cancer cells on either 2-dimensional ( 2D ) or 3D and non-malignant cells on 2D plastic do not produce NO and upregulate negative players: NFκB , EIF5A2 , SCA1 and MMP-9 — that disrupt the network . Introducing exogenous NO , LN5 or individual miRNAs to cancer cells reintegrates these pathways and induces phenotypic reversion in 3D . These findings uncover the essential elements of breast epithelial architecture , where the balance between positive- and negative-players leads to homeostasis . p53 is an extensively characterized regulator of gene expression in the context of malignant transformation and is aberrant in almost all cancer types . Many p53 studies have been performed in cells cultured in 2D conditions . Despite the extensive literature on p53 and its myriad of functions , little is known about what regulates p53 activity in higher organisms in vivo or about how p53 might regulate physiological tissue functions in 3D cultures ( Barcellos-Hoff et al . , 1989; Petersen et al . , 1992; Bissell et al . , 2005; Lee et al . , 2007 ) . ECM proteins , in particular LNs ( Miner and Yurchenco , 2004 ) , compose another important class of regulators that play a role in glandular tissue morphogenesis . Whether or how these two crucial regulators of gene expression intersect in tissue morphogenesis and homeostasis has not been examined . To explore the possibility of such an interaction as an element of tissue-specificity , we utilized the HMT3522 cancer progression series of human mammary epithelial cells ( MECs ) ( Briand et al . , 1987; Briand et al . , 1996; Rizki et al . , 2008 ) . This unique series comprise both primary normal epithelial cells or non-malignant cells ( S1 ) derived from reduction mammoplasty , and their malignant counterpart ( T4-2 ) , which were derived without external oncogenic agents after prolonged cultivation in defined medium that lacked epidermal growth factor ( EGF ) , followed by xenografts in animals ( Briand et al . , 1987 ) . Non-malignant and malignant MECs and organoids are readily distinguished by their colony structures in 3D LN1-rich ECM gels ( lrECM ) ( Petersen et al . , 1992 ) . Non-malignant mammary cells form polarized colonies resembling normal acini of the breast ( Barcellos-Hoff et al . , 1989 ) , whereas malignant cells form disorganized , tumor-like structures ( Petersen et al . , 1992; Lee et al . , 2007; Rizki et al . , 2008 ) . However , if the architecture of colonies is restored in LN1 gels by downmodulating receptors such as integrins and EGFR , or other involved oncogenic pathways to a level found in normal cells , every single malignant cell would form polarized growth-arrested colonies – by a process we call phenotypic reversion – through a novel movement we have termed ‘coherent angular motion’ ( CAMo ) ( Tanner et al . , 2012 ) . Here , we aimed to delineate core regulators of proper ECM-chromatin communications that establish normal breast acinar architecture , a feature that is aberrant in cancer cells in 3D . Using S1 cells , T4-2 cells and T4-2 cells reverted to ‘normal’ phenotype ( T4-2 Rev ) by five different signaling inhibitors , we identified a subset of 60 genes that had similar expression patterns in S1 and in all of the T4-2 Rev cells ( Bissell et al . , 2005; Becker-Weimann et al . , 2013 ) , as well as 10 miRNAs that could potentially target these 60 genes . Among the 10 miRNAs , we specifically focused on miR-34c-5p , −30e , and −144 , which are dramatically downmodulated in many breast tumors ( Lu et al . , 2005 ) . Restoration of the miRNA caused phenotypic reversion of T4-2 cells in lrECM . While studying the signaling cascades that involve these three miRNAs , we identified a reciprocal regulatory network – comprising LN1 and LN5 , NO , p53 , HOXD10 , NFκB , the three miRNAs , EIF5A2 , SCA1 , and MMP-9 – which connects the ECM-laminins and the nuclear transcription factors ( TFs ) , most possibly via a newly discovered nuclear tunnel ( Jorgens et al . , 2017 ) , to execute breast morphogenetic programs . Our results shed light on a completely novel and intricate reciprocal loop for breast acinar morphogenesis through a reiterative activation and suppression of regulatory molecules necessary to maintain the differentiated state in 3D and to prevent malignant conversion . Non-malignant S1 cells form apico-basally polarized acini in lrECM while conversely , malignant T4-2 cells form disorganized colonies ( Petersen et al . , 1992 ) . We showed initially that inhibitory antibodies to beta-1 integrin reverted the malignant cells to ‘normal’ phenotype ( Figure 1a ) ( Weaver et al . , 1997 ) . Inhibiting any of a dozen different oncogenic pathway components , including EGFR , PI3K and MMP-9 , could revert breast cancer cells ( Figure 1a–1c ) ( Bissell et al . , 2005; Beliveau et al . , 2010; Becker-Weimann et al . , 2013 ) . Such cross-modulation suggested the existence of central common integrators . Array analyses of the five most prominent reverting pathways identified 60 genes that were low in S1 , and co-downregulated in T4-2 Rev cells ( Figure 1d , Table 1 ) ( Bissell et al . , 2005 ) , leading us to suspect that the common regulators would be miRNAs . miRNA expression profiling of the S1 , T4-2 , and T4-2 Rev cells in lrECM identified a list of 30 miRNAs , the expression of which was anti-correlated with that of the 60 genes ( Figure 1d , Table 2 ) . Using a miRNA target database ( microRNA . org ) , we predicted miRNAs that could potentially target the 60 genes . By combining these two lists , we chose 10 validated miRNAs ( Figure 2a ) each of which could potentially target at least 10 out of the 60 genes ( Table 3 ) . Using published patient sample analyses , we selected three miRNAs: miR-34c-5p , −30e and −144 , that were found to be downmodulated significantly in breast tumors and tumor cell lines ( Figure 1—figure supplement 1 ) ( GSE25464 ) ( Lu et al . , 2005 ) . By in situ hybridization of tissue arrays containing 40 breast tumors vs . normal tissues , we confirmed a significant reduction of the three miRNAs in tumors ( Figure 2b , c ) . Re-expression of each of the three miRNAs in T4-2 cells led to dramatic growth inhibition in soft agar ( Figure 2d , Figure 2—figure supplement 1b ) and caused phenotypic reversion in lrECM ( Figure 2e , Figure 2—figure supplement 1c ) . Introduction of each of the three miRNAs into metastatic MDA-MB-231 breast cancer cells also led to severe growth impairment in lrECM ( Figure 2f , Figure 2—figure supplement 1d ) . These results suggest that the three miRNAs are involved in inhibiting tumor cell growth and , by implication , in the maintenance of non-malignant cell behavior . A search of the miRNA target database ( microRNA . org ) identified EIF5A2 and SCA1 as the only common target genes of the three miRNAs among the 60 genes that were modulated by each of five reverting agents ( Table 4 , Figure 1 ) . To validate this , we performed RT-PCR for EIF5A2 and SCA1 in T4-2 cells before and after miRNA expression . Endogenous levels of the two proteins were high in T4-2 cells compared to those in S1 cells , but as expected , were downmodulated in T4-2 Rev cells that were reverted either with a reverting agent or upon restoration of any of the three miRNAs ( Figure 3a ) . Thus , each miRNA acted like a reverting agent , similar to the five other reverting agents we have reported on previously ( Figure 3a; Figure 1a–c ) ( Bissell et al . , 2005; Beliveau et al . , 2010; Becker-Weimann et al . , 2013 ) . Importantly , depletion of either EIF5A2 or SCA1 in T4-2 cells with shRNA ( Figure 3—figure supplement 1a ) also caused phenotypic reversion ( Figure 3b , Figure 3—figure supplement 1b ) . To ensure that this is not an off-target effect , we restored EIF5A2 and SCA1 in T4-2 cells that were overexpressing the miRNAs . In these T4-2 cells we overexpressed cDNAs of EIF5A2 or SCA1 that lacked miRNA binding sites because the three miRNAs bind only to the 3’UTR of the two target genes ( Table 5 ) . Overexpression was confirmed by western analysis ( Figure 3—figure supplement 1c ) . Restoration of EIF5A2 or SCA1 severely impaired tumor-cell reversion , validating the importance of the inactivation of these two target genes for normal functional differentiation of breast acini ( Figure 3—figure supplement 1d and e ) . These results demonstrate that the miRNA database correctly predicted EIF5A2 or SCA1 as the target genes of the three miRNAs . To determine the regulators of the three miRNAs , we generated reporter constructs in which the luciferase gene was fused to the miRNA gene promoters , containing 3–0 , 2–0 and 1–0 kb regions from the transcription start site ( Figure 3—figure supplement 2a ) . The activity of the 1–0 kb region for miR-34c and the 3–0 kb region for both miR-30e and −144 was high in S1 and T4-2 Rev cells , but not in T4-2 cells ( Figure 3c ) . In addition , we generated reporter constructs containing non-overlapping 3–2 , 2–1 and 1–0 kb fragments of the miRNA promoters from the transcription start site ( Figure 3—figure supplement 2b ) . The activity of the 1–0 kb region for miR-34c and the 3–2 kb region for both miR-30e and −144 was high in S1 and T4-2 Rev cells ( Figure 3d ) . To determine which TFs bound to these critical regions , we analyzed the PROMO database ( Farré et al . , 2003 ) and identified multiple high-confidence binding sites for HOXD10 and NFκB ( % dissimilarity <15%; genomic frequency <1×10−4 ) ( Figure 4—figure supplement 1a , Table 6 ) ( Farré et al . , 2003 ) . We had shown previously that overexpression of HOXD10 or downmodulation of NFκB phenotypically reverts T4-2 cells ( Becker-Weimann et al . , 2013; Chen et al . , 2009 ) . As predicted , HOXD10 was high in S1 and T4-2 Rev cells compared to T4-2 cells ( Figure 4a ) . By contrast , activation of NFκB , as measured by Ser536 phosphorylation of the p65 subunit that causes its nuclear translocation ( Sasaki et al . , 2005 ) , was elevated in T4-2 cells and downmodulated in S1 and T4-2 Rev cells ( Figure 4a ) . To show that these two TFs regulate the miRNAs in opposite directions , we generated T4-2 cells that were depleted of either p65 or p50 , and its unprocessed precursor , p100 , a subunit of NFκB . We also overexpressed HOXD10 in T4-2 cells ( Figure 4—figure supplement 1b ) . In all these conditions , the activity of the miRNA promoters was elevated in the same regions as those described above ( Figure 4b , Figure 3—figure supplement 1d ) . Northern analysis confirmed the increase of miRNA expression , allowing the formation of basally polarized colonies in lrECM ( Figure 4c and d , Figure 4—figure supplement 1c ) , which were analogous to colonies of miRNAs-expressing T4-2 cells ( Figure 2c ) . These results highlight the importance of the ratios and balance of different regulatory genes in maintaining normal architecture . To prove that HOXD10 and NFκB do indeed bind the promoters of the three miRNAs , we performed chromatin immunoprecipitation ( ChIP ) analyses . We found that HOXD10 bound the promoters of the three miRNAs in S1 and T4-2 Rev cells , but not in T4-2 cells , whereas the NFκB p65 subunit bound the same regions in T4-2 cells , but not in S1 and T4-2 Rev cells ( Figure 4e , Figure 4—figure supplement 1d ) . To ascertain the functional consequence of the above experiment , we used the decoy technology described by Osako et al . ( Osako et al . , 2012 ) . These decoys were derived from their respective binding sequences in each miRNA promoter ( Table 7 ) . For T4-2 cells , which have a high level of endogenous NFκB ( Figure 4a ) , we expressed NFκB decoys; for T4-2 cells that we overexpressed HOXD10 ( Figure 4—figure supplement 1b ) , we employed HOXD10 decoys . Any alteration in the promoter activity after expressing a particular decoy would indicate that the TF was bound and sequestered by the decoy . To test for sequence-specific binding of the TFs , we engineered decoys harboring point mutations in T4-2 cells . The expression of wild-type NFκB decoys , but not mutant decoys , derepressed the promoter activities , showing that the wild-type decoys bound and sequestered NFκB , whereas the mutant decoys did not . The procedure was repeated for HOXD10 with similar conclusions ( Figure 4—figure supplement 1e ) . Collectively , these results demonstrate that HOXD10 and NFκB directly bind the specific sequences in miRNA promoters in a mutually exclusive manner to regulate miRNA expression for restoration of breast acinar architecture . p53 is a potent inhibitor of NFκB ( Webster and Perkins , 1999; Murphy et al . , 2011 ) . Because p53 activity in tumors is extremely high , it is often assumed that little or no p53 is present in normal tissues . We found appreciable levels of wild-type p53 in the epithelial compartment of sections of normal breast tissues but not in the stroma ( Figure 9—figure supplement 1 ) . In 3D cultures of S1 and T4-2 Rev cells , we found appreciable levels of Ser20-phosphorylated p53 ( pSer20-p53 ) , which stabilizes ( Chehab et al . , 1999 ) and enhances the transactivation activity of p53 ( Jabbur et al . , 2000 ) . This was also the case when either of the miRNAs were overexpressed in T4-2 cells or when their inhibitory target , EIF5A2 or SCA1 , was depleted ( Figure 5a ) . The expression of the p53-regulated genes , p21 , GADD45 and DRAM , was elevated in S1 and all T4-2 Rev cells ( Figure 5a ) . Whether p53 is both the direct inhibitor of NFκB and an activator of HOXD10 was examined by overexpressing the dominant-negative p53 ( DNp53 ) ( Harvey et al . , 1995 ) in S1 cells . This particular mutant of p53 was reported to effectively abolish tumor suppression and transcriptional activity of the endogenous wild-type p53 , leading to enhanced tumor growth , even in heterozygous mice . In S1 cells that overexpressed DNp53 , HOXD10 level plummeted as NFκB activity , measured by Ser536 phosphorylation of the p65 subunit , increased over the levels seen in control S1 cells or S1 cells overexpressing the wild-type p53 ( Figure 5b ) . As expected , expression of DNp53 prevented S1 cells from forming polarized quiescent acini in lrECM ( Figure 5c , Figure 5—figure supplement 1a ) . Similarly , RNAi-mediated depletion of the wild-type p53 in S1 or MCF10A cells abrogated acinar formation ( Figure 5d and e , Figure 5—figure supplement 1b and c ) . Furthermore , inhibition of p53 activity with a specific inhibitor , α-pifithrin ( Komarov et al . , 1999 ) , rendered T4-2 cells resistant to phenotypic reversion by any of the reverting agents tested ( Bissell et al . , 2005; Lee et al . , 2007 , 2012 ) or by re-expression of any of the three miRNAs ( Figure 5f and g , Figure 5—figure supplement 1d and e ) . Likewise , MCF10A cells that overexpressed DNp53 , were resistant to reverting agents ( Figure 5h , Figure 5—figure supplement 1f ) . It is known that the basement membrane ( BM ) of the mammary gland includes not only LN1 but also LN5 . To maintain tissue architecture , signaling pathways need to regulate each other directly or indirectly ( Bissell et al . , 1982 , 2005 ) . We had shown previously that even after placing cells in lrECM , formation of acini and production of milk proteins still required an endogenously formed BM ( Streuli and Bissell , 1990 ) . Accordingly , we measured the levels of human LNs in the conditioned media ( CM ) and in cell lysates of S1 and T4-2 cells grown in lrECM . Using a human-specific pan-LN antibody , we observed a significant increase in human LNs in both CM and lysates of S1 and T4-2 Rev cells reverted by expression of miRNAs or depletion of the two target genes ( Figure 6—figure supplement 1a ) . Functional LN proteins are heterotrimers of αβγ chains ( Miner and Yurchenco , 2004 ) . To determine which LN trimers were upregulated , we analyzed the CM of cells grown in lrECM cultures using antibody arrays against human ECM proteins . The α3 , β3 and γ2 chains of LN5 were highly elevated in S1 and T4-2 Rev cells that expressed the miRNAs or that were depleted of their two targets . By contrast , parental T4-2 cells did not produce LN5 , suggesting that LN5 is only expressed in MECs capable of forming acinar-like polarized structures ( Figure 6—figure supplement 1b–d ) . To test the possibility , we depleted one of the LN5 subunits , LAMA3 , with shRNA . Loss of LAMA3 abrogated reversion of T4-2 cells with any of the different reverting agents including any of the three miRNAs ( Figure 6a and b , Figure 6—figure supplement 2a , b ) . Depletion of LAMA3 could be rescued by addition of ectopic LN5 , confirming the specificity of the reaction ( Figure 6a and b , Figure 6—figure supplement 2a and b ) . To follow how LN5 was elevated in acinar formation and tumor cell reversion , we postulated that it could be due to LN5 protein stabilization due to suppression of MMP-9 transcription . We previously had shown that MMP-9 , a metalloproteinase secreted to degrade LNs , is elevated in T4-2 , but downmodulated in T4-2 Rev cells , leading to stabilization of secreted LNs ( Beliveau et al . , 2010 ) . We measured the level of secreted MMP-9 in lrECM cultures and showed that MMP-9 was significantly reduced in T4-2 cells that expressed any of the three miRNAs- or were depleted of the two target genes , EIF5A2 and SCA1 ( Figure 6c ) . It had been shown previously that both EIF5A2 and SCA1 lie downstream of the PI3K/AKT pathway and are involved in positive regulation of MMP transcription ( Liu et al . , 2000; Park et al . , 2013; Khosravi et al . , 2014 ) . We concluded that expression of the miRNAs inactivates both EIF5A2 and SCA1 and thus downmodulates MMP-9 leading to stabilization of LN5 . We searched for possible explanations of how LNs activate p53 . In older literature , LN was reported to induce NO production in neuronal and endothelial cells as part of mechanotransduction pathways ( Gloe and Pohl , 2002; Rialas et al . , 2000 ) . As NO is reported to be a potent activator of p53 ( Forrester et al . , 1996; Wang et al . , 2002 ) , we hypothesized that LNs might also be instrumental in inducing NO production in breast cells , which in turn would activate p53 . We applied purified LN5 ( Figure 7a ) or lrECM ( Figure 7—figure supplement 1a ) to MCF10A cells and observed an increase in pSer20-p53 after 30 min , along with increases in Ser1981-phosphorylated ATM and total level of p14 ARF , the known p53 activators ( Canman et al . , 1998; Zhang et al . , 1998 ) . Under the same conditions , Ser1417 phosphorylation of nitric oxide synthase 1 ( NOS-1 ) was also elevated ( Figure 7a , Figure 7—figure supplement 1a ) , suggesting its role in NO production . In contrast , DNp53 overexpressed in MCF10A cells was not activated in response to LNs , whereas ATM , p14 ARF and NOS-1 were all activated ( Figure 7a , Figure 7—figure supplement 1a ) . When MCF10A cells were treated with a NOS inhibitor , L-NAME , that inhibits NO production , LN5-mediated activation of p53 , as well as of ATM and p14 ARF , were severely impaired ( Figure 7b ) . We measured the level of NO in CM after addition of LNs using a fluorescence probe , DAN , against NO metabolites . S1 and MCF10A cells produced NO as a function of time in response to LN5 or lrECM ( Figure 7c and d ) . By contrast , T4-2 cells failed to do so ( Figure 7c and d ) . Addition of another ECM protein collagen-1 ( COL1 ) did not induce NO production by S1 or MCF10A cells ( Figure 7e ) , suggesting a unique role of LNs . We then monitored the intracellular NO level after addition of lrECM using a fluorescence probe DAF-FM DA . NO level peaked at around 1 hr after lrECM addition and declined thereafter in S1 and MCF10A cells , whereas it remained low in T4-2 cells ( Figure 7f , Figure 7—figure supplement 1b ) . To confirm the biological relevance of NO production by MECs , we stained 3D colonies for S-nitrosocysteine ( SNOC ) , an indicator of NO production ( Gould et al . , 2013 ) and localization ( Iwakiri et al . , 2006 ) . S1 acini showed strong basolateral SNOC staining , whereas T4-2 cells showed weak and dispersed staining . However , T4-2 Rev cells restored the strong basolateral SNOC staining analogous to S1 , suggesting the recovery of NO production upon phenotypic reversion ( Figure 7g , Figure 7—figure supplement 1c ) . We then stained normal ( n = 8 ) vs . cancerous ( n = 32 ) breast tissue sections for SNOC . Normal mammary epithelia were distinctively stained for SNOC , whereas the majority of tumor samples were only weakly and diffusely stained [positive staining ( intensity >+1 ) : 8/8 vs . 8/32 , respectively] ( Figure 7h ) . These results support the relevance of NO production to the biology of the normal breast . NO is known to play a role in the differentiation and morphogenesis of neurons , muscles and immune cells ( Rialas et al . , 2000; Stamler and Meissner , 2001; Niedbala et al . , 2002 ) . To test the involvement of NO in mammary morphogenesis , we inhibited NO production with L-NAME in two different non-malignant breast epithelial cells; this led to the formation of disorganized proliferative structures in lrECM ( Figure 8a and b , Figure 8—figure supplement 1a and b ) . Alternatively , the induction of NO production in T4-2 cells with a NO donor , SNAP , induced phenotypic reversion ( Figure 8c , Figure 8—figure supplement 1c ) . Also , application of L-NAME to T4-2 cells , even in the presence of a reverting agent ( e . g . , an inhibitor of EGFR or β1 integrin ) ( Bissell et al . , 2003 ) , abrogated phenotypic reversion in lrECM ( data not shown ) . To determine whether the activity of NO is necessary for human mammary gland morphogenesis , we monitored the alveologenesis of breast organoids treated with L-NAME in ex vivo 3D cultures . L-NAME treatment dramatically reduced the percentage of colonies capable of alveologenesis ( vehicle-treated: 28% vs . L-NAME-treated: 1 . 2% ) ( Figure 8d , Videos 1 and 2 ) . We tracked movement of L-NAME-treated S1 cells in lrECM for 48 hr by live cell imaging . We and others have shown previously that acinar forming non-malignant breast cells undergo CAMo in lrECM , whereas cancer cells exhibit random amoeboid motion ( Tanner et al . , 2012; Wang et al . , 2013 ) . S1 cells treated with L-NAME are defective in CAMo and form disorganized masses ( Figure 8e , Videos 3 and 4 ) . We showed above that NO production in response to lrECM is critical for p53 activation and the formation of mammary acini ( Figures 7a–g and 8a–e ) . This process involves de novo synthesized LN5 ( Figure 6a and b ) . We also showed that p53 upregulates the expression of HOXD10 and downregulates activation of NFκB . This dual action allows expression of the three miRNAs that inhibit TFs , SCAI and EIF5A2 , to downmodulate MMP-9 expression . The result is inhibition of laminin protein degradation , leading to the closure of the morphogenetic loop ( Figure 9a ) . To demonstrate reciprocity in 3D , we selected the interaction between p53 and LN5 , where a single manipulation at any part of the cycle allowed integration of all the pathways examined , PROMO analysis of the promoter of LAMA3 chain of LN5 revealed over 20 high-confidence p53 binding sites within 1 kb length of the CpG island around the transcription start site ( % dissimilarity <8%; genomic frequency <1×10−3 ) ( Figure 9b , Table 8 ) ( Farré et al . , 2003 ) . Consistently , LAMA3 expression in S1 cells , could be abrogated by p53 inhibition with α-pifithrin ( Figure 9c , Figure 9—figure supplement 1a ) , whereas ectopic addition of LN5 or lrECM , elevated LAMA3 transcription ( Figure 9d , Figure 9—figure supplement 1b ) in parallel to the activation of wild-type p53 ( Figure 7a , Figure 7—figure supplement 1a ) . To see whether there is a correlation between the wild-type p53 and LAMA3 levels in vivo , we performed immunohistochemical analyses of primary breast tissues using antibodies against the wildtype p53 ( Clone pAb1620 ) and LAMA3 ( Clone 546215 ) . All normal breast tissue sections were stained strongly with both antibodies ( Figure 9—figure supplement 1c ) . The reciprocity between LN5 and wild-type p53 remains strong even as cells progress to malignancy . The levels of the two proteins fell in parallel in the tumor samples ( R = 0 . 51 , p<0 . 0001 , n = 117 ) ( Figure 9e ) . The essential and prominent steps of the acinar circuitry are shown in the schematic presented in Figure 10 . The ability to phenotypically revert breast cancer cells by inhibiting a single signaling pathway in 3D lrECM has provided us with the means to identify additional major signaling pathways that must integrate for the formation of ‘phenotypically normal’ human breast acini ( Weaver et al . , 1997; Muschler et al . , 2002; Beliveau et al . , 2010; Bissell and Hines , 2011; Lee et al . , 2012; Tanner et al . , 2012; Becker-Weimann et al . , 2013 ) . Here , we set out to develop a blueprint for how the breast cells interpret their interactions with the ECM proteins LN1 and LN5 . The LNs trigger the signaling cascade leading to reciprocal communications between the ECM and TFs essential for mammary morphogenesis . To do this , we used a unique breast cancer progression series , HMT3522: non-malignant S1 , malignant T4-2 and T4-2 reverted to non-malignant phenotype ( using five signaling inhibitors of oncogenic pathways , where addition of single inhibitors could revert the malignant phenotype ) . We observed that , although T4-2 Rev cells have similar phenotypes , their gene expression patterns were very different ( Becker-Weimann et al . , 2013 ) . Nevertheless , a comparison of the gene arrays of the five T4-2 revertants identified a group of 60 similar genes that are also expressed in S1 cells ( Figure 1d ) ( Becker-Weimann et al . , 2013 ) . This led us to propose that the common denominator of reversion had to contain a number of miRNAs that regulate this gene subset . We thus devised miRNA expression arrays and identified 10 miRNAs that fit the above category ( Figure 2a ) . This result , together with the literature search ( Lu et al . , 2005 ) and our analysis of miRNA expression in normal- vs . cancerous- breast tissues ( Figure 2b ) , identified three miRNAs ( miR-34c-5p , −30e , and −144 ) that were shown to be severely downmodulated in primary breast tumors ( Figure 2b and c , Figure 2—figure supplement 1 ) ( Lu et al . , 2005 ) . As expected , restoration of any of these three miRNAs in T4-2 cells led to phenotypic reversion in lrECM ( Figure 2e ) . We utilized these miRNAs as the focal starting point to dissect the fundamental reciprocal pathways necessary for the formation and maintenance of breast tissue architecture . It has been long known that diverse biological activities in development are regulated by tissue–tissue and tissue–microenvironment interactions and signaling ( Wessells , 1977; Chiquet-Ehrismann et al . , 1986; Howlett and Bissell , 1993; Hogan , 1999; Bhat and Bissell , 2014 ) . During development , different cell types communicate and coordinate with each other through negative and positive feedback regulations . Within a given tissue , there are also negative and positive operators that must be regulated constantly to maintain homeostasis and quiescence as we demonstrated here . In addition , similar to movements that are being discovered in the formation of embryos during development ( Haigo and Bilder , 2011 ) , tissue formation starts with cells moving within a soft microenvironment such as lrECM , as we and others observed for mammary acini; we termed this ‘coherent angular motion’ ( Tanner et al . , 2012 ) . CAMo creates polarity and adhesion by interacting with exogenous ECM to lay down its own endogenous tissue-specific ECM ( Tanner et al . , 2012 ) . The balance and integration of the different signaling pathways and dynamic interactions between epithelial cells and the ECM drive the remodeling of the ECM , including formation of the BM that helps to anchor the epithelia ( Weaver et al . , 2002 ) and that protects the cells within the tissues from apoptosis . Such changes in the ECM regulate cell proliferation , survival , migration , shape and adhesion , ultimately sculpting and maintaining tissue architecture ( Wessells , 1977; Chiquet-Ehrismann et al . , 1986; Howlett and Bissell , 1993; Hogan , 1999; Bhat and Bissell , 2014; Haigo and Bilder , 2011; Weaver et al . , 2002; Daley and Yamada , 2013 ) . There are important differences , however , between developmental processes and tissue maintenance and renewal ( Howlett and Bissell , 1993; Hogan , 1999; Bhat and Bissell , 2014; Daley and Yamada , 2013 ) . Unlike the signaling pathways in development , the stability of the differentiated state does not appear to be hierarchical . Instead , it reflects the balance between growth and differentiation , between the negative and positive signaling pathways , and between the formation of a BM and the destruction of ECM by degrading enzymes that determines the stability of the differentiated state in the tissues . Another novel finding here is that NO is a pivotal player in reciprocal cell–ECM interactions in breast morphogenesis , but tumor cells produce only a small amount of NO unless the architecture is re-established and the cells have reverted to a ‘dormant state’ ( Figure 7c–e and g ) . This is a mimicry of differentiation-dependent tissue architecture . These findings demonstrate that NO production is a mechanistic link between proper architecture and proper function in breast tissues . Please see also the accompanying paper of Ricca et al . , 2018 , which describes how the reversion of T4-2 cells induced by a short period of compression in laminin is also mediated by NO production . There are a few papers in the literature on connections between LN1 and NO in other tissues ( Rialas et al . , 2000; Gloe et al . , 1999 ) , and there are other reports of activation of p53 by high levels of exogenous NO ( Forrester et al . , 1996; Gordon et al . , 2001; Wang et al . , 2003 ) . To our knowledge , however , there are no reports of endogenous NO as a critical link in the formation of mammary epithelium and its role in stability of the tissue architecture . It is crucial to note that the levels of NO produced endogenously in response to LNs in our studies , as well as the exogenous NO levels required for the reciprocal loop we describe here , are at least 500-fold lower than those used in the literature ( Forrester et al . , 1996; Gloe et al . , 1999; Gordon et al . , 2001 ) . As stated long ago , differences in quantity of such magnitudes becomes a change in quality and hence have appreciable consequence ( Bissell , 1981 ) . NO has been reported to play an important role during lactation . Increased levels of NO are produced by the mammary gland of postpartum mammals ( Akçay et al . , 2002 ) . NO promotes blood flow and the nutrient uptake of mammary glands for milk production ( Kim and Wu , 2009 ) . NO is also proposed to facilitate milk ejection by inducing contraction of myoepithelial cells ( MEPs ) in mammary glands as well as smooth muscle cells in the stroma ( Iizuka et al . , 1998; Adriance et al . , 2005; Tezer et al . , 2012 ) . In addition , NO is secreted into the breast milk as an essential component for immunity in neonatal growth ( Hord et al . , 2011 ) . Using 3D cultures and ex vivo cultures of human mammary glands , we showed here that NO also plays additional and significant roles in breast morphogenesis ( Figure 8a , b and d ) . Importantly , NO production was specific to LNs and was not induced by collagen ( Figure 7c-e ) . We and others had shown previously that LNs and COL1 elicit opposite actions on epithelia ( Gudjonsson et al . , 2002; Oktay et al . , 2000; Chamoux et al . , 2002 ) . We showed here that LNs activate NOS-1 ( Figure 7a , Figure 7—figure supplement 1a ) , supporting previous observations by others that NOS-1 is expressed in the mammary tissue at appreciable levels — in particular in MEPs during pregnancy and lactation in humans ( Tezer et al . , 2012 ) and rodents ( Iizuka et al . , 1998; Islam et al . , 2009; Wockel et al . , 2005 ) . As the molecule that appears to be responsible for linking LNs to NOS-1 , we speculate the involvement of the LN receptor , dystroglycan ( DG ) , which is known to form a multi-protein complex involving LNs and NOS-1 , in mediating the mechanotransduction of muscle cells ( Rando , 2001; Garbincius and Michele , 2015 ) . We had shown previously that DG also plays a critical signaling role in breast epithelial cells ( Muschler et al . , 2002 ) . DG anchors the BM protein , in particular LNs , to the cell surface , allowing for LN polymerization and transduction of signals for the formation of polarized colonies ( Weir et al . , 2006 ) . Such DG–LN interaction is impaired in different types of cancer cells and correlates with poorer patient prognosis ( Akhavan et al . , 2012; Esser et al . , 2013 ) . Form and function are maintained in adult organs throughout most of the life of the organism , despite constant mutations and damage from environmental assaults and aging . To maintain the correct tissue function throughout the lifetime of the organisms , signaling pathways have to integrate in order to prevent chaos and malfunction . Evolution has packed much wisdom and specificity onto the ECM , which appears to instruct the chromatin to change shape and thus also gene expression , as seen in Figure 4e . When cells on flat surfaces receive LNs , not only their shape , but also many of their signaling pathways are altered ( Figure 1 ) ( Bissell et al . , 2005 ) ; growth must stop in many tissues ( Spencer et al . , 2011; Fiore et al . , 2017 ) and differentiation and cell death must be coordinated . It is now clear that narratives that are based solely on linear and irreversible regulatory dynamics cannot satisfactorily explain the reality in vivo ( Hogan , 1999 ) . It is also clear that , at the last analysis , it is the 3D architecture of the tissue itself that is the message ( Hagios et al . , 1998 ) . Cell lines of the HMT3522 breast cancer progression series ( S1 and T4-2 ) were provided by O . W . Petersen ( Laboratory of Tumor Endocrinology , The Fibiger Institute , Copenhagen , Denmark ) ( Briand et al . , 1996 ) . The cell lines were authenticated by genome sequencing by the provider . Mycoplasma testing was negative . MCF10A cells were obtained from the Karmanos Cancer Institute ( Detroit , MI , USA ) under a Material Transfer Agreement . The cell lines was authenticated by the provider . Mycoplasma testing was negative . The isogenic cell lines of the HMT3522 human breast cancer progression series , non-malignant S1 and malignant T4-2 cells , were maintained as described previously ( Briand et al . , 1996 ) . This cell line series was established in an attempt to recapitulate the stochastic and prolonged nature of breast cancer progression by continuously culturing S1 cells , derived from reduction mammoplasty , in the absence of serum , followed by EGF removal and injection into mice , to give rise to T4-2 cells ( Briand et al . , 1996 ) . For 3D culture experiments , S1 and T4-2 cells were seeded at the density of 2 . 5 × 104 cells/cm2 and 1 . 8 × 104 cells/cm2 , respectively , in growth factor reduced Matrigel ( Corning , NY , USA ) and maintained for 10 days with the addition of fresh medium on alternate days . For T4-2 reversion , EGFR inhibitor AG1478 ( EMD Millipore , Burlington , MA , USA ) was used at 350 nM , PI3K inhibitor LY294002 at 8 μM , and MEK inhibitor PD98059 at 20 μM ( Lee et al . , 2012 ) . For p53 inhibition , 30 μM α-PFT ( α-pifithrin , Sigama-Aldrich , St . Louis , MO , USA ) was used . For inhibition of NO production , cells were treated with 2 . 5 mM L-NAME ( Nω-Nitro-L-arginine methyl ester hydrochloride , Sigma- Aldrich ) ; for induction of NO production , 10 μM SNAP ( S-Nitroso-N-acetyl-DL-penicillamine , Sigma- Aldrich ) was used . miRNA expression profiling was performed using the RT2miRNA PCR Array System ( Qiagen , Inc . USA , Germantown , MD , USA ) on the MyiQ Single-Color Real-Time PCR platform ( Bio-Rad , Hercules , CA , USA ) . Briefly , 1 . 0 × 106 cells were grown in 1 . 2 ml Matrigel in 30 mm-plates for 10 days ( for T4-2 Rev , 350 nM AG1478 was added ) . The medium was removed and cells were scraped off from the dish with 2 ml phosphate-buffered saline ( PBS ) with 5 mM EDTA . Cells were spun down to harvest pellets , which were repetitively washed with ice-cold PBS + EDTA until the Matrigel was dissolved . The total RNA was extracted with 1 ml Trizol ( Life Technologies ) and purified with an RNeasy plus mini kit ( Qiagen , Inc , USA ) according to the manufacturers’ protocols . cDNA was generated from 4 μg of RNA using the RT2miRNA First Strand Kit ( SABiosciences ) , mixed with SYBR Green Master Mix ( SABioseicences ) and loaded onto an array with 98 wells . Real-time PCR was performed according to the manufacturer’s instructions , and data analysis was performed using the manufacturer’s PCR Array Data Analysis Web Portal ( Qiagen , Inc , USA ) . Northern analysis of miRNAs was performed using the DIG detection system from Roche . Briefly , 1 . 0 × 106 cells/30-mm plate were grown in 1 . 2 ml Matrigel in triplicates for 10 days ( for T4-2 Rev , 350 nM AG1478 was added ) . Cells were scraped off from the dish with PBS with 5 mM EDTA , spun down and washed with PBS + EDTA until the Matrigel was dissolved . The total RNA was extracted with 1 ml Trizol ( ThermoFisher Scientific , Waltham , MA , USA ) . 20 μg of RNA was separated by denaturing polyacrylamide TBE-Urea gel electrophoresis ( ThermoFisher Scientific ) and electroblotted onto Bright-Star nylon membrane ( Ambion ) with 0 . 5% TBE for 2 hr . The membrane was rinsed in 2xSSC buffer , UV cross-linked at 120 mJ/cm2 , dried and stored between filter papers . LNA-modified DNA oligonucleotides complementary to the mature miRNA sequences ( Table 9 ) were obtained from IDT and DIG-labeled using the DIG Oligonucleotide Tailing Kit ( Roche Diagnostics , USA , Indianapolis , IN , USA ) . Using DIG Easy Hyb ( Roche Diagnostics , USA ) , the membrane was prehybridized and hybridized with DIG-labeled probe at room temperature overnight . The membrane was washed in 10xSSC + 0 . 1% SDS four times and processed for DIG detection using the DIG Luminescent Detection Kit ( Roche ) according to the manufacturer’s protocol . miRNA in situ hybridization ( ISH ) was performed using the miRCURY LNA miRNA ISH Optimization kit for formalin-fixed paraffin embedded ( FFPE ) tissues ( Qiagen , Inc , USA ) and double-DIG-labeled detection probes for miR-34c-5p , miR-30e and miR-144 ( EXIQON ) on breast cancer tissue arrays containing paraffin-embedded sections of normal and malignant ( stages II and III ) tissues ( US Biomax , Inc , Rockville , MD , USA ) . Briefly , the tissue slides were heated at 60°C for 1 hr , deparaffinized in xylene and hydrated in alcohol series ( 100% to 70% ) . Slides were deproteinated with proteinase K for 20 min , fixed in 4% paraformaldehyde for 10 min and washed with 0 . 2% glycine in PBS for 5 min . Then , slides were incubated in imidazole buffer ( 0 . 13 M 1-methylimidazole , 300 mM NaCl , pH 8 . 0 ) for 10 min twice , in EDC solution ( 0 . 16M 1-ethyl-3-[3-dimethylaminopropyl] carbodiimide [EDC] , pH 8 . 0 ) for 1 . 5 hr and washed with 0 . 2% glycine in PBS . Then , slides were dehydrated in an alcohol series ( 70% to 100% ) , hybridized with heat-denatured probes at room temperature overnight , and processed for DIG detection according to the manufacturer’s protocol ( Qiagen , Inc , USA ) . The slides were counterstained with Nuclear Fast Red and mounted with permount . Photomicrographs were taken with the Zeiss Axioskop Imaging Platform and Axion Vision software ( Version 4 . 7 ) . Lentivector-based precursor constructs for miR-34c-5p , miR-30e and miR-144 co-expressing copGFP were obtained from System Biosciences Palo Alto , CA , USA , and the virus particles to express each miRNA were produced according to the manufacturer’s guideline . For construction of HOXD10-overexpressing lentivirus , the full-length human HOXD10 cDNA clone was obtained from Open Biosystems ( Lafayette , CO , USA , Clone ID: 7262455 ) . For construction of p53-overexpressing lentivirus , both wild-type and dominant-negative ( A135V ) p53 expression plasmids were obtained from Clontech . The coding region was PCR-amplified using the respective primers ( Table 9 ) . The PCR product was ligated into the AscI/EcoR1 site ( for HOXD10 ) or the BamHI/EcoR1 site ( for p53 ) of the PCDF1-MCS2-EF1-puro lentiviral vector ( System Biosciences ) . For the SCA1- and EIF5A2-overexpression lentivirus construct , the cDNA clones were obtained from Origene ( Rockville , MD , USA , Cat#RC222862 and Cat#RC206249 , respectively ) and cloned into PCDF1-MCS2-EF1-puro lentiviral vector at the BamHI/EcoR1 site using the Gibson assembly system and a DNA assembly kit ( Cat# E5520S , NEB ) with the primers designed on the NEB Builder Assembly Tool website as shown in Table 9 . For shRNA production , a double-stranded DNA oligonucleotide was generated from the respective sequences ( Table 9 ) . Sense and antisense oligonucleotides were annealed and ligated into BamH1/EcoR1 site of pGreen puro lentival vector which co-expresses copGFP ( System Biosciences ) . Lentivirus production and transduction of target cells were conducted following the guideline by System Biosciences . Briefly , lentivirus vector and packaging plasmid mix ( System Biosciences ) were transfected into 293FT cells ( ThermoFisher Scientific ) using Lipofectamine 2000 . After 48 hr , medium was harvested , filtered and used to infect target cells with the addition of polybrene ( 10 μg/ml ) . The medium was replaced after 24 hr . At 72 hr post-infection , puromycin ( 0 . 5 μg/ml ) was added for selection and maintained throughout the culturing period . One million cells were grown in 1 . 2 ml Matrigel on a 30-mm plate for 10 days ( for T4-2 Rev , 350 nM AG1478 was added ) . The medium was removed and cells were scraped off from the dish with 2 ml PBS with 5 mM EDTA . They were then spun down to harvest the cell pellet and repeatedly washed with PBS + EDTA until Matrigel was dissolved . The total RNA was extracted with 1 ml Trizol ( ThermoFisher Scientific ) . cDNA was synthesized from 2 μg RNA using the SuperScript Double-Stranded cDNA Synthesis Kit ( Invitrogen ) and served as a template for PCR amplification with the respective primers ( Table 9 ) . Immunofluorescence was performed as described previously ( Weaver et al . , 1997 ) . Samples were incubated with primary antibody for 2 hr at room temperature in a humidified chamber . After intensive washing ( three times , 15 min each ) in 0 . 1% BSA , 0 . 2% Triton-X 100 , 0 . 05% Tween 20 , 0 . 05% NaN3 in PBS , fluorescence-conjugated secondary antibodies ( Molecular Probes ) were added for 1 hr at room temperature . Nuclei were stained with 0 . 5 ng/ml DAPI . One percent agar was mixed with the equivalent volume of 2x DMEM/F12 medium supplemented with all the additives necessary for culturing T4-2 cells ( Briand et al . , 1996 ) plus 20% FBS and 2% penicillin or streptomycin . 1 ml of the agar solution was poured into a 35 mm plate in triplicate and solidified . 0 . 7% agar solution equilibrated to 40°C was mixed with 2x growth medium and breast cancer cells at 7000 cells/ml and poured onto the base agar at 1 ml/plate . The solidified agar was covered with 500 μl growth medium and maintained in a 37°C humidified incubator for 14 d . The plates were stained with 0 . 01% crystal violet for 30 min , and colonies were counted under a dissecting microscope . For generation of miRNA reporter constructs , the promoter regions of miRNA genes were obtained by PCR-amplifying BAC genomic clones [miR-34c ( Ch11 ) , PR11-794P6; miR-30e ( Ch1 ) , RP11-576N9; miR-144 ( Ch17 ) , RP11-832J20] using the respective primers ( Table 9 ) and inserted into the pGL3 luciferase expression vector . Cells seeded at 5 × 105 cells/60-mm plate were transfected with 7 μg of luciferase reporter and 0 . 5 μg of β-galactosidase plasmids using Xfect transfection reagent according to the manufacturer’s protocol ( Clontech , Mountain View , CA , USA ) . After 24 hr post transfection , the medium was replaced with the fresh medium containing 5% Matrigel and cells were maintained for another 24 hr ( for T4-2 Rev , 350 nM AG1478 was added ) . Luciferase and β-galactosidase reporter activities were measured using a reporter assay kit ( Promega , Madison , WI , USA ) . Wild-type miRNA decoy sequences ( Table 7 ) were derived from the binding sites of NFκB or HOXD10 within the miRNA promoters predicted by AlGGEN PROMO software ( see below ) . The sequence-specific binding of the two TFs was tested using mutant decoys ( Table 7 ) that had point mutations in their core binding sequences . The forward and reverse oligonucleotides of decoys at 100 μM each were annealed in Duplex buffer ( Integrated DNA Technologies , Coralville , IA , USA , CAT#11-05-01-12 ) , and the same group of decoys was pooled . T4-2 cells were plated at 0 . 5 × 105/12 wells the day before transfection . NFκB decoys ( scramble , WT or MT ) , along with miRNA promoters fused to luciferase ( see above ) , were transfected into control T4-2 cells that had a high endogenous level of NFκB . HOXD10 decoys ( scramble , WT or MT ) , along with promoter constructs , were transfected into T4-2 cells that overexpressed HOXD10 . Transfection was performed with 1 μl XFect transfection reagent ( Clontech , cat# 631318 ) , 1 . 5 μg of promoter DNA and 200 nM of decoy oligonucleotides according to the manufacturer’s protocol . Cells were harvested at 48 hr post transfection . The luciferase activity was analyzed using the Bright-Glo Luciferase assay system ( E2610 , Promega ) according to the manufacturer’s protocol , and the activity was normalized using protein concentration . TF binding sites within the promoter regions were predicted by AlGGEN PROMO software ( http://alggen . lsi . upc . es/cgi-bin/promo_v3/promo/promoinit . cgi ? dirDB=TF_8 . 3 ) ( Farré et al . , 2003 ) . The feasibility of these predicted sites was indicated as the ‘Dissimilarity’ to the canonical sequence ( 0% as the best match ) . The significance of the predicted site was indicated as the ‘Frequency’ in the genomic background ( ‘Random Expectancy’ ( RE ) value x 10–3 ) ( Farré et al . , 2003 ) . ChIP assays were performed as described by Saccani et al . ( 2001 ) with a minor modification . Cells were plated at 2 × 106/100-mm plate and maintained overnight . Then , cells were maintained in the fresh medium containing 5% Matrigel for 24 hr ( for T4-2 Rev , 350 nM AG1478 was added ) . Cells placed in fresh medium with 1% formaldehyde for 10 min , scraped off from the dish with PBS and processed for nuclear extraction . Chromatins were sonicated to ∼500 bp fragments and immunoprecipitated with control rabbit IgG , HOXD10 and p65 antibodies at 4°C overnight . Chromatin-antibody complexes were washed with buffer 1 [0 . 1% SDS , 0 . 5% Triton X-100 , 2 mM EDTA , 20 mM Tris-HCl ( pH 8 . 0 ) , 150 mM NaCl] , buffer 2 [0 . 1% SDS , 2 mM EDTA , 20 mM Tris-HCl ( pH 8 . 0 ) , 500 mM NaCl] then TE buffer [10 mMTris-HCl ( pH 8 . 0 ) 1 mM EDTA] . After reversal of cross-linking by heating at 65°C overnight , immunoprecipitated chromatin was subjected to PCR reaction for ~300 bp fragments around HOXD10/NFκB binding sites in miRNA promoters ( miR-34c: −1 ~ 0 kb , miR-30e: −3~−2 kb , miR-144: −3~−2 kb ) with the appropriate primers ( Table 9 ) . The relative abundance of the secreted laminin chains was determined with ImmunoCruz Cell Adhesion-2 MicroArray ( sc-200006 , Santa Cruz Biotechnologies , Santa Cruz , CA , USA ) according to the manufacturer’s protocol . Briefly , cells were plated at 2 × 106/100-mm plate and maintained overnight . Cells were maintained in the fresh medium containing 5% Matrigel for 24 hr . The CM was harvested and spun to remove the Matrigel drip . The medium was concentrated to 1 ml using Amicon Ultra-15 centrifugal filter units ( 3 kDa cut off , Millipore ) . The protein concentration was determined with DC Protein Assay reagent ( Bio-Rad ) and normalized to 1 mg/ml . 250 μg protein was labeled with Cy3 dye ( Cy3 Mono-Reactive Dye Pack , GE Healthcare , Milwaukee , WI , USA ) . The labeled protein was dissolved in 1 . 5 ml desalting buffer , and unbound dye was removed by using Amicon Ultra-15 centrifugal filter units that concentrated the protein to 500 μl . The labeled protein was hybridized with array slides , and slides were scanned and analyzed by the CruzScan Scanning service ( sc-200215 , Santa Cruz ) . Breast cancer tissue arrays containing 150 paraffin-embedded sections of normal and malignant tissues with pathological information ( stages I through III ) were obtained from US Biomax , Inc ( BR1503b ) . Slides were deparaffinized , hydrated , and treated with antigen unmasking solutions ( Vector Laboratories , Inc . ) . After being blocked with 0 . 3% H2O2 and nonimmune goat serum , sections were incubated at room temperature with an antibody against S-nitrosocysteine ( Abcam , Cambridge , MA , USA , clone HY8E12 ) , human LAMA3 ( R&D Systems , Minneapolis , MN , USA , , clone 546215 ) or wild-type human p53 ( EMD Millipore , clone pAb1620 ) and link antibodies , followed by peroxidase-conjugated streptavidin complex and diaminobenzidine tetrahydroxy chloride solution as the peroxidase substrate ( Vector Laboratories , Burlingame , CA , USA ) . The sections were counterstained with hematoxylin . Photomicrographs were taken with the Zeiss Axioskop Imaging platform and Axion Vision software ( Version 4 . 7 ) . MMP-9 secreted into CM was measured using the MMP-9 ELISA Kit ( ThermoFisher Scientific ) according to the manufacturer’s protocol . Assay samples were prepared in the dark . Briefly , cells were plated at 1 × 106/60-mm plate and maintained overnight . Cells were maintained in 2 ml of the fresh medium containing 5% Matrigel for 24 hr . The CM was harvested and spun to remove the Matrigel drip . The cleared CM was diluted 100-fold and analyzed for MMP-9 concentration using MMP-9 standards based on the optical density values at 450 nm . To quantify the cumulative level of NO produced , the more stable oxidation product nitrite/nitrate was measured using the Measure-IT High-Sensitivity Nitrite Assay Kit ( ThermoFisher Scientific ) according to the manufacturer’s protocol . Assay samples were prepared in the dark . Briefly , cells were plated at 1 × 106/60-mm plate and maintained overnight . Cells were maintained in 2 ml of the fresh medium containing 5% Matrigel for the designated time periods . The CM was harvested and spun to remove the Matrigel drip . 10 μl of the cleared CM was analyzed for nitrite concentration using nitrite standards at the excitation/emission maxima of 340/410 nm . To capture a snap shot of NO level in live cells after laminin addition , a dye DAF-FM DA ( 4-amino-5-methylamino-2' , 7'-difluorofluorescein diacetate , ThermoFisher Scientific ) was used according to the manufacturer’s protocol . The signal intensity/area/cell was measured with ImageJ . Breast tissues from reduction mammoplasties were obtained from the Cooperative Human Tissue Network ( CHTN ) , a program funded by the National Cancer Institute . All specimens were collected with patient consent and were reported negative for proliferative breast disease by board-certified pathologists . Use of these anonymous samples was granted exemption status by the University of California at Berkeley Institutional Review Board according to the Code of Federal Regulations 45 CFR 46 . 101 . Upon receipt , the tissues were rinsed with PBS , minced and incubated overnight with 0 . 1% collagenase as previously described ( with gentle agitation ) ( Hines et al . , 2015 ) . The resulting divested tissue fragments ( organoids ) were rinsed with PBS and collected by centrifugation ( 100 g × 2 min ) . Lactiferous ducts and terminal ductal lobular units ( TDLU ) were individually isolated using a micromanipulator and drawn glass needles using a screw-actuated micrometer driven hamilton syringe for suction/injection pressure . Single organoids were subsequently embedded in 50% growth factor reduced Matrigel ( BD Biosciences ) and overlayed with M87 growth medium . At 2 hr post seeding , medium was refreshed with L-NAME ( NO inhibitor ) containing medium at 5 mM . Cells were incubated at 37°C/5% CO2 . Medium was refreshed every other day for the length of the experiment ( 14 d ) . Three-dimensional live cell imaging was performed using a Zeiss LSM 710 Meta confocal microscope and Zen Version 8 . 1 software . Cells were mixed with lrECM , seeded and covered along with complete growth media in a Lab-Tek 4-well chambered coverglass 2 hr prior to image capturing . Samples were placed in a 37 oC humidified microscope stage incubator with 5% CO2 . Images of 512 × 512 pixels in XY coordinates with a maximum Z-axis displacement of 75 μm were acquired using a 0 . 8 NA 20 × air objective at one frame/second . Images were captured successively at 20 min intervals for 48 hr . Samples were simultaneous excited by the 488 nm light ( argon ion laser ) at a power of <3% maximum and 546 nm light ( a solid-state laser ) at a power of <10% maximum . A secondary dichroic mirror was used in the emission pathway to separate the red ( band-pass filters 560–575 nm ) and green ( band-pass filters 505–525 nm ) channels . Gain was set between 100 and 180 . Processed data were imported into Imaris ( Bitplane , South Windsor , CT , USA ) , and nuclei were modeled ( detection diameter: 5 , 800~6 , 500 nm ) . The nuclei were tracked over time using the tracking function of Imaris with the maximum distance of 2 , 500–20 , 000 nm and the maximum gap size of 1 . Unless otherwise indicated , statistical analyses were performed using Graph Pad Prism Version 5 software and an unpaired two-tailed Student’s t-test for parametric tests and Spearman correlation analysis for non-parametric tests . P-values of 0 . 05 or less were considered significant . Average results of multiple experiments ( n > 3 ) are presented as the arithmetic mean ± SEM .	Most animal cells can secrete molecules into their surroundings to form a supportive meshwork of large proteins , called the extracellular matrix . This matrix is connected to the cell membrane through receptors that can transmit signals to the cell nucleus to change the levels of small RNA molecules called microRNAs . These , in turn , can switch genes on and off in the nucleus . In the laboratory , cells that build breast tissue and glands can be grown in gels containing extracellular matrix proteins called laminins . Under these conditions , ‘normal’ cells form organized clusters that resemble breast glands . However , if the communication between healthy cells and the extracellular matrix is interrupted , the cells can become disorganized and start to form clumps that resemble tumors , and if injected into mice , can form tumors . Conversely , if the interaction between the extracellular matrix and the cells is restored , each single cancer cell can – despite mutations – be turned into a healthy-looking cell . These cells form a normal-looking tissue through a process called reversion . Until now , it was not known which signals help normal breast tissue to form , and how cancerous cells revert into a ‘normal’ shape . To investigate this , Furuta et al . used a unique series of breast cells from a woman who underwent breast reduction . The cells taken from the discarded tissue had been previously grown by a different group of researchers in a specific way to ensure that both normal and eventual cancer cells were from the same individual . Furuta et al . then put these cells in the type of laminin found in extracellular matrix . The other set of cells used consisted of the same cancerous cells that had been reverted to normal-looking cells . Analysis of the three cell sets identified 60 genes that were turned down in reverted cancer cells to a level found in healthy cells , as well as 10 microRNAs that potentially target these 60 genes . A database search suggested that three of these microRNAs , which are absent in cancer cells , are necessary for healthy breast cells to form organized structures . Using this as a starting point , Furuta et al . discovered a signaling loop that was previously unknown and that organizes breast cells into healthy looking tissue . This showed that laminins help to produce nitric oxide , an important signaling molecule that activates several specific proteins inside the breast cells and restores the levels of the three microRNAs . These , in turn , switch off two genes that are responsible for activating an enzyme that can chop the laminins . Since the two genes are deactivated in the reverted cancer cells , the laminins remain intact and the cells can form organized structures . These findings suggest that if any of the components of the loop were missing , the cells would start to form cancerous clumps again . Reverting the cancer cells in the presence of laminins , however , could help cancer cells to form ‘normal’ structures again . These findings shed new light on how the extracellular matrix communicates with proteins in the nucleus to influence how single cells form breast tissues . It also shows that laminins are crucial for generating signals that regulate both form and function of specific tissues . A better understanding of how healthy and cancerous tissues form and re-form may in the future help to develop new cancer treatments .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "cancer", "biology" ]	2018	Laminin signals initiate the reciprocal loop that informs breast-specific gene expression and homeostasis by activating NO, p53 and microRNAs
Determining the interconverting conformations of dynamic proteins in atomic detail is a major challenge for structural biology . Conformational heterogeneity in the active site of the dynamic enzyme cyclophilin A ( CypA ) has been previously linked to its catalytic function , but the extent to which the different conformations of these residues are correlated is unclear . Here we compare the conformational ensembles of CypA by multitemperature synchrotron crystallography and fixed-target X-ray free-electron laser ( XFEL ) crystallography . The diffraction-before-destruction nature of XFEL experiments provides a radiation-damage-free view of the functionally important alternative conformations of CypA , confirming earlier synchrotron-based results . We monitored the temperature dependences of these alternative conformations with eight synchrotron datasets spanning 100-310 K . Multiconformer models show that many alternative conformations in CypA are populated only at 240 K and above , yet others remain populated or become populated at 180 K and below . These results point to a complex evolution of conformational heterogeneity between 180-–240 K that involves both thermal deactivation and solvent-driven arrest of protein motions in the crystal . The lack of a single shared conformational response to temperature within the dynamic active-site network provides evidence for a conformation shuffling model , in which exchange between rotamer states of a large aromatic ring in the middle of the network shifts the conformational ensemble for the other residues in the network . Together , our multitemperature analyses and XFEL data motivate a new generation of temperature- and time-resolved experiments to structurally characterize the dynamic underpinnings of protein function . Current structural biology methods provide only incomplete pictures of how proteins interconvert between distinct conformations ( Motlagh et al . , 2014; van den Bedem and Fraser , 2015 ) . Although X-ray crystallography reveals atomic coordinates with relatively high accuracy and precision , the resulting electron density maps contain contributions from multiple alternative conformations reflecting the ensemble average of 106–1015 copies of the protein in one crystal ( Rejto and Freer , 1996; Smith et al . , 1986; Woldeyes et al . , 2014 ) . At high resolution , it is often possible to detect and discretely model these alternative conformations ( Burnley et al . , 2012; Davis et al . , 2006; Lang et al . , 2010; van den Bedem et al . , 2009 ) . Structural characterization of alternative conformations by X-ray crystallography can complement NMR ( Baldwin and Kay , 2009; Fenwick et al . , 2014 ) and computational simulations ( Dror et al . , 2012; Ollikainen et al . , 2013 ) in defining the structural basis of protein dynamics and ultimately in linking dynamics to function ( Henzler-Wildman and Kern , 2007 ) . However , more than 95% of crystal structures are determined at cryogenic temperatures ( ∼100 K ) to reduce radiation damage by minimizing diffusion of reactive intermediates and chemical-damage-induced structural relaxations ( Garman , 2010; Holton , 2009; Warkentin et al . , 2013 ) . Unfortunately , cryocooling can modify main chain and side chain conformational distributions throughout the protein , including at active sites and distal regions important for allosteric function ( Fraser et al . , 2011; Halle , 2004; Keedy , et al . , 2014 ) . Recent studies have instead used room temperature data collection to reveal a multitude of previously ‘hidden’ alternative conformations that are not evident at cryogenic temperatures , many of which have important ramifications for determining molecular mechanisms ( Deis et al . , 2014; Fraser et al . , 2009; Fukuda and Inoue , 2015; van den Bedem et al . , 2013 ) . Between these temperature extremes , protein conformational heterogeneity changes in complex ways . Previous studies using a wide variety of biophysical probes including NMR , X-ray crystallography , and neutron scattering have revealed a change in the character of conformational heterogeneity and/or protein dynamics around 180–220 K ( Doster , 2010; Frauenfelder et al . , 2009; Lewandowski et al . , 2015; Ringe and Petsko , 2003 ) however , the molecular origins of this ‘glass’ or ‘dynamical’ transition remain incompletely understood . Classic work has examined the temperature dependence of protein conformational heterogeneity across individual X-ray structures determined at temperatures from ∼80 to 320 K ( Frauenfelder et al . , 1979 , 1987; Tilton et al . , 1992 ) . These studies used atomic B-factors as a proxy for conformational heterogeneity and identified a global inflection point around 180–220 K . This inflection point was interpreted in terms of a transition driven by dynamical arrest of the coupled hydration layer-protein system ( Doster et al . , 1989; Frauenfelder et al . , 1979 , 1987; Tilton et al . , 1992 ) . By contrast , solution NMR studies of picosecond–nanosecond ( ps–ns ) timescale methyl side chain order parameters showed heterogeneous changes in motional amplitudes at temperatures between 288 and 346 K . Thermal deactivation of these motions was suggested to predict a transition near 200 K without invoking solvent arrest ( Lee and Wand , 2001 ) . Recent solid-state NMR ( ssNMR ) experiments suggest that protein motions are coupled to solvent , and that three transitions at ∼195 , 220 , and 250 K mark the onset of distinct classes of motions as temperature increases ( Lewandowski et al . , 2015 ) . Unfortunately , these studies used either globally averaged data ( as with ssNMR or neutron scattering ) or imprecise atomic-level models of conformational heterogeneity ( as with B-factors in X-ray crystallography or NMR order parameters ) , thus preventing an all-atom understanding of the complex temperature response of protein crystals . New crystallographic and computational techniques now enable a more detailed investigation of the temperature dependence of protein conformational heterogeneity at the atomic level . First , the program Ringer ( Lang et al . , 2014 , 2010 ) evaluates low-level electron density traditionally considered noise to uncover statistically significant ‘hidden’ alternative conformations , which may become populated or depopulated as a function of temperature . Second , multiconformer models with explicit alternative conformations of both backbone and side chain atoms , as created by manual building or methods such as the program qFit ( Keedy et al . , 2015 , van den Bedem et al . , 2009 ) , can account for non-harmonic motions across separate energy wells ( encoded by discrete alternative conformations with distinct occupancies and coordinates ) and harmonic motions within energy wells ( encoded by B-factors ) . Third , crystallographic order parameters ( S2 ) weigh these harmonic and non-harmonic contributions in a single metric that quantifies the disorder of each residue in a multiconformer model , allowing direct comparison with NMR-determined order parameters ( Fenwick et al . , 2014 ) . Finally , methodological advances based on the physics of ice formation have enabled variable-temperature crystallographic data collection at temperatures between 300 and 100 K with modest or no use of potentially conformation-perturbing cryoprotectants ( Warkentin et al . , 2012; Warkentin and Thorne , 2009 ) . Together , these methods overcome many of the limitations of previous X-ray-based approaches and will contribute to an integrated view of how protein conformational heterogeneity and dynamics evolve with temperature . The human proline isomerase cyclophilin A ( CypA ) is an excellent model system for deploying these tools to study the structural basis of functional conformational dynamics and , in particular , to use temperature to understand the extent of correlated motions during an enzyme’s catalytic cycle . Previous NMR relaxation data for CypA ( Eisenmesser et al . , 2005 , 2002 ) indicated a single common exchange process , both in the apo state and during catalysis , for a network of dynamic residues extending from the core to the active site . Room temperature crystallography later suggested the precise alternative conformations that collectively interconvert during catalysis ( Fraser et al . , 2009 ) . However , subsequent NMR relaxation experiments of mutants designed to perturb the dynamics suggested that multiple exchange processes occur within this network ( Schlegel et al . , 2009 ) . Here , we analyze multitemperature synchrotron experiments to examine the temperature-dependent conformational heterogeneity of CypA . Additionally , we report X-ray-free electron laser ( XFEL ) data , which are free of conventional radiation damage ( Kern et al . , 2014; Spence et al . , 2012 ) , to validate previous connections between alternative conformations determined by synchrotron crystallography and NMR experiments performed in solution ( Eisenmesser et al . , 2005; Fraser et al . , 2009 ) . Our analysis shows that the temperature dependence of alternative protein conformations is heterogeneous and that the character of this heterogeneity bridges previous models for protein dynamical transitions . Our results also suggest new ways to use variable temperature with both synchrotron and XFEL crystallography to probe the dynamic underpinnings of protein function . To probe the conformational landscape of CypA , we collected eight high-resolution ( 1 . 34 –1 . 58 Å ) synchrotron crystallographic datasets across a wide range of temperatures from 100 to 310 K ( Table 1 ) with no added cryoprotectants . For each dataset , we initially refined single-conformer models . Although the single-conformer models are very similar to each other , the accompanying electron density maps reveal differences throughout the protein . In the active-site network , the mFo-DFc difference electron density maps are relatively featureless below 200 K , suggesting that a single conformation is a valid fit below this temperature . By contrast , positive and negative mFo-DFc peaks become gradually more prevalent as temperature increases above 200 K , suggesting that multiple conformations are increasingly required to explain the data as temperature increases ( Figure 2—figure supplement 1 ) . 10 . 7554/eLife . 07574 . 003Table 1 . Crystallographic statistics for multitemperature synchrotron datasets collected on a single crystal per dataset . Statistics for the highest resolution shell are shown in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 003100 K150 K180 K240 K260 K280 K300 K310 KPDB ID4YUG4YUH4YUI4YUJ4YUK4YUL4YUM4YUNWavelength ( Å ) 0 . 97670 . 97670 . 97670 . 97670 . 97670 . 97670 . 97670 . 9767Resolution range ( Å ) 33 . 58–1 . 48 ( 1 . 53–1 . 48 ) 16 . 95–1 . 34 ( 1 . 39–1 . 34 ) 16 . 12–1 . 38 ( 1 . 43–1 . 38 ) 34 . 05–1 . 42 ( 1 . 47–1 . 42 ) 33 . 98–1 . 48 ( 1 . 53–1 . 48 ) 25 . 23–1 . 42 ( 1 . 47–1 . 42 ) 22 . 67–1 . 5 ( 1 . 55–1 . 50 ) 22 . 66–1 . 58 ( 1 . 64–1 . 58 ) Space groupP212121P212121P212121P212121P212121P212121P212121P212121Unit cell ( a , b , c ) 42 . 24 , 51 . 91 , 88 . 0642 . 45 , 51 . 82 , 88 . 0142 . 42 , 51 . 96 , 88 . 2143 . 04 , 53 . 22 , 88 . 6343 . 09 , 52 . 79 , 88 . 8143 . 00 , 52 . 61 , 89 . 1243 . 01 , 52 . 61 , 89 . 3242 . 85 , 52 . 58 , 89 . 41Total reflections160 , 129 ( 15 , 842 ) 160 , 780 ( 7 , 437 ) 154 , 202 ( 11 , 295 ) 152 , 578 ( 13 , 600 ) 134 , 699 ( 13 , 381 ) 168 , 932 ( 15 , 019 ) 144 , 734 ( 14 , 433 ) 125 , 225 ( 12 , 326 ) Unique reflections32 , 657 ( 3 , 240 ) 42 , 288 ( 3 , 471 ) 39 , 548 ( 3 , 820 ) 38 , 881 ( 3 , 710 ) 34 , 411 ( 3 , 391 ) 38 , 763 ( 3 , 794 ) 32 , 999 ( 3 , 254 ) 28 , 291 ( 2 , 760 ) Multiplicity4 . 9 ( 4 . 9 ) 3 . 8 ( 2 . 1 ) 3 . 9 ( 3 . 0 ) 3 . 9 ( 3 . 7 ) 3 . 9 ( 3 . 9 ) 4 . 4 ( 4 . 0 ) 4 . 4 ( 4 . 4 ) 4 . 4 ( 4 . 5 ) Completeness ( % ) 99 ( 100 ) 95 ( 80 ) 97 ( 95 ) 99 ( 96 ) 100 ( 100 ) 100 ( 100 ) 99 ( 100 ) 100 ( 100 ) Mean I/sigma ( I ) 14 . 07 ( 1 . 57 ) 25 . 95 ( 3 . 24 ) 16 . 47 ( 1 . 64 ) 12 . 86 ( 1 . 66 ) 10 . 09 ( 1 . 46 ) 15 . 51 ( 1 . 52 ) 16 . 90 ( 1 . 63 ) 13 . 26 ( 1 . 45 ) Wilson B-factor ( Å2 ) 16 . 0713 . 1216 . 9515 . 5516 . 0617 . 6219 . 7521 . 44R-merge ( % ) 6 . 8 ( 99 . 4 ) 3 . 0 ( 29 . 4 ) 4 . 2 ( 71 . 8 ) 6 . 2 ( 99 . 2 ) 8 . 1 ( 104 . 3 ) 4 . 9 ( 100 . 0 ) 4 . 7 ( 101 . 7 ) 6 . 7 ( 127 . 3 ) R-measurement ( % ) 7 . 6 ( 111 . 0 ) 3 . 4 ( 36 . 9 ) 4 . 8 ( 85 . 6 ) 7 . 2 ( 116 . 8 ) 9 . 4 ( 120 . 8 ) 5 . 6 ( 115 . 3 ) 5 . 4 ( 115 . 9 ) 7 . 6 ( 144 . 5 ) CC1/21 . 00 ( 0 . 62 ) 1 . 00 ( 0 . 90 ) 1 . 00 ( 0 . 60 ) 1 . 00 ( 0 . 50 ) 1 . 00 ( 0 . 52 ) 1 . 00 ( 0 . 52 ) 1 . 00 ( 0 . 59 ) 1 . 00 ( 0 . 56 ) CC1 . 00 ( 0 . 88 ) 1 . 00 ( 0 . 97 ) 1 . 00 ( 0 . 87 ) 1 . 00 ( 0 . 82 ) 1 . 00 ( 0 . 83 ) 1 . 00 ( 0 . 83 ) 1 . 00 ( 0 . 86 ) 1 . 00 ( 0 . 85 ) Refinement resolution range ( Å ) 33 . 085–1 . 48 ( 1 . 558–1 . 48 ) 19 . 117–1 . 34 ( 1 . 394–1 . 34 ) 16 . 995–1 . 38 ( 1 . 435–1 . 38 ) 34 . 055–1 . 42 ( 1 . 477–1 . 42 ) 33 . 98–1 . 48 ( 1 . 547–1 . 48 ) 25 . 23–1 . 42 ( 1 . 477–1 . 42 ) 22 . 67–1 . 5 ( 1 . 579–1 . 5 ) 25 . 2221 . 58 ( 1 . 679 –1 . 58 ) Reflections used in refinement32 , 627 ( 4 , 654 ) 42 , 278 ( 3 , 932 ) 39 , 545 ( 4 , 265 ) 38 , 879 ( 4 , 161 ) 34 , 411 ( 4 , 237 ) 38 , 762 ( 4 , 256 ) 32 , 999 ( 4 , 643 ) 28 , 287 ( 4 , 632 ) Reflections used for R-free1 , 028 ( 147 ) 1 , 325 ( 125 ) 1 , 238 ( 133 ) 1 , 218 ( 130 ) 1 , 080 ( 133 ) 1 , 217 ( 133 ) 1 , 036 ( 145 ) 889 ( 146 ) R-work ( % ) 13 . 3 ( 20 . 4 ) 12 . 4 ( 16 . 4 ) 13 . 3 ( 25 . 4 ) 12 . 6 ( 26 . 3 ) 13 . 1 ( 26 . 0 ) 11 . 1 ( 22 . 6 ) 10 . 8 ( 20 . 0 ) 11 . 7 ( 21 . 8 ) R-free ( % ) 18 . 3 ( 26 . 8 ) 15 . 6 ( 21 . 3 ) 17 . 5 ( 33 . 0 ) 15 . 6 ( 30 . 4 ) 16 . 8 ( 31 . 2 ) 14 . 3 ( 25 . 5 ) 14 . 4 ( 24 . 8 ) 15 . 0 ( 28 . 8 ) Number of non-hydrogen atoms2 , 2792 , 4331 , 9691 , 9932 , 0352 , 1202 , 0962 , 172Macromolecule atoms1 , 9332 , 1321 , 7451 , 7501 , 8371 , 9241 , 9522 , 061Protein residues165164164163163163163163RMS ( bonds ) ( Å ) 0 . 0090 . 0080 . 0080 . 0090 . 0090 . 0080 . 0090 . 009RMS ( angles ) ( ° ) 1 . 161 . 201 . 231 . 201 . 161 . 161 . 141 . 14Ramachandran favored ( % ) 9794979697969796Ramachandran allowed ( % ) 3 . 35 . 72 . 74 . 134 . 23 . 33 . 9Ramachandran outliers ( % ) 00000000Rotamer outliers ( % ) 2 . 41 . 30 . 531 . 11 . 51 . 91 . 40 . 88Clashscore0 . 571 . 080 . 001 . 240 . 270 . 780 . 520 . 00Average B-factor ( Å2 ) 21 . 7417 . 2521 . 8520 . 1420 . 0021 . 4824 . 0925 . 77Macromolecule average B-factor ( Å2 ) 18 . 4814 . 6719 . 9917 . 9518 . 1719 . 6122 . 8224 . 94Solvent average B-factor ( Å2 ) 39 . 9935 . 5436 . 3435 . 8937 . 0139 . 8941 . 2341 . 30PDB: Protein Data Bank . CC: correlation coefficient . We monitored the shift from single-conformation to multiple conformations both visually ( Figure 1A , B ) and using the automated electron density scanning program Ringer ( Figure 1C , D ) . Briefly , Ringer identifies alternative conformations at low levels of electron density by evaluating the density value for the γ atom at each possible position about the χ1 dihedral angle , given a fixed main chain conformation ( Lang et al . , 2014 , 2010 ) . We focused on two residues , Ser99 and Leu98 , which are key markers of the conformational exchange by NMR ( Eisenmesser et al . , 2002 , 2005 ) and were implicated in our previous room-temperature X-ray and mutagenesis experiments ( Fraser et al . , 2009 ) . For both Ser99 ( Figure 1A ) and Leu98 ( Figure 1B ) , a dominant peak is evident at all temperatures . The reduced height of this peak as temperature increases is accompanied by the increase in a secondary peak corresponding to the electron density of the minor conformation . To quantify this trend , we computed correlation coefficients between the electron density versus dihedral angle curves for each residue ( Figure 1C , D ) . Pairs of curves for similar temperatures have higher correlations than those for different temperatures . In particular , pairs of curves for temperatures that span the low-temperature ( 100–180 K ) and high-temperature ( 240–310 K ) regimes are more poorly correlated than are curves from the same temperature regime . The dynamical transitions observed in previous studies ( Doster , 2010; Lee and Wand , 2001; Lewandowski et al . , 2015; Ringe and Petsko , 2003; Schiro et al . , 2015 ) generally occur between these two temperature regimes . 10 . 7554/eLife . 07574 . 004Figure 1 . Automated electron density sampling reveals increased conformational redistribution . Ringer curves of 2mFo-DFc electron density versus χ1 dihedral angle for ( A ) Ser99 and ( B ) Leu98 show large peaks for modeled major conformations and smaller peaks for additional minor conformations ( dashed vertical lines ) . These secondary peaks become more evident as temperature increases ( color gradient from blue to purple to red ) . A backrub motion was used for Ser99 . For ( C ) Ser99 and ( D ) Leu98 , a Pearson correlation coefficient was calculated between each pair of Ringer curves from the corresponding panel in ( A ) or ( B ) . Circles in diagonal elements are colored as in ( A ) or ( B ) ; circles in off-diagonal elements are all gray but scaled by pairwise correlation coefficient ( see legend ) . Pairs of curves from similar temperatures are generally more correlated to each other ( larger circles ) than are pairs of curves from more different temperatures ( smaller circles ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 00410 . 7554/eLife . 07574 . 005Figure 1—Figure supplement 1 . Radiation damage is minimal across data collection temperatures . Plots of Rd versus frame-number difference for each dataset in the multitemperature trajectory reveal only minimal radiation damage . The datasets around 180– 260 K exhibit higher Rd in later frames , which may reflect either a time-dependent cryocooling artifact or a radiation damage at these intermediate temperatures . Although the rate of X-ray damage varies strongly with temperature , the data collection strategy was adjusted to yield a comparable amount of damage per frame . Therefore , there is no correlation between data collection temperature and the overall extent of radiation damage; the highest temperature datasets are equally undamaged as the lowest temperature datasets . By contrast , we observe a strong correlation between data collection temperature and conformational heterogeneity . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 005 To ground this conformational redistribution in all-atom detail , we built a multiconformer model with qFit ( Keedy et al . , 2015; van den Bedem et al . , 2009 ) for each multitemperature dataset . We then finalized the model by manually editing alternative conformations and refining to convergence , resulting in models that were improved relative to the single-conformer models ( Table 2 , Video 1 ) . 10 . 7554/eLife . 07574 . 006Table 2 . Improvements in validation statistics from finalizing raw qFit models . Statistics calculated with phenix . molprobity . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 006RT synchrotronXFEL100 K150 K180 K240 K260 K280 K300 K310 KRfree ( % ) Raw qFit16 . 725 . 219 . 016 . 918 . 517 . 517 . 915 . 716 . 316 . 1Final14 . 624 . 918 . 315 . 617 . 515 . 616 . 814 . 314 . 415 . 0Δ–2 . 1–0 . 3–0 . 7–1 . 3–1 . 0–1 . 9–1 . 1–1 . 4–1 . 9–1 . 1MolProbity scoreRaw qFit1 . 471 . 801 . 791 . 311 . 211 . 181 . 451 . 280 . 951 . 19Final1 . 081 . 391 . 191 . 290 . 631 . 140 . 911 . 250 . 990 . 76Δ–0 . 39–0 . 41–0 . 80–0 . 02–0 . 58–0 . 04–0 . 54–0 . 030 . 04–0 . 43RT: Room temperature; XFEL: X-ray-free electron laser . 10 . 7554/eLife . 07574 . 007Video 1 . Animated interpolation between electron density maps in temperature trajectory . For each pair of adjacent temperatures ( e . g . 100 and 150 K ) , the temperature regime between them was bisected and an average 2mFo-DFc electron density map was calculated in reciprocal space using CCP4 utilities , until temperature points were spaced by <1 K . A new multiconformer model is shown when the animation reaches the corresponding temperature . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 007 At 180 K and below , the active-site network is best modeled as a single state , with electron density corresponding to ordered water molecules clearly evident adjacent to Phe113 ( Figure 2 , top row ) . At 240 K and above , by contrast , multiple conformations provide a better explanation of the data . Interestingly , some partial-occupancy water molecules are still present and likely co-occur with the major conformations ( Figure 2 , middle and bottom rows ) . Met61 appears to populate additional conformations above 180 K , although it is difficult to precisely define changes in its conformational ensemble as temperature increases . This residue bridges Phe113 and the catalytic residue Arg55 via steric contacts between alternative conformations in both directions , emphasizing the importance of modeling multiple conformations in all-atom detail for understanding inter-residue coupling . 10 . 7554/eLife . 07574 . 008Figure 2 . Multiconformer modeling across temperatures captures increasing conformational heterogeneity . Residues extending from the core to the active site of cyclophilin A ( CypA ) adopt a single conformer at low temperatures , but gradually transition to increasing occupancy of secondary conformations as temperature increases . These conformations are well supported by 2mFo-DFc electron density contoured at 0 . 6 σ ( cyan mesh ) and 3 . 0 σ ( dark blue mesh ) . This is corroborated by the room-temperature X-ray free-electron laser ( XFEL ) model ( gray ) , which is free from conventional radiation damage and features the same secondary conformations . Water molecules ( red spheres ) are more fully ordered at low temperatures , but become only partially occupied at higher temperatures because they are mutually exclusive with the secondary Phe113 conformation . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 00810 . 7554/eLife . 07574 . 009Figure 2—figure supplement 1 . Single-conformer models cannot explain the crystallographic data at higher temperatures . . The CypA dynamic network is shown after molecular replacement and refinement ( including automated water placement ) in PHENIX , before any manual rebuilding . The major state is well supported by 2mFo-DFc electron density contoured at 0 . 6σ ( cyan mesh ) and 3 . 0σ ( dark blue mesh ) for all datasets , but mFo-DFc difference electron density becomes more negative for the major state ( −3 . 0σ , red mesh ) and more positive for the unmodeled minor state ( 3 . 0σ , green mesh ) as temperature increases across the synchrotron datasets ( blue to red ) , especially at and above 240 K . Full-occupancy water molecules ( red spheres ) are automatically placed by PHENIX near the Phe113 minor state in lower temperature , but not in higher temperature synchrotron models because they are mutually exclusive with the secondary Phe113 conformation . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 009 Quantifying radiation damage versus exposure dose ( Figure 1—figure supplement 1 ) and limiting exposure dose per dataset ensured that the conformational heterogeneity observed in multitemperature synchrotron datasets was not dominated by radiation damage . However , XFELs can generate data that are entirely free from conventional radiation damage by diffraction-before-destruction data collection ( Kern et al . , 2014; Spence et al . , 2012 ) . To compare the distribution of alternative conformations between synchrotron and XFEL data , we collected two ambient-temperature datasets: a 1 . 75 Å resolution radiation-damage-free dataset using serial femtosecond rotation crystallography ( Table 3 ) ( Hirata et al . , 2014; Schlichting , 2015; Suga et al . , 2015 ) and an additional 1 . 2 Å resolution synchrotron dataset ( Table 4 ) . For the XFEL experiment , we collected 1 , 239 individual diffraction images , translating to unique unexposed regions of 71 crystals between each shot ( Video 2 ) , and processed them using cctbx . xfel ( Hattne et al . , 2014 ) with post-refinement in PRIME ( Uervirojnangkoorn et al . , 2015 ) . Automated molecular replacement yielded interpretable electron density maps that allowed us to refine a single-conformer structural model with reasonable quality statistics . Electron density sampling analysis using Ringer and multiconformer refinement using qFit were performed as for the multitemperature synchrotron data . 10 . 7554/eLife . 07574 . 010Table 3 . Crystallographic statistics for room-temperature XFEL dataset collected across 71 crystals . Statistics for the highest resolution shell are shown in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 010XFELPDB ID4YUPResolution range ( Å ) 43 . 98 –1 . 75 ( 1 . 81 –1 . 75 ) Space groupP212121Unit cell ( a , b , c ) 42 . 42 , 51 . 82 , 87 . 96Unique reflections19 , 942 ( 1894 ) Completeness ( % ) 99 ( 96 ) Wilson B-factor ( Å2 ) 21 . 12Refinement resolution range ( Å ) 43 . 98 –1 . 75 ( 1 . 93 –1 . 75 ) Reflections used in refinement19 , 936 ( 4 , 811 ) Reflections used for R-free625 ( 151 ) R-work ( % ) 20 . 0 ( 34 . 3 ) R-free ( % ) 24 . 9 ( 36 . 1 ) Number of non-hydrogen atoms1 , 762Macromolecular atoms1 , 559Protein residues164RMS ( bonds ) ( Å ) 0 . 017RMS ( angles ) ( ° ) 1 . 44Ramachandran favored ( % ) 96Ramachandran allowed ( % ) 3 . 6Ramachandran outliers ( % ) 0Rotamer outliers ( % ) 1 . 8Clashscore1 . 92Average B-factor ( Å2 ) 29 . 03Macromolecule average B-factor ( Å2 ) 26 . 52Solvent average B-factor ( Å2 ) 48 . 25Number of TLS groups3PDB: Protein Data Bank; TLS: translation libration screw; XFEL: X-ray-free electron laser . 10 . 7554/eLife . 07574 . 011Table 4 . Crystallographic statistics for room-temperature synchrotron dataset collected on a single crystal . Statistics for the highest resolution shell are shown in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 0111 . 2 Å SynchrotronPDB ID4YUOWavelength ( Å ) 0 . 9795Resolution range ( Å ) 44 . 60 –1 . 20 ( 1 . 24 –1 . 20 ) Space groupP212121Unit cell ( a , b , c ) 42 . 9 , 52 . 43 , 89 . 11Total reflections307 , 722 ( 18 , 999 ) Unique reflections58 , 118 ( 5 , 122 ) Multiplicity5 . 3 ( 3 . 7 ) Completeness ( % ) 91 ( 82 ) Mean I/sigma ( I ) 10 . 99 ( 5 . 93 ) Wilson B-factor ( Å2 ) 15 . 22R-merge ( % ) 11 . 2 ( 20 . 4 ) R-measurement ( % ) 12 . 2 ( 23 . 4 ) CC1/20 . 99 ( 0 . 96 ) CC1 . 00 ( 0 . 99 ) Refinement resolution range ( Å ) 45 . 19 –1 . 20 ( 1 . 23 –1 . 20 ) Reflections used in refinement58 , 108 ( 3 , 657 ) Reflections used for R-free2 , 000 ( 126 ) R-work ( % ) 12 . 7 ( 31 . 3 ) R-free ( % ) 14 . 6 ( 33 . 5 ) Number of non-hydrogen atoms2327Macromolecular atoms2143Protein residues163RMS ( bonds ) ( Å ) 0 . 009RMS ( angles ) ( ° ) 1 . 16Ramachandran favored ( % ) 96Ramachandran allowed ( % ) 4 . 1Ramachandran outliers ( % ) 0Rotamer outliers ( % ) 0 . 84Clashscore0 . 98Average B-factor ( Å2 ) 19 . 62Macromolecule average B-factor ( Å2 ) 18 . 40Solvent average B-factor ( Å2 ) 33 . 86PDB: Protein Data Bank . 10 . 7554/eLife . 07574 . 012Video 2 . Fixed-target X-ray-free electron laser ( XFEL ) data collection from cyclophilin A ( CypA ) crystals at the LCLS-XPP end station . Screen capture image of the Blu-Ice GUI showing a video display of a CypA crystal . After each shot , a new damage line appears and the crystal is translated . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 012 In agreement with our previous room-temperature studies ( Fraser et al . , 2009 ) , the XFEL and synchrotron mFo-DFc difference maps reveal evidence for the rate-limiting alternative conformations extending from the active site into the core of the protein ( Figure 3A , B ) . For example , the backrub-coupled ( Davis et al . , 2006 ) rotamer jump of Phe113 is apparent from a large positive mFo-DFc peak in both maps . Alternative conformations for core residue Ser99 are also evident from mFo-DFc peaks ( Figure 3A , B ) and Ringer electron density sampling curves ( Figure 3E ) . We did not conclusively observe a secondary peak in the electron density sampling curve corresponding to a discrete alternative conformation of Leu98 ( Figure 3F ) , but that is likely due to the lower resolution of the XFEL dataset . Multiconformer models for both datasets ( Figure 3C , D ) again feature alternative conformations across the active-site network and are strongly supported by 2mFo-DFc electron density . These results provide an important positive control on the observation of conformational heterogeneity in our synchrotron studies by establishing that electron density corresponding to the alternative conformations of CypA is not an artifact of radiation damage . The ability of XFEL crystallography to reveal native and functionally important alternative conformations at high resolution may be especially useful for other systems that are presently intractable for room- or variable-temperature synchrotron crystallography due to the small size of available crystals . 10 . 7554/eLife . 07574 . 013Figure 3 . The active-site conformational ensemble of cyclophilin A ( CypA ) determined without radiation damage at room temperature . ( A ) Electron density maps for room-temperature synchrotron ( red ) and ( B ) X-ray-free electron laser ( XFEL ) ( silver ) single-conformer models reveal conformational heterogeneity extending from the protein core ( Leu98 and Ser99 ) to the active site ( Arg55 ) of CypA . The primary conformation is well supported by 2mFo-DFc electron density contoured at 0 . 6 σ ( cyan mesh ) and 3 . 0 σ ( dark blue mesh ) . mFo-DFc difference electron density contoured at 3 . 0 σ ( green mesh ) and − 3 . 0 σ ( red mesh ) suggests unmodeled alternative conformations . ( C , D ) Finalized multiconformer models explicitly model these alternative conformations , which are well-supported by 2mFo-DFc electron density . ( E , F ) Ringer electron density sampling for the single-conformer models shows peaks representing alternative conformations for ( E ) Ser99 and ( F ) Leu98 . The primary conformations of both residues are obvious as peaks for both models , but the minor conformations ( dashed vertical line; as modeled in 3k0n ) are also evident , with 2mFo-DFc values well above the 0 . 3σ ( darker gray horizontal line ) threshold , except for the Leu98 in the XFEL model ( due to the lower resolution ) . A backrub motion of −10° positions the backbone properly for Ringer to best detect the minor conformation for Ser99 , but not for Leu98 . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 01310 . 7554/eLife . 07574 . 014Figure 3—figure supplement 1 . The 1 . 2 Å room-temperature CypA synchrotron data show no signs of radiation damage . A plot of ‘decay R-factor’ ( Rd ) as a function of frame-number difference ( as in Figure 1—figure supplement 2 ) has a slope of zero , indicating the absence of radiation damage . Rd is calculated using pairwise observations of unique reflections ( hkl ) with centroids on frames i and j , and the frame-number difference is given by i-j . The calculations were performed using a 2 . 0 Å resolution cutoff . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 014 Although more conformational heterogeneity is expected with our higher temperature synchrotron datasets , and is evident in the active site of CypA , cooling can also stabilize new conformations ( Halle , 2004 ) . For example , the loop containing residues 79–83 ( Figure 4 , Video 3 ) exhibits conformational heterogeneity only at cryogenic temperatures . This region is well fit by a single conformation at 240 K and above , but a secondary loop conformation is necessary to explain the electron density at 100 , 150 , and 180 K . Additionally , the loop is clearly single-state in the highest resolution ( 1 . 2 Å ) dataset ( Figure 4—figure supplement 1 ) , demonstrating that the slightly lower resolution of the elevated-temperature datasets does not obscure a secondary conformation . 10 . 7554/eLife . 07574 . 015Figure 4 . Alternative loop conformations can appear at lower temperatures . The surface loop containing residues 79–83 adopts alternative conformations at low temperatures ( top row ) but not at high temperatures ( bottom two rows ) . The secondary loop conformation is separated from the body of the protein by an ordered water molecule ( red sphere ) ; the van der Waals interactions between the loop and the water may reflect an enthalpic stabilization that is more dominant at low temperatures . The electron density peak to the right of the water corresponds to the backbone carbonyl oxygen of Glu81 . 2mFo-DFc electron density contoured at 0 . 6σ ( cyan mesh ) and 2 . 0σ ( dark blue mesh ) . XFEL: X-ray-free electron laser . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 01510 . 7554/eLife . 07574 . 016Figure 4—figure supplement 1 . Alternative loop conformations are not present in the highest-resolution ( 1 . 2 Å ) dataset . The surface loop containing residues 79–83 does not adopt alternative conformations in the 1 . 2 Å synchrotron dataset . 2mFo-DFc electron density contoured at 0 . 6σ ( cyan mesh ) and 2 . 0σ ( dark blue mesh ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 01610 . 7554/eLife . 07574 . 017Video 3 . Animated rotation around the 100 K ( blue ) and 310 K ( red ) models and electron density maps from Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 017 In the primary conformation , the 79–83 loop is not involved in any main chain–main chain hydrogen bonds to the rest of CypA , suggesting that the barrier to forming the secondary conformation does not involve breakage of cooperative secondary-structure-like interactions . The observation of a secondary state for residues 79–83 at 100–180 K , but not at 240–310 K , suggests that it is enthalpically stabilized at lower temperatures ( Halle , 2004; Keedy et al . , 2014 ) . Consistent with this mechanism , the secondary conformation of the 79–83 loop is accompanied by an ordered , partial-occupancy water molecule ( Figure 4 , top row ) . This water molecule , which is clearly distinct from the carbonyl oxygen of the primary conformation of Glu81 , wedges between the loop and the rest of the protein . The surprising appearance of specific solvent-linked protein conformational heterogeneity exclusively below 240 K emphasizes the complex and heterogeneous changes in protein–solvent energetics that can occur at cryogenic temperatures . Despite counter examples such as the 79–83 loop , most residues in CypA , especially in the active site , exhibit increases in discrete conformational heterogeneity above 180 K . To quantify these changes in regions implicated by NMR relaxation experiments , we measured the 2mFo-DFc electron density in the volumes occupied by the alternative conformations of Ser99 and Phe113 . By contrast , B-factors , which can model the harmonic motions near any single conformation , are poor proxies for the non-harmonic change between discretely separated conformations . To quantify the change in minor state occupancy as a function of temperature , we summed the electron density in the volume that is occupied exclusively by the minor conformation and avoided any voxels that overlap with the van der Waals radii of atoms of the major conformation ( Figure 5A ) . The resulting curves of minor-state electron density versus temperature have a shallow slope at 180 K and below , but a much steeper slope at 240 K and above ( Figure 5B , C ) . Additionally , the electron density for the XFEL data is consistent with the data collection temperature ( 273 K ) and the overall trends . 10 . 7554/eLife . 07574 . 018Figure 5 . Quantifying temperature titration of conformational heterogeneity in multiconformer models . ( A ) 2mFo-DFc electron density was summed over the volume occupied by the minor conformation but not the major conformation ( blue grid points ) for Ser99 and Phe113 . ( B , C ) Minor-state 2mFo-DFc electron density increases with temperature . Electron density sums were normalized for each residue . Multitemperature points from synchrotron data are shown in colors corresponding to temperature . The X-ray-free electron laser point is shown as a gray triangle . Best-fit lines are shown for 180 K and below ( blue ) versus 240 K and above ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 018 However , most residues that populate alternative conformations do not have such easily characterized and separable regions of electron density . To quantify how conformational heterogeneity throughout CypA varies as a function of temperature , we used B-factor-dependent crystallographic order parameters ( S2 ) ( Fenwick et al . , 2014 ) . These order parameters include both harmonic contributions , which reflect conformational heterogeneity near one conformation ( encoded by B-factors ) , and non-harmonic contributions , which reflect conformational heterogeneity between multiple discretely separated conformations ( encoded by occupancies and displacements in coordinates ) . Importantly , these order parameters account for both conformational heterogeneity within energy wells , whether it is modeled by B-factors or by subtly different alternative conformations , as well as discretely separated alternative conformations that occupy distinct rotamers . Similar to the 2mFo-DFc electron density integration results for Phe113 , we observed a large change in χ1 bond order parameters at 240 K and above ( Figure 6A ) . 10 . 7554/eLife . 07574 . 019Figure 6 . Diversity in temperature dependences of side chain disorder across cyclophilin A ( CypA ) does not predict the observed average arrest of disorder . ( A ) The complement of B-factor-influenced side chain order parameter for the bond most closely associated with the χ1 dihedral angle for Phe113 . Lines reflect least-squares fits to synchrotron models at 180 K and below ( blue ) versus 240 K and above ( red ) . Multitemperature synchrotron points in colors; X-ray-free electron laser ( XFEL ) point ( not included in fits ) as gray triangle . ( B ) Distribution of the intersection temperature between the <200 and >200 K lines fitted with kernel density function . The peak is near 250 K , although there is a tail toward lower temperatures . Intersection temperatures were <170 K for four residues and >330 K for five residues . ( C ) Predicted and observed values for the complement of side chain order parameter , averaged over all residues in CypA . The predicted values were obtained by extrapolating each residue’s fit line for 240 K and above ( red curve ) or for the full 100–300 K ( purple curve ) , flooring the result to 0 , then averaging across all residues in CypA . Observed values , similarly floored and averaged , are shown as points . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 01910 . 7554/eLife . 07574 . 020Figure 6—figure supplement 1 . Heterogeneous response of side chain disorder to temperature . The complement of B-factor-influenced side chain order parameter for the bond most closely associated with the χ1 dihedral angle for all residues in CypA . Lines reflect least-squares fits to synchrotron models for 180 K and below ( blue ) versus 240 K and above ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 02010 . 7554/eLife . 07574 . 021Figure 6—figure supplement 2 . Residues with persistent disorder across temperatures are interconnected in the crystal lattice . Several residues with high 1 – S2 values ( Val2 , Glu15 , Gly80 , Glu81 , Lys82 , Pro105 , Ala117 , Glu120 , Lys125 , Met142 , Ser147 , and Lys151 ) are shown for the central molecule ( blue-to-red sticks and backbone ) and also in symmetry mates ( green sticks , gray backbone ) . Many of these residues appear to interact with each other via lattice contacts . Frustration in these interactions may lead to persistent disorder . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 02110 . 7554/eLife . 07574 . 022Figure 6—figure supplement 3 . Diversity in temperature dependences of side chain-end disorder across CypA does not predict the observed average arrest of disorder . Each panel is as in Figure 6 , but the order parameter now models the final heavy-atom to heavy-atom bond for each side chain ( see ‘Methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 02210 . 7554/eLife . 07574 . 023Figure 6—figure supplement 4 . Heterogeneous response of side chain-end disorder to temperature . The small multiple plots are as in Figure 6—figure supplement 1 , but the order parameter now models the final heavy-atom to heavy-atom bond for each side chain ( see ‘Methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 02310 . 7554/eLife . 07574 . 024Figure 6—figure supplement 5 . Different residues extrapolate to different maximal-order temperatures . Fit lines for temperature points at 240 K and above were used to extrapolate to the maximal-order temperature , at which 1 − S2 = 0 . ( A ) Order parameter modeling the bond most closely associated with the χ1 dihedral angle . ( B ) Order parameter modeling the final heavy-atom to heavy-atom bond for each side chain . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 02410 . 7554/eLife . 07574 . 025Figure 6—figure supplement 6 . Both harmonic and non-harmonic flexibilities contribute to changes in order parameters with temperature . ( A– D ) Contributions to χ1 order parameter versus temperature for four representative residues in CypA . Ser99 and Phe113 are in the active-site network , Glu81 is surface-exposed and adopts alternative conformations at all temperatures ( Figure 4 ) , and Phe8 is buried in the protein core and is single-rotamer at all temperatures . Similar conformations within the same rotameric well were grouped together for this analysis . ( A ) Occupancy of minor alternative conformations . ( B ) Intra-residue heavy-atom-average B-factor . ( C ) Complement of the S2ang component of the χ1 order parameter , which uses occupancy-weighted angles between bond vectors across alternative conformations . ( D ) Complement of the S2ortho component of the χ1 order parameter , which uses occupancy-weighted B-factors . Placement of XFEL points and coloring as in Figure 6A . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 02510 . 7554/eLife . 07574 . 026Figure 6—figure supplement 7 . Globally averaged disorder exhibits an apparent transition near 250 K . Points indicate observed values for the complement of side chain order parameters , floored at 0 and then averaged over all residues in CypA , as in Figure 6 . The blue and red lines represent fits to the ≤180 K and ≥240 K floored and averaged points , respectively . The fits to these globally averaged data suggest a transition at ∼250 K , even though the underlying heterogeneity of the individual residue responses does not indicate there is a transition near this temperature . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 026 Next , we applied the order parameter analysis to all side chain χ1 angles in CypA . Although conformational heterogeneity generally increases with temperature throughout the enzyme , we observed a diverse set of conformational responses ( Figure 6—figure supplement 1 ) . The trends for the majority of residues suggested a transition somewhere between our data points at 180 and 240 K , below which the change in conformational heterogeneity with temperature is reduced . To quantify this trend , we performed separate fits for the low-temperature ( ≤180 K ) and high-temperature ( ≥240 K ) data points for all residues . The slopes of conformational heterogeneity ( 1 – S2 ) versus temperature were significantly different ( p=1 × 10–62 , paired T-test ) on either side of this transition range: the average slope for the low-temperature fit lines ( 2 . 5 × 10–4 K–1 ) was an order of magnitude smaller than for the high-temperature fit lines ( 2 . 6 × 10–3 K–1 ) . This is consistent with the idea that heterogeneity is much less dependent on temperature below the 180–240 K ‘transition’ range . However , some residues behaved differently from the rest of the protein . Val2 retains its conformational heterogeneity at all temperatures , which is expected based on its weakly constrained position at the N terminus . Many of the remaining outlier residues ( Glu15 , Glu81 , Pro105 , Ala117 , Glu120 , Lys125 , Met142 , Ser147 , Lys151 ) appear to be involved in a spatially contiguous set of crystal contacts across symmetry mates in the context of the crystal lattice ( Figure 6—figure supplement 2 ) . This cluster includes Glu81 , which adopts alternative backbone conformations only at low temperatures ( Figure 4 ) . The variability of these residues can likely be explained by distinct sets of conformations across crystal contacts that are differentially , but somewhat stochastically , favored during the cooling process ( Alcorn and Juers , 2010 ) . Our data suggest that CypA does not undergo a single global transition from having strongly temperature-dependent changes in side chain conformational heterogeneity to relatively temperature-independent behavior . An ‘intersection’ or ‘transition’ temperature for each bond angle can be estimated from the intersection of the low-temperature and high-temperature fit lines of 1 – S2 versus temperature . The distribution of these intersection temperatures is broad and asymmetrical , with an elongated tail from the peak near 250 K toward 200 K ( Figure 6B ) . Furthermore , the distribution of intersection temperatures is more complex for order parameters reporting on the terminal heavy-atom bond of the side chain than for χ1 ( Figure 6—figure supplement 3 , 4 ) . This increase likely occurs because side chain end orientations are subject to more degrees of freedom and therefore temperature changes may redistribute them in a greater variety of ways . Our data provide insight into models for the origin of the temperature dependence of protein conformational heterogeneity and into proposed dynamical transitions . In one model , deactivation of different internal protein motions at high temperatures ( near 300 K ) is sufficient to predict a dynamical transition near 200 K ( Lee and Wand , 2001 ) . In a second model , solvent-coupled arrest of protein motions produces a transition in a similar temperature range ( Ringe and Petsko , 2003 ) . To distinguish between these two models , we analyzed the average side chain disorder across all residues in CypA , focusing on the bond most closely associated with the χ1 dihedral angle , at each of the eight temperatures we studied ( Figure 6C ) . These averaged disorder values drop as temperature is decreased from 310 K , then flatten out somewhere between 240 and 180 K , with some scatter due to the variability in the cryocooling process ( data points in Figure 6C and Figure 6—figure supplement 7 ) . Next , we used two different linear fits to extrapolate 1 – S2 across all temperatures for each residue , floored the result at maximum order ( 1 – S2 = 0 ) , and then averaged across all residues to obtain predictions for the residue-averaged disorder versus temperature . In the first fit , the linear function was fit to data for each residue at all temperatures . Consistent with the necessity of using separate fit lines for the low-temperature and high-temperature data ( Figure 5 and Figure 6—figure supplement 1 ) , the resulting prediction gives a poor account of the averaged experimental data and does not indicate a transition ( purple line in Figure 6C ) . In the second fit , only the high-temperature ( 240 K and above ) data for each residue were fit . The resulting prediction is more consistent with the averaged high-temperature experimental data and does indicate a transition ( red line in Figure 6C ) . The flattening of this predicted curve at low temperatures occurs as more individual residues achieve maximal predicted order ( S2→1 ) ( Figure 6—figure supplement 5 ) . This latter prediction , which is extrapolated from high-temperature crystallographic data , is reminiscent of predictions based on NMR relaxation experiments conducted at 288–346 K ( Lee and Wand , 2001 ) . Our observations are consistent with the idea that thermal depopulation of protein alternative conformations is sufficient to predict the existence of an average inflection without invoking a transition of the solvent . However , the low temperature of the predicted inflection ( ∼100 K ) , as well as the large separation in low-temperature disorder between our experimental data ( data points in Figure 6C ) and predictions from high temperature ( red line in Figure 6C ) , suggest that thermal depopulation of protein alternative conformations cannot by itself account for the observed ∼200 K transition . This large separation also indicates that data collected at high temperatures ( >260 K ) cannot be reliably extrapolated to predict conformational heterogeneity at low temperatures . It also follows that data from low temperatures ( ≤180 K ) cannot be simply extrapolated to predict the features of the energy landscape that may be important above ∼200 K . Many effects may contribute to the discrepancy between the observed data and the behavior projected from the high-temperature fits . To gain additional insight into this discrepancy , we decomposed the order parameters into their B-factor versus discrete-conformers components and examined their temperature dependences ( Figure 6—figure supplement 6 ) . Roughly 67% of residues ( e . g . Phe8 ) remain within one rotamer well across all temperatures . Approximately 13% of residues ( e . g . Ser99 and Phe113 ) populate clearly separable multiple rotameric states at high temperatures , and then show complete depopulation of minority states on cooling so that only a single rotameric state remains at 180 K and below . However , 6% of residues ( e . g . Thr5 ) continue to populate multiple rotameric states at or below 180 K . An additional 6% of residues ( e . g . Lys91 ) populate new rotameric states only at or below 180 K . These results help explain the excess residual disorder in our experimental structures below 240 K compared to projections based on high-temperature fits . Although slopes of crystallographic B-factors with temperature remain nearly flat below 240 K ( Figure 6—figure supplement 6B ) , we expect that true harmonic thermal disorder does subtly decrease from 180 to 100 K; these thermal effects could be more detectable if even higher resolution data were collected at even lower temperatures , perhaps by using liquid helium as a cryogen to cool to ∼15 K or below ( Chinte et al . , 2007 ) . While the results above show a variety of thermal and non-thermal conformational responses , it remains unclear whether these responses involve coupled conformational shifts of multiple residues . In particular , the network of alternative side chain conformations spreading from the core of the protein ( Ser99 ) into the active site ( Arg55 ) across multiple β-strands exhibits qualitatively similar behavior of increasing occupancy above 240 K . In previous work ( Fraser et al . , 2009 ) , the collective presence of these alternative conformations at room temperature , but not at cryogenic temperatures , and the close contacts between these residues had suggested a concordance with the single exchange process fit by NMR relaxation dispersion for this dynamic active-site network . However , using our new multitemperature data , this network appears subdivided based on the apparent intersection or transition temperatures of the constituent residues , with Ser99 and Phe113 behaving most similarly to each other ( Figure 7 , Video 4 ) . 10 . 7554/eLife . 07574 . 027Figure 7 . The temperature dependence of side chain disorder is non-homogenously spatially distributed in CypA . Intersection temperatures from ( A ) χ1 order parameters as in Figure 6B or ( B ) side chain terminus order parameters as in Figure 6—figure supplement 3 B are mapped to the 1 . 2 Å room-temperature synchrotron model . Each residue is marked with a sphere colored based on its apparent transition temperature , from low ( blue ) to high ( red ) . The active-site network is subdivided: Ser99 and Phe113 ( left of boxed region ) both transition at a low temperature regardless of order parameter bond vector , but Met61 and Arg55 transition at higher , different temperatures . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 02710 . 7554/eLife . 07574 . 028Video 4 . 360° rotation around Figure 7B . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 028 Here , we have mapped the conformational landscape of the dynamic enzyme CypA by analyzing multiconformer models from multitemperature crystallography . Unlike previous temperature-dependent analyses of X-ray crystallography , here we consider both harmonic disorder ( B-factors ) and non-harmonic displacements ( alternative conformations and occupancies ) , characterized using crystallographic bond order parameters . We have four primary findings: Our results provide new insight into the relationship between energy landscapes , the glass transition , and protein function . Glasses and other disordered systems have complex energy landscapes , in part due to the large number of microenvironments and the extensive frustration that disorder generates in the intermolecular interactions . Proteins at biological temperatures are ‘glassy’ in this sense—they have complex energy landscapes , due to their large size , heterogeneous amino acid composition , and many degrees of conformational freedom ( Frauenfelder et al . , 1991 ) . The extensive heterogeneity in the temperature response of individual residues in CypA that we observe here provides additional direct evidence for this underlying energetic heterogeneity . In addition to the inherent ‘glassiness’ of proteins at biological temperatures , dynamical transitions , some of which have been called ‘glass transitions’ , have been reported at lower temperatures , including 180 , 200 , 220 , 240 , and 250 K , based on Mössbauer spectroscopy , X-ray crystallography , liquid and solid-state NMR , neutron scattering , and other techniques ( Lewandowski et al . , 2015; Schiro et al . , 2015 ) . These transitions typically manifest as a change in slope of some measurement in the vicinity of the suggested transition temperature . However , many of these measurements are sensitive to motions only within some timescale window ( often ps-ns ) , monitor only a subset of amino acid types , and/or spatially average over all residues in the protein . By contrast , multitemperature crystallography with multiconformer models has many advantages by providing a time-independent and fully site-resolved measurement of ensemble-averaged atomic displacements , including both harmonic and discrete conformational heterogeneity , within the crystal . This combined methodology lets us examine dynamical and glass transitions in protein crystals from a new perspective . Glass transitions are by definition non-equilibrium phenomena that arise when the kinetics of relaxation toward equilibrium slow so dramatically that equilibrium cannot be reached on experimental timescales . One signature of a true glass transition in proteins would be if occupancies of minority alternative conformations were arrested at non-zero values below some temperature . Indeed , here we see no appreciable temperature evolution of individual conformer occupancies or B-factors at 180 K and below , and the average disorder at these temperatures is far in excess of what is predicted from high-temperature extrapolations . These observations are consistent with the falling out of equilibrium expected in a glass transition , but not with a transition driven by the thermal freeze-out of alternative side chain conformations ( Lee and Wand , 2001 ) . Moreover , the persistence of multiple rotameric states at low temperatures is consistent with solvent arrest that impedes further changes in side chain disorder . Although the details of protein–solvent interactions may differ in crystals versus in solution , local variability at different protein–solvent interface microenvironments in the crystal ( Teeter et al . , 2001 ) likely contributes to the heterogeneity of temperature responses that we observe . The critical importance of site resolution is evident in the results of Figure 6 . Averaging over side chain disorder in all residues yields an apparent transition near 250 K ( Figure 6 and Figure 6—figure supplement 7 ) . Perhaps coincidentally , this same temperature has been associated with a transition for protein side chains in site-averaged solid-state NMR measurements ( Lewandowski et al . , 2015 ) . However , the ‘transition’ in our residue-averaged result obscures the highly heterogeneous temperature dependence of the individual side chains . We find no evidence of tight cooperativity or of a collective global response near 250 K that would be expected in the case of a ‘true’ dynamical transition . Instead , our data are consistent with local , non-cooperative freeze-out of conformational states defined by the energy landscape over a broad temperature range . The heterogeneous response of side chain order parameters to temperature is driven largely by the changes to the populations of alternative conformations , which ‘flat-line’ at different temperatures across CypA . This diversity of ‘flat-lining’ temperatures is present even within the dynamic active-site network , even though the constituent residues have similar occupancies for their major and minor states at high temperatures ( Figure 7 ) . This result contrasts with previous NMR and X-ray experiments that hypothesized correlated motions of this network as rate-limiting for the catalytic cycle ( Eisenmesser et al . , 2005; Fraser et al . , 2009 ) ( Figure 8A ) . 10 . 7554/eLife . 07574 . 029Figure 8 . The dynamic active-site network of cyclophilin A ( CypA ) has a complex energy landscape . ( A ) The previous simple model in which Ser99 , Phe113 , and Arg55 ( Met61 omitted for clarity ) interconvert from one macrostate ( blue ) to the other ( red ) completely collectively . NMR data suggest this process occurs on a millisecond timescale . ( B ) A more nuanced model in which network microstates are populated differently depending on the network macrostate , defined by the Phe113 rotameric state . In the left macrostate , Ser99 rotamer changes are disfavored because of steric overlaps with Phe113 , but Arg55 rotamer changes are accommodated; the reverse is true ( perhaps to a lesser extent ) in the right macrostate . Within each microstate , rapid thermal motions occur ( bottom right ) , and may alleviate some minor steric overlaps . Timescales are estimates consistent with NMR observables for CypA and other systems . ms: millisecond; ns: nanosecond; ps: picosecond . DOI: http://dx . doi . org/10 . 7554/eLife . 07574 . 029 To bridge these views , we propose that the active-site network adopts two substates , which are primarily distinguished by Phe113 rotamer interconversion . Each of these substates adopts a differently weighted ensemble of conformations for other residues ( Figure 8B ) . In this model , Met61 and Arg55 can switch rotamers more easily than Ser99 when Phe113 is in its χ1 gauche ( p ) rotamer pointed toward Ser99 , whereas Ser99 can switch rotamers more easily than Met61 and Arg55 when Phe113 is in its χ1 gauche ( m ) rotamer pointed toward Met61 . Additionally , thermal ‘breathing’ motions within rotameric wells may relieve minor steric overlaps within some of these macro- and microstates ( Figure 8B , bottom right ) . This model is consistent with Phe113 having the lowest ‘flat-lining’ temperature of the network ( Figure 7 ) , and makes sense sterically because of the large size of the aromatic ring . These hypothesized motions are consistent with the timescales and temperature dependencies of motion assigned by solid-state NMR studies of crystalline protein GB1 ( Lewandowski et al . , 2015 ) . Furthermore , the model helps explain the difficulty of fitting NMR relaxation data for perturbed versions of the active-site network as a single collective exchange process ( Schlegel et al . , 2009 ) . The aromatic ring of Phe113 could play a dominant role in determining the chemical shift changes of the surrounding residues . Each of these residues could also populate multiple rotamers in the excited state measured by NMR . Our hierarchical perspective evokes the ‘population shuffling’ model of ( Smith et al . , 2015 ) , in which a protein macrostate ( in CypA , defined by the Phe113 rotamer ) also determines the different relative populations of rotamers for a subset of other residues ( in CypA , the other residues in the active-site network ) . In this model , the interconversion between macrostates , and not the collective motion of all residues between distinct rotamers , is correlated with the rate-limiting step of the CypA catalytic cycle . Diversity in the temperature dependences of alternative conformations as we see here is inevitable given the limitations of the amino acid alphabet , yet its spatial pattern within a protein may provide insight into selective pressures . Evolutionary optimization must ensure that functionally important alternative conformations are robustly populated and interconvert appreciably at physiological conditions . However , the energy landscapes of individual residues are coupled to varying extents , such that some subsets of residues must be collectively optimized to preserve some , but not perfect , collectivity in functional motions . For proteins with large sequence alignments , evolutionary covariation has been used to predict ‘sectors’ of functionally cooperative residues , which are often dispersed in primary sequence but strikingly contiguous in tertiary structure ( Halabi et al . , 2009 ) . By contrast , temperature-dependent crystallography has the potential to unveil couplings in atomic detail by identifying sets of residues whose conformational ensembles respond concertedly to temperature change . Based on our results with CypA , we expect this coupling to be weak , but measurable . Serial femtosecond XFEL crystallography combined with ultra-fast temperature jumps could enable a temporal view of these coupled conformational changes . Novel static and time-resolved multitemperature crystallographic approaches will provide powerful tools for resolving concerted motions to explore how proteins function and evolve . Wild-type CypA was produced and crystallized as previously reported ( Fraser et al . , 2009 ) . Briefly , crystals were grown by mixing equal volumes of well solution ( 100 mM ( 4- ( 2-hydroxyethyl ) -1-piperazineethanesulfonic acid ) HEPES pH 7 . 5 , 23% PEG 3350 , 5 mM Tris ( 2-carboxymethyl ) phosphine [TCEP] ) and protein ( 60 mg mL –1 in 20 mM HEPES pH 7 . 5 , 100 mM NaCl , 0 . 5 mM TCEP ) in the hanging-drop format . For the multitemperature synchrotron datasets at 100 , 150 , 180 , 240 , 260 , 280 , 300 , and 310 K , we collected data at the Cornell High Energy Synchrotron Source ( CHESS ) at beamline A1 with a 100 µm collimator using a wavelength of 0 . 9767 Å . Crystals were looped by hand , stripped of excess mother liquor ( 100 mM HEPES pH 7 . 5 , 23% PEG 3350 , 5 mM TCEP ) using NVH oil ( Warkentin and Thorne , 2009 ) , and placed directly into the nitrogen-gas cryostream pre-set to the desired temperature at the beamline . Water inside protein crystals is nanoconfined so that ice nucleation is dramatically suppressed , but water outside crystallizes readily and rapidly . Careful removal of all excess solvent from the crystal surface is essential to obtaining ice-free diffraction between 260 K and 180 K without using large cryoprotectant concentrations . For the XFEL experiment , we collected multiple diffraction images per crystal using a 10-µm X-ray beam with each irradiation point separated by at least 25–40 µm to avoid collateral radiation damage . A total of 1 , 239 still diffraction images were collected from 71 CypA crystals over the course of two experiments using a goniometer setup and a Rayonix MX325HE detector at LCLS-XPP ( Cohen et al . , 2014 ) ( Video 2 ) . All data were collected at ambient temperature ( approximately 273 K ) . To prevent dehydration , crystals were coated with paratone oil immediately after looping and mounted on the goniometer at the XPP endstation of LCLS using the SAM sample exchange robot ( Cohen et al . , 2002 ) . For the new 1 . 2 Å room-temperature synchrotron dataset , paratone oil was applied to cover a 2 μL hanging drop containing a single large crystal of CypA . The crystal was harvested through the paratone and excess mother liquor was removed using a fine paper wick . Attenuated data were collected at SSRL beamline 11-1 at 273 K controlled by the cryojet on the PILATUS 6M PAD detector . The synchrotron datasets were indexed , integrated , and scaled using XDS and XSCALE , and intensities were subsequently converted to structure factor amplitudes using XDSCONV . All datasets were from single crystals . Data reduction statistics for the highest resolution room-temperature dataset and the multitemperature datasets can be found in Tables 1 , 4 respectively . The XFEL data were processed using cctbx . xfel ( Hattne et al . , 2014 ) . Of the 1 , 239 images collected , 772 were indexed and their intensities were integrated . Post-refinement , as implemented by PRIME ( post-refinement and merging , version date: November 11 , 20:22:51 2014 ) ( Uervirojnangkoorn et al . , 2015 ) , was used to correct the intensity measurements and merge the data . We optimized over the uc_tolerance , n-postref_cycle , sigma_min , partiality_min , and gamma_e values to obtain the final structure factor amplitudes . Data reduction statistics for the XFEL data are provided in Table 3 . To promote consistency between models derived from different datasets , Rfree flags were generated using PHENIX for the highest resolution ‘reference’ ( 1 . 2 Å , 273 K ) dataset first and were subsequently copied to all other multitemperature and XFEL datasets for the automated molecular replacement and refinement pipeline . For each dataset , we calculated initial phases by performing molecular replacement with phenix . auto_mr using PDB ID 2cpl as a search model . We next refined XYZs and ADPs of the initial model with phenix . refine for 4 macrocycles with XYZ and ADP weight optimization turned on; identified translation libration screw ( TLS ) groups with phenix . find_tls_groups; and refined optimized XYZs , ADPs , and TLS parameters for six more macrocycles . These single-conformer models and associated electron density maps were used as input for two subsequent steps . First , the single-conformer models were analyzed with Ringer ( Lang et al . , 2010 ) via mmtbx . ringer using default settings . A coupled side chain–backbone ‘backrub’ motion ( Davis et al . , 2006 ) of −10° for Ser99 ( see Figure 5A ) was necessary to match the Cα and Cβ positions of the minor conformation as modeled in PDB ID 3k0n; using this modified backbone indeed yielded maximal minor-conformation Ringer peaks for our multitemperature datasets . No backrub motion was necessary for Leu98 due to the different type of backbone displacement ( Fraser et al . , 2009 ) . Correlation coefficients between pairs of Ringer electron density versus dihedral angle curves were calculated using the cor function in R ( Team , 2014 ) . Second , the single-conformer models were used as input to qFit ( Keedy et al . , 2015; van den Bedem et al . , 2009 ) . Subsequent to the automated model building , we manually deleted ill-fitting waters and edited alternative protein side chain conformations based on fit to the electron density in Coot ( Emsley et al . , 2010 ) and refinement with phenix . refine . For example , at 240 K , qFit automatically modeled Phe113 as single-state , but significant mFo-DFc peaks remained , so we decided on a two-state model . Met61 was particularly difficult to model accurately across temperatures due to convolved issues of χ3 non-rotamericity for Met in general ( Butterfoss et al . , 2005 ) , the relatively high electron count for sulfur , and likely temperature-modulated Met-specific radiation damage . For these reasons , visual inspection of the maps and manual building is currently essential for alternative backbone conformations with moderate displacements , as observed in residues 79–83 ( Figure 4 ) . We are currently developing new methods to automatically detect and model such backbone excursions in multiscale multiconformer models . These efforts improved Rfree and MolProbity scores across datasets ( Table 2 ) . Because of the lower resolution , the XFEL model was refined with three TLS groups and with optimization of X-ray versus geometry and ADP weights . For minor-state electron density sums , 2mFo-DFc ( Fc filled ) map values were summed across a grid of points defined by superimposing each model onto PDB ID 3k0n using all Cα atoms , summing the 2mFo-DFc value at each point with 0 . 25 Å of a target minor-state heavy atom ( Oγ for Ser99; Cδ1 , Cε1 , Cε2 , or Cζ for Phe113 ) , and normalizing to unity across datasets for each residue being analyzed . This procedure allowed a strictly common reference set of map evaluation points . Results were very similar when using unfilled maps ( data not shown ) . We calculated B-factor-influenced order parameters ( S2 ) as previously reported ( Fenwick et al . , 2014 ) except that we monitored one of two different types of bond vector . For the χ1 order parameter , we used Cβ-Xβ ( where X = C or O ) for most amino acids , Cα-Cβ for Ala , and Cα-Hα for Gly . For the side chain-end order parameter , we used the heavy-atom to heavy-atom bond vector for each amino acid that was closest to the side chain terminus , with ties broken by the number in the atom name ( e . g . Cγ-Cδ1 instead of Cγ-Cδ2 for Leu ) . All negative order parameters ( caused by high B-factors ) were floored to 0 . χ1 order parameters were floored for 7 residues , and side chain-end order parameters were floored for 23 residues . Per-residue ‘apparent dynamic transition temperatures’ were then calculated as the intersection between the <200 K and >200 K fit lines in order parameter versus temperature plots and floored to 0 K if necessary . The kernel density curve was fit with the density function in R ( Team , 2014 ) . For extrapolation of fit lines in Figure 6 , we used a fit to all data points or to just the high-temperature data points ( ≥240 K ) for each residue , and extrapolated to the temperature at which order would be maximized ( 1 – S2 = 0 ) . To predict global behavior , at each temperature we averaged across all residues the predicted 1 – S2 values from the fit , making sure to floor non-physical predicted values of 1 – S2 < 0 to 0 , as in ( Lee and Wand , 2001 ) .	Proteins are the workhorses of the cell . The shape that a protein molecule adopts enables it to carry out its role . However , a protein’s shape , or 'conformation' , is not static . Instead , a protein can shift between different conformations . This is particularly true for enzymes – the proteins that catalyze chemical reactions . The region of an enzyme where the chemical reaction happens , known as the active site , often has to change its conformation to allow catalysis to proceed . Changes in temperature can also make a protein shift between alternative conformations . Understanding how a protein shifts between conformations gives insight into how it works . A common method for studying protein conformation is X-ray crystallography . This technique uses a beam of X-rays to figure out where the atoms of the protein are inside a crystal made of millions of copies of that protein . At room temperature or biological temperature , X-rays can rapidly damage the protein . Because of this , most crystal structures are determined at very low temperatures to minimize damage . But cooling to low temperatures changes the conformations that the protein adopts , and usually causes fewer conformations to be present . Keedy , Kenner , Warkentin , Woldeyes et al . have used X-ray crystallography from a very low temperature ( -173°C or 100 K ) to above room temperature ( up to 27°C or 300 K ) to explore the alternative conformations of an enzyme called cyclophilin A . These alternative conformations include those that have previously been linked to this enzyme’s activity . Starting at a low temperature , parts of the enzyme were seen to shift from having a single conformation to many conformations above a threshold temperature . Unexpectedly , different parts of the enzyme have different threshold temperatures , suggesting that there isn’t a single transition across the whole protein . Instead , it appears the way a protein’s conformation changes in response to temperature is more complex than was previously realized . This result suggests that conformations in different parts of a protein are coupled to each other in complex ways . Keedy , Kenner , Warkentin , Woldeyes et al . then performed X-ray crystallography at room temperature using an X-ray free-electron laser ( XFEL ) . This technique can capture the protein’s structure before radiation damage occurs , and confirmed that the alternative conformations observed were not affected by radiation damage . The combination of X-ray crystallography at multiple temperatures , new analysis methods for identifying and measuring alternative conformations , and XFEL crystallography should help future studies to characterize conformational changes in other proteins .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics" ]	2015	Mapping the conformational landscape of a dynamic enzyme by multitemperature and XFEL crystallography
Human papillomaviruses ( HPVs ) are the most common sexually transmitted infectious agents . Because of the species specificity of HPVs , study of their natural transmission in laboratory animals is not possible . The papillomavirus , MmuPV1 , which infects laboratory mice ( Mus musculus ) , can cause infections in the female cervicovaginal epithelium of immunocompetent mice that progress to cancer . Here , we provide evidence that MmuPV1 is sexually transmitted in unmanipulated , immunocompetent male and female mice . Female 'donor' mice experimentally infected with MmuPV1 in their lower reproductive tract were housed with unmanipulated male mice . The male mice were then transferred to cages holding 'recipient' female mice . One third of the female recipient mice acquired cervicovaginal infections . Prolonged infections were verified by histopathology and in situ hybridization analyses of both male and recipient female mice at the study endpoint . These findings indicate that MmuPV1 is a new model animal papillomavirus with which to study sexually transmission of papillomaviruses . Human papillomaviruses ( HPVs ) are the most common sexually transmitted infection in the United States ( Satterwhite et al . , 2013 ) , and oncogenic high-risk HPVs alone account for approximately 5% of cancers worldwide ( zur Hausen , 2009 ) . Despite the significant public health burden caused by these small DNA tumor viruses , research on papillomavirus sexual transmission has been severely limited due to the paucity of small animal preclinical models resulting from strict virus species-specific tropism . Mucosal tissues of the female reproductive tracts of rabbits ( Harvey et al . , 1998 ) , multimammate rats ( Nafz et al . , 2007; Nafz et al . , 2008 ) , and rhesus macaques ( Wood et al . , 2007 ) are susceptible to rabbit oral papillomavirus ( ROPV ) , mastomys coucha papillomavirus 2 ( McPv2 ) , and rhesus macaques papillomavirus ( RhPV ) , respectively . However , natural transmission of a papillomavirus through sexual contact has only been reported in rhesus macaques ( Ostrow et al . , 1990 ) . Unfortunately , cost and lack of suitable molecular biology tools and reagents for these models have generally deterred their broad scale use ( Christensen et al . , 2017; Uberoi and Lambert , 2017 ) . The recent discovery of a murine papillomavirus ( MmuPV1 or MusPV1 ) ( Ingle et al . , 2011 ) alleviates many of these limitations , allowing the study of papillomavirus infection and disease in laboratory mice , which are tractable , genetically modifiable , and relatively affordable . MmuPV1 infects and causes disease at both cutaneous and mucosal sites of several common strains of laboratory mice ( Uberoi and Lambert , 2017; Cladel et al . , 2015; Cladel et al . , 2017a; Cladel et al . , 2013; Handisurya et al . , 2014; Handisurya et al . , 2013; Hu et al . , 2015; Jiang et al . , 2017; Joh et al . , 2012; Sundberg et al . , 2014; Wang et al . , 2015 ) . We recently published that MmuPV1 infection of the female reproductive tract causes neoplastic disease in immunocompetent FVB/N mice ( Spurgeon et al . , 2019 ) . The severity of disease is exacerbated by treatment with estrogen ( E2 ) alone or in combination with ultraviolet B radiation ( UVB ) , which induces prolonged systemic immunosuppression ( Uberoi et al . , 2016 ) , leading to precancerous lesions and squamous cell carcinoma ( SCC ) . Here , we describe application of our MmuPV1 infection cervicovaginal model to study MmuPV1 sexual transmission . We report natural papillomavirus sexual transmission in immunocompetent , unmanipulated male and female mice . By 4 months following experimental infection with MmuPV1 in their lower reproductive tract and treatment with E2 and UVB , immunocompetent FVB/N female mice develop high-grade precancerous cervicovaginal lesions and SCCs ( Spurgeon et al . , 2019 ) . These lesions were associated with highly productive MmuPV1 infections throughout the cervicovaginal epithelia , as evidenced by strongly positive immunohistochemical staining for the major viral capsid protein L1 within the female reproductive tract ( Spurgeon et al . , 2019 ) ( see also Figure 1A ) . This observation prompted us to test whether MmuPV1 can be sexually transmitted . Cohorts of female mice ( referred to as ‘Donors’ ) that were either mock-infected or experimentally infected with MmuPV1+UV+E2 were held for 4 months ( Figure 1B ) . The female Donors were then used to establish monogamous breeding pairs with uninfected male mice ( referred to as ‘male Breeders’ ) and breeding allowed for at least 3 weeks . Male Breeders were then transferred into a cage with an uninfected female mouse ( referred to as a ‘Recipient’ ) for at least 3 weeks . While the female Donors were treated with medroxyprogesterone acetate ( Depo-Provera ) and nonoxynol-9 to potentiate MmuPV1 infection ( Spurgeon et al . , 2019; Roberts et al . , 2007 ) , it is important to emphasize that none of the male Breeders were experimentally manipulated prior to or during matings and the female Recipients were not experimentally manipulated unless indicated below . We performed four separate transmission experiments summarized in Figure 1B and Table 1 using various conditions . In Experiments 1 and 2 , breeding occurred for 3 weeks with both the Donor and Recipient , and Recipient female mice were treated with E2 for 2 months starting at 8 weeks post-breeding . In Experiment 3 , a fraction of Recipients ( n = 4 ) were pretreated with Depo-provera 5 days prior to breeding , and in Experiment 4 , male Breeders remained with Donors and Recipients for 8 weeks each instead of 3 weeks . For Experiment 4 , the Donors from Experiment 3 were used as the source of MmuPV1 . All experiments were conducted with wild-type FVB/N mice , totaling 9 mock-infected and 22 MmuPV1 Donor-positive breeding pairs . Prior to housing with male Breeders , we first assessed whether the female Donors harbored infections in their reproductive tracts by performing cervicovaginal lavage ( CVL ) . DNA recovered from the CVLs were subjected to PCR to detect MmuPV1 DNA and the host gene , p53 , as a positive control ( Figure 1C ) . All female Donors were found to have MmuPV1 infections based upon the CVL/PCR tests . This confirmed our previously published results that MmuPV1+UV+E2-infected mice efficiently establish infections that persist for at least 4 months ( Spurgeon et al . , 2019 ) . Indeed the infections of these female Donors persisted for up to 10 months post-infection ( Figure 1C ) . To monitor for evidence for sexual transmission , we monitored the MmuPV1 infection status of the reproductive tracts of the female Recipient mice . CVL/PCR was performed on these mice starting approximately 3 weeks following introduction of the male Breeder mouse and approximately every month thereafter ( Figure 2A ) . All female Recipient mice whose matings resulted in pregnancy were allowed to deliver offspring prior to their first CVL/PCR screen . Using this screening method , we identified 32% ( n = 7/22 ) of female Recipient mice to harbor infections within their reproductive tracts ( Figure 2B ) . These infections were observed across all four experiments ( Table 1 ) . Of the MmuPV1-positive female Recipient mice , 57% ( n = 4/7 ) established prolonged MmuPV1 infections ( MmuPV1 positive for at least 2 CVLs ) while 43% ( n = 3/7 ) had transient infections ( MmuPV1 positive for only one CVL ) ( Figure 2B , Table 1 ) . Prolonging the exposure of male Breeders to both the female Donors and female Recipients from 3 weeks to 8 weeks in Experiment 4 resulted in a higher percentage of MmuPV1-positive Recipients ( 50%; n = 3/6 ) than observed in Experiments 1 ( 33%; n = 1/3 ) , 2 ( 33%; n = 2/6 ) , or 3 and 3* ( 14%; n = 1/7 ) . Preconditioning female recipient mice with Depo-Provera , a contraceptive drug , did not appear to influence susceptibility of mice to MmuPV1 infection ( 25% MmuPV1 positive: n = 1/4 ) . In a separate set of experiments , we determined that co-habitation of female mice experimentally infected with MmuPV1 in their reproductive tracts with uninfected female mice did not lead to transmission of MmuPV1 infections to the reproductive tracts of the latter mice based on negative CVL/PCR results ( Figure 2C ) , consistent with the premise that the infections arising in reproductive tracts of Recipient females housed with male Breeders ( Figure 2A and B ) resulted from sexual activity between the males and females . One obvious positive readout for sexual activity is pregnancy . Of the 7 Recipient females that acquired MmuPV1 infections in their reproductive tract , 6 became pregnant during the course of being housed with the male breeders ( Table 1 ) . The single Recipient female that did not become pregnant had been pre-treated with the contraceptive Depo-Provera ( Mouse #16 in Experiment 3* ) . Pregnancy in the Donor females was less penetrant with 4 out of the 7 Donor female mice initially housed with these same male Breeders having become pregnant ( Table 1 ) . This lower penetrance likely reflects that the Donor Females were originally experimentally infected with MmuPV1 using a protocol that involves treatment with Depo-Provera , which can prevent estrus cycling for an extended period of time . We were interested in learning whether pups born to infected mums would acquire MmuPV1 infections . While our analysis was not exhaustive , we did not find evidence for MmuPV1 infections in the skin of several offspring of MmuPV1-positive Donor Females that we screened for E4 mRNA using in situ hybridization . To confirm that the MmuPV1-positive PCR results from the CVLs reflect persistent infections of the cervical/vaginal epithelium , we performed endpoint histopathological and MmuPV1-specific in situ hybridization ( RNAscope ) analyses on the reproductive tract of a female Recipient mouse ( Recipient #3 ) , which was MmuPV1-positive at the endpoint by CVL/PCR ( Figure 2A and D ) . RNAscope used probes to detect viral transcripts containing the E4 region because that region is present in most early and late transcripts ( Xue et al . , 2017 ) . Several discrete regions of epithelia were positive for MmuPV1 viral transcripts . These regions correlated with histopathological signs of MmuPV1 infection ( Spurgeon et al . , 2019 ) , including disorganization of the stratified epithelium , areas of hyperkeratinization , karyomegaly , perinuclear halos similar to koilocytes , and condensed chromatin . We also observed evidence for a productive viral infection as indicated by cells staining positively for the viral capsid protein L1 by immunofluorescence ( Figure 2D ) , albeit at levels of detection that are much lower than that afforded by RNAscope-based detection of viral transcripts . The infected regions of epithelia were pathologically scored as having low-grade or mild dysplasia . This particular mouse was MmuPV1-positive by 6 weeks post-breeding , and treated for 2 months with estrogen starting at 8 weeks post-breeding . Our previous results indicate that neoplastic disease worsens in MmuPV1 and MmuPV1+E2-infected mice upon extended duration ( 4 or 6 months ) ( Spurgeon et al . , 2019 ) . It is therefore possible that MmuPV1-infected female Recipients may develop moderate to high-grade disease or even SCC if the infection is allowed to proceed for a longer period of time . We analyzed the reproductive tracts of additional Recipient female mice that were positive for MmuPV1 by CVL/PCR at the endpoint and found them to have sites of infections based upon MmuPV1 E4-specific in situ hybridization ( data not shown ) . These results confirm that sexual transmission of MmuPV1 can lead to persistent infections in the absence of genetic or environmental manipulation . Because many female Recipient mice contracted MmuPV1 infections of their reproductive organs after being housed with male Breeders ( Figure 2 ) , we evaluated the reproductive organs of the male Breeders for the presence of MmuPV1 . Attempts to detect the MmuPV1 by lavage of the male genitalia were not successful ( insufficient DNA was retrieved based upon an inability to detect mouse p53 DNA by PCR; data not shown ) . Therefore we resorted to in situ hybridization analysis of male reproductive organs obtained at the time of euthanasia . We identified several male Breeders with MmuPV1-positive foci of infection by RNAscope ( Figure 3A ) . All foci of infections were detected in epithelia of the penis , including the glans epithelium , mump ridge groove , and prepuce ( foreskin ) /preputial space ( Phillips et al . , 2015; Rodriguez et al . , 2011 ) ( Figure 3B and C ) . Notably , many of these sites are anatomical locations infected by HPV in men ( Giuliano et al . , 2007 ) . We also observed evidence for productive viral infections in the penis using L1 immunofluorescence ( Figure 3B ) . Similar to our observation in the MmuPV1-positive Recipient females ( Figure 2D ) , L1 was detected albeit at a lower level than that observed for E4 transcripts , which uses the sensitive RNAscope technology . Two of the male Breeders ( Male #3 and Male #5 ) that had detectable MmuPV1 infections on their penis by the endpoint RNAscope analysis were associated with MmuPV1-positive female Recipients ( Table 1 ) . The other two males that had detectable MmuPV1 infections on their penis did not appear to have transmitted MmuPV1 to Recipient females , based upon CVL/PCR . It remains possible that they did transmit , but the infections in the female Recipients were transient in nature and not caught by the intermittent CVL/PCR tests . Other males that were negative for MmuPV1-infections based upon endpoint RNAscope analysis did transmit MmuPV1 to female Recipients , suggesting they either had transient infections or else their foci of persistent infections were missed by the RNAscope analysis , which is very possible as only one section per male mouse was subjected to in situ hybridization . Of the 4 MmuPV1-positive male Breeders we identified using in situ hybridization , only 1 Donor-Breeder mating resulted in pregnancy , whereas all 4 Breeder-Recipient matings resulted in pregnancies ( Table 1 ) . The data presented in this study provide strong evidence that MmuPV1 is sexually transmitted . MmuPV1 becomes the first model for studying sexual transmission of papillomaviruses in laboratory mice ( Mus musculus ) . Our immediate goals are to use this natural sexual transmission model in immunocompetent mice to study the dynamics of sexual transmission , the role of host immunity , and methods for prevention and treatment . Immunocompetent , wild-type FVB/N mice ( Taconic Biosciences; Albany , NY ) mice were used in this study . All animal experiments were performed in full compliance with standards outlined in the ‘Guide for the Care and Use of Laboratory Animals’ by the Laboratory Animal Resources ( LAR ) as specified by the Animal Welfare Act ( AWA ) and Office of Laboratory Animal Welfare ( OLAW ) and approved by the Governing Board of the National Research Council ( NRC ) . Mice were housed at McArdle Laboratory Animal Care Unit in strict accordance with guidelines approved by the Association for Assessment of Laboratory Animal Care ( AALAC ) , at the University of Wisconsin Medical School . All protocols for animal work were approved by the University of Wisconsin Medical School Institutional Animal Care and Use Committee ( IACUC , Protocol number: M005871 ) . At 6–8 weeks of age , female virgin FVB/N mice were infected with MmuPV1 virus as described previously ( Spurgeon et al . , 2019 ) . Briefly , mice were injected subcutaneously with 3 mg medroxyprogesterone acetate ( Amphastar Pharmaceuticals , Rancho Cucamongo , CA ) 4–7 days prior to MmuPV1 infection to induce diestrus . On the day of the infection , mice were pre-treated vaginally with 50 µL Conceptrol ( Options , #247149 ) containing 4% nonoxynol-9 to induce chemical injury to the cervicovaginal epithelium . At 4 hr post-treatment with Contraceptrol , mice were infected intravaginally with 108 VGE ( viral genome equivalents ) MmuPV1 virions suspended in 25 µL 4% carboxyl methylcellulose ( Sigma , #C4888 ) . The MmuPV1 virus stock used for infection was generated by isolating virions from papillomas arising on infected FoxN1nu/nu mice . To treat mice with estrogen , a continuous-release estrogen ( E2 ) tablet ( 17β-estradiol; 0 . 05 mg/60 days; Innovative Research of America , Sarasota , FL ) was inserted subcutaneously in the shoulder fat pads of the dorsal skin . For those mice receiving estrogen , treatment began 5 days following MmuPV1 infection . A new tablet was inserted every 2 months as needed . Infection and estrogen treatment were performed while mice were anesthetized with 5% isoflurane . Animals were exposed to a single dose of UVB at 1000 mJ/cm2 . UVB was administered using a custom designed Research Irradiation Unit ( Daavlin , Bryan , OH ) with lamps controlled using Daavlin Flex Control Integrating Dosimeters . Donors were lavaged prior to breeding to confirm infection , and then introduced to male FVB/N to establish monogamous breeding pairs . After breeding with infected female Donor mice , the males were isolated for 2–5 days , then introduced to uninfected female Recipient mice . In all experiments except Experiment 3 , female recipient mice were 6–8 weeks old virgin mice that were not treated with depoprovera or nonoxynol-9 . In Experiment 3 , female Recipient mice were pre-treated with depoprovera 5 days prior to introduction of Male Breeder . Male breeding with female Recipient mice was allowed to proceed for 3 weeks in Experiments 1–3 . In Experiment 4 , breeding with female Donors and Recipients was allowed to proceed for 8 weeks . The method for detecting MmuPV1 DNA by PCR in vaginal lavages was modified from that described in Hu et . al . and Cladel et . al . ( Hu et al . , 2015; Cladel et al . , 2017b ) . Briefly , 25 µL of sterile PBS was introduced intravaginally with a pipette tip , triturating 4–5 times prior to retrieval using the pipetteman . The vaginal lavages were stored at −20° C and DNA isolated using spin-columns ( DNeasy Blood and Tissue Kit; Qiagen #69506 , Hilden , Germany ) . Eluted DNA was analyzed by PCR using primers specific to the MmuPV1 E2 gene: MmuPV1_E2_1 ( 5’-GCCCGAAGACAACACCGCCACG-3’ ) and MmuPV1_E2_2 ( 5’-CCTCCGCCTCGTCCCCAAATGG-3’ ) and analyzed using agarose gel electrophoresis . The presence of DNA suitable for PCR amplification was verified by performing PCR for the p53 gene . The primers for the p53 gene were as follows: p53-1 ( 5′-TATACTCAGAGCCGGCCT-3′ ) , p53-2 ( 5′-ACAGCGTGGTGGTACCTTAT-3′ ) , and p53-3 ( 5′-TCCTCGTGCTTTACGGTATC-3′ ) . Reproductive organs were harvested , fixed in 4% paraformaldehyde , and paraffin-embedded . Serial sections ( 5 μm ) were cut and every 10th section was stained with H and E and evaluated by histopathology . The scoring system for worst stage of disease has been described previously ( Spurgeon et al . , 2019 ) . A detailed protocol for detecting MmuPV1 L1 using a tyramide-based signal amplification ( TSA ) method is available at: dx . doi . org/10 . 17504/protocols . io . i8cchsw . MmuPV1 viral transcripts were detected using RNAscope 2 . 5 HD Assay-Brown ( Advanced Cell Diagnostics , Newark , CA ) according to manufacturer instructions with probes specific for MmuPV1 E4 ( Catalog #473281 ) as described previously ( Spurgeon et al . , 2019; Xue et al . , 2017 ) . Tissue sections were treated following protease treatment and prior to probe hybridization with 20 units of DNase I ( Thermo Fisher Scientific , #EN0521 ) , or DNase I combined with 500 ug RNase A ( Qiagen , #1006657 ) plus 2000 units RNase T1 ( Fermentas , Waltham , MA , #EN0542 ) for 30 min at 40°C . Slides were counterstained with hematoxylin before mounting and coverslipping . High resolution wide-field fluorescent images were acquired using Leica SP8 3X STED microscope ( Xue et al . , 2017 ) by means of a 20X objective lens ( Specifications: HC PL APO 20x/0 . 75 CS2 , Dry ) LAS-X suite ( version: 2 . 0 . 1 ) . Full slide scans of tissues were performed using Aperio ScanScope XT Slide Scanner using 20x/0 . 75 Plan Apo objective . All other images were captured using a Zeiss AxioImager M2 microscope and AxioVision software version 4 . 8 . 2 ( Jena , Germany ) .	Human papillomaviruses are responsible for about 5% of all cancers: infections with high-risk strains of the virus lead to the vast majority of cervical cancers , other cancers of the anus and genital area , as well as a growing fraction of head and neck cancers . In humans , these viruses are transmitted through sexual contact . Other animals do not get infected by human papillomaviruses , and this makes it difficult to study in the laboratory how these viruses pass from one individual to another . However , a mouse papillomavirus has recently been identified: known as MmuPV1 , it also causes cervical cancer in rodents , but it was unknown whether it was transmitted sexually . To investigate this question , Spurgeon and Lambert experimentally infected female mice with MmuPV1 and allowed them to have intercourse with healthy males . When the males then were mated to healthy females , approximately a third of these female mice became infected with MmuPV1 . Males that transmitted the virus were also found to have penile infections . These results show that , like the human papillomavirus , MmuPV1 spreads through sexual interactions . Knowledge gathered by studying MmuPV1 could help to understand sexually transmitted human papillomaviruses that cause cancer . Additional work could look into how the virus leads to cancer and investigate the viral and host factors that contribute to sexual transmission . Further studies may also focus on testing drugs that prevent transmission or eliminate the persistent infections that can lead to cancer .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "short", "report", "microbiology", "and", "infectious", "disease" ]	2019	Sexual transmission of murine papillomavirus (MmuPV1) in Mus musculus
To move the body , the brain must precisely coordinate patterns of activity among diverse populations of motor neurons . Here , we use in vivo calcium imaging , electrophysiology , and behavior to understand how genetically-identified motor neurons control flexion of the fruit fly tibia . We find that leg motor neurons exhibit a coordinated gradient of anatomical , physiological , and functional properties . Large , fast motor neurons control high force , ballistic movements while small , slow motor neurons control low force , postural movements . Intermediate neurons fall between these two extremes . This hierarchical organization resembles the size principle , first proposed as a mechanism for establishing recruitment order among vertebrate motor neurons . Recordings in behaving flies confirmed that motor neurons are typically recruited in order from slow to fast . However , we also find that fast , intermediate , and slow motor neurons receive distinct proprioceptive feedback signals , suggesting that the size principle is not the only mechanism that dictates motor neuron recruitment . Overall , this work reveals the functional organization of the fly leg motor system and establishes Drosophila as a tractable system for investigating neural mechanisms of limb motor control . Dexterous motor behaviors require precise neural control of muscle contraction to coordinate force production and timing across dozens to hundreds of muscles . This coordination is mediated by populations of motor neurons , which translate commands from the central nervous system into dynamic patterns of muscle contraction . Although motor neurons are the final common output of the brain , the scale and complexity of many motor systems have made it challenging to understand how motor neuron populations collectively control muscles and thus generate behavior . For example , a human leg is innervated by over 150 , 000 motor neurons and a single calf muscle is innervated by over 600 motor neurons ( Kernell , 2006 ) . How can the nervous system coordinate the activity of such large motor neuron populations to flexibly control the force , speed and precision of limb movements ? One way to streamline motor control is to establish a hierarchy among neurons controlling a particular movement , such as flexion of a joint . This hierarchy allows premotor circuits to excite different numbers of motor neurons depending on the required force: motor neurons controlling slow or weak movements are recruited first , followed by motor neurons that control progressively stronger , faster movements ( Denny-Brown , 1929 ) . A recruitment order for vertebrate motor neurons innervating a single muscle was first postulated over 60 years ago ( Henneman , 1957 ) . Subsequent work identified mechanisms associated with the recruitment order and synthesized these findings as the size principle , which states that small motor neurons , with lower spike thresholds , are recruited prior to larger neurons , which have higher spike thresholds ( Henneman et al . , 1965a; Henneman et al . , 1965b; Henneman and Olson , 1965; Mcphedran et al . , 1965; Mendell , 2005; McPhedran et al . , 1965 ) . Evidence for the size principle has been provided by electrophysiological analysis of motor neurons in a number of species , from crayfish to humans ( Gabriel et al . , 2003; Hill and Cattaert , 2008; McLean and Dougherty , 2015; Milner-Brown et al . , 1973; Sasaki and Burrows , 1998 ) . These studies have also described systematic relationships between motor neuron electrical excitability , recruitment order , conduction velocity , force production , and strength of sensory feedback and descending input ( Bawa et al . , 1984; Binder et al . , 1983; Fleshman et al . , 1981; Heckman and Binder , 1988; Kernell and Sjöholm , 1975; Zengel et al . , 1985 ) . A simplifying assumption of the size principle is that all motor neurons within a pool receive identical presynaptic input , such that recruitment order is entirely dictated by the gradient of motor neuron excitability . However , recordings in vertebrates have found that presynaptic inputs may vary within a motor pool ( McLean and Dougherty , 2015 ) and that recruitment order can change during specific movements ( Desnedt and Gidaux , 1981; Kishore et al . , 2014; Menelaou and McLean , 2012; Smith et al . , 1980 ) . Are these violations of the size principle a fundamental feature of motor control circuits or rare exceptions to the rule ? Answering this question would be greatly aided by a tractable system in which it was possible to genetically target identified motor neurons for recording , manipulation , and mapping of presynaptic inputs . The leg of the fruit fly , Drosophila melanogaster , contains 14 muscles which are innervated by just 53 motor neurons ( Baek and Mann , 2009; Brierley et al . , 2012; Maniates-Selvin et al . , 2020; Soler et al . , 2004 ) . In spite of this tiny scale , the fly leg supports a variety of fast and flexible behaviors . During forward walking , the fly grips the substrate with the distal segment of its front leg ( the tarsus ) and flexes the femur-tibia joint , pulling the body forward ( Video 4 ) . The femur-tibia joint of a walking fly flexes and extends 10–20 times per second , reaching swing speeds of several thousand degrees per second ( DeAngelis et al . , 2019; Gowda et al . , 2018; Mendes et al . , 2013; Strauss and Heisenberg , 1990; Wosnitza et al . , 2013 ) . Flies also use their legs to target other body parts during grooming ( Hampel et al . , 2015; Seeds et al . , 2014 ) , for social behaviors like aggression and courtship ( Clowney et al . , 2015; Hoopfer et al . , 2015 ) , and to initiate flight take-off ( Card and Dickinson , 2008; Zumstein et al . , 2004 ) and landing ( Ache et al . , 2019 ) . These behaviors require a wide range of muscle force production across multiple timescales . However , little is currently known about the organization and function of leg motor control circuits in the fly . While a great deal of progress has been made on understanding the processing of sensory signals in the Drosophila brain , we lack an understanding of how this information is translated into behavior by the fly’s ventral nerve cord ( VNC ) . Investigating motor control in Drosophila is important because the fly’s compact nervous system and identified cell types make it a tractable system for comprehensive circuit analysis ( Tuthill and Wilson , 2016a ) . In this study , we investigate motor control of the Drosophila tibia . We first mapped the organization of tibia flexor motor units using calcium imaging from leg muscles in behaving flies ( Figure 1 ) . With electrophysiology , we then discovered that motor neurons controlling tibia flexion ( Figure 2 ) lie along a gradient of anatomical and physiological properties ( Figure 3 ) that correlate with muscle force production ( Figure 4 ) . Slow motor neurons produce <0 . 1 µN per spike , while fast motor neurons produce ~10 µN per spike , approximately equal to the fly’s weight . Recordings during spontaneous leg movements revealed a recruitment hierarchy: slow motor neurons typically fire first , followed by intermediate , then fast neurons ( Figure 5 ) . Interestingly , all tibia flexor motor neurons receive feedback from proprioceptors at the femur/tibia joint , but these sensory signals vary in amplitude , sign , and dynamics across the different motor neuron types ( Figure 6 ) . Optogenetic manipulation of each motor neuron type had unique and specific effects on the behavior of walking flies ( Figure 7 ) , consistent with their roles in controlling distinct force regimes . Together , these data establish the organization and function of a key motor control module for the fly leg . Overall , we found that motor neurons controlling the fly tibia exhibit many features consistent with the size principle . However , we also observed that tibia motor neurons receive distinct proprioceptive feedback signals and that recruitment order is occasionally violated . Thus , in addition to the size principle , heterogeneous input from premotor circuits is likely to play an important role in coordinating neural activity within a motor pool . We first sought to understand how muscles in the fly femur control movement of the tibia ( Figure 1 ) . Fly leg muscles are each composed of multiple fibers ( Soler et al . , 2004 ) and innervated by distinct motor neurons ( Baek and Mann , 2009; Brierley et al . , 2012 ) . A motor neuron and the muscle fibers it innervates are referred to as a motor unit . In most invertebrate species , multiple motor neurons can innervate the same muscle fiber , so motor units may be overlapping ( Hoyle , 1983 ) . We used widefield fluorescence imaging of muscle calcium signals with GCamp6f ( Chen et al . , 2013; Lindsay et al . , 2017 ) to map the spatial organization of motor units in the fly femur during spontaneous tibia movements ( Figure 1C , Figure 1—figure supplement 1 , and Video 1 ) . We observed that spatial patterns of calcium activity were different for different movements ( e . g . , tibia flexion vs . extension ) . To quantify this spatial organization , we performed unsupervised clustering so that pixels with correlated activity were grouped together into activity clusters . We assigned arbitrary numbers to these clusters , from ventral to dorsal ( Figure 1—figure supplement 1 ) . Three clusters significantly increased their fluorescence when the leg was flexed ( Figure 1C ) ; we refer to these clusters as Flexors 1 , 2 , and 3 . Flexors 2 and 3 were active during most periods of tibia flexion , whereas Flexor 1 was only active only during large , fast movements ( Figure 1E ) . This organization was similar across flies and robust to changes in clustering parameters ( Figure 1—figure supplement 1 ) . Likely because of the orientation of the leg in these experiments , we did not observe any clusters whose activity increased specifically during tibia extension ( see Figure 1—figure supplement 1 and Materials and Methods for further discussion ) . For the remainder of this study , we focus our efforts on motor control of tibia flexion . For Flexors 1 and 2 , we could record and identify extracellular motor neuron spikes in the femur ( electromyogram: EMG ) , which confirmed that the activity clusters identified through calcium imaging correspond to distinct motor units ( Figure 1E and see Figure 2D , below ) . In other words , we propose that activity clusters 1 and 2 reflect the firing patterns of specific motor neurons . Below , we elaborate on this relationship further when discussing specific motor neurons . Muscles controlling the same joint may exert different levels of force , depending on fiber type composition and musculoskeletal biomechanics . To understand the relationship between motor unit activity and force production , we allowed the fly to pull on a flexible probe with its tibia ( Figure 1D ) . We calibrated the force required to deflect the probe a given distance and measured a spring constant of 0 . 22 μN/μm ( Figure 1—figure supplement 2 ) . Flies were able to move the force probe up to 400 μm , reaching speeds of 8 mm/s ( Figure 1F ) . That is , the fly was capable of producing close to 100 μN of force at the tip of the tibia and changes in force of ~1 . 3 mN/s . For comparison , the mass of the fly is ~1 mg for a weight of ~10 μN; this means that the femur-tibia joint can produce enough force to lift approximately ten times the fly’s body weight . We used the activity clusters computed from unloaded leg movements ( Figure 1C ) to examine whether different flexor motor units control different levels of tibia force production . We observed that Flexor 2 activity increased across a range of tibia velocities and forces , but Flexor 1 activity increased only during the fastest , most powerful movements . To quantify this relationship , we compared probe force and velocity when clusters were recruited either alone or together ( Figure 1G–J ) . The kinetics of GCaMP6f are slow relative to fly tibia movements , so we examined only periods when the derivative of the fluorescence signal was positive , i . e . muscle contraction was increasing ( Figure 1E ) . The highest probe forces and velocities were achieved only when both Flexors 1 and 2 were active together ( Figure 1G–H ) . When Flexor 2 was active alone , probe velocities were always lower ( Figure 1J ) . Occasionally , the derivative of Flexor 1 fluorescence alone was high ( Figure 1I ) , but in these rare instances the intensity of Flexor 2 was also high ( Figure 1—figure supplement 3 ) , indicating that Flexor 2 was contracting . These results indicate that distinct motor units control distinct levels of force production , and that they are recruited in a specific order , with motor units controlling low forces firing prior to motor units controlling higher forces . The results of Figure 1 illustrate two organizational features of fly leg motor control . First , fly tibia flexion is controlled by a number of distinct motor units in the femur that are active at different levels of force production . Second , the sequential activity of tibia flexor motor units is consistent with a hierarchical recruitment order . The spatial organization of tibia flexor motor units also provides a template that can be used to identify genetic driver lines that label specific motor neurons . We visually screened a large collection of Gal4 driver lines ( Jenett et al . , 2012 ) for expression in motor neurons that innervate the activity clusters we identified through calcium imaging ( Figure 1 ) . We focus our analysis on three lines that each label a single tibia flexor motor neuron in the femur . The first line ( R81A07-Gal4 ) labels a motor neuron that innervates the high-force motor unit ( Figure 2 , left ) that we identified through calcium imaging ( Flexor 1 ) . The second ( R22A08-Gal4 ) labels a neuron that projects to the proximal femur ( Figure 2 , middle ) , one of several likely candidates that controls Flexor 2 ( Baek and Mann , 2009 ) . The muscle fibers innervated by these two neurons connect to the same tibia flexor tendon ( Soler et al . , 2004 ) . The third line ( R35C09-Gal4 ) labels a single motor neuron that projects to the distal part of the femur , where the signals in our calcium imaging experiments were weak and noisy ( Figure 2 , right ) . The muscle fibers in this distal region have been collectively referred to as the reductor muscle ( Baek and Mann , 2009; Brierley et al . , 2012; Soler et al . , 2004 ) , but their alignment and attachment points suggest they control flexion of the tibia ( see Materials and Methods for further details ) . For consistency with the literature on other insects ( Burrows , 1996 ) , and based on their functional properties described below , we refer to these three motor neurons as fast , intermediate , and slow . Previous studies of motor neuron development suggest that the fast motor neuron is a unique , embryonically born neuron that persists through metamorphosis , whereas the intermediate and slow motor neurons belong to larger cohorts born during metamorphosis ( Baek and Mann , 2009; Brierley et al . , 2012 ) . Those studies estimated that 2–5 neurons innervate the proximal muscle fibers in the femur , similar to the intermediate motor neuron , and 8–9 neurons innervate distal muscle fibers , similar to the slow motor neuron . Using in vivo whole-cell patch-clamp electrophysiology , we recorded the membrane potential of each motor neuron while the fly pulled on the force probe . The recording configuration was similar to that used for calcium imaging ( Figure 1 ) . We observed increases in the firing rate of each motor neuron type during tibia flexion ( Figure 2D ) . Simultaneous EMG recordings from motor neuron axons in the femur confirmed that the fast motor neuron corresponds to Flexor 1 , identified via calcium imaging ( Figure 1 ) . The intermediate neuron contributes to the activity of Flexor 2 , likely along with one to two other similar intermediate neurons . Perhaps because of the small size of its axon , we were unable to identify slow motor neuron spikes with extracellular recordings . The three identified motor neurons have conspicuous differences in the size of their axons , dendrites , and cell bodies ( Figure 3A ) . The fast motor neuron has an exceptionally large soma ( for a Drosophila neuron ) and thick dendritic branches . Its axon has a wide diameter and branches extensively in the femur ( Figure 2B ) . The intermediate motor neuron has a smaller cell body , thinner dendritic branches , and smaller axonal arborization in the femur . The slow motor neuron has the smallest cell body , dendrites , and axon of the three . Each motor neuron has a unique dendritic branching pattern within the VNC neuropil . For example , the intermediate neuron projects toward the midline , and the arborizations of the fast and intermediate neurons branch more extensively than the slow neuron ( Figure 2C , Figure 2—figure supplement 1 ) . Thus , each motor neuron is anatomically positioned to receive distinct presynaptic inputs . The intrinsic electrical properties of the tibia flexor motor neurons varied along a continuum ( Figure 3B–E ) . The average resting potential of fast motor neurons was lower ( −68 mV ) than that of intermediate ( −60 mV ) and slow ( −48 mV ) motor neurons ( Figure 3C ) . While the fast and intermediate neurons were silent at rest , the slow neuron had a resting spike rate of approximately 30 Hz ( Figure 3D ) . We also observed a gradient in input resistance , which we calculated from the voltage responses to small hyperpolarizing current injections before each trial ( −5 pA ) . The input resistance was 150 MΩ for fast motor neurons , 300 MΩ for intermediate motor neurons , and 700 MΩ for slow motor neurons . Current injection in the fast and intermediate neurons failed to reliably trigger action potentials , based on recordings from both the cell body ( whole-cell ) and axon ( EMG ) . This is likely because of intrinsic morphological or electrophysiological properties that electrically isolate the soma from the spike initiation zone ( Sasaki and Burrows , 1998 ) . By contrast , injecting current into the slow neuron effectively modulated the spike rate ( Figure 3B ) . Overall , our electrophysiological characterization of tibia flexor motor neurons ( Figures 2 , 3 ) revealed a systematic relationship between motor neuron anatomy , input resistance , resting membrane potential , and spontaneous firing rate . These properties accompany differences in the force generated by a spike in each neuron , which we quantify next . How much force does a single spike in a motor neuron produce ? How does force-per-spike , or gain , vary across neurons within a motor pool ? To answer these questions , we measured probe displacement as a function of firing rate in each motor neuron type . To evoke spikes in the fast and intermediate neurons , we optogenetically stimulated motor neuron dendrites in the VNC using expression of Chrimson ( Klapoetke et al . , 2014 ) , a red-shifted channelrhodopsin ( Figure 4A–B ) . Brief ( 10–20 ms ) flashes of increasing intensity produced increasing numbers of spikes , corresponding spikes in the EMG , and movement of the tibia detected by small displacements of the force probe ( Video 2 ) . Aligning probe movement to spike onset showed that a single fast motor neuron spike produced ~10 μN of force , resulting in a 50 μm movement of the force probe ( Figure 4A ) . An intermediate neuron spike produced ~1 μN that moved the tibia 5 μm ( Figure 4B ) . For comparison , during take-off , the peak force production of the fly leg is ~100 μN ( Zumstein et al . , 2004 ) . Increasing the spike rate of the slow motor neuron with current injection produced small ( ~1 µm ) and slow tibia movements ( Figure 4C ) . Decreasing slow motor neuron firing produced tibia extension ( Figure 4C , E ) , suggesting that spontaneous firing in this cell contributes to the resting force on the probe . Bath application of 1 μM MLA , an antagonist of nicotinic acetylcholine receptors , led to a decrease in the spontaneous firing rate and reduced the resting force on the probe by ~1 . 5 μN , or ~15% of the fly’s weight ( Figure 4C ) . This suggests that the spontaneous firing rate in slow motor neurons is set by excitatory synaptic input . A comparison of force production as a function of firing rate revealed that fast , intermediate and slow motor neurons occupy distinct force production regimes which span three orders of magnitude ( Figure 4D ) . The shapes of these force production curves were also distinct . For fast and intermediate neurons , the force produced by two spikes was ~1 . 6X the force produced by a single spike ( Figure 4E ) and the force-per-spike curves saturated at ~10 spikes ( Figure 4D ) . These observations suggest that the fast and intermediate muscle fibers fatigue during repeated stimulation . In contrast , the resting spike rate of slow motor neurons maintains constant force on the probe ( Figure 4C , F ) . Thus , slow motor neurons may be used to maintain body posture , while intermediate and fast motor neurons are transiently recruited to execute body movements . Consistent with this hypothesis , we did not observe that flies maintained sufficient resting force on the probe ( >5 s ) to require sustained firing of fast or intermediate neurons ( Figure 1E ) . The rates of force generation for the fast and intermediate motor units were greater than the slow motor unit . For example , the effect of a spike in the fast and intermediate motor neurons reached half maximal force in ~8 . 5 ms ( Figure 4H ) , whereas the effect of increasing the spike rate in slow motor neurons was gradual and did not reach its peak within 500 ms . Hyperpolarizing slow motor neurons required ~100 ms to reach its maximal effect on force production , which could reflect slow release of muscle tension or could be due to activity of other motor neurons in the population that influence resting force on the probe . We recorded from several other motor neurons controlling tibia flexion , none of which had more extreme properties than the fast and slow motor neurons in Figure 4 . Importantly , these cells exhibited similar relationships in morphology , intrinsic properties , and force production ( Figure 4—figure supplement 1 ) , which provides confidence that these correlations are not an artifact of the motor neurons we have chosen to highlight in this study . Calcium imaging from leg muscles suggested that distinct motor units are active during distinct tibia movement regimes , and that this activity follows a recruitment order ( Figure 1 ) . To test these hypotheses more rigorously , we examined recordings from single cells and pairs of tibia flexor motor neurons during spontaneous leg movements ( Figure 5A–C ) . The membrane potential of each tibia flexor motor neuron reflected probe movement , even during periods of purely subthreshold activity ( Figure 5A–C ) . This shows that motor neurons that are not currently spiking are receiving synaptic input that , if shared across the motor pool , could drive other motor neurons to spike ( Gabriel and Büschges , 2007 ) . This synaptic input could arise from premotor neurons or proprioceptive feedback . To understand the relationship between motor neuron spiking and tibia movement , we compared force and velocity production in the 25 ms following each motor neuron spike ( Figure 5D–E ) . Spikes in each motor neuron type tended to occur during specific regimes of force and velocity . Plots of probe position and movement during example epochs ( Figure 5D ) illustrate that fast motor neurons typically spiked when the tibia was already flexed and moving . In comparison , the only period during which slow motor neurons were silent was when the tibia was extending ( negative probe velocity ) or fully extended ( zero probe position ) . As a result , the spike-triggered distribution of probe dynamics for the slow motor neuron is broad ( Figure 5E , right ) . Large spike rate modulations in the slow motor neuron often coincided with gradual changes in the position of the probe , consistent with the slow kinetics of force production measured using current injection ( Figure 4 ) . In comparison , spikes in fast motor neurons caused the probe to rapidly accelerate and approach maximum velocity ( Figure 5D–E ) . Intermediate neuron spikes could also produce small but measurable increases in force ( Figure 5B , inset ) . Overall , this analysis shows that each motor neuron type has a different regime of force probe position and velocity in which it is likely to spike ( Figure 5—figure supplement 1 ) . These regimes are overlapping . For example , in the regime where fast motor neuron spikes are likely , both intermediate and slow spikes are likely as well . Similarly , the slow motor neuron is likely to be spiking throughout the intermediate neuron’s regime . These relationships between motor neuron spiking and force production are consistent with the existence of a hierarchical recruitment order ( Figure 5F ) . We tested this model by examining paired recordings of somatic ( whole-cell ) and axonal ( EMG ) spikes from different motor neurons . Neurons lower in the recruitment hierarchy were consistently active when a neuron producing more force was recruited ( Figure 5G ) . In other words , activity in slow neurons preceded that of intermediate neurons , and activity in intermediate neurons preceded that of fast neurons . Interestingly , we sometimes observed violations of this recruitment order ( Figure 5—figure supplement 1 ) . The exceptions occurred when the fly was rapidly waving its leg , rather than pulling on the force probe . In these cases , the slow motor neuron spike rate appeared to be suppressed by inhibition . This phenomenon is similar to rapid paw shaking behavior in cats ( Smith et al . , 1980 ) and rapid swimming in zebrafish ( Menelaou and McLean , 2012 ) . Such violations suggest that motor neuron excitability is not the only factor that determines recruitment order . Rather , differences in presynaptic input may selectively de-recruit slow motor neurons through inhibition . So far , we have shown that tibia flexor motor neurons exhibit a gradient of functional properties ( Figures 2–4 ) and a recruitment order during natural leg movements ( Figure 5 ) . We next asked how motor neurons respond to proprioceptive sensory feedback . A key function of proprioceptive feedback is to maintain stability by counteracting sudden perturbations , such as when an animal stumbles during walking ( Tuthill and Azim , 2018 ) . Larger perturbations require more corrective force , so as feedback increases it should result in the recruitment of additional motor neurons with increasing force production capacity . We first compared the amplitude and dynamics of proprioceptive feedback in flexor motor neurons in response to ramping movements of the tibia ( Figure 6A–B ) . Passive extension of the tibia caused excitatory postsynaptic potentials ( PSPs ) in all three flexor motor neurons , a response known as a resistance reflex . PSP amplitude was largest in slow motor neurons and smallest in fast neurons ( Figure 6C–D ) . The slope of the rising phase of the PSP increased with the rate of extension . In the fast and intermediate neurons , sensory-evoked PSPs typically failed to elicit spikes , whereas the firing rate of slow motor neurons was significantly modulated . Slow motor neurons were exceptionally sensitive to passive tibia movement: a 1° change in tibia angle resulted in a significant change in firing rate ( Figure 6F ) . Extremely fast extension movements were sufficient to evoke single spikes in intermediate neurons ( Figure 6—figure supplement 1 ) , but we never observed feedback-evoked spikes in fast motor neurons . The amplitude of the PSP evoked by proprioceptive feedback did not vary across different resting femur-tibia joint angles ( Figure 6—figure supplement 1 ) . In addition to differences in amplitude , the sign and dynamics of proprioceptive feedback varied across fast , intermediate , and slow motor neurons . Fast and slow motor neurons both hyperpolarized slightly at flexion onset , while intermediate neurons depolarized once the flexion event ceased ( Figure 6E ) . These results demonstrate that the different tibia flexor motor neurons receive distinct feedback signals from leg proprioceptors . As reported in recordings from other insect species ( Bässler and Büschges , 1998 ) , we occasionally observed a reflex reversal , in which the sign of the proprioceptive PSP reversed ( Figure 6—figure supplement 1 ) . In most instances , tibia extension led to excitatory responses in flexor motor neurons . But occasionally , typically when the fly was actively moving as the stimulus was delivered , extension produced an inhibitory PSP . Reflex reversal is an important mechanism that could allow the fly to switch between active and passive motor states ( Bässler and Büschges , 1998; Tuthill and Wilson , 2016a ) . A major source of proprioceptive information about tibia position is the femoral chordotonal organ ( FeCO ) . The Drosophila FeCO is comprised of ~150 mechanosensory neurons that encode position , movement , and vibration of the femur-tibia joint ( Mamiya et al . , 2018 ) . We used optogenetic activation ( iav-LexA > Chrimson ) to measure proprioceptive feedback from the FeCO to each motor neuron type ( Figure 6—figure supplement 2 ) . Stimulating FeCO neurons caused the fly to extend and then flex its tibia in a rapid and stereotyped manner . Motor neuron responses to FeCO stimulation increased from slow motor neurons to fast motor neurons , consistent with responses to passive stimulation . In interpreting these experiments , it is important to note that Chrimson expression in FeCO neurons reduced motor neuron responses to passive leg movements compared to control flies ( Figure 6—figure supplement 2 ) . We speculate that this side-effect of Chrimson expression is caused by a homeostatic decrease in the gain of proprioceptive feedback pathways following increased excitability . Overall , these data demonstrate that the FeCO provides proprioceptive feedback to tibia flexor motor neurons , and that the amplitude and dynamics of proprioceptive feedback signals vary across the different motor neurons . This gradient of sensory feedback amplitude is likely shaped by the concurrent gradient of motor neuron intrinsic properties ( Figure 3 ) . We observed differences in the sign and dynamics of proprioceptive feedback , which indicate that tibia flexor motor neurons receive distinct presynaptic inputs . These differences also suggest that proprioceptive feedback signals may be specialized for controlling particular motor neurons . We have thus far measured force production and recruitment order of slow , intermediate , and fast tibia flexor motor neurons under controlled conditions . We next used optogenetics to alter motor neuron activity in tethered , walking flies , in order to test how sensorimotor feedback loops respond to perturbations of normal activity patterns . Flies flex the tibia of the front leg to pull the body forward during walking ( Video 4 ) , so we examined how each motor neuron contributes to this movement . We focused a green laser at the ventral thorax , at the base of the left front ( T1 ) leg to optogenetically activate ( Gal4 >CsChrimson ) ( Klapoetke et al . , 2014 ) or silence ( Gal4 >gtACR1 ) ( Mohammad et al . , 2017 ) each motor neuron type . As a control , we used an empty-Gal4 driver line , which has the same genetic background but lacks Gal4 expression . To verify our optogenetic manipulations and characterize basic leg reflexes , we first measured the movement of the femur-tibia joint in headless flies with their legs unloaded ( i . e . , the fly was suspended in air , Figure 7A ) . Activation of fast motor neurons caused rapid flexion of the tibia ( Figure 7B , C , Video 3 ) . We observed repeated patterns of tibia flexion and extension , suggesting that the fast neuron fired single spikes and that the resulting flexion was immediately countered by a resistance reflex . Activation of intermediate neurons caused slower tibia flexion , while activation of slow neurons caused even slower contractions , often followed by unexpected , prolonged tibia extension . Silencing each motor neuron type did not consistently evoke leg movements , though silencing the full motor neuron population resulted in paralysis ( Video 4 ) . These results are consistent with our measures of force production during electrophysiology ( Figures 4 , 5 ) and provide a useful baseline for interpreting results from walking flies . To optogenetically manipulate motor neuron activity during walking , we positioned tethered flies on a spherical treadmill ( i . e . , a Styrofoam ball ) within a visual arena ( Reiser and Dickinson , 2008; Figure 7D ) . Previous studies have found that walking speed and gait on the treadmill are similar to freely walking flies ( Szczecinski et al . , 2018 ) . In our setup , the fly was motivated to walk forward with a wiggling stripe ( Figure 7E ) , and we tracked the treadmill and the fly’s behavior with multiple high-speed cameras . We first asked how flies would respond to transient optogenetic activation of the different tibia flexor motor neurons . Activation of the fast motor neuron caused the tibia of the front leg to persistently flex , thus , interrupting walking ( Figure 7E , Video 5 ) . Consequently , flies appeared to slow down , as measured by a decrease in forward walking velocity . Activation of the slow motor neurons also interrupted walking ( Figure 7E ) , but this was because the fly reacted by extending its tibia , as when the legs were unloaded ( Video 7 ) . Based on the muscle fibers innervated by the slow motor neuron , we speculate that optogenetic activation of this neuron may increase twisting at the joint , which the fly reacts to with a compensatory extension of the leg . Unlike the other two cells , activation of the intermediate motor neuron appeared to cause the flies to transiently increase their walking speed . Activation of the intermediate motor neuron in stationary flies also caused them to start walking ( Figure 7E–F , Video 6 ) . Extra spikes in the intermediate tibia flexor motor neuron may accelerate the rotation of the treadmill , forcing the other legs to move faster in response . Overall , these results show that activation of different motor neurons drive distinct behavioral responses that depend on the fly’s behavioral state ( e . g . , walking vs . standing ) . Optogenetically silencing motor neurons had less pronounced effects on walking speed ( Figure 7E ) . However , silencing of fast and intermediate neurons decreased the probability that a stationary fly would initiate walking compared to controls ( Figure 7H , Videos 6 and 7 ) . These results suggest that motor neurons with a high force production capacity ( Figure 4 ) may be dispensable for normal walking behavior , but their activity may be important for producing the high levels of force required to initiate walking . A key assumption of the original size principle is that all motor neurons within a pool receive the same presynaptic input ( Henneman et al . , 1965a ) ; the alternative is that different premotor neurons provide input to different subsets of motor neurons . Our results suggest the motor system controlling the fly tibia operates in a middle ground between these two extremes . Although tibia motor neurons generally follow a recruitment order in accordance with the size principle ( Figure 5 ) , we found that the dynamics of proprioceptive feedback vary across motor neurons within the pool ( Figure 6 ) . We also observed a small but significant proportion of behavioral events in which recruitment order was violated ( Figure 5—figure supplement 1 ) . These data suggest that although the tibia flexor motor neuron pool may share some presynaptic inputs , they are not identical . Thus , motor control of the fly tibia is more complex than a straightforward implementation of the size principle . Similar exceptions to the strict dogma of the size principle have been observed in vertebrate species . For instance , the effect of proprioceptive feedback varies across a pool of motor neurons controlling the cat leg ( Heckman and Binder , 1988; Heckman and Binder , 1991 ) . During rapid escape behaviors in zebrafish , slow motor neurons can completely drop out of the population firing pattern ( Menelaou and McLean , 2012 ) . Muscles controlling human fingers can change recruitment order based on movement direction ( Desmedt and Godaux , 1977 ) . These examples suggest that some behaviors exhibit more degrees of freedom than can be supported by recruitment of motor neurons based on their intrinsic excitability alone . Investigations of the vertebrate spinal cord have also identified circuit motifs that support flexible control of motor neurons within a pool ( Kiehn , 2016; McLean and Dougherty , 2015 ) . For example , recordings in turtles and zebrafish have shown that motor neurons receive coincident excitation and inhibition , which could underlie selective de-recruitment ( Berg et al . , 2007; Kishore et al . , 2014 ) . Zebrafish spinal cord premotor neurons provide patterned inhibition to speed-specific circuits ( Callahan et al . , 2019; Menelaou et al . , 2019 ) and can flexibly switch between different motor neuron recruitment patterns ( Bagnall and McLean , 2014 ) . Our characterization of identified tibia motor neurons provides a handle to investigate similar premotor circuit motifs for flexible limb motor control in the fly . Little is currently known about the leg premotor circuitry in Drosophila . However , the hypotheses generated by this work should soon be testable using connectomic reconstruction of identified motor neurons and their presynaptic inputs in the Drosophila VNC ( Maniates-Selvin et al . , 2020 ) . The muscles that control flexion of the Drosophila tibia are innervated by approximately 15 motor neurons ( Baek and Mann , 2009; Brierley et al . , 2012 ) . Here , we chose to analyze three identified neurons that span the extremes of this motor pool . Each of these motor neurons was stereotyped across flies in its anatomy , physiology , and function . The fast tibia flexor motor neuron labeled by R81A07-Gal4 is the only motor neuron that innervates the large tibia flexor muscle fibers in the middle of the femur ( Figure 2B ) . Because it has the largest axon of any motor neuron in the femur , the fast motor neuron also produces the largest extracellular spikes ( Figure 2D ) . The slow tibia flexor motor neuron labeled by R35C09-Gal4 is one of 8–9 motor neurons that innervates tibia flexor muscle fibers located at the distal tip of the femur . Measurements of force production from other motor neurons that innervated this distal region ( Figure 4—figure supplement 1 ) suggest that the R35C09-Gal4 neuron is among the weakest within this group . Finally , we studied an intermediate motor neuron labeled by R22A08-Gal4 that innervates muscle fibers separate from the fast and slow neurons . Previous anatomical studies suggest that 2–5 neurons innervate nearby muscle fibers in the proximal region of the femur . Our recordings of other tibia flexor motor neurons ( Figure 4—figure supplement 1 ) were consistent with the gradient of properties we describe in detail for three identified motor neurons , which suggests that the relationship between anatomy , physiology , and force production applies to other neurons in the tibia flexor motor pool . The structure of the leg motor system in Drosophila has several similarities to other well-studied walking insects . In the metathoracic leg of the locust , a single tibia flexor muscle is innervated by nine motor neurons that were also classified into three groups: fast , intermediate , and slow ( Burrows and Hoyle , 1973; Phillips , 1981; Sasaki and Burrows , 1998 ) . Different motor neurons within this pool lie along a gradient of intrinsic properties ( Sasaki and Burrows , 1998 ) and are sensitive to different types of proprioceptive feedback , such as position vs . velocity or fast vs . slow movements ( Field and Burrows , 1982; Newland and Kondoh , 1997 ) . The femur of the stick insect is innervated by flexor motor neurons that were also described as slow , semi-fast , and fast , based on their intrinsic properties and firing patterns during behavior ( Bässler , 1993; Schmidt et al . , 2001 ) . Much of the work in bigger insects and crustaceans has focused on the most reliably identifiable neurons , the fast and slow extensor tibiae ( FETi and SETi ) , antagonists to the more diverse tibia flexor motor neurons . Consequently , details of how sensory feedback and local interneurons recruit specific subsets of flexor motor neurons has been relatively understudied ( Clarac et al . , 2000 ) ( but see Gabriel et al . , 2003; Gabriel and Büschges , 2007; Hill and Cattaert , 2008 ) . Our work exemplifies the advantage of using Drosophila genetics to identify cell types , and even individual cells , within a diverse motor pool . Genetic markers have also recently been identified for tibia extensor motor neurons in flies ( Venkatasubramanian et al . , 2019 ) , which will enable future investigation of premotor input to antagonist motor neurons . Flies differ from these other invertebrates in one major respect: while most arthropods possess GABAergic motor neurons that directly inhibit leg muscles , holometabolous insects such as Drosophila do not ( Schmid et al . , 1999; Witten and Truman , 1998 ) . Indeed , we found that a transgenic line for GABAergic neurons ( Gad1-Gal4; Diao et al . , 2015 ) does not label any axons in the Drosophila leg ( data not shown ) . The presence of inhibitory motor neurons has been proposed as a key underlying reason why insects have been able to achieve flexible motor control with small numbers of motor neurons ( Belanger , 2005; Wolf , 2014 ) . That flies lack this capability means that other mechanisms must be at play . Fly legs are innervated by neurons that release neuromodulators , such as octopamine , which could underlie flexible tuning of muscle excitability ( Zumstein et al . , 2004 ) . Drosophila melanogaster were raised on cornmeal agar food on a 14 hr dark/10 hr light cycle at 25°C and 70% relative humidity . We used female flies , 1–4 days post eclosion , for all experiments except tethered behavior experiments . Both male and female dark-reared flies , between 2–10 days post-eclosion , were used for tethered walking behavior experiments . For experiments involving optogenetic reagents ( Chrimson variants and gtACR1 ) , adult flies were placed on cornmeal agar with all-trans-retinal ( 100 µL of 35 mM ATR in 95% EtOH , Santa Cruz Biotechnology ) for 24 hr prior to the experiment . Vials were wrapped in foil to reduce optogenetic activation during development . Flies were positioned in a custom steel holder as described in Tuthill and Wilson , 2016b , with modifications to allow us to image movement of the fly leg . Each fly was anesthetized on ice for two minutes , so that she could be positioned ventral side up with her head and thorax fixed in place with UV-cured glue . The front legs were glued to the horizontal top of the holder , the coxa aligned with the thorax , and the femur positioned at a right angle to the coxa and body axis . In this configuration , the fly could freely wave her tibia in an arc at an angle of ~50–65° to the top surface of the holder . The holder was placed in the imaging plane of a Sutter SOM moveable objective microscope . All recordings were performed in extracellular fly saline ( recipe below ) at room temperature . We used a water immersion 40X objective ( Nikon ) for patch-clamping and a 5X air objective ( Nikon ) to view the fly’s right front femur and tibia through the saline during spontaneous movements of the leg . Videos of the preparation were acquired at 170 Hz through the 5X objective with a Basler acA1300-200um machine vision camera . Custom acquisition code written in MATLAB ( Azevedo , 2020a; copy archived at https://github . com/elifesciences-publications/FlySound ) controlled generation and acquisition of digital and analog signals through a DAQ ( National Instruments ) . Input signals were digitized at 50 kHz . To measure forces produced by leg movements , we imaged the position of a flexible ‘force probe’ as the fly pulled against it . The force probe was a PBT filament fiber from a synthetic paint brush ( Proform CS2 . 5 AS ) , threaded through the end of a glass micropipette ( 1 . 5 mm OD , 1 . 1 mm ID , WPI ) . To create a force probe , UV-cured glue ( KOA 300 , KEMXERT ) was sucked up into the micropipette , the fiber was threaded into the glue , leaving 1–2 cm protruding out from the tip of the glass , and the glue was cured . The micropipette allowed us to mount the force probe in a custom holder and to couple it to a micromanipulator ( Sutter MP-285 ) . Videos of the force probe were acquired at 170 Hz through the 5X objective with a Basler acA1300-200um machine vision camera . One pixel equaled 1 . 03 μm2 . We wrote custom machine vision code that detects the position of the force probe in each frame of the video by 1 ) allowing the user to draw a line along the probe in the video , 2 ) rotate the image of the probe perpendicular to the line , 3 ) average down the rows of the rotated image to get a single intensity profile , with a peak at the probe’s location , and then 4 ) find the center of mass of the intensity peak , ±FWHM above baseline . A similar technique employing a probe to measure force has been used in Drosophila in previous studies ( Elliott and Sparrow , 2012 ) . At steady state , the position of the force probe was related to the force through a spring constant , k , F = -kx ( Figure 1—figure supplement 2 ) . We measured the spring constant by positioning the force probe over an analytical balance . A glass coverslip was oriented vertically on edge in a piece of wax on the balance , and the tip of the force probe was positioned at the top edge of the coverslip . We then moved the micromanipulator to different positions . The ‘mass’ of the force probe was multiplied by gravity to give the force at that position . We then fit a linear relationship between force and position to measure the spring constant ( Figure 1—figure supplement 2A ) . The force probe we used for experiments in this study had a spring constant of k = 0 . 2234 µN/µm and protruded approximately 1 . 5 cm past the end of the micropipette . The force probe was not only a spring . It also had mass and was placed in saline , so inertia ( m ) and drag ( c ) affected its dynamics: F = m d2x/dt2 + c dx/dt+kx . To measure these properties , we ‘flicked’ the force probe by moving it to different positions with a glass micropipette , abruptly letting go , and allowing the probe to relax back to rest ( Figure 1—figure supplement 2B ) . We imaged the position of the probe at 1 . 2 kHz with a restricted region of interest , and extracted dynamical parameters m = 0 . 1702 mg and c = 0 . 1377 kg/s . While not zero , the effective mass and drag were negligible ( Figure 1—figure supplement 2C ) . The probe was slightly under-damped with a relaxation time constant of 2 . 5 ms and oscillation period of 5 . 8 ms , such that when imaging spontaneous movements at 170 Hz , the probe would effectively come to rest within one frame . The relaxation time constant was much faster than the fly’s spontaneous movements , even when the fly was attempting to let go of the force probe . In Figure 4D–F , we calculate force by including drag and inertia , but in other figures we report leg displacement and the approximation of force , assuming that drag and inertia are negligible . We easily captured the lateral movement of the force probe across the frame but avoided estimating the vertical movement as the fly pulled the probe closer to its leg . As a result , the displacement ( and thus the force ) measured by the probe in Figures 1 , 5 may be slightly underestimated . To move the leg and passively stimulate proprioceptive feedback , we mounted the force probe perpendicularly on a piezoelectric actuator with a 60 μm travel range ( Physik Instrumente ) . The axial position of the probe was controlled by an amplifier ( Physik ) , with voltage commands generated in MATLAB and delivered through the DAQ board ( National Instruments ) . The output of the actuator’s strain gauge was used to control the position of the actuator through closed-loop feedback . The strain gauge sensor output was sampled at 50 kHz . The probe tip was positioned near the end of the tibia , giving a lever arm of 417 ± 7 ( s . d . ) μm across flies ( n = 8 ) . We then moved the tibia through its range of motion until it was approximately at 90° to the femur . To measure the effect of leg angle on the amplitude of sensory feedback , we then moved the probe to a range of axial positions ( −150 μm = −21° , −75 μm = −10° , 75 μm = 10° and 150 μm = 21° , negative direction is extension ) and repeated the stimuli ( Figure 6—figure supplement 1 ) . We delivered ramp stimuli that moved the leg 60 μm ( 8° ) with varying speeds , in both flexion ( + ) and extension ( - ) directions . We measured the actual speed of step stimuli by finding the maximum derivative of the strain gauge signal during the step onset . The range of angular velocities produced by the probe span the range shown to activate position- and velocity-sensitive femoral chordotonal neurons in the femur ( Mamiya et al . , 2018 ) . Though the force probe was flexible , when we imaged the displacement of the force probe we saw that the probe tip matched the strain gauge feedback ( errors < 5% ) , suggesting that passive or muscle forces did not impact these small stimuli . To generate larger , faster movements than we could deliver with the probe , we whacked the leg by flicking the probe , similar to how we calibrated the probe dynamics ( Figure 6—figure supplement 1 ) . In trials where the fly was free to wave its leg rather than pull on the force probe , we tracked the leg using DeepLabCut ( Mathis et al . , 2018 ) . For a training set , we labeled the tibia position for ~45 frames for three different videos from each fly using custom labeling code ( Azevedo , 2020b; copy archived at https://github . com/elifesciences-publications/LabelSelectedFramesForDLC ) . We labeled six points on the stationary femur , six points on the tibia ( Figure 1—figure supplement 1B ) , as well as prominent bright objects that would otherwise often be falsely identified as part of the legs , such as the EMG electrode , the force probe , and several specular creases in the steel holder . The resnet50 network used in DeepLabCut served as the starting point for training , but as we added more flies to the training set , we initiated further training from the previously trained network . We found that the networks failed to generalize across flies but that ~150 labeled frames were sufficient to ensure >99% accuracy on other frames for that fly ( Figure 1—figure supplement 1B ) . In post-processing , we measured the distribution of pairwise nearest neighbor distances between the six detected tibia points and assumed that outliers indicated that a point was poorly identified . If only a single point was misidentified ( ~0 . 7% ) , we filled in the point with random draws from the nearest-neighbor pairwise distance distributions . The network misidentified more than one point 0 . 2% of the time , typically when the fly moved its leg particularly quickly , causing the image to blur . We excluded such frames . We median-filtered the x , y coordinates across video frames , and found the centroid of the six points , approximately the middle of the tibia . The centroid points traced out an ellipse that was the projection of the circular arc of the leg in the plane of the camera . Fitting an ellipse to the centroids allowed us to calculate the azimuthal angle of the leg arc ( ~50–65° ) and the real angle between the stationary femur and the moving leg . We then used the real angle of the leg to detect when the leg was extended ( >120° ) or flexed ( <30° ) ( Figure 1—figure supplement 1E ) . We imaged the calcium influx into muscles with GCaMP6f ( Figure 1 , Figure 1—figure supplement 1A ) , driven by MHC-LexA expression in muscles . Epifluorescent 488 nm illumination excited GCaMP6f fluorescence . A long-pass dichroic ( 560 nm , Semrock ) passed IR wavelengths to the leg imaging camera and reflected GCaMP6f emission to a second Basler camera imaging at 50–60 Hz . The imaging window of the GCaMP camera was restricted to the femur . Video frames were registered ( Guizar-Sicairos et al . , 2008 ) to remove vibrations due to movements of the saline meniscus ( Video 1 ) . In the dark , the fly tended not to move its leg . However , the fly began to struggle and move its leg as soon as the blue epifluorescent light turned on to excite the calcium sensor . We imaged spontaneous movements under two conditions . 1 ) The fly could pull on the force probe with its tibia or 2 ) the fly waved its leg around spontaneously without the force probe . In the second case , a glass hook was placed near the femur as a barrier to prevent the fly from completely flexing the tibia and obscuring the calcium signals in the femur . We used the k-means algorithm in MATLAB to segment calcium signals into clusters . We used trials in which the fly waved its leg with no force probe . We drew an ROI around the femur , excluding only points near the femur tibia joint where intensity changes were dominated by the tibia obscuring the femur . We used the correlation of pixel intensity as the distance metric and varied the number of clusters , k = 3–8 . Once pixels were assigned to a cluster , we applied a Gaussian kernel to the pixel cluster assignments and excluded pixels which fell below 0 . 75 , indicating that less than ¾ of the surrounding pixels were of the same cluster . This resulted in clearly defined clusters with separation between them but no major gaps , indicating that similar clusters were also grouped anatomically ( Figure 1—figure supplement 1C ) . To confirm that the clustering indicated changes in calcium influx rather than muscle movement , we also ran the same clustering routines on flies expressing GFP in the muscles ( Figure 1—figure supplement 4 ) . In this case , clusters were dispersed , fluctuated very little in intensity and bore no similarity to musculature . We found that six clusters produced broadly similar clusters across flies . We numbered the six clusters as follows: Cluster one was large and distal/ventral; Cluster two was proximal and ventral; Cluster three ran down the center of the leg , neighboring Cluster 2; Clusters 4 , 5 , six were assigned from proximal to distal . With fewer than five clusters , the proximal cluster tended to be much larger , incorporating much of the region labeled as cluster 3 ( Figure 1—figure supplement 1F ) . With more than six clusters , the smallest and least modulated clusters tended to divide , not giving us any further information . Six clusters potentially allowed for pixels that were most correlated with extension of the leg to cluster together , but we did not see large increases in fluorescence with extension . We used the same clusters to measure fluorescence changes in trials where the fly pulled on the force probe . When the leg was flexed , the force probe could obscure the femur ( e . g . Figure 1—figure supplement 4 and Figure 1—figure supplement 2 ) , so we included only proximal portions of the clusters . If the force probe still happened to obscure >40% of the pixels in a cluster , that cluster intensity for that frame was excluded from the analysis . The time constant of the calcium indicator was slower than the fly’s movements , such that fluorescence built up over subsequent contractions . Thus , pixel intensity ( ΔF/F ) was not directly related to contraction of the muscle . We took positive increases in cluster intensity to indicate muscle activation , i . e . neural activity . We applied a Sovitzky-Golay filter to interpolate cluster calcium signals ( 50 Hz ) to the leg movements ( 170 Hz ) , which also computed the time-derivative of the local spline for each cluster ( sgolay_t function in MATLAB , by Tiago Ramos , N = 7 , F = 9 ) . GCaMP6f decay was slow relative to leg movements ( Figure 1E and Figure 1—figure supplement 1E ) , so negative derivatives reflected noise in the cluster fluorescence . We used this estimate of the noise ( two standard deviations ) to threshold the positive cluster derivatives , and thus find cluster ‘activations’ . Surprisingly , we did not identify any clusters that increased their calcium activity during tibia extension . Flies occasionally held their legs extended ( Video 1 ) , at which point we expected to see some clusters increase fluorescence ( Figure 1—figure supplement 1D–E ) . On average , the leg musculature was dim during these periods Figure 1—figure supplement 1E ) , whereas the fluorescence of superficial flexors muscle fibers could increase more than six-fold during flexion events . Diffuse emission from the bright and slowly fading flexors may have obscured small increases in extensor fluorescence . We still found it curious that contractions of extensors did not produce brighter events . We speculate that calcium influx and contractile forces may be larger in flexor muscles than extensors because flies use flexion of the forelimb tibia to support their weight , to hold onto the substrate , and to pull their body during walking , whereas extensors generally lift up unloaded limbs when swinging them forward . To perform whole-cell patch clamp recordings , we first covered the fly in a drop of extracellular saline and dissected a window in the ventral cuticle of the thorax to expose the VNC . The perineural sheath surrounding the VNC was ruptured manually with forceps , near the midline , anterior to the T1 neuromeres . We first used a large bore cleaning pipette ( ~7–10 μm opening ) to remove debris and gently blow cell bodies apart , clearing a path from the ruptured hole in the sheath to the targeted motor neuron soma . The recording chamber was then transferred to the microscope and perfused with saline at a rate of 2–3 mL/min . The extracellular saline solution was composed of ( in mM ) 103 NaCl , 3 KCl , 5 TES , eight trehalose , 10 glucose , 26 NaHCO3 , 1 NaH2PO4 , 4 MgCl2 , 1 . 5 CaCl2 . Saline pH was adjusted to 7 . 2 and osmolality was adjusted to 270–275 mOsm . Saline was bubbled with 95% O2/5% CO2 . Whole-cell patch pipettes were pulled with a P-97 linear puller ( Sutter Instruments ) from borosilicate glass ( OD 1 . 5 mm , ID 0 . 86 mm ) to have approximately 5 MOhm resistance . Pipettes were then pressure-polished ( Goodman and Lockery , 2000 ) using a microforge equipped with a 100X inverted objective ( ALA Scientific Instruments ) . Polished pipettes had resistances of approximately 12 MOhms . The polished surface allowed for high seal resistances ( >50 GΩ ) to limit the impact of seal conductance on Vrest ( <1 mV ) . We used a Multiclamp 700A amplifier ( Molecular Devices ) for all recordings . The bridge resistance was balanced before sealing onto a soma . The pipette capacitance was compensated after the seal was made . The internal solution for whole-cell recordings was composed of ( in mM ) 140 KOH , 140 aspartic acid , 10 HEPES , 2 mM EGTA , 1 KCl , 4 MgATP , 0 . 5 Na3GTP , 13 neurobiotin , with pH adjusted using KOH to 7 . 2 and osmolality adjusted to 268 mOsm . The neurobiotin was a generous gift from Rachel Wilson at Harvard Medical School . The liquid junction potential for the whole cell recordings was −13 mV ( Gouwens and Wilson , 2009 ) . We corrected the membrane voltages reported in the paper by post hoc subtraction of the junction potential . We applied 1 μM methyllycaconitine citrate ( MLA , Tocris ) to block cholinergic transmission in the VNC . To encourage spontaneous self-driven movements , we sometimes bath applied 0 . 5 mM caffeine ( Sigma-Aldrich ) during whole cell recordings , which prolonged periods of struggling in response to the epifluorescent light . Figure 5 includes trials recorded in caffeine: 45 of the 90 trials from one fast cell , and 50 of the 100 trials from a second fast cell . Caffeine’s effects have been reported to be due to dopamine receptor agonism ( Nall et al . , 2016 ) . Caffeine application did not increase activity on the leg electrodes during periods of rest . We screened the Janelia FlyLight collection ( Jenett et al . , 2012 ) to find Gal4 lines that sparsely label leg motor neurons . We obtained flies from the Bloomington Drosophila Stock Center ( BDSC ) , and imaged leg expression of GFP to characterize muscle innervation patterns . Soler et al . , 2004 described the leg musculature in Drosophila , which also serves as the basis for the leg motor neuron nomenclature ( Baek and Mann , 2009; Brierley et al . , 2012 ) . For clarity , however , we refer to muscles of the femur as flexors or extensors , rather than as depressors or levators , which refer to the natural stance of the insect . Soler et al . in turn based their nomenclature on Miller , 1950 . There appears to be a discrepancy between the two: the muscle named the tibia reductor muscle by Soler et al . is described as one of two depressor muscles by Miller , muscles 40 and 41 . Miller applied the nomenclature of Snodgrass , 1935 to Drosophila leg muscles . Muscles that control the tibia are located in the femur . Snodgrass described a reductor of the femur that was actually located in the trochanter , perhaps leading to this misreading . The trochanter-femur joint in most insects has limited mobility . Presumably , Snodgrass named it a reductor muscle because depressor or levator would be inaccurate . We agree with the characterization of Miller and Soler that muscles 40 and 41 are two distinct muscles as they attach via distinct tendons , as seen in X-ray images of the leg musculature ( Pacureanu et al . , 2019 ) . Because the function of muscle 41 is to flex the tibia , we refer to it here as a tibia flexor . We recorded electrical activity in the leg by inserting finely pulled glass electrodes ( OD 1 . 5 mm , ID 0 . 86 mm ) into the cuticle of the femur , taking care to avoid impaling the muscle . The electrodes were filled with extracellular saline . Extracellular currents were recorded in voltage clamp to improve the signal-to-noise . We confirmed that injecting currents of similar size did not move the leg nor produce additional electrical activity . The currents we recorded likely reflect spikes from the motor neuron axons , commensurate with the short latencies between somatic spikes and the events on the leg electrode ( Figure 4 ) . If we observed extracellular leg spikes that were time locked with whole-cell-recorded somatic spikes , we then never observed unmatched somatic spikes . If we were observing muscle action potentials or muscle EPSPs , we would have expected some failures . For simplicity we refer to activity recorded in the leg as the electromyogram ( EMG ) , consistent with the use of the term in other organisms . The content of EMG signals recorded in the leg depended on the placement of the electrode . To improve our chances of recording spikes from specific neurons , we used femur bristles ( macrochaetes ) as landmarks to place the electrode near the branched axon terminals . These include four large distal macrochaetes ( ‘bristles 0–3’ ) and one very proximal bristle which we call here the ‘terminal bristle’ . Fast motor neuron spikes were most likely found by impaling the leg between bristle 2 and 3 , while large intermediate spikes were often found near the terminal bristle . Even still , the polarity , amplitude , and shape of events from identified neurons could vary substantially . When the electrode was placed near the third macrochaete , we tended to record very large units of >200 pA from the fast motor neuron . Fast neuron units were by far the largest amplitude events in the femur . When the electrode was placed in the proximal part of the femur , near the terminal bristle , the spikes from the fast neuron tended to be smaller but still identifiable . We could not unambiguously detect EMG units associated with Flexor 3 in Figure 1 . We ran our spike detection routines ( see below ) on EMG records only in cases where they could be clearly identified or when EMG spikes aligned with the somatic spikes . To measure force production as a function of motor neuron activity , we drove the expression of CsChrimson ( Klapoetke et al . , 2014 ) with 81A07-Gal4 , and the expression of Chrimson88 ( Strother et al . , 2017 ) with 22A08-Gal4 . CsChrimson expression in 22A08-Gal4 prevented straightening of the wings and caused the front legs to be bent midway through each segment . We drove expression of Chrimson ( Klapoetke et al . , 2014 ) in sensory neurons with iav-LexA . We activated Chrimson by placing a fiber-coupled cannula ( 105 μm diameter , Thorlabs ) next to the ventral window in the cuticle and illuminating the T1 neuropil with a 625 nm LED ( Thorlabs ) . We used short flashes of 10 or 20 ms to activate neurons , increasing intensity by increasing the voltage supplied to the LED driver ( Thorlabs ) . We measured the power output of the LED for each voltage we used . To detect spikes in current clamp recordings of membrane potential , we applied the following analysis steps to our electrophysiology traces ( digitized at 50 kHz ) : 1 ) filter , 2 ) identify events with large peaks above a threshold , 3 ) compare the shape of the filtered events to a template ( distance metric ) , 4 ) threshold events based both on the shape and on the amplitude of the unfiltered spike . The parameter space for each of these steps was explored in an interactive spike detection interface ( Azevedo , 2020c; copy archived at https://github . com/elifesciences-publications/spikeDetection ) . We inspected records by eye to reject occasional false positives , such as changes in membrane voltage caused by current injection . The algorithm was generally effective once the parameters were tuned for each cell . However , two cases typically caused spikes to become very difficult to unambiguously identify in slow motor neurons . First , during large current injections and the resulting depolarization , the spikes became very small and difficult to identify ( Figure 4 ) . High frequency spikes were clear in the raw voltage , and the leg moved , so we do not believe the neurons entered depolarization block . In such cases we hand tuned the parameters and inspected the identified spikes by eye to estimate the spike rate . Second , prolonged self-generated flexion could also depolarize slow motor neurons to the point that spike detection was difficult , we speculate because of large synaptic conductances that decreased input resistance . By contrast , spike detection worked perfectly during high spike rates evoked by sensory feedback ( Figure 6 ) . Since we were interested in measuring spike latencies and conduction velocities , we calculated the second derivative of the raw voltage trace , smoothed over five samples , for each identified spike . We found the peak of the second derivative , using that point as the onset/acceleration of the spike . We used the same algorithm to detect spikes from EMG current records , with different thresholding parameters in Step four for each recording due to the variability in EMG event size . After a whole-cell motor neuron recording , we dissected the VNC but left the legs and head attached . We placed the tissue in 4% paraformaldehyde in phosphate-buffered saline ( PBS ) for 20 min , then separated and retained the VNC and the right leg with the filled motor neuron . The VNC tissue was washed in PBST ( PBS + Triton , 0 . 2% w/w ) , placed in blocking solution ( PBST + 5% normal goat serum ) for 20 min , and then placed for 24 hr in blocking solution containing a primary antibody for neuropil counterstain ( 1:50 mouse anti-Bruchpilot , Developmental Studies Hybridoma Bank , nc82 s ) and a streptavidin-Alexa Fluor conjugate to label the neurobiotin-filled motor neuron ( Invitrogen ) . We washed the tissue again in PBST , and then placed the VNC in blocking solution containing secondary antibodies for 24 hr ( 1:250 goat anti-mouse Alexa Fluor conjugate from Invitrogen; streptavidin-Alexa Fluor ) . The leg was first incubated in blocking solution , like the VNC . Then it was placed in PBST containing 0 . 01% sodium azide ( Thermo Fisher ) , 1 unit of phalloidin ( Thermo Fisher ) and the streptavidin-Alexa Fluor , and allowed to incubate for two weeks at 4°C , with occasional nutation . Following staining , the tissue was mounted in Vectashield ( Vector Labs ) and imaged on a Zeiss 510 confocal microscope ( Zeiss ) . Cells were traced in FIJI ( Rueden et al . , 2017; Schindelin et al . , 2012 ) , using the Simple Neurite Tracing plug-in ( Longair et al . , 2011 ) . Images in Figure 2 show the results of the filling function in the FIJI plug-in . In some cases , the neuropil counterstain ( anti-Bruchpilot ) was omitted and the native autofluorescence of the tissue ( along with nonspecific binding of streptavidin and GFP fluorescence ) was used as reference . To quantify morphology ( Figure 3 ) , we measured the soma diameter , the width of the neurite entering the neuropil , and the width of the axon , as close to the exit of the neuropil as possible . We made two measurements for each image and location and averaged the values . Fly wings were clipped under cold anesthesia ( <4 mins ) 24 hr before walking experiments . the fly’s dorsal thorax was attached to a tungsten wire ( 0 . 1 mm diameter ) with UV-curing glue ( KOA 300 , KEMXERT ) . Tethered flies were given only water for 2–5 hr prior to being placed on the treadmill . In some experiments , the tethered flies were then decapitated under cold anesthesia and allowed to recover for 5–20 min prior to the experiment . Intact tethered flies were positioned on a hand-milled foam treadmill ball ( density: 7 . 3 mg/mm3 , diameter: 9 . 46 mm ) that was suspended on a stream of air ( 5 l/min ) and that freely rotated under the fly’s movement ( Figure 7D ) . The ball and fly were illuminated by three IR lights . In experiments on headless flies , we removed the spherical treadmill , leaving the flies suspended in air . For all trials , the temperature in the chamber was maintained between 26–28°C with a relative humidity of 58–65% . We coaxed flies to walk on the ball by displaying visual stimuli on a semi-circular green LED display ( Götz and Wenking , 1973; Reiser and Dickinson , 2008 ) . To elicit forward walking , we displayed a single dark bar ( width 30° ) on a light background which oscillated across 48 . 75° about the center of the fly’s visual field , at 2 . 7 Hz . During periods between trials , the LED panels displayed a fixed dark stripe ( 30° ) on a bright background in front of the tethered fly . To characterize the role of the motor neurons in behaving tethered flies , we optogenetically activated or silenced genetically targeted motor neurons . A green laser ( 532 nm , CST DPSS laser , Besram Technology , Inc ) , pulsed at 1200 Hz with a 66% duty cycle , passed through a converging lens and a pinhole ( 50 µm diameter ) with a resulting power of 87 mW/mm2 at the target . It was aimed at the fly’s left thoracic coxa-body wall joint , thus targeting the motor neuron axons and the T1 neuromere below the cuticle . Experiments using a driver line labeling all motor neurons ( OK371-Gal4 ) indicated that optogenetic stimulation primarily effected neurons innervating the left prothoracic leg , though we cannot rule out effects on other VNC neurons ( Video 8 ) . Each trial was four seconds long . We presented walking flies with the visual stimulus , the flies reached a steady running speed at ~1 . 5 s ( Figure 7E ) , and the laser stimulus began at 2 s . We omitted the laser in some trials ( 0 ms ) , or the laser stimulus was either 90 ms or 720 ms in duration , interleaved in random order . Trials were separated by a 25 s period during which video data were written to disk and the LED panels displayed a fixed , stationary stripe . We used Fictrac ( Moore et al . , 2014 ) to calculate fly walking trajectories ( position , speed , and rotational velocity ) from live video of the spherical treadmill’s rotation ( Point Grey Firefly camera , imaging at 30 Hz ) . Trajectories were then converted from pixels to mm using the spherical treadmill’s diameter of 9 . 46 mm . Leg movements were captured from six simultaneously triggered cameras ( Basler acA800-510µm , imaging at 300 Hz ) that were spatial distributed around the fly’s body . Digital and analog data signals were collected with a DAQ ( PCIe-6321 , National Instruments ) sampling at 10 kHz and recorded with custom MATLAB scripts . When headless flies were suspended above the ball , we manually tracked the position of the left femur and tibia during the 720 ms optogenetic stimulus period using manual annotation of video frames in FIJI ( Rueden et al . , 2017; Schindelin et al . , 2012 ) . We then calculated the femur-tibia joint angle from the position measurements in MATLAB . When flies were in contact with the ball , we visually categorized the fly’s behavior in the 200 ms preceding the optogenetic stimulus as ‘Stationary’ , ‘Walking/turning’ , ‘Grooming’ or ‘Other’ . Flies that took no steps for the duration of the categorization period were classified as Stationary . Flies that took at least four coordinated steps over the duration of the 200 ms period were classified as Walking/turning , irrespective of any distance traveled . Trials in which the fly switched behaviors , groomed or did not display clear markers for walking/turning during the categorization period were classified as Other/Grooming and excluded from analyses . For Walking data , we calculated the average forward velocity over time , for each stimulus length , for each fly . We computed the percent change in walking speed for each fly by averaging walking speed during the stimulation ( 90 ms and 720 ms ) and the subsequent 200 ms period , subtracting the average walking speed at stimulus onset , then dividing by the walking speed at stimulus onset . For Stationary trials , we calculated the percent of trials in which stationary flies reached steady state walking , i . e . sustained walking ( >3 mm/s ) in the half second following laser stimulation . We noticed that the laser stimulus caused the empty Gal4 flies ( BDP-Gal4 ) to decrease their walking speed slightly , likely because the flies could see the stimulus . This was particularly noticeable when we first used a fiber-coupled LED with a 105 µm diameter cannula . This prompted us to focus the laser on the fly’s left leg in order to further reduce the spot size and minimize the behavioral artifact . Even so , the optogenetic stimulus increased the probability that stationary flies would start walking ( Figure 7H ) . To control for this effect , we compared the change in speed in motor neuron lines for each stimulus duration , to the change in speed in the control empty Gal4 flies ( see statistical analysis below ) . In no-light trials , walking initiation ( Figure 7H , black and gray bars ) and changes in speed ( Figure 7E , black traces ) did not vary across different genotypes , although baseline walking speed varied slightly across the different lines . For electrophysiology and calcium imaging results in Figures 1–6 , no statistical tests were performed a priori to decide upon sample sizes , but sample sizes are consistent with conventions in the field . Unless otherwise noted , we used the non-parametric Mann-Whitney-Wilcoxon rank-sum test to compare two populations ( e . g . Figure 4 ) and 2-way ANOVA with Tukey-Kramer corrections for multiple comparisons across three populations ( e . g . Figure 3 ) . To compare changes in fluorescence across multiple clusters and extension vs flexion ( Figure 1 ) , we used a 2-way ANOVA modeling an interaction between clusters and flexion vs . extension , with Tukey-Kramer corrections for multiple comparisons . To compare cluster ΔF/F of multiple clusters ( Figure 1—figure supplement 1 ) , we used a 2-way ANOVA with Tukey-Kramer corrections for multiple comparisons . All statistical tests were performed with custom code in MATLAB . For fly walking behavior in Figure 7 , we used bootstrap simulations with 10 , 000 random draws to compare both the likelihood of stationary flies to start walking , as well as changes in walking speed ( Saravanan et al . , 2019 ) . Stationary trials were assigned a binary value to indicate that the fly began walking ( 1 ) or not ( 0 ) . For a given stimulus duration and optogenetic condition ( Chrimson or gtACR ) , the binary values for the empty Gal4 control flies and a motor neuron line were combined and then drawn randomly with replacement in proportion to the number of trials for each genotype . As a metric , we measured the difference in the fraction of flies that began walking . The p-value was the fraction of instances in which the randomly drawn distribution produced a value of our metric more extreme then we saw in the data ( two-tailed ) ( Figure 7H ) . For trials in which the fly was already walking at the onset of the laser stimulus ( Walking trials , Figure 7E ) , we compared the relative change in speed following the stimulus for a given motor neuron line to the empty Gal4 line . We randomly assigned trials to each genotype and calculated the average speed change as above . We used the Benjamini-Hochberg procedure to calculate the false-discovery-rate for either activation or silencing . Figure 1AW[*]; P{w[+mC]=Mhc-GAL4 . K}2 , P{y[+t7 . 7] w[+mC]=20XUAS-IVS-mCD8::GFP}attP2Figure 1Bw[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/+Figure 1C–Iw[1118]; MHC-LexA , w[13XLexAop2-IVS-GCaMP6f-p10}su ( Hw ) attP5/Berlin WT; +/Berlin WT;Figure 2w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/+Figure 3w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/+Figure 4w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS-CsChrimson . mVenus}attP2 w[1118] , 10XUAS-syn21-Chrimson88-tDT3 . 1 ( attP18 ) ; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/+Figure 5w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/+Figure 6A–Ew[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/+ w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/+; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/+Figure 6G–Hw[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/iav-LexA; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/13XLexAop2-IVS-Syn-21-Chrimson::tdTomato ( attP2 ) w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/iav-LexA; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/13XLexAop2-IVS-Syn-21-Chrimson::tdTomato ( attP2 ) w[1118]; P{JFRC7-20XUAS-IVS-mCD8::GFP} attp40/iav-LexA; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/13XLexAop2-IVS-Syn-21-Chrimson::tdTomato ( attP2 ) Figure 7w[1118]; +/+; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS-CsChrimson . mVenus}attP2 w[1118]; +/+; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS-CsChrimson . mVenus}attP2 w[1118]; +/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS-CsChrimson . mVenus}attP2 w[1118]; +/+; P{y[+t7 . 7] w[+mC]=BDP-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS-CsChrimson . mVenus}attP2 w[1118]; +/+; P{y[+t7 . 7] w[+mC]=GMR81A07-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS- gtACR1}attP2 w[1118]; +/+; P{y[+t7 . 7] w[+mC]=GMR35 C09-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20XUAS-IVS-gtACR1}attP2 w[1118]; +/+; P{y[+t7 . 7] w[+mC]=GMR22A08-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS- gtACR1}attP2 w[1118]; +/+; P{y[+t7 . 7] w[+mC]=BDP-GAL4}attP2/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS- gtACR1}attP2 w[1118]; P{w[+mW . hs]=GawB}VGlut[OK371]/+; +/P{y[+t7 . 7] w[+mC]=20 XUAS-IVS- gtACR1}attP2 Data will be made available from the authors website . Acquisition code is available at https://github . com/tony-azevedo/FlySound . Analysis code is available at https://github . com/tony-azevedo/FlyAnalysis .	In the body , spindly nerve cells called motor neurons connect the brain to the muscles . Their role is to control movement , as they translate the electrical signals from the brain into instructions to the muscles . In humans , it takes over 150 , 000 motor neurons to control the movement of one leg; in contrast , fruit flies only need 50 neurons to operate a leg , despite also executing a variety of movements . Fruit flies are commonly used in laboratories to study an array of biological processes , yet little is known about how their motor neurons direct movements . In particular , it was unclear whether the same principles that control how muscles contract in mammals also applied in the tiny fruit fly . To begin investigating , Azevedo et al . mapped out the arrangement of motor neurons that control muscles in the fruit fly leg . As the leg moved , the activity of both the neurons and the muscles they controlled was recorded , as well as the force that had been generated . The experiments showed that each motor neuron controls a certain range of leg force and speed: some produced small , slow motion important for posture and dexterity , while others created large , fast movements essential to running or escape . In addition , the neurons activate in a particular order – cells that control slow movements fire first , and those that direct fast maneuvers later . These processes are also found in other organisms , but the difference is that flies have so few neurons , allowing scientists to reliably identify each motor neuron . Future experiments will therefore be able to test how flies recruit the right neurons to create specific movement sequences . Fruit flies are often used to research human illnesses that affect movement , such as motor neuron disease . A better understanding of the way their neural circuits coordinate the body could help reveal how these conditions emerge .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2020	A size principle for recruitment of Drosophila leg motor neurons
T follicular helper cells ( Tfh ) are crucial for the initiation and maintenance of germinal center ( GC ) reactions and high affinity , isotype-switched antibody responses . In this study , we demonstrate that direct TGF-β signaling to CD4 T cells is important for the formation of influenza-specific Tfh cells , GC reactions , and development of isotype-switched , flu-specific antibody responses . Early during infection , TGF-β signaling suppressed the expression of the high affinity IL-2 receptor α chain ( CD25 ) on virus-specific CD4 T cells , which tempered IL-2 signaling and STAT5 and mammalian target of rapamycin ( mTOR ) activation in Tfh precursor CD4 T cells . Inhibition of mTOR allowed for the differentiation of Tfh cells in the absence of TGF-βR signaling , suggesting that TGF-β insulates Tfh progenitor cells from IL-2-delivered mTOR signals , thereby promoting Tfh differentiation during acute viral infection . These findings identify a new pathway critical for the generation of Tfh cells and humoral responses during respiratory viral infections . During acute viral infections , CD4 T cells differentiate into primarily T helper 1 ( Th1 ) and T follicular helper ( Tfh ) effector cells ( Marshall et al . , 2011; Johnston et al . , 2012; Hale et al . , 2013 ) . Similar to CD8 T cells , Th1 cells express the transcription factors ( TF ) T-bet and Blimp1 , the effector molecules IFN-γ , TNFα , ( and in many cases granzyme B [GrzB] and perforin ) and migrate to sites of viral replication to eliminate infected cells . In contrast , Tfh cells primarily remain in secondary lymphoid tissues where they communicate with B cells in germinal centers ( GC ) to facilitate antibody affinity maturation and isotype switching . Tfh cells express substantially lower levels of T-bet , and instead of Blimp1 they express the TF Bcl6 . Some pro-inflammatory cytokines induced during infection , such as IL-12 , IFN-γ , IFN-αβ , and IL-2 , promote Th1 differentiation; however , the signals required for Tfh differentiation during viral infection have not been as well characterized . Tfh cells must first encounter their cognate peptide-MHC with proper costimulation from professional antigen presenting cells such as dendritic cells . Following activation , Tfh precursor cells start to express the TF Bcl6 and the chemokine receptor CXCR5 , as they downregulate CCR7 and P-selectin glycoprotein ligand 1 ( PSGL1 ) ( Johnston et al . , 2009; Poholek et al . , 2010; Choi et al . , 2011; Pepper et al . , 2011 ) . These events allow for the migration of activated Tfh precursor cells toward the interfollicular zone and T-B border where they again meet peptide-MHC as well as other costimulatory ligands such as Inducible T cell Costimulator ligand ( ICOSL ) from B cells ( Breitfeld et al . , 2000; Schaerli et al . , 2000; Hardtke et al . , 2005; Kerfoot et al . , 2011 ) . These interactions result in the further upregulation of Bcl6 , migration into the GC , and ability to assist B cells in affinity maturation and proper isotype switching ( Poholek et al . , 2010; Baumjohann et al . , 2011; Choi et al . , 2011 ) . In addition to these cell surface ligand-receptor pairings , cytokines play critical roles in the full differentiation of effector Tfh cells during infection . Cytokines utilizing STAT3 signaling pathways including IL-6 , IL-21 , and IL-27 have been implicated in driving Tfh differentiation , but may have overlapping or compensatory effects depending on the immunizing agent and inflammatory environment ( Ma et al . , 2012; Ray et al . , 2014 ) . For example , IL-6 appears to act on early anti-viral Tfh precursors ( Choi et al . , 2013a ) , and while it is not absolutely required for fully differentiated Tfh effector cells during acute lymphocytic choriomeningitis ( LCMV ) infection ( Poholek et al . , 2010; Eto et al . , 2011 ) , it does promote the sustained activation of Tfh cells during chronic LCMV infection ( Harker et al . , 2011 ) . Further , IL-27 is required for Tfh differentiation during protein immunization ( Batten et al . , 2010 ) , while IL-21 is sometimes also involved ( Nurieva et al . , 2008; Vogelzang et al . , 2008; Eto et al . , 2011; Karnowski et al . , 2012 ) . In addition to STAT3 , STAT4 signaling via IL-12 may also promote early Tfh progenitor cells during infection ( Nakayamada et al . , 2011 ) and appears to be critical for the differentiation of human Tfh cells ( Schmitt et al . , 2009 , 2013 ) . However , STAT4 signals are absolutely required for the differentiation of Th1 cells , suggesting that additional signals are needed to repress the expression of the Th1 TFs T-bet and Blimp1 in Tfh progenitor cells . Th1 and Tfh identities can be discerned within the first few days of viral infection indicating that early cytokine signals are involved in the initial stages of the Tfh/Th1 cell fate decision . Increased expression of the high affinity IL-2Rα chain CD25 on early effector CD4 T cells correlates with enhanced expression of the Th1 TFs T-bet and Blimp1 and lower levels of the Tfh TF Bcl6 and this is driven largely by IL-2-STAT5 signaling ( Choi et al . , 2011; Pepper et al . , 2011; Choi et al . , 2013b ) . In contrast , CD25lo early effectors have greater potential to generate Tfh cells ( Ballesteros-Tato et al . , 2012; Johnston et al . , 2012; Nurieva et al . , 2012; Choi et al . , 2013b ) . Intriguingly , IL-2 signals are also important for the homeostasis of regulatory T cells ( Treg ) . Therefore , understanding how effector and regulatory CD4 T cells listen to IL-2 will unveil pathways and targets to modulate CD4 T cell responses during infection , autoimmunity , and cancer . Another important signal at the interface of balancing effector and regulatory CD4 T cells is the cytokine TGF-β . As an immune-suppressive factor , TGF-β promotes the differentiation of peripherally derived regulatory T cells ( pTreg ) and inhibits the development of autoreactive T cell responses . In contrast , TGF-β can also serve a pro-inflammatory role by inducing the differentiation of effector Th17 cells . T cell-specific ablation of TGF-β signaling , either via TGF-βRII deletion or expression of a dominant negative receptor , has demonstrated that direct TGF-β signals are important for both Treg homeostasis and suppression of effector T cell activation and proliferation ( Li et al . , 2006; Marie et al . , 2006; Sanjabi et al . , 2009 ) . The aberrant activation of effector cells in the absence of TGF-β signals cannot be rescued by addition of Treg ( Li et al . , 2006 ) , indicating that direct TGF-β signaling on effector CD4 T cells is required to maintain their homeostasis . Furthermore , TGF-β suppresses T-bet expression ( Gorelik et al . , 2002; Park et al . , 2005 ) and the exuberant proliferation of T cells display Th1 attributes ( Ishigame et al . , 2013 ) , demonstrating that TGF-β has the capacity to suppress Th1 differentiation . In this study , we have identified a new role for TGF-β in balancing the development of Th1 and Tfh cells during acute viral infection . Specifically , we found that CD4 T cell-directed TGF-β was a critical signal for anti-viral Tfh differentiation , GC B cell reactions , and isotype-switched antibody response during influenza infection . TGF-β suppressed the expression of the high affinity IL-2Rα chain CD25 , which restricted IL-2 signaling via STAT5 and mTOR in Tfh progenitor cells early during infection in vivo . Finally , we show that blockade of the mTOR signaling pathway can rescue Tfh differentiation of anti-viral CD4 T cells generated in the absence of TGF-β . Thus , we have identified that T cell-directed TGF-β insulates Tfh precursor cells from IL-2 signals and plays a critical role in the generation of effector Tfh cells and high affinity , class-switched antibodies—an essential source of protective immunity to this global health burden . To better understand the specification of diverse CD4 T cell subtypes during viral infection , we compared the gene expression profiles of Tfh and Th1 effector CD4 T cell subsets that formed during acute LCMV infection ( Marshall et al . , 2011 ) . This analysis revealed a number of TGF-β-associated genes commonly found in Treg cells , including Nt5e ( CD73 ) , Folr4 ( folate receptor 4 ) , Foxp3 , and Ikzf2 ( Helios ) ( Hill et al . , 2007 ) , to be more highly expressed in PSLG1lo Ly6Clo T-betlo CXCR5hi Tfh cells relative to the PSGL1hi Ly6Chi T-bethi CXCR5lo Th1 cells ( Marshall et al . , 2011; Hale et al . , 2013 ) ( Figure 1A ) . We first sought to determine if these results indicated that T follicular regulatory ( Tfr ) cells , a recently described immune-suppressive Tfh population ( Chung et al . , 2011; Linterman et al . , 2011; Wollenberg et al . , 2011 ) , formed during acute LCMV infection . To assess this , we infected B6 or TCR transgenic Smarta ( Stg ) chimeras with acute LCMV Armstrong and monitored Tfh and Treg properties in either GP66–77 tetramer+ or Stg CD4 T cells by flow cytometry . Although we detected enhanced FoxP3 mRNA in the Tfh cells from our microarray analysis , we did not identify any LCMV-specific CD4 T cells that expressed FoxP3 protein or other Treg-associated markers such as GITR , to the level of canonical CD25+ FoxP3+ Tregs ( Figure 1B and Figure 1—figure supplement 1 ) . This suggested that LCMV-specific CD4 T cells do not differentiate into Tfr cells ( Marshall et al . , 2011; Srivastava et al . , 2014 ) . However , in agreement with the differential mRNA expression , we did find enhanced expression of several of the TGF-β- or Treg-associated proteins including CD73 , folate receptor 4 , and Helios on Tfh cells relative to the Th1 cells ( Figure 1C ) ( Hill et al . , 2007; Iyer et al . , 2013 ) . These observations suggested that conventional Tfh cells bear some similarities in their gene expression profiles with Treg cells , despite having little-to-no FoxP3 expression . 10 . 7554/eLife . 04851 . 003Figure 1 . TGF-β-associated gene expression signature in Tfh cells . ( A ) Bar graph shows a selected set of genes upregulated in d8 LCMV-specific Stg PSGL1lo Ly6Clo Tfh cells relative to PSGL1hi Ly6Chi Th1 cells isolated and sorted from the spleen as measured using Illumina DNA microarrays ( Marshall et al . , 2011 ) that have been described to be induced by TGF-β or associated with Treg cells ( Hill et al . , 2007 ) . ( B ) Representative histogram plot ( top ) shows amount of intracellular FoxP3 in total host splenic CD4 T cells ( shaded gray ) and LCMV-specific Th1 ( hatched line ) and Tfh ( black line ) Stg CD4 T cells from the spleen at day 8 p . i . Region gated identifies FoxP3+ nTregs . Bar graphs ( bottom ) depict the cumulative frequency ( left ) of FoxP3+ CD4 T cells or gMFI averages ( right ) of the indicated CD4 T cell populations . ( C ) Expression of the indicated Treg-associated proteins in ( A ) was compared between LCMV-specific Th1 ( hatched line ) and Tfh cells ( black line ) , and FoxP3+ Treg cells gated on total host CD4 T cells ( shaded gray ) from the spleen at day 8 p . i . Histogram plots ( top ) are representative examples of individual mice and bar graphs ( bottom ) depict the gMFI averages of each protein in the indicated CD4 T cell populations . Graphs in B and C are representative of one of five independent experiments ( n = 4–5 mice/group/experiment ) . p < 0 . 05 , *p < 0 . 0005 . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 00310 . 7554/eLife . 04851 . 004Figure 1—figure supplement 1 . LCMV-specific Stg CD4 T cells do not form canonical regulatory T cells nor T follicular regulatory cells . 1 × 104 Stg CD4 T cells were adoptively transferred into congenic C57BL/6 recipient mice , which were infected with LCMV Armstrong the following day . Eight days p . i . , splenocytes were assessed for intracellular FoxP3 expression and the other indicated proteins . Top row depicts LCMV-specific Stg CD4 T cells , while the bottom row is gated on the host-derived CD4 T cells . Data depict an individual mouse representative of more than three independent experiments with 10+ total mice . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 004 We hypothesized that the expression of these Treg-associated gene products may be an indication of TGF-β signaling in the virus-specific Tfh cells . In order to assess the contribution of direct TGF-β signals on the formation of anti-viral CD4 T cell subsets , we crossed TGF-βRIIf/f CD4-cre mice to the Stg TCR transgenic mice . Fixing the TCR delays the onset of autoimmunity in the TGF-βRIIf/f CD4-cre mice ( Sanjabi and Flavell , 2010 ) ; however , activated CD44hi CD4 T cells do emerge over time ( data not shown ) . Therefore , when making chimeras , we adoptively transferred naïve CD44lo TGF-βRII+/+ CD4-cre+ Stg cells ( herein referred to as WT ) or naïve CD44lo TGF-βRIIf/f CD4-cre+ Stg cells ( KO ) into congenic C57BL/6 recipients and 1 day later infected the mice with the acute Armstrong strain of LCMV . Intriguingly , we found that direct TGF-β promoted the differentiation of Tfh precursor cells at day 3 post infection ( p . i . ) , such that there were about 1/3 fewer CD25lo CXCR5+ Tfh precursor cells in the absence of direct TGF-β signals ( WT = 60 . 25% ± 4 , KO = 42% ± 3 . 3 ) ( Figure 2—figure supplement 1A ) . Additionally , the TGF-βRII KO early effector CD4 T cells expressed more Th1 proteins Ly6C and T-bet and slightly lower Tfh TF Bcl6 ( Figure 2—figure supplement 1B ) . However , by day 8 there was no phenotypic difference between TGF-βRII WT and KO CD4 T cells in the spleen ( Figure 2—figure supplement 1C–D ) . These data indicated that TGF-β played a role in the early specification of splenic Tfh progenitor cells , but that other signals compensated for TGF-β signaling over the course of a systemic LCMV infection . Because TGF-β is a dominant regulator of T cells in mucosal tissues , we speculated that it may play a larger role in controlling anti-viral effector T cell responses during infection at mucosal sites , such as the lung . Moreover , respiratory influenza infection induces transcription of TGF-β and the influenza neuraminidase enzyme promotes the cleavage of latent TGF-β complex into its bioactive form in the lung mucosa ( Schultz-Cherry and Hinshaw , 1996; Carlson et al . , 2010; Roberson et al . , 2012 ) . To assess the contribution of TGF-β on the anti-viral CD4 T cell response during a respiratory infection , we infected TGF-βRII WT and KO Stg chimeras i . n . with a recombinant influenza virus expressing the LCMV GP66–77 epitope recognized by the Stg TCR ( WSN-GP33/66 ) ( Marsolais et al . , 2008 ) . First , we confirmed that the phenotypic properties of influenza-specific CD4 T cells closely mirrored that of LCMV-specific CD4 T cell populations , and importantly , verified that the influenza-specific Stg cells were neither FoxP3+ Treg nor Tfr cells ( Figure 2—figure supplement 2 ) . Moreover , we found that the proportion and total number of PSGL1lo Ly6Clo and PD-1hi CXCR5hi Tfh cells in the lung-draining mediastinal lymph node ( MLN ) were markedly reduced in the absence of direct TGF-β signals ( Figure 2A ) . Furthermore , there was an increase in number of PSGL1hi Ly6Chi Th1 cells in all tissues examined ( Figure 2A–C ) . Concomitant with the cell surface phenotypes , we found increased expression of the Th1 TF T-bet and reduced expression of the Tfh TF Bcl6 as well as increased production of IFN-γ and IL-2 in the TGF-βRII KO influenza-specific CD4 T cells compared to their WT counterparts ( Figure 2D ) . Finally , and potentially most importantly given their function to help B cells in the GC , we also detected fewer TGF-βRII KO Stg cells localized in PNA+ GC in the MLN ( Figure 2E ) , suggesting that TGF-β is important for optimal trafficking of Tfh cells to GCs . Together , these data suggest that direct TGF-β signaling was important for the generation of Tfh cells , while it suppressed Th1 differentiation during respiratory influenza virus infection . 10 . 7554/eLife . 04851 . 005Figure 2 . Direct TGF-β is required for influenza-specific Tfh differentiation . 2 × 105 CD44lo TGF-βRII+/+ CD4-cre+ ( WT ) or TGF-βRIIf/f CD4-cre+ ( KO ) Stg cells were adoptively transferred into C57BL/6 congenic recipients infected with WSN-GP33/66 the following day . ( A–D ) On day 8 p . i , Stg cells in the MLN , spleen , and lung were stained with antibodies against the indicated proteins to distinguish Th1 and Tfh cells . Cells were also stimulated with GP66-peptide for 6 hr to assess IFN-γ and IL-2 production by intracellular cytokine staining and flow cytometry . ( E ) Stg cells ( blue , highlighted by white arrows ) located within MLN PNA+ GCs ( green ) were assessed using immunofluorescent microscopy and their numbers were enumerated using Imaris software . Graphs in A–D are representative of one of five independent experiments ( n = 4–5 mice/group/experiment ) . Panel E shows representative microscopy images and the cumulative data from two independent experiments with 8 total mice/group were graphed . p < 0 . 05 , *p < 0 . 005 , and colored asterisks correspond to the color in the stacked graphs . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 00510 . 7554/eLife . 04851 . 006Figure 2—figure supplement 1 . Direct TGF-β restricts anti-viral TH1 precursor formation but does not impact overall effector CD4 T cell differentiation during LCMV . ( A ) Stg chimeric mice ( 1 × 106 CD44lo TGF-βRII+/+ CD4-cre+ ( WT ) or TGF-βRIIf/f CD4-cre+ ( KO ) ) were infected with LCMV Armstrong and 3 days p . i . , spleens were collected and analyzed by flow cytometry for the indicated proteins . ( B–D ) Stg chimeric mice ( 1 × 104 CD44lo TGF-βRII+/+ CD4-cre+ ( WT ) or TGF-βRIIf/f CD4-cre+ ( KO ) ) were infected with LCMV Armstrong and the phenotype of Stg cells was assessed 8 days p . i . Data are representatives of five independent experiments with 15–20 total mice/group . p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 00610 . 7554/eLife . 04851 . 007Figure 2—figure supplement 2 . Expression of PSGL1 and Ly6C distinguish between influenza-specific Th1 and Tfh CD4 T cells . 2 × 105 Stg CD4 T cells were adoptively transferred into congenic C57BL/6 recipient mice , which were infected with influenza WSN-GP33/66 i . n . the following day . 8 days p . i . , lymphocytes in the MLN were assessed for the indicated proteins . ( A and C ) Representative example of Stg cells in the MLN . In the histograms , red lines denote PSGL1hi Ly6Chi Stg cells , blue is PSGL1hi Ly6Clo Stg cells , and green is gated on PSGL1lo Ly6Clo Stg cells . ( B ) Stg cells from the MLN , spleen , lung , and airways . ( C ) includes FoxP3+ Tregs in the shaded gray histograms . Data depict an individual mouse representative of more than 30 total mice from 10+ independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 007 Due to the reduced Tfh differentiation in the absence of direct TGF-β signals , we sought to investigate whether this led to defects in B cell help and formation of antiviral antibodies . Our Stg CD4 T cell adoptive transfer model , as described above , was not suitable to examine GC B cell and antibody responses because the endogenous host-derived CD4 T cells could provide B cell help in these chimeric mice . Therefore , we set up a ‘B cell helpless system’ using OT-II TCR transgenic mice as recipients because they have a fixed non-influenza-specific CD4 T cell compartment that cannot provide help to GC B cells during the infection . We chose to adoptively transfer polyclonal CD4 T cells in this system to provide a broad repertoire of influenza-specific naïve precursors to provide B cell help . It should also be noted that we switched to the distal Lck-cre deletion strain , which is superior to the CD4-cre strain because of a small number of TGF-βR+ ‘escapees’ that differentiate into nTreg during neonatal development and prevent autoimmunity in these mice ( Zhang and Bevan , 2012 ) . Regardless , we set up the experiments similarly to previously described system by adoptively transferring polyclonal naïve CD44lo Thy1 . 2+ TGF-βRIIf/f Lck-cre− ( WT ) or TGF-βRIIf/f Lck-cre+ ( KO ) CD4 T cells into congenic Thy1 . 1+ OT-II recipient mice and 1 day later infected mice with influenza WSN-GP33/66 . First , we confirmed that the host-derived OT-II CD4 T cells remained naïve ( CD44lo ) throughout the influenza infection ( Figure 3A–B ) . In accord with our findings for the TGF-βRII KO Stg cells , the donor ( Thy1 . 2+ ) polyclonal activated ( CD44hi ) CD4 T cells lacking TGF-βRII also displayed impaired Tfh cell development . Specifically , in the MLN and spleen , there was a profound reduction in Ly6Clo CXCR5+ cells and a moderate reduction in PSGL1lo Ly6Clo Tfh cells in the TGF-βRII KO cells compared to the WT controls ( Figure 3A–B ) . Conversely , there was a concomitant increase in Ly6Chi Th1 cells that lacked TGF-βRII relative to the WT controls . 10 . 7554/eLife . 04851 . 008Figure 3 . Direct TGF-β signaling is required for polyclonal influenza-specific Tfh differentiation . 5 × 106 CD44lo Thy1 . 2+ CD4 T cells from TGF-βRIIf/f Lck-cre− ( WT ) or TGF-βRIIf/f Lck-cre+ ( KO ) mice were adoptively transferred into congenic Thy1 . 1+ OT-II TCR transgenic mice and infected with WSN-GP33/66 the following day . 14 days p . i . , host OT-II ( Thy1 . 1+ ) and donor ( Thy1 . 2+ ) CD4 T cells in the MLN ( A ) and spleen ( B ) were assessed for expression of CD44 ( to distinguish activated T cells , see histogram plots left ) and CXCR5 , PSGL1 , and Ly6C to distinguish Tfh and Th1 attributes . FACS plots are from representative mice and bar graphs are representative of one of four independent experiments ( n = 4–5 mice/group/ experiment ) . p < 0 . 05 and colored asterisks correspond to the color in the stacked graphs . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 008 Importantly , we found that the adoptive transfer of WT CD4 T cells largely restored B cell help in the OT-II recipient mice such that there was an enhanced number Fas+ GL7+ GC B cells and IgMlo B cells in the MLN 14 days p . i . However , the TGF-βRII KO CD4 T cells were unable to rescue the formation of GC B cells in these chimeric mice ( Figure 4A ) . Further , WT CD4 T cells partially rescued GC B cell class switching to the IgG1 subtype , while the TGF-βRII KO CD4 T cells did so less efficiently , particularly in the spleen ( Figure 4B ) . In addition to the enumeration of GC B cells by flow cytometry , we also observed fewer and smaller GCs from the spleens of mice receiving TGF-βRII KO CD4 T cells by fluorescent microscopy ( Figure 4C ) . Finally , T cell-directed TGF-β was also important for influenza-specific IgG and IgA in the airways of infected mice ( Figure 4D ) . Taken together , these data demonstrate a previously unappreciated requirement of intrinsic TGF-β signaling in antiviral T cells for Tfh cell function as B cell helpers , and thus , GC reactions and isotype-switched antibody responses during respiratory influenza virus infection . 10 . 7554/eLife . 04851 . 009Figure 4 . T cell-directed TGF-β is required for GC B cell and isotype-switched antibody responses during influenza infection . 5 × 106 CD44lo Thy1 . 2+ CD4 T cells from TGF-βRIIf/f Lck-cre− ( WT ) or TGF-βRIIf/f Lck-cre+ ( KO ) mice were adoptively transferred into congenic Thy1 . 1+ OT-II TCR transgenic mice and infected with WSN-GP33/66 the following day . 14 days p . i . , GC B cells ( Fas+ , GL7+ IgM− ) ( A ) and IgG1+ GC B cells ( B ) in the MLN and spleen ( SPL ) were assessed by flow cytometry ( left plots ) and enumerated in bar graphs ( right ) . ( C ) Splenic PNA+ GCs ( green , highlighted by red ellipses ) were assessed by immunofluorescent microscopy of frozen sections and the numbers of GC/tissue section was calculated using Imaris software . ( D–E ) Influenza-specific IgG and IgA were measured from bronchoaviolar lavage fluid ( BAL ) by ELISA at day 10 p . i . ( D ) or longitudinally at the indicated time points ( E ) . Data in panels A–B are representative of four independent experiments ( n = 3–5 mice/group/experiment ) . The bar graphs in panels C–D show cumulative data from two independent experiments ( n = 3–5 mice/group/experiment ) , the images in panel C are from representative mice of these cohorts . p < 0 . 05 , p < 0 . 005 , p < 0 . 0005 . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 009 Because CD4 T helper subsets begin to diverge within the first few days of viral infection ( Choi et al . , 2013b ) and we found fewer Tfh and more Th1 precursor cells in the absence of TGF-β signals during LCMV infection ( Figure 2—figure supplement 1A ) , we questioned when TGF-β was required for Tfh differentiation during influenza virus infection . In order to assess this , we generated chimeras with 2 × 106 CD44lo TGF-βRII WT or KO Stg CD4 T cells , infected the mice 1 day later with WSN-GP33/66 i . n , and assessed the phenotypes of the early effector CD4 T cells in the lung-draining MLN at days 4–5 p . i . Although a difference in Bcl6 expression in the TGF-βRII KO CD4 T cells was not observed at this time point , the TGF-βRII KO CD4 T cells displayed enhanced Th1 attributes including enhanced expression of CD25 , Ly6C , T-bet , IFN-γ , and IL-2 ( Figure 5 ) . These findings suggested that direct TGF-β suppressed Th1 precursor formation within the first few days of influenza virus infection . 10 . 7554/eLife . 04851 . 010Figure 5 . Direct TGF-β suppresses early influenza-specific Th1 precursor formation in the lung-draining MLN . 2 × 106 TGF-βRII+/+ CD4-cre+ ( WT ) or TGF-βRIIf/f CD4-cre+ ( KO ) Stg cells were adoptively transferred into C57BL/6 congenic recipients that were infected with WSN-GP33/66 the following day . On day 4–5 p . i . , CD44hi Stg cells in the MLN were assessed for the indicated proteins ( CD25 , Ly6C , T-bet , IFN-γ , IL-2 , Bcl6 ) by flow cytometry . Cells were stimulated with PMA and ionomycin for 4 hr to assess IFN-γ and IL-2 . Bar graphs are representative of three independent experiments ( n = 3–4 mice/group/experiment ) . p < 0 . 05 , p < 0 . 005 , p < 0 . 0005 . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 010 Due to the enhanced expression of CD25 in the absence of TGF-βRII signaling , we questioned whether TGF-β may modulate the IL-2 responsiveness of early effector CD4 T cells . Addition of recombinant TGF-β ( 10 ng/ml ) to Stg CD4 T cell cultures stimulated with GP66 peptide in vitro did not inhibit T cell activation because the upregulation of CD25 and the proliferation rates of the CD4 T cells were comparable between cultures containing or lacking exogenous TGF-β ( Figure 6A ) . However , TGF-β profoundly affected the ability of the activated CD4 T cells to sustain CD25 expression and IL-2 responsiveness . That is , at day 2 post activation , the amount of surface CD25 and intracellular phospho-STAT5 ( pSTAT5 ) after IL-2 stimulation was the same whether or not TGF-β was present , but 1 day later , the T cells exposed to TGF-β displayed considerably less CD25 and lower IL-2 responsiveness compared with those that were not ( Figure 6A ) . Next , we assessed whether TGF-β modulates CD25 expression and IL-2 signaling in virus-specific CD4 T cells in vivo by comparing CD25 , pSTAT5 , and pS6 levels in TGF-βRII WT and KO Stg CD4 T cells isolated directly ex vivo from day 3 post LCMV infection , a setting in which Th1 progenitor cells are more frequent to facilitate analysis ( Figure 2—figure supplement 1A ) . We observed a considerably enhanced CD25+ pSTAT5+ population from TGF-βRII KO early effector cells relative to the WT cells ( Figure 6B ) , suggestive of heightened IL-2 signaling in the TGF-βRII KO CD4 T cells in vivo during infection . Further , we also found enhanced ex vivo CD25+ pS6+ early effector cells ( Figure 6B ) , indicating that both STAT5 and AKT/mTOR signaling arms are amplified in the absence of TGF-β signals . Together , these data demonstrate that TGF-β directly suppresses the expression of CD25 and phosphorylation of STAT5 and S6 in early effector CD4 T cells in vivo and thus , likely promotes Tfh cell differentiation by limiting IL-2 signaling in Tfh precursor cells . 10 . 7554/eLife . 04851 . 011Figure 6 . TGF-β restricts IL-2 responsiveness and insulates early Tfh progenitor cells from mTOR signaling . ( A ) Stg CD4 T cells were labelled with Cell Trace dye and cultured in vitro with 0 . 1 μM GP66 peptide ± 10 ng/ml TGF-β and stained for surface expression of CD25 ( top ) or restimulated with IL-2 to assess pSTAT5 ( bottom ) . Data are representative of four independent experiments . ( B ) Stg chimeric mice ( 1 × 106 CD44lo TGF-βRII+/+ CD4-cre+ ( WT ) or TGF-βRIIf/f CD4-cre+ ( KO ) ) were infected with LCMV Armstrong and 3 days p . i . , spleens were fixed immediately in 2% PFA and stained with antibodies to measure surface expression of CD25 , Ly6C , and intracellular phosphorylation of STAT5 ( pSTAT5 ) and pS6 directly ex vivo . Data are representative of three independent experiments including 3–5 total mice/group . ( C–D ) 2 × 105 CD44lo TGF-βRII+/+ CD4-cre+ ( WT ) or TGF-βRIIf/f CD4-cre+ ( KO ) Stg cells were adoptively transferred into congenic C57BL/6 recipients infected with WSN-GP33/66 the following day . Mice were treated with PBS or 75 mg/kg rapamycin i . p . daily . On day 8 p . i , Stg cells in the MLN were assessed for expression of PSGL1 and Ly6C ( C ) or T-bet or GzmB ( D ) . Data are representative of three independent experiments encompassing a total of 9–15 mice/group . p < 0 . 05 , *p < 0 . 005 and colored asterisks correspond to the color in the stacked graphs . DOI: http://dx . doi . org/10 . 7554/eLife . 04851 . 011 Since we found enhanced pS6 directly ex vivo , we questioned whether we could rescue Tfh differentiation in the absence of TGF-βRII signaling by blocking mTOR activity . To do this , we treated WT and TGF-βRII KO Stg chimeric mice with the mTORC1 inhibitor rapamycin ( Rapa ) daily throughout influenza infection . It should be noted that IL-2 is not the sole factor that activates mTOR in T cells and that the TCR and a variety of other pro-inflammatory cytokines and costimulatory ligands also utilize this signaling pathway . Interestingly , we found that rapamycin treatment rescued the differentiation of PSGL1lo Ly6Clo Tfh cells and suppressed the aberrant expansion of the PSGL1hi Ly6Chi Th1 cells that arise in the absence of direct TGF-β signals ( Figure 6C ) . Likewise , rapamycin suppressed the over-expression of T-bet and GrzB in the TGF-βRII KO CD4 T cells ( Figure 6D ) . Taking into account that mTOR blockade with rapamycin affects many other cell types , these results suggest that TGF-β dampens mTOR activity in CD4 T cells to allow for Tfh cell differentiation . Together with the findings above , these data strongly support that direct TGF-βRII signaling restricts CD25 expression and IL-2 responsiveness in virus-specific CD4 T cells to maximize the development of Tfh cells , GC B cell reactions , and isotype-switched antibody responses during influenza virus infection . In this study , we demonstrate that T cell-directed TGF-β signals are critical for insulating early effector CD4 T cells from Th1-promoting IL-2 signals , thereby allowing for virus-specific Tfh cell differentiation . Consequently , T cell intrinsic TGF-βRII signaling is also required for GC reactions and isotype-switched antibody production during respiratory influenza virus infection . In support of these findings , TGF-β was also recently shown in human T cells to promote the differentiation of certain Tfh properties in vitro ( Schmitt et al . , 2014b ) . T cell-directed TGF-β restricted the expression of the high affinity IL-2Rα chain CD25 on early effector CD4 T cells in vitro and in vivo , while , early anti-viral effector cells generated in the absence of TGF-β signals displayed evidence of enhanced IL-2 signaling in vivo . Since IL-2 promotes Blimp1 expression and Th1 differentiation ( Pepper et al . , 2011; Ballesteros-Tato et al . , 2012; Johnston et al . , 2012; Nurieva et al . , 2012; Oestreich et al . , 2012 ) , these data suggest that TGF-β works to insulate early effector CD4 T cells from IL-2 signals and may play a central role in limiting Blimp1 expression in Tfh progenitor cells . Interestingly , TGF-β signaling in B cells also promotes IgA class switching and mucosal immunity ( Cazac and Roes , 2000; Borsutzky et al . , 2004; Seo et al . , 2013 ) . Since TGF-β is important for the differentiation of both mucosal Tfh and B cells , which also must interact for proper antibody responses , this provides a unique example of a coordinated signaling pathway to maximize humoral immunity to respiratory influenza virus infection . Cytokines utilizing the STAT3 signaling pathway including IL-6 , IL-21 , and IL-27 have been implicated in driving Tfh differentiation ( Nurieva et al . , 2008; Batten et al . , 2010; Eto et al . , 2011; Ma et al . , 2012; Choi et al . , 2013a; Harker et al . , 2013; Ray et al . , 2014 ) , but due to their partially compensatory pathways , separating the requirements for individual signals throughout infection has proved challenging . Furthermore , it is unclear whether other signals induced locally or globally during viral infections may contribute to Tfh cell differentiation . Herein , we identify TGF-β as an additional signal that may restrict IL-2-induced Blimp1 and directly suppress T-bet expression to allow for Tfh cell differentiation . STAT3 signaling in the presence of TGF-β is also a well-described inducer of Th17 effector cells . In fact , a subset of human Tfh cells share multiple properties with that of Th17 cells including co-expression of Bcl6 and Rorγt and the ability to provide B cell help ( Schmitt et al . , 2014a , 2014b ) . However , we did not detect Rorγt expression in murine anti-viral effector CD4 T cells ( data not shown ) . Moreover , addition of TGF-β to murine T cell cultures in the presence of IL-6 and IL-21 induced IL-17-producing T cells as expected , but actually suppressed Tfh properties including ICOS and IL-21 expression ( Schmitt et al . , 2014b ) . These data suggest that although TGF-β and STAT3 signals are required for murine Tfh differentiation during viral infection in vivo , they are not sufficient to induce this cell fate in vitro and that the specification of Tfh cells by these factors is under tight regulation in vivo . Since STAT4 signaling via IL-12 can promote early Tfh properties including Bcl6 expression ( Nakayamada et al . , 2011 ) , it is possible that IL-12-STAT4 signaling is also required to induce a low level of T-bet in murine Tfh precursor cells to prevent Th17 development , but this will have to be formally evaluated . In the absence of TGF-β , blocking mTOR with rapamycin during the first week of infection was sufficient to restore Tfh cell differentiation and suppress T-bet and GrzB expression . However , rapamycin treatment had no overt affect on the differentiation of WT Stg cells during this phase of influenza infection . Interestingly , rapamycin was recently reported to promote heterosubtypic immunity to influenza by reducing GC reactions and switched antibody responses , resulting in enhanced levels of protective influenza-specific IgM antibodies ( Keating et al . , 2013 ) . This study showed that the protective effects of rapamycin depended on both CD4 T and B cells , and that B cell-intrinsic mTORC1 activity was responsible for enhancing Aicda expression and class switching . Although the generation of Tfh cells was not examined in this prior study , our new findings would suggest that suppression of mTOR in the virus-specific CD4 T cells would likely suppress Th1 in favor of Tfh cell differentiation , possibly contributing to the protective effects of rapamycin on heterosubtypic immunity to influenza . Intriguingly , Treg cells appear to play a central role in anti-viral effector CD4 T cell fate decisions during viral infections due to their production of TGF-β and ability to consume IL-2 . In support of this hypothesis , Treg cells were recently shown to be required for Tfh differentiation and GC reactions during influenza infection ( Leon et al . , 2014 ) . Although this group did not find a role for TGF-β by treating mice with an anti-TGF-β blocking antibody , we have demonstrated a requirement for direct TGF-β using genetic ablation of the signaling receptor . Thus , Treg cells appear to play a central role in effector CD4 T cell fates during viral infection , likely by controlling the local bioavailability of both IL-2 and TGF-β . Another potential consideration is the differentiation of influenza-specific pTreg cells , which can temper pulmonary inflammation upon secondary or heterologous infections ( Brincks et al . , 2013; Kraft et al . , 2013 ) . Future studies will determine whether influenza-specific pTreg , naturally occurring nTreg cells , or other cell types altogether , are a physiologically relevant source of TGF-β to promote Tfh cell differentiation . Further , it is unclear at this time how TGF-β may be interpreted differently by anti-viral Tfh cells and Treg cells , which can suppress CD25 in the former context , but not in the latter . It is likely that IL-2 itself is playing an important role in this setting by initiating a dominant positive feedback loop involving IL-2 , STAT5 , Blimp1 , and FoxP3 to maintain the full Treg program in pTreg cells ( de la Rosa et al . , 2004; Fontenot et al . , 2005 ) and that TGF-β may only be required for the initial induction of FoxP3 , but this remains to be addressed . Infectious pathogens such as influenza virus are a global health burden , and despite annual modifications to seasonal flu vaccines that induce protective antibody responses , we have stumbled in our quest for a universal vaccine . Ideally , a universal flu vaccine would elicit broadly protective circulating and lung-resident memory T cells as well as circulating IgM , IgG , and mucosal IgA antibodies that target conserved viral structures to maximize the potential for immunological cross-reactivity to drift variants and different viral subtypes . Since TGF-β plays a critical role in mucosal tissues in the formation of tissue-resident memory CD8 T cells ( Mackay et al . , 2013; Zhang and Bevan , 2013 ) , IgA-producing B cells ( Cazac and Roes , 2000; Borsutzky et al . , 2004 ) , and as revealed here Tfh cells , it is an attractive target to consider in the development of a broadly protective universal influenza vaccine . C57BL/6Ncr mice were purchased from the National Cancer Institute ( Frederick , MD ) . TGF-βRIIf/f CD4-cre mice were obtained from R Flavell and crossed to Stg TCR transgenic mice . TGF-βRIIf/f Lck-cre mice were a gift from M Bevan . Mice were infected with 2 × 105 pfu LCMV i . p . or ∼50 pfu recombinant influenza WSN-GP33/66 provided by Dr Michael Oldstone ( Marsolais et al . , 2008 ) i . n . after anesthetizing with ketamine hydrochloride and xylazine . Rapamycin-treated mice were administered ∼75 mg/kg rapamycin i . p . daily . All animal experiments were done with approved Institutional Animal Care and Use Committee protocols . At various time points post infection , SPL , MLN , bronchoaviolar lavage ( BAL ) , and lungs were dissected . SPL and LN were processed as previously described ( Marshall et al . , 2011 ) . BAL samples were collected by flushing the lungs twice with PBS and collecting both the supernatant for antibody ELISA and cell fraction for flow cytometry . Lymphocyte isolation from the lung tissue was achieved with the Miltenyi Biotec MACSDissociator using their published protocols . Direct ex vivo phospho-staining was performed by homogenizing the spleen in 2% paraformaldehyde immediately after isolation and permeabilizing the splenocytes in ice-cold methanol . To make Stg chimeras , splenocytes were isolated from WT or TGF-βRII KO Stg mice . CD44hi cells were depleted by staining the cells with an anti-CD44 biotin antibody , followed by labeling with the EasySep Biotin selection reagent ( Stem Cell Technologies , Vancouver , Canada ) . The CD44hi bead-bound fraction was removed by placement in a magnet . Purity of depletion was assessed by streptavidin staining , and cells were used only if the CD44hi fraction was <5% . 1 × 104 Stg cells were adoptively transferred via retro-orbital injection for day 8 LCMV infection , 1 × 106 for day 3 LCMV , 2 × 105 for day 8 influenza infection , or 2 × 106 for days 4–5 influenza infection . Polyclonal chimeras were made in the same fashion and 5 × 106 CD44lo CD4 T cells were adoptively transferred into OT-II recipient mice . Lymphocyte isolation and surface and intracellular staining were performed as described previously ( Marshall et al . , 2011 ) . For in vitro stimulation , lymphocytes were stimulated with GP66–77 peptide ( 1 μg/ml ) for 6 hr with Brefeldin A and 10 ng/ml IL-2 or 1 µg/ml PMA and ionomycin . GP66–77 MHC class II tetramer ( NIH tetramer core facility , Emory University , Atlanta , GA ) staining was performed at 37°C for 1 . 5 hr . CXCR5 staining was achieved by incubating the cells at 25°C for 1 hr . TF staining was performed after permeabilization with the FoxP3 fixation and permeabilization kit ( eBioscience ) . Phospho-flow was conducted by stimulating lymphocytes with soluble rIL-2 for 25 min at 37°C , fixing cells in 2% paraformaldehyde , and permeabilizing with ice-cold methanol . Antibodies were purchased from Biolegend ( San Diego , CA ) , BD Pharmingen ( San Diego , CA ) , eBioscience ( San Diego , CA ) , or Cell Signaling Technology ( Danvers , MA ) . Flow cytometry was acquired with a BD LSRII with Diva software and analyzed with Flow Jo software ( Treestar , San Carlos , CA ) . Total splenocytes from Stg mice were cultured in RPMI supplemented with 10% FCS , 1% L-glutamine , 1% pen-strep , and 50 μM β-mercaptoethanol plus 0 . 1 μg/ml GP66–77 , 10 ng/ml recombinant human TGF-β ( Peprotech ) was added to specified wells . Tissues were fixed in 4% paraformaldehyde overnight , followed by sinking in sequentially greater sucrose solutions ( 10% , 20% , and 30% ) . Fixed tissues were embedded in OCT compound ( Sakura ) , and tissue blocks were frozen in 2-methylbutane ( Sigma–Aldrich ) chilled by dry ice . Eight micrometer sections were cut with a cryostat , air-dried , and fixed with cold acetone . Sections were stained with 1–5 μg/ml antibodies against Ly5 . 1 , CD4 , PNA , IgD , and ProLong Gold antifade reagents ( Invitrogen ) was added after washing . Images were captured on a Zeiss LSM 510 Meta confocal microscope , mounted on an Axiovert 100 M with automated XYZ control equipped with an argon laser with emissions at 458 , 488 , and 514 nm and two HeNe lasers with emission wavelengths at 543 and 633 nm . Image analysis was performed using Imaris suite ( Bitplane , South Windsor , CT ) . 96-well Polysorp microtiter plates ( Nunc ) were coated overnight with UV-inactivated WSN-GP33/66 in carbonate buffer . AP-conjugated goat anti-mouse IgG and IgA secondary antibodies were used for detection ( Southern Biotech ) . ODs were converted to units based on standard curves with sera from C57BL/6 mice infected with influenza ( Softmax Pro 3 . 1 software; Molecular Devices ) . Where indicated , p values were determined by two-tailed unpaired Student's t test . p values <0 . 05 were considered significant and denoted as p < 0 . 05 , p < 0 . 005 , and *p < 0 . 0005 . All error bars represent standard deviation .	The influenza virus is thought to cause illness in up to 10% of adults and 30% of children each year worldwide . Most of these cases resolve on their own and don’t require treatment , but three to five million people are hospitalized and up to half a million people die each year . Unfortunately , the vaccines currently available to protect against influenza only target particular varieties or “strains” of the virus . The strains that circulate vary from year-to-year so it is necessary to develop new influenza vaccines every year . However , it is difficult to correctly predict which strains will circulate , so a more effective solution would be to develop a new vaccine that can help the body defend itself against many , or ideally any influenza strain . During a viral infection , a type of immune cell in the host can specialize into two different types of cells to help fight the virus: T helper 1 cells and CD4 T follicular helper cells . T helper 1 cells help to kill host cells that have become infected . CD4 T follicular helper cells promote the production of proteins called antibodies , which identify and neutralize the virus . Here , Marshall et al . studied how T helper 1 cells and CD4 T follicular helper cells form in mice suffering from a lung infection similar to influenza . It was already known that a protein called transforming growth factor beta ( TGF-β ) helps the immune response to mount an effective defense against an infection without causing too much harm to the host . Marshall et al . show that TGF-β increases the number of CD4 T follicular helper cells in the mice by suppressing the production of another protein—called IL-2—on the surface of CD4 T cells . Treating mice lacking the ability to detect TGF-β with a drug that blocks a protein controlled by IL-2 also allows more CD4 T follicular helper cells to be produced . Marshall et al . ’s findings reveal that TGF-β is involved in controlling the balance of T helper 1 cells and CD4 T follicular helper cells produced during viral infections of the respiratory tract . Since TGF-β also has other roles in immune responses against viruses , it is now an attractive target for the development of a vaccine that may protect us against all strains of the influenza virus .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2015	The transforming growth factor beta signaling pathway is critical for the formation of CD4 T follicular helper cells and isotype-switched antibody responses in the lung mucosa
Do we expect periodic grid cells to emerge in bats , or perhaps dolphins , exploring a three-dimensional environment ? How long will it take ? Our self-organizing model , based on ring-rate adaptation , points at a complex answer . The mathematical analysis leads to asymptotic states resembling face centered cubic ( FCC ) and hexagonal close packed ( HCP ) crystal structures , which are calculated to be very close to each other in terms of cost function . The simulation of the full model , however , shows that the approach to such asymptotic states involves several sub-processes over distinct time scales . The smoothing of the initially irregular multiple fields of individual units and their arrangement into hexagonal grids over certain best planes are observed to occur relatively quickly , even in large 3D volumes . The correct mutual orientation of the planes , though , and the coordinated arrangement of different units , take a longer time , with the network showing no sign of convergence towards either a pure FCC or HCP ordering . Where does our internal representation of space come from ? And how does it code for space extending in three dimensions ? New findings about space-related activity in the bat have recently raised this question again ( Ulanovsky and Moss , 2011; Yartsev et al . , 2011; Yartsev and Ulanovsky , 2013; Finkelstein et al . , 2014 ) . The similarity in the place cell and , most remarkably , in the grid cell representation between rodents and bats suggests a common neural substrate for spatial navigation , shared across these mammals ( Andersen and Buneo , 2002; Jacobs et al . , 2010; Sereno and Lehky , 2011; Killian et al . , 2012; Indovina et al . , 2013; Jacobs et al . , 2013; Thurley et al . , 2014 ) , and it provides an indication possibly valid also for other animals living and moving extensively in three dimensions , like for example dolphins , monkeys and even non-mammalian species ( Healy et al . , 2005; Dacke and Srinivasan , 2007; Wu and Dickman , 2012; Burt de Perera and Holbrook , 2012 ) . At the same time , the obvious difference in the behavior of these species requires a system that flexibly adapts to perform computations as different as mapping two- or three-dimensional space ( Knierim et al . , 2000; Hayman et al . , 2011; Taube and Shinder , 2013 ) . Here we describe a model of grid cell formation that accounts for both these aspects of spatial cognition , in a unitary perspective on the mEC network ( Figure 1 ) . Figure 1 . Scheme of the simulations . ( A ) We simulate the trajectory of a bat exploring a volume of space over a prolonged period of time . At each step , the bat moves forward at a constant speed and chooses a new direction of movement close to the previous one . ( B ) The feed-forward network ( equivalent to that shown in [Kropff and Treves , 2008] ) , including here only two mEC neurons . The would-be grid units receive inputs from place cell-like units with firing fields similar to those reported in Yartsev and Ulanovsky ( 2013 ) . ( C ) Snapshots of the evolution of the firing rate map of a single unit . The figures correspond to ( from left to right ) 2 , 5 , 10 and 20 million time steps of learning . © 2013 , M Yartsev , N Ulanovsky . 2013M Yartsev , N UlanovskyFigure 1 is reproduced from M Yartsev , N Ulanovsky . 2013 . Representation of three-dimensional space in the hippocampus of flying bats . Science 340:367–372 . Reprinted with permission from AAAS . Grid cells seemingly require some clever engineering design that generates the common periodicity among neighboring units while keeping them distinct in terms of spatial phase ( Zilli , 2012 ) . While place cell and head direction cells have been shown to directly generalize to three dimensions ( Yartsev and Ulanovsky , 2013; Finkelstein et al . , 2014 ) , the form that grid cells will exhibit in higher dimensionality ( currently tested in flying bats [Ginosar et al . , 2014] ) is still not clear . Further , the question is still open of how the mechanism producing such a complex periodic pattern on a surface can , in the case of bats , extend to a volume ( Jeffery et al . , 2013 ) , even when accepting the information-theoretic optimality of a regular lattice ( Mathis et al . , 2014 ) . In the self-organization perspective that we propose , the spatial responses first emerge spontaneously , at the single unit level , with no engineering required . In the simplest version of the model , which we have explored in a series of studies ( Kropff and Treves , 2008; Si et al . , 2012; Stella et al . , 2013 ) , the periodicity of the grid pattern is a result of firing rate adaptation during exploration sessions that span a considerable developmental time ( Figure 1 , c ) . It is fixated gradually by means of synaptic plasticity in the feed-forward connections , which convey broad spatial inputs , for example but not necessarily from ‘place units’ ( Figure 1 , b ) . Contrary to other models limited to the explanation of grid cell expression , this model delves into the issue of their induction and , most importantly , can account for the effects of the geometry of the explored environment on the outcome of the self-organization process . We have shown how , for example , the model produces pentagonal grids on a spherical surface and heptagonal ones on a hyperbolic one ( Urdapilleta et al . , 2015 ) . The nature of our model allows us to now investigate the process of grid cell self-organization in three dimensions without the need to modify any of its features . We use bats as our reference , as it is the species currently available for experiments during roughly homogeneous navigation along the three dimensions of physical space ( Figure 1 , a ) . In our simulations a virtual bat explores a volume of side L with a constant speed v . The position of the animal is sampled at time steps of constant Δt . We temporarily leave these quantities unspecified . We will discuss their actual values at the end of the paper as they are critical for the interpretation of the results . For the moment they should be understood as expressed in arbitrary units . The path the animal performs is generated as a correlated random walk in which the direction of movement at any time step depends on the previous one . For simplicity , the change in running direction between two consecutive steps of the virtual bat is sampled from a Gaussian distribution with zero mean and standard deviation σh = 0 . 15 radians . The self-organization process we consider at the single-unit level can be described in analytical terms as an unsupervised optimization process , if one neglects the collateral interactions that are presumed to align the grids ( Si et al . , 2012 ) . The simplified version of the model which can be analyzed mathematically is very abstract , and does not specify most of the parameters necessary to the simulations . Nevertheless , it indicates which are the asymptotic states that should be approached by the system after having evolved for a long time . The asymptotic states are defined in terms of a variational principle , amounting to the minimization of a cost function of the form: ( 13 ) H=HK+HA==∫dχ[▽Ψ ( χ ) ]2+γ∫dχ∫dtΨ ( χ ( t ) ) K ( t−t′ ) Ψ ( χ ( t′ ) ) , where χ is the spatial coordinate and Ψ represents the firing rate of the neuron across the environment . The functional is defined based on the hypothesis that the activity of the units reflects only two simple constraints:The minimization of the variability of the maps across space , that is , a preference for smooth maps . Such smoothness is expected to stem from the smoothness of the spatial inputs and of the neuronal transfer function . This constraint is expressed in the first term of the functional , the kinetic one . The penalization of maps that require a neuron to fire for prolonged periods of time , which is opposed by neuronal fatigue . The second term of the functional , the adaptation term , represents this constraint . The parameter γ parameterizes the relative importance of the two constraints . The dependence of the functional on time can be eliminated by taking into account the averaging effect of a long run over many trajectories and different speeds experienced during training . We therefore substitute the time-dependent kernel K ( t − t′ ) in the second term of Equation 13 with an effective space-dependent one , K ( χ′ − χ ) , using the average speed of the animal to fix the relationship between the two: ( 14 ) H=HK+HA==1V∫Γdχd[▽Ψ ( χ ) ]2+γ1V∫ΓdχΨ ( χ ) ∫Γdχ′Ψ ( χ′ ) K ( \|χ′−χ\| ) , where we have also made explicit the normalization by the area V of the d-dimensional environment Γ . We directly apply this expression to ask which is the favorite arrangement of the fields in a 3D volume V . Constructing either an FCC or an HCP arrangement of fields is a rather articulated endeavor that requires assembling a hierarchy of elements of increasing complexity . The three-dimensional structure described by the two arrangements implies the establishment of long-range relationships between the level of activity at distant points of the environment and involves determining the position of a large number of fields at the same time . Both FCC and HCP are described by unitary cells of 13 fields and the difference between the two lies in the different positioning of just three of them with respect to the others . It is evident by looking at Figure 2 , however , that this long-range order is constructed from a set of building blocks that express symmetries and regularities at a local level , involving fewer fields and a smaller set of constraints . Understanding the outcome of the self-organization of grid cells in three dimensions can be thus approached bottom-up , starting from basic features of the representation and then following the learning process up towards their combination into overarching structures . As in the two-dimensional case , a description of the grid can start from computing the mean distance between first-neighbor fields and the mean angle formed by triplets of adjacent fields . These two measures involve , respectively , two and three fields at a time and are not informative about correlations extending beyond these boundaries . At this level order emerges almost immediately ( Figure 6 ) . The mean angle among neighboring fields ( calculated over all the triplets of all the cells of a simulated population ) is close to π/3 from the very beginning of learning and the real effect of continuing exploration is the reduction of the variability over the course of about 4 million time steps . 10 . 7554/eLife . 05913 . 007Figure 6 . Emergence of local regularities in the arrangement of fields . Left: Mean angle formed by triplets of fields as a function of learning time ( see the ‘Local gridness measure’ section in ‘Materials and methods’ for details on the measure ) . Right: Mean spacing between fields extracted from the autocorrelograms , for various conditions as a function of learning time . DOI: http://dx . doi . org/10 . 7554/eLife . 05913 . 007 A similar behavior is observed when plotting the value of the mean spacing of the grids in time ( calculated from the unit autocorrelograms ) ( Figure 6 , right: blue line ) . Also in this case , after a short transient the value stabilizes after around 3–4 million time steps . Our choice of model parameters ( and specifically of the adaptation parameter ) leads to a spacing of 0 . 55 × L . We run simulations in different conditions to test the sensitivity of this quantity to specific components of the model . We consider the case of having no internal connectivity in mEC , removing any interaction between different mEC units ( Figure 6 , right: red line ) that are therefore developing grids independently , and the case in which rather than having a single value of the adaptation time course , common to all the units , the population expresses a range of possible values , drawn from a uniform distribution ranging from 0 . 85 × b1 to 1 . 2 × b1 , where b1 = 0 . 1 is the value otherwise used ( Figure 6 , right: green line ) . In both cases we see that the time course of the development of a common grid spacing is not affected by the modifications of the standard model . Stabilization is obtained in the same time interval and while removing collateral connections appears to have absolutely no effect , the variability in the adaptation parameter results in a slightly different final value of the spacing ( 0 . 5 × L ) . These results indicate that the three-dimensional grid develops from the same ingredients of its lower dimensional equivalent . Mean spacing and mean angle are quickly fixed over the entire network almost simultaneously and are the first recognizable signs of the emergence of an ordered structure from the initial random distribution of activity . The equivalence between this process and that observed in a model of two-dimensional grid development is due to the same principles driving self-organization . The presence of an additional dimension does not affect the way in which fields are initially brought by adaptation to homogeneously and regularly cover the entire space . Using the measure described in the ‘Local gridness measure’ section in ‘Materials and methods’ , we can evaluate the difference between the distribution of activity of a unit and a random arrangement of fields . Plotting the average across units of this index , which reflects the decrease of the variance in the angles between triplets , already observed in Figure 6 , we see again ( Figure 7 , top left ) , that after roughly 4 million time steps the system is already arranged in a stable ordered manner , with equilateral triangles among neighboring fields that dominate the activity pattern . This ordering can be generated by the system independently of higher order symmetries , and it provides a first step for further arranging fields in more articulated structures . 10 . 7554/eLife . 05913 . 008Figure 7 . Three distinct time courses for the emergence of long-range ordering . The panels present the time evolution of the measure of different symmetries in the arrangement of fields . Top row: Fast convergence of neighboring triplets of fields towards equilateral triangles ( left ) and of group of fields into planes with a hexagonal arrangement ( right ) . Middle row: Slow convergence of the different units in the population towards a common orientation of the layers of fields . Left: Angle between the principal plane expressed by the various units . Right: Angle between fields arranged over mutually aligned planes . Bottom row: No convergence of global , inter-planar order . The measures of similarity with face centered cubic ( FCC ) and hexagonal close packed ( HCP ) ordering do not evolve with the extension of the learning time and remain close to intermediate values indicating an even distribution of the values over the population of grid cells . See ‘Materials and methods’ for details on the measures . DOI: http://dx . doi . org/10 . 7554/eLife . 05913 . 008 A second step taken by the system is the coordination of multiple field triplets to arrange them in a hexagonal pattern . This process corresponds to the formation of a grid on the plane , but in three dimensions it involves the creation of not just one hexagon but of multiple superimposed planes each of which contains hexagonally arranged fields . To investigate the structure of the activity on the single layers , we take multiple slices of the autocorrelogram matrix . We take sections passing through the center , with different angles of azimuth and elevation . We then compute the autocorrelogram values on each of these planes , and we compare it with a hexagonal template of equidistant peaks with π/3 periodicity . The plane most resembling a hexagonal pattern according to a correlation measure ( the ‘best plane’ ) is selected together with its similarity score . This method provides us with an equivalent of the traditional grid score used to judge the quality of planar grid cells . In Figure 7 ( top right ) we plot the time evolution of the average over the population of this score . Starting from very low values , indicative of a still unorganized ensemble of fields , the score steadily rises to reach a value of about 0 . 85 ( out of a maximum of 1 ) after 6 million time steps and then remains stable over the rest of the simulation . The previous analysis considers each unit separately and does not provide a measure of the coordination across the population of the formation of the field layering . But the presence of collateral interactions should induce some coordination in the way these layers are arranged in different cells , that is , in the direction along which these best planes align in space . We looked for the dispersion of their orientations , calculating the angles formed by the best planes of different cells ( β , Figure 7 , middle left ) , and the angle between the hexagonal grid axis of cells sharing a similar best plane ( ω , Figure 7 , middle right ) . We indeed see that collateral interaction tends to align in time the principal layer of different cells , defining a preferred orientation for the global structure of the grid . The emergence of the common alignment of the layers is slower than the formation of the layers themselves . It takes about 12 million time steps of exploration time to significantly reduce the dispersion of the angles ( note that β is slightly faster , maybe indicating a tendency to first define the layers and then arrange shifts within them ) . Having hexagonal layers tiled upon each other along the same direction still leaves to each cell the degree of freedom of setting the relative phase between pairs of these layers . It is clear at this point that a proper choice of these phases would result in the reproduction of either an FCC arrangement or an HCP arrangement . This higher level in the hierarchy of order for three-dimensional grids , which when attained would provide a completely regular tiling of the volume , does not have a correspondence in the two-dimensional case . It is thus a completely new level of complexity that can be only expressed when producing grids in three dimensions . In the bottom panels of Figure 7 we show the time course of the scores for FCC and HCP similarity ( see the ‘Measure of long-range order’ section in ‘Materials and methods’ ) . Their values are almost unchanged over the duration of the simulations and after 30 million time steps , a time largely exceeding that necessary for the other quantities to converge , both of them are still close to 0 . 5 , which reflects the presence in the population of a large distribution of values . Our simulations are able to distinguish a hierarchy in the time course leading to the formation of three-dimensional grids . Different levels of complexity appear in the arrangement of fields with different speed . Fast converging quantities like the formation of equilateral triangles of fields and successively of layers of fields with hexagonal symmetry appear first , framing the activity of single units in the network . These initial structures are then modified on a slower time scale to obtain a global coordination among cells . Planes of fields are rotated to align them across the population , generating a common tiling of the fields of different cells that conserve their unique spatial phase . This global ordering is only partial though , as the phases of the units are only partially overlapping with those necessary to reproduce a perfectly regular tiling of the volume ( either with an FCC or an HCP ) . If we exclude the very initial phase of network dynamics , self-organization does not appear to affect these aspects of network activity that therefore remain loosely determined even after a very extended period of time . The critical question is how would these timescales scale up with the size of the environment . At least for the most rapid self-organizing processes , their very nature , dependent on plasticity in the feed-forward connections , would appear at first sight to require the pairing of each activity field of each unit to the specific configuration of sensory inputs which impinge at the same time on the feed-forward connections , therefore implying times for the formation of the grid that scale up with the number of fields in the volume . A volume of linear size L includes roughly N3=2 ( L/a ) 3 fields of spacing a in either an FCC or HCP arrangement ( and 62 ( L/a ) 3 trajectories connecting neighboring fields ) . A square of side L on a plane would include roughly N2= ( 2/3 ) ( L/a ) 2 fields . The emergence of equilateral triangles in 3D appears to require roughly a factor 2 more time than in 2D ( Si et al . , 2012 ) , in approximate agreement with the factor N3/N2= ( 3/2 ) ( L/a ) ≃1 . 2×1 . 8≃2 coming from the above argument . Note that if that were correct , equilateral triangles in an environment roughly four times as large , as can be argued to be the one used in the bat experiments in the Ulanovsky laboratory ( Ginosar et al . , 2014 ) , would emerge in roughly 40–50 hr of continuous flight . Although the developmental maturation of the bat encompasses longer cumulative flying hours , what is likely relevant for structuring the feed-forward connections is time spent flying in the environment of the actual experiment . Not only can the mechanisms leading to grid-like activity only unfold while navigating and not during rest periods , they also appear to require , in our model , the exact configuration of input activity at each location in the environment . Therefore this constraint is likely to put the time scale of grid cell formation well above the feasible time duration of the experiment . In fact , however , we find that the time for self-organization lengthens only a little , and clearly sublinearly , with the volume flown by the virtual bat . Given the multiple sub-processes involved in the self-organization of the grid units , we focus on a summary measure , derived from the analytical model: the cost function ( Equation 14 ) . Each of the two terms of the cost function , the kinetic and the adaptation kernel , can be calculated for each model grid unit at each time step of the simulation , and average values can be extracted and fit , for example , with sums of exponential functions . What cannot be calculated from the simulations themselves is the value of the γ factor that , in the cost function , would determine the weight of the adaptation kernel with respect to the kinetic term . We find that the population-averaged data points for both terms can be well fit by a sum of two exponentials , plus a constant ( Figure 8 , inset ) and with the same time parameter for the first exponential in each term: ( 29 ) HK ( t ) ≃Av exp ( −t/τvS ) +Bv exp ( −t/τvL ) +K . ( 30 ) HA ( t ) ≃Cv exp ( −t/τvS ) −Dv exp ( −t/τM ) +Ev . with A , B , C , D , E volume-dependent positive fit parameters , and τS , M , L short , medium and long relaxation time scales ( K turns out not to depend on the volume; nor , it seems , does τM ) . 10 . 7554/eLife . 05913 . 009Figure 8 . Effect of environment volume on grid developmental time . Temporal evolution of the cost function calculated for environments of different size . Lines from green to red correspond to environments of increasing size: 1 ( green ) , 1 . 2 , 1 . 4 , 1 . 44 , 1 . 68 and 1 . 96 ( red ) times the basic volume , respectively . With the choice of parameters reported in the discussion , these volumes would range from a 15 . 625 m3 room to a 30 . 625 m3 room . The inset shows the breakdown of the contribution to the total cost of the kinetic part and of the adaptation part for the cubic environment of size 2 . 5 × 2 . 5 × 2 . 5 m . Dots correspond to data points , lines to a fit . The constant of the kernel term , which varies with the volume , is not included for clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 05913 . 009 The short term relaxation is therefore a joint decrease of both terms , while later the adaptation term rises , whereas the kinetic term continues to decrease . An empirical ansatz can be defined for γ as the largest value that still keeps the sum HK ( t ) + γHA ( t ) monotonically decreasing . With this ansatz , we plot the estimate of the cost function ( without the Ev term , for clarity ) for varying volume sizes , where we have multiplied either one or two of the three linear dimensions by either 1 . 2 or 1 . 4 , obtaining volumes larger than the standard one by factors 1 . 2 , 1 . 4 , 1 . 44 , 1 . 68 and 1 . 96 . We can see from Figure 8 that the relaxation of this estimated cost function is mainly determined by the most rapid exponential terms , and is virtually complete by 3–4 million time steps , with a limited volume dependence . Consequently , apart from the slight prolongation of the initial transient , the time evolution of our measure for volumes of different size appears to be quite similar . These results suggest that the time required for the complex dynamics of grid development depends only weakly on the number of fields that have to be arranged in a orderly manner in the volume . The initial relaxation , which accomplishes most of the rearrangement and probably centers on adjusting the angles between triplets of fields ( cp . Figure 7 , top left and Figure 8 ) , occurs in a time that increases sublinearly with system size . This is followed by some bouncing back of the adaptation term , which may have to do with the finer adjustment of most field distances , leading to planar hexagonal grids ( see Figures 6 and 7 , top right ) , with a time constant which can be taken to be independent of the volume . It is protracted later by continued but minor smoothing of the fields , now in place on the best planes , concurrent , if collateral interactions are included , with the adjustment of the planes with respect to each other ( Figure 7 , middle ) , which extends over longer times . We do note that we have observed considerable variability in the degree of smoothness of the individual fields obtained at the end of the simulation , with a tendency for the larger volumes to require longer time and end up with rougher fields . Given the variability from simulation to simulation , however , it remains to be determined whether this trend is robust and whether it points at a significant bifurcation in trajectories of grid development . How are these results relevant to predict the grid configuration expressed in 3D , and that can be tested in a flying bat ? Our model points towards a hierarchy of timescales , associated with the emergence of periodical spatial activity of increasing complexity . To establish a relation between our results and a real bat , it is necessary to specify the actual values of the temporal and spatial parameters of our model , to obtain a time scale for the development of the grids that we can then compare with experimental findings . If we take time steps of size Δt = 10 ms and an average bat velocity of v = 1 m/s , the small environment used in most of our simulations will correspond to a cubic room of size L = 2 . 5 m . Then , with this choice of parameters , grids are formed with a field spacing of 2 . 5 m × 0 . 55 ≈ 1 . 4 m and an interlayer distance of 2 . 5m×0 . 55×63≈1 . 1m . The time scale of grid formation can be calculated considering that 1 million simulation time steps correspond to 10 , 000 s or nearly 3 hr . Our model then predicts , in an environment the size of ours , the presence of ( i ) triplets of fields forming roughly equilateral triangles in ≈10–12 hr of continuous flight , ( ii ) hexagons in ≈15–18 hr , and finally ( iii ) different units that achieve a common orientation after ≈30–35 hr . These time scales do not seem very different from those predicted by the same model for the development of grids in a two-dimensional environment of similar linear size ( relative to the grid spacing ) . Figure 6 in Si et al . ( 2012 ) indicates a time scale of about 20 , 000 s , or 5–6 hr , for the development of gridness in 2D . At the same time , this moderate increase in grid formation time might make it comparable to the flight time available for spatial learning during bat experiments . In these conditions , even the weak , sub-linear dependence of time scales with volume , that we do observe , may be sufficient to determine a switch between the possibility of forming regular structures and leaving them beyond reach . A regular tiling of the environment ( either in the form of an FCC or of an HCP lattice ) is a different story , even though it would be the optimal arrangement from an information-theoretic perspective ( Mathis et al . , 2014 ) . The total simulation time would correspond to a maximum of ≈80–90 hr of continuous bat flight in the ( 2 . 5 m ) 3 volume . This time is just a lower bound for the time necessary to form a regular tiling of the environment , and likely a loose one , as our simulations do not seem to be converging towards one of them . These considerations suggest that bats may form a partially regular 3D tiling of the environment at most once , and then possibly only if constrained to fly for a prolonged time in a rather small cage , while a completely regular , crystalline tiling of space seems to be hardly in the range of time available to real bats . In conclusion , the presence of an additional dimension does not seem to preclude the appearance of some orderly arrangement of the fields in mEC units of bats . Nevertheless , this order might express only a partial set of the full spectrum of potential three-dimensional symmetry properties . It might be still sufficient to distinguish the activity of these cells from a random multi-peaked pattern , but it would place it at a substantial distance from a perfectly regular pattern , too . In our model this distance varies across a population of cells: some of them show only small deviations from perfect symmetry . We thus cannot exclude the possibility of finding some of these extreme cases in real animals . At the same time , the vast majority of simulated grid cells are very far from a perfectly regular arrangement and while the number of units actually found in the tails of the distribution of scores might be strongly dependent on some specific factors of the development of grid cells , the bulk of grid-like but imperfect cells can be regarded as a robust aspect of our model , and might extend to very different situations of 3D grid cell development , possibly including other species experiencing three-dimensional navigation . The triangular tile is the minimal structure associated with regular volume tessellations . The two properties defining any regular triangle are the length of the side and the internal angle . Therefore , to characterize the local structure of the grid pattern in an individual unit we extract these two properties from the spikes it produces . Firstly , from our three-dimensional rate maps we generate a representative number of spike pairs through a Poisson process to construct the distribution of distances . Typically , this distribution is highly multi-peaked , where the first peak corresponds to distances between intra-field spikes , the second peak between spikes belonging to neighboring fields , and subsequent peaks between spikes in non-adjacent fields . Since the length of the side of the tiling triangle in a regular pattern would correspond to the location of the second peak , we define a range of distances around this peak as a filter condition to declare spikes belonging to neighboring fields . The limits of this range were defined by the surrounding troughs , if they exist , or fixed to 0 . 5 d and 1 . 4 d , if they do not , where d is the distance corresponding to the second peak , declared as the grid distance of the unit . As a control condition , we generate a distribution of pseudo-spikes from reshuffling spike–cell combinations and randomly reassigning spikes to different units , thus removing the field structure of the activity of each cell . Therefore , distances between pseudo-spikes are unimodally distributed . Secondly , triplets of spikes were putatively classified as belonging to neighboring fields based on distance filtering in the previous range , and the three internal angles determined . These three angles were pooled together and accumulated in an overall angular distribution . The distribution of angles so obtained for the spiking activity and the control condition were different and their ratio was used to characterize the angle subtended in the triangular pattern . Typically ( in the asymptotical state ) , this ratio was unimodal and distributed asymmetrically around a peak . We defined the characteristic angle as the median of the above-chance distribution ( ratio values above unity indicate an above-chance condition or , in other words , angles more frequently obtained than randomly ) and the significance of the angle as the maximum of the ratio distribution . FCC and HCP differences in the configuration of fields generate distinct symmetry properties for the two arrangements . These symmetries are reflected in the autocorrelograms that can be extracted from them . In the same way as the autocorrelogram of a hexagonal grid is a hexagonal grid , calculating the autocorrelogram of FCC ( using function 20 ) just reproduces the same configuration ( Figure 2 , bottom left ) of fields , with six symmetric pairs of equivalent peaks surrounding the central one . Indeed the symmetries of the structure are such that one can find four planes passing through the origin which contain peaks arranged in a hexagonal way; these planes form angles of 72° and are all equivalent . The case is different for HCP , where the central symmetry is missing . In this case the autocorrelogram extracted from Equation 25 does not reproduce the original form of the function . In Figure 2 , bottom right , one can see that the autocorrelogram presents nine pairs of peaks around the central one . But in this case these peaks are not all the same . The HCP structure is again periodical for translations along a plane , generating six peaks of height 1 , like those of FCC ( Figure 2 , purple peaks ) . As the structure is translated out of this preferred plane , the ABAB arrangement of the HCP layers is such that there are no translations that reproduce the exact same configuration of fields , in the autocorrelogram . The six peaks above the central one ( Figure 2 , orange peaks ) are indeed half-peaks , corresponding to an overlap of only half ( six out of 12 ) of the peaks of the basic unit . Therefore , although one can identify seven planes with hexagonally arranged peaks on this autocorrelogram , they are not all equivalent to those in FCC . Only one of them contains all the peaks of height 1 and forms an angle of 72° with the other six , which include half-peaks and form an angle of 56° between them . We can then use different measures to quantify the degree of similarity of a unit activity to the FCC and HCP prototypical field arrangements . One measure is based on the autocorrelogram . From this , we first identify the best plane , the one which yields the highest grid score , measured here as the value of the planar autocorrelogram at the origin , that is , the planar autocorrelation over all the slices passing through the origin . Once the best plane has been identified , we use the fact that the FCC has three more planes with hexagonal symmetry , at ∼72° from the best plane and between one another . HCP instead has six of them , again at ∼72° from the best plane , but at ∼56° between them . We then take the slice scores , that is , the planar autocorrelation values on any one slice . We take all the slices at an angle of ∼72° from the best plane and sum the scores of the best triplet of slices with ∼72° of separation ( ζ2 − 4 ) . We then exclude them and take a second triplet of slices again with a ∼72° distance from one another and a distance of ∼56° from the first triplet ( ζ5 − 7 ) . These two numbers tell us about the number of different planes with hexagonal symmetry that can be built from our autocorrelograms . Both scores run from −3 to 3 , as they are the sum of three correlations . We expect ζ2 − 4 to be high for both FCC and HCP arrangements , and its value should be considered as an indicator of the general quality of the grid . ζ5 − 7 instead should be high only for those grids presenting an HCP type of arrangement , but again its value might be affected by the quality of the grid . We thus define a score for the degree of FCC similarity as: ( 31 ) χFCC= ( ζ2−4−ζ5−7 ) /ζ2−4 , that should be close to 1 in the presence of FCC , and to 0 in the HCP case . On the other hand , HCP is characterized by the repetition of the same field positions every two layers , while FCC has a periodicity of three layers . Then another way to characterize the grids is to look for similarities between layers . Since the best plane we calculated indicates the direction of stacking of layers in the HCP ( along its normal vector ) , we can go back to the firing rate map , take slices along this direction ( that is , slices with the same best plane orientation ) and calculate the correlation between planes separated by a two-layer distance ( 2 × λz ) : ( 32 ) χHCP=ρauto ( 2×λz ) . Contrary to the previous score , this one should be close to 1 when HCP is expressed , and to 0 when FCC is . Here we give an example of the expression for the cost function obtained for n = 2 . ( 33 ) H ( ψ2FCC ) =71×k2162+γ× ( 1/648× ( 256K˜ ( k ) +K˜ ( 2k ) +6× ( K˜ ( 22/3k ) +K˜ ( ( 2k ) /3 ) ) ) ) , ( 34 ) H ( ψ2HCP ) =3045×kxy2+1881×kz22601+γ× ( 13468× ( 54K˜ ( 2kz ) +1160K˜ ( kxy ) +1920K˜[kxy+kz]+36×K˜ ( kxy+2kz]+2×K˜ ( 2kxy ) +48×K˜ ( 2kxy+kz ) +9×K˜ ( 2kxy+2kz ) +200×K˜ ( 3kxy ) +36×K˜ ( 3kxy+2kz ) ) ) , where k is the only parameter for spacing in FCC and kxy , kz are the two spacings of HCP , along the horizontal and vertical plane , respectively .	Our ability to navigate through our environment depends on a region of the brain called the hippocampus . In the 1990s it was shown that this structure , which takes its name from the Greek word for ‘seahorse’ owing to its shape , was larger in London taxi drivers than it was in the general population . However , as early as the 1960s , experiments in rats had revealed that specific cells within the hippocampus—called place cells—fire whenever an animal is in a particular location , and thus enable the animal to build up a map of its environment . In 2014 , the scientist who discovered place cells shared the Nobel Prize in Physiology or Medicine with two neuroscientists who had discovered an additional type of cell that is involved in navigation . These grid cells , which are located in a region of the brain that provides input to the hippocampus , ‘fire’ at multiple points in space . When the scientists who discovered grid cells plotted these points in two dimensions , they formed a grid of tessellating triangles that spanned the entire area . However , many animals , including aquatic mammals , monkeys and bats , navigate in three dimensions rather than two . This raises an obvious question: can grid cells also represent three-dimensional space ? Stella and Treves have addressed this issue by constructing a computer model that simulates grid cell activity in a virtual bat flying through a virtual room . The model reveals that grid cells switch from firing largely at random to firing in some semblance of a three-dimensional pattern relatively quickly . However , this pattern bears little resemblance to the highly ordered arrangement seen in two dimensions . Indeed , the model suggests that a bat flying at 1 metre per second around a room that measured 2 . 5 × 2 . 5 × 2 . 5 metres would need to fly continuously for a very long time ( at least 80 hours ) before such a pattern could be established in three dimensions . This suggests that the regular tessellation shown by grid cells in two dimensions might not be routinely established in three dimensions . Instead , simpler ‘precursor’ firing patterns may form over shorter periods of time , providing a looser mapping of three-dimensional space .	[ "Abstract", "Introduction", "Model", "Results", "Discussion", "Materials", "and", "methods" ]	[ "neuroscience" ]	2015	The self-organization of grid cells in 3D
Brian 2 allows scientists to simply and efficiently simulate spiking neural network models . These models can feature novel dynamical equations , their interactions with the environment , and experimental protocols . To preserve high performance when defining new models , most simulators offer two options: low-level programming or description languages . The first option requires expertise , is prone to errors , and is problematic for reproducibility . The second option cannot describe all aspects of a computational experiment , such as the potentially complex logic of a stimulation protocol . Brian addresses these issues using runtime code generation . Scientists write code with simple and concise high-level descriptions , and Brian transforms them into efficient low-level code that can run interleaved with their code . We illustrate this with several challenging examples: a plastic model of the pyloric network , a closed-loop sensorimotor model , a programmatic exploration of a neuron model , and an auditory model with real-time input . Neural simulators are increasingly used to develop models of the nervous system , at different scales and in a variety of contexts ( Brette et al . , 2007 ) . These simulators generally have to find a trade-off between performance and the flexibility to easily define new models and computational experiments . Brian 2 is a complete rewrite of the Brian simulator designed to solve this apparent dichotomy using the technique of code generation . The design is based around two fundamental ideas . Firstly , it is equation based: defining new neural models should be no more difficult than writing down their equations . Secondly , the computational experiment is fundamental: the interactions between neurons , environment and experimental protocols are as important as the neural model itself . We cover these points in more detail in the following paragraphs . Popular tools for simulating spiking neurons and networks of such neurons are NEURON ( Carnevale and Hines , 2006 ) , GENESIS ( Bower and Beeman , 1998 ) , NEST ( Gewaltig and Diesmann , 2007 ) , and Brian ( Goodman and Brette , 2008; Goodman and Brette , 2009; Goodman and Brette , 2013 ) . Most of these simulators come with a library of standard models that the user can choose from . However , we argue that to be maximally useful for research , a simulator should also be designed to facilitate work that goes beyond what is known at the time that the tool is created , and therefore enable the user to investigate new mechanisms . Simulators take widely different approaches to this issue . For some simulators , adding new mechanisms requires specifying them in a low-level programming language such as C++ , and integrating them with the simulator code ( e . g . NEST ) . Amongst these , some provide domain-specific languages , for example NMODL ( Hines and Carnevale , 2000 , for NEURON ) or NESTML ( Plotnikov et al . , 2016 , for NEST ) , and tools to transform these descriptions into compiled modules that can then be used in simulation scripts . Finally , the Brian simulator has been built around mathematical model descriptions that are part of the simulation script itself . Another approach to model definitions has been established by the development of simulator-independent markup languages , for example NeuroML/LEMS ( Gleeson et al . , 2010; Cannon et al . , 2014 ) and NineML ( Raikov et al . , 2011 ) . However , markup languages address only part of the problem . A computational experiment is not fully specified by a neural model: it also includes a particular experimental protocol ( set of rules defining the experiment ) , for example a sequence of visual stimuli . Capturing the full range of potential protocols cannot be done with a purely declarative markup language , but is straightforward in a general purpose programming language . For this reason , the Brian simulator combines the model descriptions with a procedural , computational experiment approach: a simulation is a user script written in Python , with models described in their mathematical form , without any reference to predefined models . This script may implement arbitrary protocols by loading data , defining models , running simulations and analysing results . Due to Python’s expressiveness , there is no limit on the structure of the computational experiment: stimuli can be changed in a loop , or presented conditionally based on the results of the simulation , etc . This flexibility can only be obtained with a general-purpose programming language and is necessary to specify the full range of computational experiments that scientists are interested in . While the procedural , equation-oriented approach addresses the issue of flexibility for both the modelling and the computational experiment , it comes at the cost of reduced performance , especially for small-scale models that do not benefit much from vectorisation techniques ( Brette and Goodman , 2011 ) . The reduced performance results from the use of an interpreted language to implement arbitrary models , instead of the use of pre-compiled code for a set of previously defined models . Thus , simulators generally have to find a trade-off between flexibility and performance , and previous approaches have often chosen one over the other . In practice , this makes computational experiments that are based on non-standard models either difficult to implement or slow to perform . We will describe four case studies in this article: exploring unconventional plasticity rules for a small neural circuit ( case study 1 , Figure 1 , Figure 2 ) ; running a model of a sensorimotor loop ( case study 2 , Figure 3 ) ; determining the spiking threshold of a complex model by bisection ( case study 3 , Figure 4 , Figure 5 ) ; and running an auditory model with real-time input from a microphone ( case study 4 , Figure 6 , Figure 7 ) . Brian 2 solves the performance-flexibility trade-off using the technique of code generation ( Goodman , 2010; Stimberg et al . , 2014; Blundell et al . , 2018 ) . The term code generation here refers to the process of automatically transforming a high-level user-defined model into executable code in a computationally efficient low-level language , compiling it in the background and running it without requiring any actions from the user . This generated code is inserted within the flow of the simulation script , which makes it compatible with the procedural approach . Code generation is not only used to run the models but also to build them , and therefore also accelerates stages such as synapse creation . The code generation framework has been designed to be extensible on several levels . On a general level , code generation targets can be added to generate code for other architectures , for example graphical processing units , from the same simulation description . On a more specific level , new functionality can be added by providing a small amount of code written in the target language , for example to connect the simulation to an input device . Implementing this solution in a way that is transparent to the user requires solving important design and computational problems , which we will describe in the following . We will explain the key design decisions by starting from the requirements that motivated them . Note that from now on we will use the term ‘Brian’ as referring to its latest version , that is Brian 2 , and only use ‘Brian 1’ and ‘Brian 2’ when discussing differences between them . Before discussing the requirements , we start by motivating the choice of programming language . Python is a high-level language , that is , it is abstracted from machine level details and highly readable ( indeed , it is often described as ‘executable pseudocode’ ) . In this sense , it is higher level than C++ , for example , which in this article we will refer to as a low-level language ( since we will not need to refer to even lower level languages such as assembly language ) . The use of a high-level language is important for scientific software because the majority of scientists are not trained programmers , and high-level languages are generally easier to learn and use , and lead to shorter code that is easier to debug . This last point , and the fact that Python is a very popular general purpose programming language with excellent built-in and third party tools , is also important for reducing development time , enabling the developers to be more efficient . It is now widely recognised that Python is well suited to scientific software , and it is commonly used in computational neuroscience ( Davison et al . , 2009; Muller et al . , 2015 ) . Note that expert level Python knowledge is not necessary for using Brian or the Python interfaces for other simulators . We now move on to the major design requirements . In this section , we give a high level overview of the major decisions . A detailed analysis of the case studies and the features of Brian they use can be found in Appendix 1 . Source code for the case studies has been deposited in a repository at https://github . com/brian-team/brian2_paper_examples ( Stimberg et al . , 2019a; copy archived at https://github . com/elifesciences-publications/brian2_paper_examples ) . We have described some of the key design choices we made for version 2 of the Brian simulator . These represent a particular balance between the conflicting demands of flexibility , ease-of-use , features and performance , and we now compare the results of these choices to other available options for simulations . There are two main differences of approach between Brian and other simulators . Firstly , we require model definitions to be explicit . Users are required to give the full set of equations and parameters that define the model , rather than using ‘standard’ model names and default parameters ( cf . Brette , 2012 ) . This approach requires a slightly higher initial investment of effort from the user , but ensures that users know precisely what their model is doing and reduces the risk of a difference between the implementation of the model and the description of it in a paper ( see discussion above ) . One limitation of this approach is that it makes it more difficult to design tools to programmatically inspect a model , for example to identify and shut down all inhibitory currents ( although note that this issue remains for languages such as NeuroML and NineML that are primarily based on standard models as they include the ability to define arbitrary equations ) . The second main difference is that we consider the complete computational experiment to be fundamental , and so everything is tightly integrated to the extent that an entire model can be specified in a single , readable file , including equations , protocols , data analysis , etc . In Neuron and NEST , model definitions are separate from the computational experiment script , and indeed written in an entirely different language ( see below ) . This adds complexity and increases the chance of errors . In NeuroML and NineML , there is no way of specifying arbitrary computational experiments . One counter-argument to this approach is that clearly separating model definitions may reduce the effort in re-using models or programmatically comparing them ( as in Podlaski et al . , 2017 ) . A consequence of the requirement to make model definitions explicit , and an important feature for doing novel research , is that the simulator must support arbitrary user-specified equations . This is available in Neuron via the NMODL description format ( Hines and Carnevale , 2000 ) , and in a limited form in NEST using NESTML ( Plotnikov et al . , 2016 ) . NeuroML and NineML now both include the option for specifying arbitrary equations , although the level of simulator support for these aspects of the standards is unclear . While some level of support for arbitrary model equations is now fairly widespread in simulators , Brian was the first to make this a fundamental , core concept that is applied universally . Some simulators that have since followed this approach include DynaSim ( Sherfey et al . , 2018 ) , which is based on MATLAB , and ANNarchy ( Vitay et al . , 2015 ) . Other new simulators have taken an alternative approach , such as Xolotl ( Gorur-Shandilya et al . , 2018 ) which is based on building hierarchical representations of neurons from a library of basic components . One aspect of the equation-based approach that is missing from other simulators is the specification of additional defining network features , such as synaptic connectivity patterns , in an equally flexible , equation-oriented way . Neuron is focused on single neuron modelling rather than networks , and only supports directly setting the connectivity synapse-by-synapse . NEST , PyNN ( Davison et al . , 2008 ) , NeuroML , and NineML support this too , and also include some predefined general connectivity patterns such as one-to-one and all-to-all . NEST further includes a system for specifying connectivity via a ‘connection set algebra’ ( Djurfeldt , 2012 ) allowing for combinations of a few core types of connectivity . However , none have yet followed Brian in allowing the user to specify connectivity patterns via equations , as is commonly done in research papers . Brian is released under the free and open CeCILL 2 license . Development takes place in a public code repository at https://github . com/brian-team/brian2 ( Brian contributors , 2019 ) . All examples in this article have been simulated with Brian 2 version 2 . 2 . 2 . 1 ( Stimberg et al . , 2019b ) . Brian has a permanent core team of three developers ( the authors of this paper ) , and regularly receives substantial contributions from a number of students , postdocs and users ( see Acknowledgements ) . Code is continuously and automatically checked against a comprehensive test suite run on all platforms , with almost complete coverage . Extensive documentation , including installation instructions , is hosted at http://brian2 . readthedocs . org . Brian is available for Python 2 and 3 , and for the operating systems Windows , OS X and Linux; our download statistics show that all these versions are in active use . More information can be found at http://briansimulator . org/ .	Simulating the brain starts with understanding the activity of a single neuron . From there , it quickly gets very complicated . To reconstruct the brain with computers , neuroscientists have to first understand how one brain cell communicates with another using electrical and chemical signals , and then describe these events using code . At this point , neuroscientists can begin to build digital copies of complex neural networks to learn more about how those networks interpret and process information . To do this , computational neuroscientists have developed simulators that take models for how the brain works to simulate neural networks . These simulators need to be able to express many different models , simulate these models accurately , and be relatively easy to use . Unfortunately , simulators that can express a wide range of models tend to require technical expertise from users , or perform poorly; while those capable of simulating models efficiently can only do so for a limited number of models . An approach to increase the range of models simulators can express is to use so-called ‘model description languages’ . These languages describe each element within a model and the relationships between them , but only among a limited set of possibilities , which does not include the environment . This is a problem when attempting to simulate the brain , because a brain is precisely supposed to interact with the outside world . Stimberg et al . set out to develop a simulator that allows neuroscientists to express several neural models in a simple way , while preserving high performance , without using model description languages . Instead of describing each element within a specific model , the simulator generates code derived from equations provided in the model . This code is then inserted into the computational experiments . This means that the simulator generates code specific to each model , allowing it to perform well across a range of models . The result , Brian 2 , is a neural simulator designed to overcome the rigidity of other simulators while maintaining performance . Stimberg et al . illustrate the performance of Brian 2 with a series of computational experiments , showing how Brian 2 can test unconventional models , and demonstrating how users can extend the code to use Brian 2 beyond its built-in capabilities .	[ "Abstract", "Introduction", "Materials", "and", "methods", "Discussion" ]	[ "tools", "and", "resources", "neuroscience" ]	2019	Brian 2, an intuitive and efficient neural simulator
The dynamic assembly of multi-protein complexes underlies fundamental processes in cell biology . A mechanistic understanding of assemblies requires accurate measurement of their stoichiometry , affinity and cooperativity , and frequently consideration of multiple co-existing complexes . Sedimentation velocity analytical ultracentrifugation equipped with fluorescence detection ( FDS-SV ) allows the characterization of protein complexes free in solution with high size resolution , at concentrations in the nanomolar and picomolar range . Here , we extend the capabilities of FDS-SV with a single excitation wavelength from single-component to multi-component detection using photoswitchable fluorescent proteins ( psFPs ) . We exploit their characteristic quantum yield of photo-switching to imprint spatio-temporal modulations onto the sedimentation signal that reveal different psFP-tagged protein components in the mixture . This novel approach facilitates studies of heterogeneous multi-protein complexes at orders of magnitude lower concentrations and for higher-affinity systems than previously possible . Using this technique we studied high-affinity interactions between the amino-terminal domains of GluA2 and GluA3 AMPA receptors . The dynamic formation of multi-protein complexes is a key step in the assembly of supramolecular structures and in the regulation of many cellular processes ( Wu , 2013; Li et al . , 2012; Gavin et al . , 2002; Krogan et al . , 2006; Wu and Fuxreiter , 2016 ) . For example , in immunological signal transduction the assembly of adaptor protein complexes into micro-clusters after T-cell activation plays a critical role in the sensitivity and specificity of activation ( Sherman et al . , 2011; Dustin and Groves , 2012 ) . Another well-known multi-protein complex is the post-synaptic density , a large structure assembled via interactions between many different scaffolding proteins , signaling proteins and ligand gated ion channels , that regulates postsynaptic neurotransmission and plasticity ( Kennedy , 2000; Ferré et al . , 2007; Kumar and Mayer , 2012 ) . Many of the protein interactions involved in the assembly of such molecular machinery are multivalent ( Li et al . , 2012; Houtman et al . , 2006; Coussens et al . , 2013 ) . This often allows structurally polymorph complexes to co-exist ( Wu and Fuxreiter , 2016 ) , posing formidable challenges for any biophysical method to elucidate basic architectural principles and driving forces , which requires the study of reversible interactions of multiple protein components with multiple states . Sedimentation velocity analytical ultracentrifugation ( SV ) is a classical technique that allows determination of the number , size , and shape of reversibly formed protein complexes , and provides information on their affinity , stoichiometry and binding kinetics ( Schuck , 2013 , 2015 ) . Though a long established technique , it is worth recapitulating the basic principles of SV ( Figure 1 ) . In SV the spatio-temporal evolution of macromolecular concentration profiles in a sample solution after application of a strong centrifugal field is optically monitored in real-time . SV has unique opportunities for studying protein interactions , since—different from separation techniques—faster sedimenting protein complexes will always remain in a bath of slower-sedimenting constituent components , such that the association/dissociation of non-covalent complexes is maintained throughout the experiment ( Figure 1 ) . Since sedimentation takes place free in solution , the analysis can be based on first principles and mathematical models for the sedimentation/diffusion process , and modern size-distribution analysis results in sedimentation coefficient distributions with high hydrodynamic resolution . Thus , SV has emerged as a powerful technique in the study of the solution state behavior of complex interacting systems of macromolecules , including ion-channels , adaptor proteins , membrane proteins , nucleic acids , and carbohydrates ( Kumar and Mayer , 2012; Houtman et al . , 2005; le Maire et al . , 2008 Niewiarowski et al . , 2010; Padrick and Brautigam , 2011; Harding et al . , 2015; Jose et al . , 2012 ) . Extended to multi-signal analysis SV can distinguish different sedimenting components from characteristic extinction properties , directly revealing binding stoichiometries and resolving co-existing complexes ( Houtman et al . , 2006; Coussens et al . , 2013; Padrick and Brautigam , 2011; Balbo et al . , 2005; Barda-Saad et al . , 2010 ) . For example , through applications of this approach an essential mechanism for the formation of signaling particles in T-cell activation ( Houtman et al . , 2006; Coussens et al . , 2013; Barda-Saad et al . , 2010 ) was discovered to be the multivalent intracellular oligomerization of LAT via three-component adaptor protein complexes ( Houtman et al . , 2006 ) . Such biophysical multi-protein solution studies naturally complement super-resolution fluorescence imaging and co-localization studies of live cells ( Sherman et al . , 2011; Houtman et al . , 2006; Coussens et al . , 2013; Barda-Saad et al . , 2010 ) . But , unfortunately , traditional SV is limited in several ways by optical detection systems that generally require the use of micromolar concentrations of purified proteins . 10 . 7554/eLife . 17812 . 003Figure 1 . Concentration profiles in a sedimentation velocity experiment . Two different macromolecular components are depicted ( blue and red ) reversibly forming a complex ( magenta ) . As a result of centrifugal force at 200 , 000–300 , 000 g , macromolecules sediment at a rate determined by their mass , density , and Stokes radius ( or translational friction coefficient ) ( Svedberg and Rinde , 1924 ) . The velocity of sedimentation normalized relative to the centrifugal field strength is expressed in the molecular sedimentation coefficient s . With time , transport clears the region of the solution column closest to the center of rotation and a moving front is formed -- the sedimentation boundary – that separates the cleared zone from a zone of constant concentration named the solution plateau region . While the boundary moves with time ( dashed vs solid line ) , the concentration in the plateau region continuously decreases , solely due the radial geometry of sedimentation resulting in an increase in intermolecular distances ( for a detailed description , see Schuck et al . , 2015 ) . If protein interactions cause complexes to form , these generally sediment faster and therefore migrate through a bath of slower sedimenting free constituent components . This allows association/dissociation reactions to continuously occur in a way that reflects equilibrium and kinetic properties of the interaction , at the same time as the complex boundaries are hydrodynamically resolved ( Schuck , 2010 ) . The temporal evolution of the boundary shapes is governed by macromolecular diffusion and polydispersity , and the latter can be extracted by mathematical modeling of experimental data in form of sedimentation coefficient distributions ( Schuck , 2000 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 003 Recently , analytical ultracentrifugation was enhanced by the availability of a commercial fluorescence optical detection system ( FDS ) , that uses confocal detection radially scanning the sample in the spinning rotor ( MacGregor et al . , 2004 ) ( Figure 2a ) . After accounting for characteristic data features , the FDS allows highly quantitative analyses of the sedimentation process ( Zhao et al . , 2013b ) , extending the sensitivity of SV by several orders of magnitude into the low picomolar range ( Zhao et al . , 2014a; Le Roy et al . , 2015 ) . This enables the measurement of binding energies in self- and hetero-association with very high affinity ( Zhao et al . , 2012; 2013a; 2014a ) , and also makes it possible to study proteins that are relatively scarce , not well purified , and in some cases even in cell extracts ( Le Roy et al . , 2015; Polling et al . , 2013; Kingsbury and Laue , 2011; Kokona et al . , 2015 ) . Unfortunately , these advantages come with a drawback of having only a single excitation wavelength , either 488 nm or 561 nm , due to the technical constraints of accommodating a moveable confocal optical system inside the evacuated rotor chamber of the ultracentrifuge . Thus , spectral discrimination of multiple components is not available in AUC fluorescence detection , which significantly limits the study of multi-protein complexes . 10 . 7554/eLife . 17812 . 004Figure 2 . Principle of fluorescence detected sedimentation velocity and optical switching of psFPs . ( a ) Schematic setup: A rotating sample solution is scanned radially in a confocal configuration with 488 nm excitation ( 13 mW unless mentioned otherwise ) , inducing slow photoswitching of psFPs . Centrifugal forces cause strongly size-dependent migration , as depicted in Figure 1 . Optionally , localized exposure at 488 nm or uniform illumination at 405 nm can further modulate the spatio-temporal signal . ( b-f ) Radial fluorescence scans ( dots , color indicating times in order purple-blue-green-yellow-red; every 2nd scan shown ) during sedimentation at 50 , 000 rpm and 20°C for different fluorophores and illumination conditions . More detailed inspection of the data is possible from the associated movies . Solid lines are the best-fit with a single-species ( c to f ) or distribution ( b ) model for the sedimentation/diffusion/photoswitching process; residuals are shown in the lower panels . ( b ) For DL488-GluA2 ( 5 nM ) only a small depletion of plateau signal with time occurs , due to sample dilution as geometrically predicted from radial migration in the sector-shaped sample solution . ( see Video 1 ) ( c ) rsEGFP ( 30 nM ) exhibits an exponential depletion of the sedimentation signal ( see Video 2 ) . ( d ) Exposure in the scanning beam causes Padron ( 20 nM ) to switch from predominantly dark to a fluorescent state , causing an exponentially saturating signal increase with time ( see Video 3 ) . ( e ) The sedimentation of 10 nM rsEGFP2-GluA3 is recorded with a 25 mW scanning beam , interrupted by 120 s exposures with 405 nm light at time points 40 min ( prior to the purple scans ) , 64 min ( prior to the blue scans ) , and 102 min ( prior to the green scans ) , each time switching fluorophores from dark state back to the fluorescent state ( see Video 4 ) . ( f ) 5 min into the sedimentation run of 5 nM rsEGFP2-GluA3 , a localized initial trough was generated by holding the scanning beam stationary at 6 . 5 cm for 20 min , locally causing strong conversion of fluorophores into the dark state . Standard scans of the sedimentation process follow , highlighting diffusion into the trough superimposed by migration and slow switching off ( see Video 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 004 In the present work , we embark on a different approach for multi-component analysis based on photophysical properties of photoswitchable fluorescent proteins ( psFPs ) . The psFPs make up a class of fluorescent proteins that can be actively switched between fluorescent and non-fluorescent states using different wavelengths of illumination . While they have been engineered for entirely different purposes in nanoscience and fluorescence imaging , we have previously observed that under the illumination conditions of FDS-SV they are induced to slowly switch by virtue of the excitation light when radially scanning the spinning sample ( Zhao et al . , 2014b ) . Even though the mechanism of photoswitching in psFPs generally may involve multiple states , in the low-power exposure that occurs during sedimentation using the FDS the process is quantitatively modeled very well as a single exponential . This is exploited and further developed in the present work . Different classes of psFPs exhibit different time-courses of photoswitching in FDS-SV , and may be switched on to a fluorescent state or switched off to a dark state . We show that this process is highly quantitative , and how this signal change can be manipulated spatially and temporally during sedimentation . The new spatio-temporal signal dimension is folded into the computational analysis of the sedimentation process , and thus offers an avenue for monochromatic multi-component ( MCMC ) detection . Using existing commercial FDS instrumentation , the MCMC approach allows us to simultaneously determine separate sedimentation coefficient distributions for each class of fluorophore , which can be used to determine the identity and binding stoichiometry of hydrodynamically resolved complexes of psFP-tagged proteins . We first demonstrate experimental proof of principle of exploiting the photo-switching kinetics as a novel aspect of fluorophore analysis . Using psFPs commonly employed in super-resolution microscopy , we show an example for the identification of binding partners in a three-component protein mixture . We then use this approach to study the high-affinity interactions of glutamate receptor GluA2 and GluA3 amino terminal domains , which engage in competitive homo-dimerization and hetero-dimerization processes that are thought to control the combination of receptor subtypes into diverse homomeric and heteromeric ion channel tetramers with different gating properties ( Kumar and Mayer , 2012; Rossmann et al . , 2011; Herguedas et al . , 2016 ) . Fluorescence SV signals of psFPs are highly unusual when compared to the temporal evolution of concentration profiles in traditional AUC ( Figure 2b ) , in that they exhibit sedimentation boundaries modulated by characteristic signal magnification or diminution on the time-scale of sedimentation ( Figure 2c , d ) . This is caused by the 488 nm excitation beam of the FDS scanner which induces photoswitching between fluorescent and dark states . But in contrast to typical power densities of ~ kW/cm2 for photoswitching on the millisecond time-scale ( Grotjohann et al . , 2012 ) , the transient exposure in FDS-SV during radial scanning and sample rotation leads to a time-averaged incident power density that is ~105-fold weaker . This slows the photoswitching kinetics down to the time-scale of hours , commensurate with the sedimentation process in SV . A prerequisite for exploiting this new temporal dimension for the multi-component decomposition of fluorescence SV data is the ability to precisely describe the signal evolution of the individual psFPs . We have developed a model Equation 4 that assumes a single-step process with constant quantum efficiency for switching , while accounting for the radially non-uniform exposure during scanning ( caused by psFPs transitioning through the beam in a shorter time as they migrate to higher radii at the same angular velocity ) . Combined with a description of molecular sedimentation and diffusion , the model predicts signal boundaries to be subject to an exponential overall signal modulation with slightly radially sloping solution plateaus , with a small positive slope for those switching off ( Figure 2c ) and a small negative slope for FPs switching on ( Figure 2d ) under 488 nm illumination . As shown in the examples of Figure 2 , fits of this model to within the noise of data acquisition can routinely be achieved for the fluorescence sedimentation data of diverse FPs , including , for example , the rapidly de-activating rsEGFP ( Grotjohann et al . , 2012 ) and the strongly activating Padron ( Andresen et al . , 2008 ) ( Figure 2c , d ) . From the fits of the single component samples , we can obtain the relative amplitude and time-constant of photoswitching for different fluorophores , which serves as a highly reproducible , characteristic temporal tag . For example , for rsEGFP with a scanning beam of 13 mW we measure a depletion rate of 5 . 08 [4 . 97–5 . 19; 95% CI]×10–4/sec , approaching a final fluorescence of 13 . 4 [13 . 0–13 . 9; 95% CI]% its initial value , associated with a particle sedimentation coefficient of 2 . 50 [2 . 45–2 . 54; 95% CI] S and apparent molar mass of 29 . 5 [26 . 5–33 . 0; 95% CI] kDa . By contrast , as previously established ( Zhao et al . , 2014b ) , no photophysical processes are detectable under this illumination for other fluorophores such as fluorescein-based DL488 ( Figure 2b ) and standard EGFP . The strikingly different signal patterns in sedimentation of psFPs can be utilized for the computational decomposition of sedimentation data of mixtures into separate sedimentation coefficient distributions for each differently tagged component . As an initial proof of principle , Figure 3a shows sedimentation data of a mixture of rsEGFP2 and FITC-BSA . The shape of the signal boundaries is governed jointly by the signal time-domain , polydispersity in the sedimentation coefficient distribution , as well as diffusion . The initially strongly decreasing plateau intensity over time is a characteristics of rsEGFP2 signals that can be readily visually discerned and distinguished from the stable fluorescein signal that causes the plateaus to stabilize at approximately half the total loading signal . In addition to the decreasing plateau , the decreasing amplitude of the sedimentation boundary also carries significant information on the time-dependent rsEGFP2 signal . This is highlighted in Figure 3b by the failure of an impostor fit with conventional boundary analysis , in which the decreasing plateau level are compensated for with ad hoc inclusion of time-dependent baseline offsets ( usually absent in FDS-SV [Schuck et al . , 2015; Zhao et al . , 2013b] ) . 10 . 7554/eLife . 17812 . 005Video 1 . Example of conventional boundaries from molecules with stable signal . Time-course of signal profiles for DL488-GluA2 sedimenting at 50 , 000 rpm , as shown in Figure 2b . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 00510 . 7554/eLife . 17812 . 006Video 2 . Example of decreasing boundaries of molecules switching off with 488 nm exposure . Time-course of signal profiles for rsEGFP sedimenting at 50 , 000 rpm while undergoing continuous slow signal depletion through the 488 nm illumination of the FDS scanner , as shown in Figure 2c . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 00610 . 7554/eLife . 17812 . 007Video 3 . Example of growing boundaries of molecules switching on with 488 nm exposure . Time-course of signal profiles for Padron sedimenting at 50 , 000 rpm while undergoing continuous slow signal amplification through the 488 nm illumination of the FDS scanner , as shown in Figure 2d . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 00710 . 7554/eLife . 17812 . 008Video 4 . Example of blinking boundaries . Time-course of signal profiles for rsEGFP2-GluA3 sedimenting at 50 , 000 rpm while undergoing continuous slow signal depletion through the 488 nm illumination of the FDS scanner , in combination with periodic signal reset through 2 min . pulses of strong 405 nm illumination . ( See also Figure 2e . ) DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 00810 . 7554/eLife . 17812 . 009Video 5 . Example of FRAP-like sedimentation . Initial localized illumination with the stationary 488 nm excitation beam of the FDS creates a trough in the signal of rsEGFP2-GluA3 . The sedimentation/diffusion process at 50 , 000 rpm causes the relative trough to diminish , a process that is superimposed by the standard overall signal depletion from the 488 nm scanner . ( See also Figure 2f ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 00910 . 7554/eLife . 17812 . 010Figure 3 . Mono-chromatic multi-component ( MCMC ) decomposition of mixtures . ( a ) Evolution of radial fluorescence profiles of 20 nM rsEGFP2 and 30 nM FITC-BSA sedimenting at 50 , 000 rpm and 20°C . Raw data are shown as points with higher color temperature indicating the passage of time ( only every second scan shown ) , and solid lines are the best-fit based on the MCMC distribution model with two components , producing residuals as shown in the lower panel attached . ( b ) An impostor fit with a conventional c ( s ) analysis not accounting for time-dependent signal increments , but compensating their effect on plateau levels by artificial inclusion of time-dependent baseline offsets , creates large misfit in the boundary region . ( c ) The resulting sedimentation coefficient distributions of each fluorophore component from the MCMC analysis of the mixture ( solid lines ) and in separate control experiments with individual samples ( dashed lines ) . ( d ) Fluorescence sedimentation data acquired under the same conditions for a mixture of 20 nM rsEGFP2 , 20 nM Padron , and 50 nM anti-GFP mAb , presented in the same format as ( a ) . ( e ) Analogous to ( b ) , showing the best-fit ‘conventional’ boundary model not accounting for time-dependent signal increments while including artificial baseline parameters . ( f ) Resulting sedimentation coefficient distributions from the MCMC analysis ( solid lines ) and standard control experiments ( dashed lines ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 01010 . 7554/eLife . 17812 . 011Figure 3—figure supplement 1 . Sedimentation signals of rsEGFP2-GluA3 . Evolution of radial fluorescence profiles of 5 nM rsEGFP2-GluA3 sedimenting at 50 , 000 rpm and 20°C . Raw data are shown as points with higher color temperature indicating the passage of time ( only every second scan shown ) , and solid lines are the best-fit based on the MCMC distribution model with a single component , producing residuals as shown in the lower panel . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 011 In the MCMC analysis the fluorophore photoswitching parameters were fixed to predetermined values of the individual fluorophores ( run side-by-side in a different rotor position ) . The MCMC decomposition via Equation 5 results in an excellent fit ( Figure 3a ) , with sedimentation coefficient distributions of species with temporal signal characteristics of fluorescein and rsEGFP2 as shown in Figure 3c ( solid lines ) . The decaying component is correctly found to migrate with ~2 . 5 S , and the temporally stable signal is found to sediment in the usual hydrodynamically resolved monomer , dimer , and trimer peaks of BSA . The relative errors in the distinction between rsEGFP2 and fluorescein components are calculated to be 0 . 21% over the entire range ( based on Equation 6 ) . The sedimentation coefficients extracted from the mixture are highly consistent with those obtained in control experiments of individual protein components ( dashed lines in Figure 3c ) , as expected for a mixture of non-interacting species . Fluorescence SV data from a three-protein mixture with interactions is shown in Figure 3d , consisting of two psFPs ( rsEGFP2 and Padron ) , in addition to a molar excess of unlabeled monoclonal anti-GFP antibody ( mAb ) . From visual inspection of the data one can discern a boundary structure with more complex temporal modulation , which , again , cannot be accounted for with any conventional boundary model ( Figure 3e ) . Application of the MCMC decomposition produces an excellent fit , with component sedimentation coefficient distributions ( Figure 3f solid lines ) again being highly consistent with the results from the separately measured Padron and rsEGFP2/mAb mixture ( Figure 3f dashed lines ) . Here , the sedimentation coefficient distribution of rsEGFP2 shows a 7 S peak with no 2 . 5 S peak , consistent with the formation of an antibody/rsEGFP2 complex . Thus , the binding partner of the antibody can be correctly identified by exploiting the new temporal signature dimension characteristic of the psFPs . Based on Equation 6 , the relative error in assignment of component concentrations is 2 . 6% at 2 . 5 S and 1 . 1% at 7 S . Thus , these model systems demonstrate the potential for simultaneous component discrimination and hydrodynamic resolution . When the psFPs are involved in complex formation , the question arises whether the photoswitching kinetics of psFPs may also depend on the complex state . In fact , in modeling the data of Figure 3d , the time-constants of rsEGFP2 were initially constrained to the separately measured values , but subsequent refinement led to a significantly better fit: As may already be visually discerned from comparison with Figure 3—figure supplement 1 , a portion of the fast-moving boundary of the rsEGFP2/mAb complex in Figure 3d remains fluorescent after long time ( α = 0 . 2070 [0 . 2052–0 . 2088 , 95% CI] ) , whereas free rsEGFP2 shows nearly complete conversion to the dark state ( α = 0 . 0274 [0 . 0270–0 . 0281 , 95% CI] ) . Clearly , when bound by the anti-GFP antibody , the contrast of rsEGFP2 is lower than in the unbound state . Similar observations were made with rsEGFP , which exhibits slightly altered photoswitching kinetics in the presence of the anti-GFP antibody ( data not shown ) . These observations were confirmed by rsEGFP and rsEGFP2 fluorescence measurements in a benchtop fluorimeter ( Figure 4a ) , where we found marked alterations in the off rates and on/off contrast ratios for both rsEGFP and rsEGFP2 in the presence of antibody as compared to their unbound state . At the same time , their emission spectral properties changed little with the addition of the antibody ( Figure 4b ) . These effects point to possible conformational changes induced by antibody binding impacting the equilibrium of photophysical states ( see Discussion ) . In general , however , such effects should be irrelevant in applications of this approach where the psFP will be designed as an inert tag rather than offering a binding site . 10 . 7554/eLife . 17812 . 012Figure 4 . Photoswitching behavior and fluorescence emission spectra of EGFP variants in the absence and presence of GFP antibody in a benchtop spectrofluorometer . ( a ) Normalized fluorescence data of 500 nM rsEGFP in the presence ( red ) and absence ( black ) of equimolar anti-GFP mAb , monitored at 508 nm during excitation at 490 nm . Equivalent experiments for rsEGFP2 in the presence ( magenta ) and absence ( blue ) of the same antibody , measured at 502 nm during 483 nm excitation . The incident power density is 3 mW/cm2 . The data shown here represent the average of at least four measurements for each sample , with the line width ( or patch width ) representing the data acquisition errors . ( b ) Normalized fluorescence spectra of 500 nM rsEGFP with excitation at 490 nm in the presence ( red ) and absence ( black ) of equimolar anti-GFP mAb , before ( solid lines ) and after ( dotted lines ) photoswitching . Equivalent emission spectra for rsEGFP2 with 483 nm excitation in the presence ( magenta ) and absence ( blue ) of the same antibody , before ( solid lines ) and after ( dotted lines ) photoswitching . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 012 We have found that several psFPs are suitable for FDS-SV with well-defined photophysical characteristics under the current illumination conditions . Besides rsEGFP , rsEGFP2 and Padron these also include Dronpa ( Zhao et al . , 2014b ) . While older FDS instruments are equipped with a fixed-power 13 mW solid-state laser , newer models operate with an adjustable 50 mW diode laser that offers additional flexibility for modulating photoswitching kinetics by using different power settings for the scanning beam . The optimal choice of psFP and excitation power will be determined by consideration of the temporal conversion characteristics of all psFPs in the mixture , in combination with the expected time-course of sedimentation , as dictated by s-value of the protein complex and the rotor speed . We have implemented Equation 6 in the data analysis software as a design tool that – prior to data acquisition – can guide the psFP selection to maximize the contrast for MCMC based on known psFP characteristics in FDS and the sedimentation coefficient range of the expected protein complexes . For example , the black line in Figure 5 shows the theoretically best switching rate constant for a 2 . 8 S species under standard SV conditions to be between 1–5×10–4/sec . 10 . 7554/eLife . 17812 . 013Figure 5 . Calculated relative concentration error for distinguishing photoswitching from non-switching FPs in SV , with and without blinking . The calculations are based on a decaying signal contribution of the psFP with different rate constants , assuming standard sedimentation conditions ( 12 mm solution column , 20°C , 50 , 000 rpm , similar to data shown in Figure 2c ) and an s-value of 2 . 8 S . Shown are the predictions of Equation 6 for constant scanning ( black ) and blinking ( magenta ) with 2 min reset events occurring at 26 min and 69 min . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 013 It is evident that switching of psFPs is required for discrimination , but at the same time the decaying boundary amplitudes for psFPs that switch off during sedimentation will diminish the information content on such psFPs-tagged species , as the signal disappears at later times . However , rapid photoswitching can be advantageous if we exploit the reversibility of the process to switch it back on . Stopping the 488 nm exposure will cause rsEGFP to spontaneously relax back into the equilibrium fluorescent state on the time-scale of an hour . Thus , passive restoration of the fluorescent signal by simply pausing the data acquisition is possible ( data not shown ) but presents a viable approach only for slowly sedimenting species . However , rapid and virtually complete restoration of the initial psFPs signal is possible by illumination at 405 nm . In this way , for example , rsEGFP and rsEGFP2 can undergo hundreds of on/off cycles with very little loss ( Grotjohann et al . , 2012 ) . To this end , we took advantage of the window in the rotor chamber that usually forms part of the interference optical system , and modified it to allow illumination of the spinning rotor with a high-power 405 nm LED at ~10 mW/cm2 . To achieve a quantitatively uniform reset of fluorescence , a sector-shaped mask was used to produce constant 405 nm exposure times for molecules rotating through the beam at all radii . SV experiments were carried out with several cycles of rapid signal depletion under 488 nm illumination at 25 mW ( producing a depletion rate of 9 . 7×10–4/s for rsEGFP2 ) , alternating with a 2 min pulse of 405 nm illumination ( leading to > 98% recovery of the fluorescence signal of rsEGFP2 ) . The sequence of exposure and scanning events is schematically represented in Figure 6a . An example for the resulting ‘blinking’ boundary for rsEGFP2 is shown in the Video 4 and Figure 2e . An experiment with Padron exhibiting the opposite blinking behavior is shown in Figure 6b . With signal recovery time-points included into the model , blinking SV data could be fitted within the noise of data acquisition , and blinking has been fully implemented in the MCMC decomposition of signals from mixtures with different fluorophores . However , on the basis of the theoretical analysis of information content in Figure 5 ( magenta line ) , for switching processes at rates below 3×10–4/sec , blinking can be detrimental to component discrimination , because it does not allow the fluorophore to ever significantly populate the dark state . 10 . 7554/eLife . 17812 . 014Figure 6 . Blinking fluorescence SV data . ( a ) Cartoon illustrating the timing of events in standard vs . blinking experiments . Dependent on sample history , an initial brief illumination with 405 nm may be applied prior to start of centrifugation . In standard time-domain experiments ( top ) a series of 488 nm scans at constant frequency is initiated shortly after full rotor speed is reached . It causes a gradual depletion of signals for molecules switching off , depicted by the fading blue bar , such as shown in Figure 2c . For blinking experiments ( middle and bottom ) the sequence of scans was briefly interrupted at select time points ( indicated by purple arrows ) by short pulses of high-power 405 nm illumination , switching on molecules that switch off at 488 nm such as rsEGFP2 in Figure 2e , or switching off molecules that switch on at 488 nm such as Padron in ( b ) , respectively . ( b ) Example of blinking data from 30 nM Padron sedimenting at 50 , 000 rpm and 20°C . Fluorescence scans were acquired with 488 nm excitation in ~5 min intervals ( only every second scan shown ) , interrupted three times by 2 min exposure of the spinning rotor to 405 nm light . During the data acquisition cycles the fluorescence signal increases , due to switching on of Padron , but the increase is reversed each time in the 405 nm light that switches Padron to the off state . At these time points , gaps in the boundary overlay can be visually discerned; these appear largely due to the decreasing signal magnitude of the boundary . The complete fluorescence or dark state is not reached in these cycles , and the fluorescence signal magnitude after 405 nm exposure is treated as a fitting parameter . The solid line is the best-fit combined sedimentation/photoswitching model of a single species with apparent molar mass of 26 . 9 kDa and s-value of 2 . 58 S . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 014 For the slow switching case , we followed a different approach to enhance discrimination by creating a spatially localized trough in the psFP fluorescence signal . In contrast to blinking , this can be executed without any instrument modification , simply by positioning the existing FDS detector at a fixed radius for an extended period of time at the beginning of the sedimentation process . Based on the size of the focal spot we estimate the stationary exposure to be on the order of 100-fold stronger as compared to the time-averaged exposure during scanning across the standard radial scan range . Localized switching to the dark state causes a trough in the fluorescence signal ( or a peak for molecules switching on at 488 nm ) . Much like the well-known technique of fluorescence recovery after photobleaching ( FRAP ) , molecules in the dark state that are located initially in the trough will diffuse and exchange with fluorescent molecules from outside , diminishing the trough . In contrast to FRAP , the diffusion processes are coupled to size-dependent migration from sedimentation , which translates and stretches the shape of the trough as it diminishes . Simultaneously , this process is superimposed by the same uniform photoswitching of psFPs that sets in with radial scanning ( Video 5 ) . Even though this appears to be a more complicated process , the evolution of the trough is naturally modeled in the same framework developed above for the sedimentation in quasi-uniform exposure: Since SV models already account for diffusional spread superimposed to sedimentation , the FRAP-like SV model solely requires the additional application of appropriate boundary conditions ( Equation 8 ) at early times . However , in comparison to standard sedimentation boundaries , the resulting data are enriched in several ways , as ( 1 ) additional information on diffusion becomes available; ( 2 ) two additional sedimentation boundaries are generated to improve the precision of sedimentation coefficients; and ( 3 ) in data from mixtures of psFPs with stable fluorescent molecules , the depth of the trough will provide a landmark at early times reporting directly on the concentration of psFPs-tagged components . In applications to the analysis of experimental data , excellent fits were achieved with this model with residuals close to the noise of data acquisition . For example , with the single-species model fitted to the rsEGFP2-GluA3 data shown in Figure 2f , the s-value is 4 . 59 [4 . 50–4 . 68 , 68% CI] S , consistent with the value obtained from blinking and from constant illumination experiments . Glutamate receptor ion channels ( iGluRs ) assemble into homo- and hetero-tetramers with their subtype composition dictating the properties of synaptic transmission ( Geiger et al . , 1995 ) . It is thought that the strength of homo- and hetero-dimerization of the amino-terminal domains of different iGluRs governs their assembly and tetrameric architecture ( Rossmann et al . , 2011; Kumar et al . , 2011 ) . Therefore , the binding affinities of iGluRs have been of significant interest ( Zhao et al . , 2012; Rossmann et al . , 2011; Karakas et al . , 2011; Clayton et al . , 2009; Kumar et al . , 2009 ) . In a pioneering study , homo- and hetero-dimerization affinities for different AMPA receptor amino terminal domains ( ATDs ) were determined by standard FDS-SV ( Rossmann et al . , 2011 ) . However , for analysis of hetero-oligomerization only one component could be monitored , and the FAM-label used in that study was later found to induce artificially strong homo-dimerization of GluA2 ATD ( Zhao et al . , 2013c ) , with its effect on interactions with the other AMPA receptor subunit ATDs unknown . Recent work , which has resulted in structures of full length heteromeric AMPA receptors assembled from GluA2 and GluA3 ( Herguedas et al . , 2016 ) provides a compelling reason for accurately measuring the affinity for assembly of GluA2/GluA3 ATD heterodimers . Using Dylight488-labeled GluA2-ATD ( DL488-GluA2 ) and a fusion protein of GluA3-ATD and rsEGFP2 ( rsEGFP2-GluA3 ) , we take advantage of the component discrimination in MCMC to monitor simultaneously the homomeric and heteromeric complexes , which consist of GluA2 monomers and homodimers , GluA3 monomers and homodimers , and GluA2/GluA3 heterodimers ( Figure 7 ) . In the initial control experiments , a dilution series of DL488-GluA2 ( Figure 8a , b ) yielded a homo-dimerization Kd , 22 of 24 [9–59; 95%CI] nM; no homo-dimerization was detected for rsEGFP2-GluA3 at concentrations up to 30 nM ( Figure 8c ) . These results are consistent with previously measured Kd22 and Kd33-values for the homo-dimerization of GluA2 and GluA3 , 20 . 5 nM [15 . 9–26 . 4; 95% CI] ( Zhao et al . , 2013a ) and 5 . 6 [1 . 7–14] μM ( Zhao et al . , 2012 ) , respectively , while by contrast , the homo-dimerization Kd , 22 of 1 . 8 nM for FAM-labeled GluA2 was 11-fold lower ( Rossmann et al . , 2011 ) . 10 . 7554/eLife . 17812 . 015Figure 7 . Cartoon of the mixed homo- and hetero-dimerization of GluA2-ATD and GluA3-ATD . DL488-GluA2 ( magenta ) exhibits moderately strong homo-dimerization , while rsEGFP2-GluA3 ( blue ) undergoes weak homo-dimerization . Both are competitive with a strong hetero-association . DL488-GluA2 provides a constant fluorescent signal contribution , whereas rsEGFP2-GluA3 signals are modulated and can be recognized in the time-domain of the signal . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 01510 . 7554/eLife . 17812 . 016Figure 8 . Analysis of the self-association of DL488-GluA2 and rsEGFP2-GluA3 . ( a ) DL488-GluA2 samples at a series of concentrations from 0 . 2 nM to 200 nM were sedimented at 50 , 000 rpm , 20°C and FDS scans were acquired with 488 nm excitation . The resulting SV data were fit to a c ( s ) model . ( b ) Isotherm of sw-values determined by integration of c ( s ) profiles in ( a ) ( circles ) . The monomer-dimer model of the isotherm ( solid line ) leads to best-fit values of s1 = 3 . 57 S , s2 = 5 . 11 S ( uncorrected s-values ) and a best-fit Kd , 22 of 24 [9–59; 95% CI] nM . ( c ) Sedimentation coefficient distributions of rsEGFP2-GluA3 at concentrations between 1 nM and 30 nM , determined in blinking configuration as shown in Figure 2e . The constant peak in this concentration range is consistent with the previously determined Kd , 33 of 5 . 6 μM ( Zhao et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 016 To analyze the heterogeneous interaction between GluA2 and GluA3 ATDs , an equimolar dilution series from 0 . 1–30 nM was studied in a configuration with a 13 mW excitation beam , distinguishing rsEGFP2-GluA3 from DL488-GluA2 solely from slow photoswitching of the former . Representative raw data sets and MCMC fits to the sedimentation boundaries are shown in Figure 9 , leading to calculated component sedimentation coefficient distributions shown in Figure 10a . At a high concentration of 30 nM , the sedimentation coefficient distributions rsEGFP2-GluA3 and DL488-GluA2 both show a single co-sedimenting peak at ~5 . 9 S , consistent with a saturated 1:1 hetero-dimer complex , moderately extended with a frictional ratio of 1 . 42 . With decreasing concentrations the peaks occur at lower sedimentation coefficients , characteristic for a rapid association/dissociation equilibrium relative to the time-scale of sedimentation ( Schuck , 2010 ) . Due to the decreasing signal/noise ratio peaks also broaden at lower concentrations . Further , at concentrations below 1 nM a separate slower sedimenting peak of DL488-GluA2 can be discerned , consistent with the s-value of free GluA2 monomer , co-existing with a faster-sedimenting dimeric fraction of GluA2 . Due to the smaller difference between the s-value of the hetero-dimeric complex and free rsEGFP2-GluA3 , the latter cannot be resolved at the low signal/noise ratios below 1 nM . However , the signal-weighted average sedimentation coefficient sw can still be determined as a function of concentration for both components ( Figure 10b ) . The Monte-Carlo error analysis shows sw to be well-defined at all concentrations for the constant signal component , but for the decaying signal component statistical errors increase at lower concentrations , with the lowest useful data point at 0 . 3 nM . A global fit with an isotherm model for the coupled system where hetero-oligomerization of GluA3 is competitive with homo-oligomerization of GluA2 leads to an estimate of the heterodimer Kd , 23 of 0 . 32 [0 . 20–0 . 46 , 95% CI] nM , with a limiting s-value for the complex of 6 . 09 ( 6 . 01–6 . 17 ) S . 10 . 7554/eLife . 17812 . 017Figure 9 . Competitive self-association and hetero-association of DL488-GluA2 and rsEGFP2-GluA3 . Scan data were acquired by FDS-SV , continuously scanning with 488 nm excitation during the sedimentation at 50 , 000 rpm , 20°C . Data shown are for equimolar concentrations at 0 . 3 nM ( a ) , 3 nM ( b ) and 10 nM ( c ) . Solid lines and residuals are from the MCMC decomposition , with pre-determined switching parameters for rsEGFP2-GluA3 , leading to distributions shown in Figure 10a . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 01710 . 7554/eLife . 17812 . 018Figure 10 . Dilution series analysis of GluA2-GluA3 interaction . Sedimentation at 50 , 000 rpm of a dilution series of DL488-GluA2 with equimolar rsEGFP2-GluA3 was observed with 13 mW excitation beam in FDS-SV , causing a decrease of rsEGFP2 signal with time-constant of 5 . 5×10–4/sec . ( a ) Sedimentation coefficient distribution of the decaying signal component of rsEGFP2-GluA3 ( blue , reporting on all blue species in Figure 7 ) and the constant signal component of DL488-GluA2 ( magenta , reporting all magenta species in Figure 7 ) at 30 nM ( solid lines ) , 3 nM ( dashed lines ) and 0 . 3 nM ( dotted lines ) . Raw data and fits are in Figure 9 . ( b ) Weighted-average sedimentation coefficients sw for the rsEGFP2-GluA3 ( blue circles ) and DL488-GluA2 ( magenta circles ) , with 68% CI error bars from Monte-Carlo analysis . The solid line is the best-fit isotherm for a linked homo- hetero-dimerization equilibrium , using a fixed Kd , 22 of 27 . 1 nM for the separately determined GluA2 homo-dimerization . The best-fit estimate for the hetero-dimerization Kd , 23 is 0 . 32 [0 . 20–0 . 46] nM . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 018 To explore an alternative MCMC design , an experiment with the FRAP-like SV configuration was carried out ( Figure 11 ) . Mixtures of rsEGFP2-GluA3 and DL488-GluA2 were exposed to localized 488 nm light from the stationary FDS for a period of 20 min . The initial illumination was assessed from the analysis of a control psFP sample spinning side-by-side in the same rotor sharing the same exposure ( Figure 2f ) , and then fixed for the analysis of the mixtures ( Figure 11 ) . Visually , the reduced trough depth directly reflects the smaller relative signal contribution of rsEGFP2 . Consistent with the results from the previous experimental series , the weight average component sedimentation coefficients series from the FRAP-like SV leads to an estimate of the heteromer Kd , 23 of 0 . 68 [0 . 36–1 . 24 , 95% CI] nM ( Figure 12 ) . 10 . 7554/eLife . 17812 . 019Figure 11 . FRAP-like FDS-SV from mixtures of rsEGFP2-GluA3 and DL488-GluA2 . Equimolar mixtures at 0 . 1 nM ( a , c ) , 2 nM ( b , d ) , 3nM ( e , g ) and 10 nM ( f , h ) were sedimented at 50 , 000 rpm , 20°C , with initial 20 min stationary exposure at 6 . 5 cm of the excitation beam at 488 nm , prior to radial scanning with 488 nm excitation ( every second scan shown ) . Initial exposure parameters were taken from the rsEGFP2-GluA3 data in Figure 2f and fixed . Solid lines and residuals are from the MCMC decomposition ( a , b , e , f ) with resulting distributions shown below the scans in panels ( c , d , g , h ) , with the rsEGFP2-GluA3 component in blue and the DL488-GluA2 component in magenta . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 01910 . 7554/eLife . 17812 . 020Figure 12 . Component sw isotherm of DL488-GluA2 and rsEGFP2-GluA3 as observed in FRAP-like FDS-SV . Equimolar mixtures were studied by FDS-SV with initial localized switching , producing data including those shown in Figure 11 . Plotted here are component sw-values for rsEGFP2-GluA3 ( blue ) and DL488-GluA2 ( magenta ) from integration of the MCMC component sedimentation coefficient distributions . Error bars shown are estimated 68% confidence intervals from Monte-Carlo analysis . The model of competitive homo- and hetero-dimerization ( lines ) results in an estimate of the heteromer Kd , 23 of 0 . 68 [0 . 36–1 . 24 , 95% CI] nM . DOI: http://dx . doi . org/10 . 7554/eLife . 17812 . 020 The present work shows the proof of principle for how different macromolecular components can be readily distinguished in fluorescence-based sedimentation velocity experiments using single wavelength excitation in a way that is naturally compatible with many fluorescent probes commonly used in super-resolution microscopy . Although psFPs were designed to serve a very different purpose , we have embarked on the quantitative analysis of their switching kinetics as a novel temporal tag to substitute for spectral discrimination . In the low-power illumination conditions of FDS-SV the time-course of photoswitching can serve as a finger-print for different psFPs and permit quantitative multi-component analysis . Thus , the study of heterogeneous interactions can be carried out with mixtures of psFP-tagged proteins , proteins labeled with stable fluorophore dyes , and/or non-fluorescent proteins . This opens the possibility for the same fusion proteins observed co-localizing in microscopy to be overexpressed and purified for in vitro binding studies in FDS-SV , to provide information of number , size , and stoichiometry of complexes . Further , there is a potential for FDS-SV to be applied to more crowded solutions such as cell extracts and serum ( Polling et al . , 2013; Kokona et al . , 2015; Hill and Laue , 2015 ) , in which the hydrodynamic interpretation of complex sizes is less quantitative , but complex stoichiometries derived from multi-component detection should be invariant . Exploiting the switchable signal further , we have demonstrated that additional spatial and temporal modulation of the psFPs signal in FDS-SV is possible . This has the potential to further improve discrimination of different psFPs and to enhance hydrodynamic analysis , for example , in the FRAP-like configurations used in the present study . The switchability of signal and localized optical manipulation afforded by psFPs can be envisioned to expand the toolbox of analytical ultracentrifugation in many other unexpected ways . For instance , the creation of lamella-like signal configurations within the concentration plateau region may enable novel assays , for example , to observe subunit exchange rates of size-resolved macromolecular complexes from the lamella migration and signal amplitude changes , or to study complex formation in crowded solutions within the plateau region of rapidly sedimenting co-solutes . Even though the present work was focused on SV , sedimentation equilibrium ( SE ) analytical ultracentrifugation would be expected to equally benefit from the use the new temporal tag: for example , once molecules are in thermodynamic equilibrium of sedimentation , psFPs may be reset or depleted , for example , by uniform 405 nm or 488 nm illumination , and the ensuing temporal signal evolution could reveal the equilibrium distribution specifically of psFPs in a mixture . Further , one could envision FRAP-like localized switching in SE to complement thermodynamic with hydrodynamic information obtained while maintaining SE . Beyond analytical ultracentrifugation , quantitatively exploiting the characteristic dynamics of photoswitching arising from different switching quantum yields may prove useful in other techniques . The new MCMC approach we developed may be compared to other fluorescence methods used to detect binding , including fluorescence correlation/cross-correlation , polarization , proximity and life time imaging methods . These generally offer greater flexibility in solution conditions and can be carried out in vivo . However , the strongly size-dependent migration of molecules in the gravitational field is a unique feature of FDS-SV that can lead to significantly higher resolution of macromolecular complexes . We believe this has great potential for the study of systems that exhibit polydisperse mixtures of different complexes , which would elude traditional structural characterization , and only display population-averaged features in spectroscopic techniques . Of interest , we discovered changes in the switching rate and final contrast of rsEGFP and rsEGFP2 induced by antibody binding . This suggests that subtle binding-induced conformational changes of the GFP β-barrel occur that can lead to changes in the equilibrium populations of photophysical states of the chromophore . For Dronpa , NMR studies have revealed flexibility of the β-barrel allowing non-radiative decay processes in the dark state ( Mizuno et al . , 2008 ) , and viscosity-mediated reduction of flexibility has been proposed as a mechanism for the viscosity sensitivity of its fluorescence ( Kao et al . , 2012 ) . Also , structural flexibility mediating conformational changes from the protein exterior to the chromophore environment inside the β-barrel has been exploited for GFP as a sensor ( Berg et al . , 2009 ) . It seems conceivable that a similar mechanism may be at work in the binding sensitivity of its photoswitchable variants . Even though the psFPs will typically be designed to be an inert tag , rather than offering binding sites , this reveals the potential for an additional , more subtle source of information in temporally modulated FDS-SV , if the equilibrium of photophysical states is altered through occupation of a binding site . Prior work on the assembly of AMPA receptor ATDs suggested an inefficient process for the formation of GluA2/GluA3 heteromers because the Kd , 22 for homodimer assembly by GluA2 , 1 . 8 nM , was only 1 . 4 fold greater than the Kd , 23 of 1 . 3 nM for heterodimer assembly ( Rossmann et al . , 2011 ) . Using the MCMC approach we find that the process is much more efficient , with a Kd , 23 for heterodimer assembly of 0 . 32 nM , 85-fold smaller than the 27 . 1 nM Kd , 22 for homodimer assembly by GluA2 , and four orders of magnitude smaller than the Kd , 33 for GluA3 homo-dimerization , which is consistent with the obligate formation of GluA2/GluA3 heterodimer in neurons ( Rossmann et al . , 2011 ) . While this first application highlights the potential of the method to determine sub-nanomolar dissociation equilibrium constants of protein complexes involving competitive or cooperative self- and hetero-association processes , other key aspects of the method are the ability to simultaneously monitor the size-distribution of individual components and their complexes formed in solution , and the potential to discriminate two or more different psFPs . The availability of size-distributions for all components distinguishes MCMC FDS-SV from methods such as SPR or ITC often used for studying bimolecular reactions . Considering that , from large-scale proteomic experiments ( Gavin et al . , 2002; Krogan et al . , 2006 ) , on average more than four components are in cellular protein complexes , it is necessary to develop methods that can elucidate their stoichiometries , architectures , and driving forces . We believe MCMC FDS-SV offers a powerful new approach to study such multi-component complexes in solution . SV experiments were carried out in an Optima XL-A ( Beckman Coulter , Indianapolis IN ) equipped with an FDS ( Aviv Biomedical , Lakewood NJ ) . The FDS light source is a 488 nm 10 mW diode laser , and emission is detected between 505 nm and 565 nm . Setup of the AUC followed standard protocols ( Zhao et al . , 2013b ) , using an 8-hole An-50 TI rotor . In order to minimize local exposure of the samples during FDS adjustment and magnet angle locking , radial calibration and PMT settings were evaluated at 3000 rpm , followed by temperature equilibration at 20°C , and acceleration to 50 , 000 rpm . If initial FDS adjustment required more than a few minutes , the samples in the rotor were illuminated with 405 nm prior to temperature equilibration . With the FDS focal depth set at 4000 µm and the PMT power at 51%–80% with gain of 8 ( dependent on sample concentration ) , sequences of radial scans were acquired continuously for 12 hr with radial intervals of 20 µm . For passive blinking experiments , pauses in data acquisition were applied . For rapid blinking experiments a modified Optima XL-A was used , where 405 nm light from a high-power LED was guided into the rotor chamber through the interference optics port in the heat sink to illuminate the spinning rotor . A sector-shaped mask concentric to the axis of rotation was used . The FDS signal is linear with concentration in the nanomolar concentration range and below ( Zhao et al . , 2013b; Lyons et al . , 2013 ) , and accordingly , we consider the fluorescence signal F ( r , t ) a product of concentration c ( r , t ) and a specific molar fluorescence signal increment E . The overall initial signal increment ϵ0 will depend , for example , on laser power , absorption coefficient , intrinsic fluorescence quantum yield , and detection efficiency . Photoswitching is modeled as a two-state process , with a single exponential transition from an initial equilibrium of photophysical states to another equilibrium after a long time of 488 nm irradiation in the FDS . The rate of transition will depend on a quantum efficiency of photoswitching and the total count of photons incident on the migrating fluorophore . The latter will be dependent on the incident laser power density , as well as the distance from the center of rotation . Briefly , since all molecules have the same angular velocity ω , they traverse a constant beam width δ more quickly at higher radii . At position r′ the sample has a transverse velocity ωr′ and ( with negligible difference between the arc and secant in the AUC geometry ) molecules are illuminated for the time δ/ωr′ at each rotation . It follows that their total photon count is ( 1 ) Φ ( r , t ) =ϕ∫0tδ2πr′ ( t′ ) dt′ where r′ ( t ) is the trajectory of the molecule through the sample . Neglecting Brownian motion , the path history of particles with sedimentation coefficient s found at time tat position r is ( 2 ) r′ ( t′ ) =reω2s ( t′−t ) which , after integration , leads to ( 3 ) Φ ( r , t ) =ϕδ2πrω2s[ eω2st−1 ] as the accumulated total photon count seen by the fluorophore found at radius r at time t sedimenting with s . Therefore , combining constants from illumination and molecular photoswitching into a single parameter β , we find the time-dependent fluorescent signal increment ( 4 ) E ( s , r , t−t405 , i ) =ε0 ( 1+α exp{−β ( r/r0 ) ω2s[eω2st ( t−t405 , i ) −1]} ) where α describes the fractional signal change after long time , and r0 is a reference radius . The reference time-point t405 , i is the last time of fluorophore reset with strong 405 nm exposure ( and t405 , I=0 without reset illumination ) . The slight radial dependence causes slopes that for individual psFPs are indistinguishable from the effects of a small mismatch between the focal plane of the scanner and the plane of rotation causing small gradients of signal magnification ( Zhao et al . , 2013b; Zhao et al . , 2014b , ) but for studying mixtures of psFPs they need to be quantitatively accounted for . Analogous to the multi-wavelength sedimentation coefficient distribution analyses ( MSSV ) ( Balbo et al . , 2005 ) , we model the evolution of experimental fluorescence scans a ( r , t ) as a superposition of signals from different components p , each exhibiting a different unknown distribution of sedimentation coefficients cp ( s ) ( 5 ) a ( r , t ) =∑p∫cp ( s ) E ( p ) ( s , r , t ) χ1 ( s , r , t ) ds where E ( p ) ( s , r , t ) denotes the temporal signal change ( Equation 4 ) , and χ1 ( s , r , t ) denotes the evolution of concentration for a species with sedimentation coefficient s in the centrifugal field , as calculated by the master partial differential equation for sedimentation and diffusion in the centrifugal field , the Lamm equation ( Lamm , 1929; Brown and Schuck , 2008 ) . Sedimentation and diffusion are both governed by the same hydrodynamic frictional coefficient ( Cheng and Schachman , 1955 ) , which allows diffusion to be estimated for all species based on a hydrodynamic scaling law D ( s ) for particles with given translational friction coefficient ( Schuck , 2000 ) . Since f/f0 measures the hydrodynamic extension of a particle relative to a compact sphere of the same mass and takes a narrow range of values for most folded proteins ( Serdyuk et al . , 2007; Cantor and Schimmel , 1980 ) , it was fixed in MCMC . The distributions cp ( s ) are calculated after discretization into a linear least squares problem with Tikhonov regularization ( Balbo et al . , 2005; Schuck , 2000 ) , and implemented in the public domain software SEDPHAT ( https://sedfitsedphat . nibib . nih . gov/software/default . aspx ) . Analogously to MSSV , the approach is compatible with mass conservation-based regularization approaches to further enhance component resolution ( Brautigam et al . , 2013 ) . After determination of cp ( s ) , integration can be carried out to determine signal weighted-average sedimentation coefficients sp , w compatible with the transport method , and further isotherm analysis of sp , w as a function of protein concentrations can take place with binding models ( Schuck et al . , 2003 , 2015 ) . For the GluA2/GluA3 interaction analysis , the isotherm modeling was based on the competitive self- and hetero-dimerization model in SEDPHAT . As an experimental design tool an analytical prediction of errors in the MCMC decomposition was implemented for two species with fluorophores p and q sedimenting at the same s-value , in the absence of initial localized switching . After discretization of Equation 5 and neglect of weak s-value dependence of the signals , it can be shown that , given n meaningful data points with relative error δa , the relative error in cp is ( 6 ) δc≈cond ( E ) 2δan , i . e . , is dependent on the condition number of the matrix E where ( 7 ) Ep , q=∑tE ( p ) ( t ) E ( q ) ( t ) w ( s , t ) contains the mutual products of signals at experimentally available scan times t , weighted with a sedimentation term w ( s , t ) =∑rχs , r , t2 . Mathematically , E plays the same role as the extinction coefficient matrix in multi-wavelength analysis . Essentially , fluorophores will be distinguishable on the basis of their temporal signature if cross-terms in E are small during observed times in the sedimentation experiment . For the error analysis of experimental data , a Monte-Carlo statistical analysis was implemented in order to determine error estimates for the distributions cp ( s ) and their integrals such as sw . In the present work , 500 iterations were chosen to estimate the sensitivity of sw to data acquisition noise , not including variation of the meniscus . In the presence of inhomogeneous illumination , the spatially uniform time-dependent extinction coefficient model of Equation 4 must be extended to account explicitly for sedimentation and diffusion of molecules in fluorescent state χf and dark state χd . In the presence of illumination with a spatial-temporal intensity profile I ( r , t ) at a wavelength of 488 nm , transition from the fluorescent to the dark state occurs with a rate kf−d ( r , t ) =qf−dI ( r , t ) ( assuming a quantum efficiency qf−d ) . The evolution of concentrations of molecules in the fluorescent and dark state is then described with the coupled Lamm equation ( which may be extended similarly for the reverse transition ) ( 8 ) ∂χf∂t=−1r∂∂r ( χfsω2r2−D∂χf∂rr ) −χfkf−d ( r , t ) ∂χd∂t=−1r∂∂r ( χdsω2r2−D∂χd∂rr ) +χdkf−d ( r , t ) The spatial illumination profile I ( r , t ) was empirically modeled as a superposition of beams with Gaussian intensity profile , accounting for the cone-shape intensity profile . The amplitudes and center radii were introduced as additional parameters refined from the fit . After initial localized illumination , spatially uniform scanning governs the further evolution of the signal of the fluorescent species and E ( s , r , t ) ×χf ( r , t ) can be inserted in the multi-component decomposition of Equation 5 . Modeling the coupled sedimentation/diffusion/photoswitching process will account for sedimentation and diffusion that takes place during strong illumination , which causes the precise shape of the trough to slightly depend on the rates of macromolecular sedimentation and diffusion . However , the initial exposure is precisely identical for all different samples in the rotor . With this approach , the precise initial illumination profile is inconsequential because the parameters of interest are only in the further evolution with time . The plasmid pQE31-rsEGFP was kindly provided by Stefan Jakobs ( Max Planck Institute for Biophysical Chemistry ) . The plasmid rsEGFP2 was produced by mutating the pQE31-rsEGFP plasmids using following primers , T65A ( 5’-CCACCCTGGCCTACGGCGTG-3’ ) , A150V ( 5’-GCCACAACGTCTATATCATGG-3’ ) , and N205S ( 5’-GCACCCAGTCCAAGCTGAGC-3’ ) . Mutations are underlined . The plasmid pQE31-Padron was also provided by Stefan Jakobs and EGFP was expressed from the pRSETA-EGFP plasmid both described previously ( Zhao et al . , 2014b ) . The GluA2 ATD cDNA , residues Met1–Ser404 , was cloned into the pRK5-IRES-EGFP mammalian expression vector with a C-terminal LELVPRGS-His8 linker , thrombin cleavage site and affinity tag . The rsEGFP–GluA3 ATD fusion construct was created as follows: The cDNA for the GluA3 signal peptide residues Met1–Gly22 , followed by an SGSG tetrapetide linker , was inserted upstream of the cDNA for rsEGFP residues Val2–Lys239 , and connected by an SGS tripeptide linker to the cDNA for GluA3 , residues Gly23–Ser408 , followed by a C-terminal LELVPRGS-His8 linker , thrombin cleavage site and affinity tag , cloned into the pRK5 mammalian expression vector . Both constructs were expressed in HEK293T cells by transient transfection of suspension cultures using PEI MAX 40000 ( Polysciences Inc . ) . The secreted constructs with native complex glycosylation were concentrated by ultrafiltration using Pellicon 10 kDa MW cutoff cassettes ( Millipore ) and purified by immobilized metal affinity chromatography using a HiTrap NTA column ( GE Healthcare ) , followed by proteolysis with thrombin , and then ion exchange chromatography using a HiTrap Q column ( GE Healthcare ) . The pooled fractions were concentrated and then dialyzed against AUC or labeling buffer as appropriate ( 150 mM NaCl , 1 mM EDTA , 20 mM NaPhosphate , pH 7 . 5 and 7 . 0 respectively ) . A 14 . 4 µM concentration of purified GluA2 ATD was mixed with a 12 M excess of N-hydroxysuccinimide ( NHS ) ester–activated Dylight488 ( Thermo Fisher Scientific ) dissolved in DMSO and then diluted with labeling buffer . The reaction was incubated in the dark at room temperature for 45 min and then loaded onto a high resolution size exclusion chromatography column ( Superdex 75 10/300 GL ) equilibrated with AUC buffer to separate free dye from labeled protein . The protein concentration and labeling ratio were then determined by UV-Vis spectrophotometry using values for ε280 of 55 , 720 M−1cm−1 for the unmodified protein , ε280 of 10 , 290 M−1cm−1 and ε493 of 70 , 000 M−1cm−1 for Dylight488 . Bovine serum albumin ( BSA ) was acquired from Sigma Aldrich ( catalog#A7030 , St . Louis ) and fluorescently labeled with FITC using the labeling kit from Thermo Fisher targeting the primary amine on BSA ( catalog#53027 ) . Monoclonal anti-GFP IgG ( D153-3 ) was purchased from MBL International ( Woburn , MA ) and subjected to exhaustive dialysis with the working buffer , phosphate buffered saline ( 5 . 62 mM Na2HPO4 , 1 . 06 mM KH2PO4 , 154 mM NaCl , pH 7 . 40 ) . To all samples , unlabeled BSA was added at 0 . 1 mg/ml to suppress surface adsorption of the proteins of interest . Benchtop fluorescence spectroscopy experiments were carried out in a Fluorolog spectrofluorometer ( Horiba Instruments Inc . , Irvine CA ) . rsEGFP or rsEGFP2 at were studied at a final concentration of 0 . 5 µM in the presence or absence of equimolar anti-GFP antibody , after 2 hr incubation of the mixture . Emission spectra were collected with slit widths set to 1 nm . For kinetic experiments of photoswitching , the 490 nm excited 508 nm fluorescence of rsEGFP or the 483 nm excited 502 nm fluorescence of rsEGFP2 was monitored , respectively .	Many proteins in cells combine to form molecular machines or complexes that carry out specific processes inside cells . Analytical ultracentrifugation is a technique commonly used to explore the physical properties of proteins and their complexes and in this way to gain insights into the biological roles of these molecules . The technique involves spinning a sample containing the molecules to generate a strong centrifugal force , while monitoring the movement of the molecules . Under these conditions , molecules with different sizes and masses sink – or “sediment” – at different rates , so individual proteins and their complexes can be clearly distinguished . Analytical ultracentrifugation was recently extended to make it possible to detect fluorescent tags added on to proteins . This advance allowed researchers to study more dilute samples or complexes that are held together especially tightly . However , only tags of a single color can be detected because of physical constraints of the fluorescent detection system . This meant that only one kind of fluorescent signal could be tracked at any one time . However , a group of fluorescent tags called photoswitchable fluorescent proteins ( psFPs ) offer new opportunities for detecting multiple signals . This is because these psFPs switch between fluorescent and non-fluorescent states while being detected in the ultracentrifuge . Zhao et al . have now exploited this unique photoswitching property by accurately measuring how fast a number of psFPs switched between fluorescent and non-fluorescent states while they were sedimenting . Each different psFPs switched in a distinct way , even for psFPs of the same color , meaning that each psFP could be identified from its switching rate , similar to identifying a person from their fingerprints . This discovery allowed Zhao et al . to distinguish different psFPs in a mixed sample as if they had different colors . Further experiments went on to demonstrate that this approach could identify the binding proteins in a protein mixture made of three components , and be used to study a biologically important protein complex that can itself exist in two distinct forms . The approach will therefore provide a valuable tool to observe different components in a complex individually and will provide researchers the opportunity to study how mixed protein complexes form at very low concentrations . Future developments of the approach may make it possible to study other properties of protein complexes such as their overall shape and their behavior under conditions that mimic those inside the cell .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics", "tools", "and", "resources" ]	2016	Monochromatic multicomponent fluorescence sedimentation velocity for the study of high-affinity protein interactions
The epigenetic inheritance of DNA methylation requires UHRF1 , a histone- and DNA-binding RING E3 ubiquitin ligase that recruits DNMT1 to sites of newly replicated DNA through ubiquitylation of histone H3 . UHRF1 binds DNA with selectivity towards hemi-methylated CpGs ( HeDNA ) ; however , the contribution of HeDNA sensing to UHRF1 function remains elusive . Here , we reveal that the interaction of UHRF1 with HeDNA is required for DNA methylation but is dispensable for chromatin interaction , which is governed by reciprocal positive cooperativity between the UHRF1 histone- and DNA-binding domains . HeDNA recognition activates UHRF1 ubiquitylation towards multiple lysines on the H3 tail adjacent to the UHRF1 histone-binding site . Collectively , our studies are the first demonstrations of a DNA-protein interaction and an epigenetic modification directly regulating E3 ubiquitin ligase activity . They also define an orchestrated epigenetic control mechanism involving modifications both to histones and DNA that facilitate UHRF1 chromatin targeting , H3 ubiquitylation , and DNA methylation inheritance . Epigenetic regulation of chromatin architecture and gene expression is driven , in large part , by proteins that write , erase , and read histone post-translational modifications ( PTMs ) and DNA methylation . These proteins and their complexes are often comprised of multiple regulatory domains , permitting intricate mechanisms that govern allosteric control of enzymatic activity and multivalent engagement of chromatin through one or more reader modules ( Du et al . , 2015; Musselman et al . , 2012; Noh et al . , 2016; Rothbart and Strahl , 2014; Ruthenburg et al . , 2007; Su and Denu , 2016 ) . The E3 ubiquitin ligase UHRF1 ( ubiquitin-like , containing PHD and RING finger domains 1 ) is one such multi-domain epigenetic regulator ( Figure 1A ) that plays a central role in DNMT1-directed DNA methylation maintenance during DNA replication ( Bostick et al . , 2007; Sharif et al . , 2007 ) . It does so in part through the reader activity of its linked TTD-PHD ( tandem Tudor and plant homeodomain ) towards the N-terminus of histone H3 when it is di- and tri-methylated at lysine 9 ( H3K9me2/me3 ) ( Arita et al . , 2012; Rothbart et al . , 2013 , 2012 ) , and through RING ( really interesting new gene ) domain-mediated catalysis of H3K18 and H3K23 ubiquitylation that promotes DNMT1 association with H3 ( Nishiyama et al . , 2013; Qin et al . , 2015 ) . 10 . 7554/eLife . 17101 . 003Figure 1 . UHRF1 binding to HeDNA is required for DNA methylation regulation but is dispensable for chromatin interaction . ( A ) Domain map of human UHRF1 with identified biochemical functions ( top ) and loss-of-function point mutations used in this study ( bottom; see also Figure 1—figure supplement 1 ) . UBL ( ubiquitin-like ) ; TTD ( tandem Tudor domain ) ; PHD ( plant homeodomain ) ; SRA ( SET and RING-associated domain ) ; RING ( really interesting new gene ) . Amino acid positions demarcating domain boundaries are also shown . ( B ) FP binding assays quantifying the interaction of wild-type , DNAmut , and HeDNAmut MBP-tagged UHRF1 with the indicated FAM-labeled DNA oligonucleotides . Error is represented as ± s . e . m . for two independent experiments . ( C ) Representative immunofluorescence staining for 5-methylcytosine ( 5mC ) in control and UHRF1 knockdown Hela cells after genetic complementation with the indicated wild-type and mutant forms of full-length UHRF1 . Error is represented as ± S . D . from at least four fields of view . Mock , no DNA control; Scale bar , 20 μm . ( D ) Chromatin association assays for FLAG-tagged UHRF1 ( wild-type ) or the indicated mutants from asynchronously growing HeLa cells . Mock , no DNA control . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 00310 . 7554/eLife . 17101 . 004Figure 1—figure supplement 1 . UHRF1 mutations characterized in this study . ( A ) Table of mutant UHRF1 proteins characterized in this study , their functional consequences , and references associated with their initial characterization . ( B ) Crystal structures of the UHRF1 TTD-PHD domain bound to an H3K9me3 peptide ( PDB:3ASK ) , the UHRF1 SRA domain bound to an HeDNA oligonucleotide ( PDB:3CLZ ) , and the RING domain ( PDB:3FLZ ) . Insets highlight structural details of these interactions that involve residues mutated in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 00410 . 7554/eLife . 17101 . 005Figure 1—figure supplement 2 . The DNA binding affinity of UHRF1 is highly sensitive to salt concentration . FP binding assays quantifying the interactions of MBP-UHRF1 with FAM-labeled HeDNA or UnDNA oligonucleotide probes . Error is represented as ± s . e . m . for two independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 005 The SRA ( SET and RING-associated domain ) of UHRF1 binds DNA with modest selectivity towards hemi-methylated CpG dinucleotides ( HeDNA ) ( Arita et al . , 2008; Avvakumov et al . , 2008; Hashimoto et al . , 2008 ) and has also been implicated in DNA methylation regulation . However , as previously studied mutations and deletions of the SRA disrupt DNA interaction regardless of DNA methylation status ( Liu et al . , 2013; Sharif et al . , 2007 ) , the specific contribution of HeDNA recognition to this epigenetic regulatory process has not been defined . We therefore sought to gain insight into the function of HeDNA recognition through the UHRF1 SRA domain and determine the relationship between the enzymatic and histone- and DNA-binding activities of this multi-domain epigenetic regulator . We first produced recombinant full-length human UHRF1 and quantified the interaction of this protein with double-stranded DNA oligonucleotides containing a single unmodified ( UnDNA ) , hemi-methylated ( HeDNA ) , or symmetrically methylated ( SyDNA ) CpG dinucleotide by fluorescence polarization ( FP ) . UHRF1 displayed a 10- to 20-fold preference for HeDNA over UnDNA , and a 5- to 10-fold preference for HeDNA over SyDNA ( Figure 1B; left panel ) . We also confirmed the binding preferences of two previously characterized single amino acid substitutions to the SRA domain ( Avvakumov et al . , 2008 ) . G448D ( DNAmut ) disrupts all DNA-binding ( Figure 1B; middle panel ) by installing a negatively charged residue at a position that contacts the DNA backbone ( Figure 1—figure supplement 1 ) , and N489A ( HeDNAmut ) , harbored within the NKR finger that contacts the unmethylated cytosine opposite the methylated base ( Figure 1—figure supplement 1 ) , disrupts only HeDNA-sensing ( Figure 1B; right panel ) . We also observed that the DNA binding affinity of UHRF1 was exquisitely sensitive to small perturbations in salt concentration; we measured a nearly 500-fold affinity difference for HeDNA between 50 mM and 150 mM NaCl ( Figure 1—figure supplement 2 ) . We next used a previously developed genetic complementation system in HeLa cells ( Rothbart et al . , 2012 ) to determine the contribution of DNA-binding and HeDNA-sensing to UHRF1 function in DNA methylation maintenance . Consistent with our previous observations ( Rothbart et al . , 2013 , 2012 ) , global DNA methylation levels were significantly reduced following stable knockdown of endogenous UHRF1 by shRNA ( Figure 1C ) . DNA methylation was restored by reintroduction of a wild-type UHRF1 transgene , but like mutations that disrupt histone interaction through the PHD finger ( H3mut; Figure 1—figure supplement 1 ) , E3 ubiquitin ligase activity ( RINGmut ) and DNA binding ( DNAmut ) , HeDNAmut could not rescue DNA methylation loss in cells despite retaining its ability to bind DNA ( Figure 1B–C ) . These results demonstrate that in addition to the well-appreciated roles of histone-binding and ubiquitin ligase activity to the DNA methylation regulatory function of UHRF1 ( Nishiyama et al . , 2013; Rothbart et al . , 2013 , 2012 ) , hemi-methylated DNA sensing is critical for DNA methylation maintenance . Notably , unlike H3mut and DNAmut , wild-type and HeDNAmut bound to bulk chromatin biochemically fractionated from HeLa cells ( Figure 1D ) . Collectively , these findings suggest that the histone- and DNA-binding domains of UHRF1 are performing complementary functions to target UHRF1 to chromatin , and that HeDNA recognition provides an additional regulatory layer in the DNA methylation program . To test this hypothesis , we first sought to determine whether the independently characterized DNA- and histone-binding activities of UHRF1 might function in concert . In agreement with previous analyses of the isolated TTD-PHD ( Rothbart et al . , 2013 , 2012 ) , full-length UHRF1 displayed a preference for H3K9me3 peptides over unmodified H3 peptides ( H3K9un ) ( Figure 2A; top panel , see also Supplementary file 1 ) . No binding was observed for H3mut ( Figure 2—figure supplement 1A ) or for wild-type protein binding to peptides containing an N-terminal 5-carboxyfluorecin ( FAM ) probe to block PHD engagement ( Figure 2A; bottom panel ) . Performing these assays with full-length UHRF1 allowed us to ask whether DNA binding affects H3 peptide binding and vice versa . Histone binding measurements in the presence of 10 μM unlabeled HeDNA , SyDNA , or UnDNA enhanced the interaction with C-terminal FAM-labeled H3K9me3 and H3K9un peptides ( Figure 2A; top panel ) . HeDNA did not enhance binding to N-terminal FAM-labeled peptides , indicating that the multivalent interaction of the TTD-PHD with a single H3 peptide ( Rothbart et al . , 2013 , 2012 ) remained intact . Reciprocally , DNA-binding measurements in the presence of 10 μM unlabeled H3K9me3 peptide enhanced DNA binding affinity 5–10 fold irrespective of the methylation status on DNA ( Figure 2B–C and Figure 2—figure supplement 1B-C , E ) . Collectively , these experiments demonstrate that the histone- and DNA-binding modules of UHRF1 are regulated by reciprocal positive allostery . 10 . 7554/eLife . 17101 . 006Figure 2 . The DNA- and histone-binding domains of UHRF1 are regulated by reciprocal positive allostery . ( A ) FP binding assays quantifying the interaction of MBP-UHRF1 with a C-terminally FAM-labeled H31-20K9me3 peptide ( see Supplementary file 1 for a full list of peptides used in this study ) in the absence or presence of the indicated unlabeled DNA oligonucleotides . Error is represented as ± s . e . m . for two independent experiments . ( B–C ) FP binding assays quantifying the interactions of wild-type and the indicated mutant MBP-UHRF1 proteins with FAM-labeled HeDNA or UnDNA in the presence and absence of the indicated unlabeled H31-20 peptides . Error is represented as ± s . e . m . for two independent experiments . See Figure 2—figure supplement 1E for Kd values associated with panel C . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 00610 . 7554/eLife . 17101 . 007Figure 2—figure supplement 1 . Quantifying the interaction of full-length UHRF1 and various mutants with histone H3 peptides and DNA oligonucleotides . ( A ) FP binding assays quantifying the interaction of MBP-tagged UHRF1 H3mut with C-terminally FAM-labeled H31-20K9un and H31-20K9me3 peptides . Error is represented as ± s . e . m . for two independent experiments . ( B ) FP binding assays quantifying the interaction of UHRF1 with FAM-labeled HeDNA and UnDNA in the presence or absence of 10 μM unlabeled H31-20K9me3 . Error is represented as ± s . e . m . for two independent experiments . ( C ) Table of the calculated Kd values for UHRF1 and MBP-UHRF1 binding to HeDNA or UnDNA in the presence or absence of unlabeled H31-20K9me3 ( see also Figure 2C ) . The DNA and peptide binding activity of these constructs were similar; however , MBP-tagged protein displayed better solubility at higher protein concentration . ( D ) FP binding assays monitoring the interaction of Linkermut with HeDNA or UnDNA in the presence or absence of 10 μM unlabeled H31-20K9me3 . Error is represented as ± s . e . m . for two independent experiments ( E ) Table of calculated Kd values for the indicated wild-type and mutant MBP-UHRF1 proteins binding to HeDNA or UnDNA in the presence or absence of unlabeled H31-20K9me3 ( see also Figure 2C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 007 In agreement with the multivalent histone engagement model of the UHRF1 TTD-PHD , the extent to which H3 peptides augmented the interaction of UHRF1 with DNA was dependent on the epigenetic signature on H3 . H3K9me3 peptide showed a three-fold enhancement of DNA binding over H3K9un , and asymmetric di-methylation of arginine 2 ( H3R2me2a ) , which blocks the UHRF1 PHD interaction with H3 ( Rajakumara et al . , 2011 ) , did not enhance DNA binding ( Figure 2B ) . Consistently , H3mut completely perturbed the ability of an H3K9me3 peptide to positively regulate DNA binding , a previously characterized double mutation to the linker connecting the TTD-PHD that uncouples multivalent engagement to H3 ( Linkermut ) ( Arita et al . , 2012; Rothbart et al . , 2013 ) exhibited a weaker enhancement of DNA binding in the presence of peptide than wild-type , and DNAmut remained unable to bind DNA in the presence of H3K9me3 ( Figure 2C and Figure 2—figure supplement 1D–E ) . Conversely , the DNA binding affinity of HeDNAmut was still enhanced by H3K9me3 , although HeDNAmut could not discriminate between UnDNA and HeDNA ( Figure 2C ) , providing a biochemical basis for HeDNAmut retention on chromatin ( Figure 1D ) . The observed positive allostery between DNA- and histone-binding suggested the possibility of a direct physical interaction between the SRA and TTD-PHD domains of UHRF1 . Consistent with this hypothesis , the UHRF1 TTD-PHD associated with the SRA and SRA-RING in pull-down experiments ( Figure 3A ) , and this association was perturbed in the presence of DNA , irrespective of methylation status ( Figure 3B , left ) . SRA-RING DNAmut maintained interaction with the TTD-PHD in the presence of DNA ( Figure 3B , right ) . However , an H3K9me2 peptide did not inhibit the interaction between the SRA and the TTD-PHD ( Figure 3C ) . These results suggest that the DNA-binding surface of the SRA contributes to an intramolecular interaction in a manner non-competitive with histone binding . To ensure that the allostery observed was due to an intramolecular rearrangement and not through oligomerization , we characterized UHRF1 in the presence and absence of ligands with several biophysical techniques . Indeed , UHRF1 remained monomeric and in good agreement with the expected molecular weight as measured by analytical size exclusion chromatography , dynamic light scattering , and atomic force microscopy ( Figure 3D–F ) . 10 . 7554/eLife . 17101 . 008Figure 3 . DNA binding disrupts a UHRF1 intramolecular interaction . ( A ) In vitro pull-down analysis of the interaction between GST-TTD-PHD and MBP or the indicated MBP fusions of UHRF1 . ( B ) Pull-down analysis of the interaction between GST-TTD-PHD and MBP-SRA-RING ( wild-type or DNAmut ) fusions of UHRF1 in the presence or absence of the indicated DNA oligonucleotides . GST Ctrl is a GST fusion of the PHD-Bromo from BPTF ( see Materials and methods ) . ( C ) Pull-down in the presence of H31-20K9me2 . ( D ) Analytical size exclusion chromatography of UHRF1 in the absence or presence of HeDNA and H31-15K9me2 . The calculated molecular weights for apo and ligand-bound UHRF1 are in agreement with the expected molecular weight of monomeric UHRF1 . ( E ) Dynamic light scattering of UHRF1 in the absence or presence of HeDNA and H31-15K9me2 . UHRF1 remains mono-dispersed ( poly-dispersity < 25% ) both in the presence and absence of the indicated ligands . The calculated mass range of 91–98 kD is in agreement with the expected molecular weight for full-length monomeric UHRF1 , 90 kD . ( F ) Atomic force microscopy histograms of the volumes for 617 apo UHRF1 particles ( left ) and 884 HeDNA-bound UHRF1 particles ( right ) . Distributions were fit to a single Gaussian peak using the peak fit function in Origin 6 . 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 008 Collectively , these results show that the histone- and DNA-binding domains of UHRF1 interact and that general DNA binding releases this physical association . The data further suggest that ligand-induced intramolecular rearrangement of UHRF1 domain connectivity results in high-affinity retention of UHRF1 on chromatin through positive regulation of histone- and DNA-binding activities . This model is generally consistent with a recent report published during the preparation of this manuscript ( Fang et al . , 2016 ) , which shows that HeDNA enhances histone interaction and suggests a closed-to-open conformational change in UHRF1 intramolecular architecture upon ligand binding . However , two key differences between our findings are that we show reciprocal positive allostery between the histone- and DNA-binding domains of UHRF1 , and that general DNA interaction ( regardless of methylation state ) can displace the TTD-PHD domain and enhance histone interaction . We note that the conclusions from Fang et al . relied upon the interpretation of qualitative in-solution pull-down experiments conducted with a 2:1 DNA:UHRF1 ratio ( see Fang et al . , 2016 ) , whereas we used quantitative FP to measure relative binding affinities and included unlabeled ligands in our experiments at concentrations at least five-fold over their measured Kd values to ensure saturation . Since UHRF1 ubiquitin ligase activity is required to support DNA methylation yet is dispensable for bulk chromatin interaction ( see Figure 1C–D and Nishiyama et al . , 2013; Qin et al . , 2015 ) , we hypothesized that there may be a functional link between HeDNA-binding and UHRF1 ligase activity . To begin testing this hypothesis , we in vitro reconstituted UHRF1-mediated ubiquitylation using recombinant UHRF1 , H3 peptides , Flag-tagged ubiquitin , and the ubiquitin conjugation enzymes E1 ( Uba1 ) and E2 ( UbcH5c ) ( for a review of the mechanism of ubiquitin activation see Schulman , 2011 ) . Surprisingly , we observed robust ubiquitylation of an H31-32K9me2 peptide ( Supplementary file 1 ) in a 20 min end-point assay in the presence of HeDNA ( Figure 4A ) . Neither apo-UHRF1 , SyDNA , nor UnDNA could stimulate this activity at concentrations well above their measured Kd values ( Figure 4A ) , despite the ability of these DNAs to positively regulate histone binding ( Figure 2A ) . Consistent with our measured Kd for HeDNA ( Figure 1B ) , we observed reduced ubiquitylation activity as HeDNA concentration fell below 300 nM ( Figure 4A ) . To our knowledge , this is the first demonstration that a DNA-protein interaction , and in particular an epigenetic modification , directly regulates enzymatic activity of an E3 ubiquitin ligase . 10 . 7554/eLife . 17101 . 009Figure 4 . UHRF1-mediated histone H3 ubiquitylation is stimulated by substrate and HeDNA recognition . ( A ) UHRF1 ubiquitylation assays on an H31-32K9me2 peptide in the absence or presence of the indicated DNA oligonucleotides: HeDNA was titrated at semi-log intervals spanning 30 μM to 1 nM . SyDNA or UnDNA was added at 30 μM or 100 μM , respectively . ( B ) Rate measurement quantifying UHRF1 auto-ubiquitylation and H31-32K9me2 ubiquitylation in the presence of HeDNA or UnDNA at the indicated time points . Rate experiments were performed three times with similar results , and a representative blot is depicted . Blots were quantified using ImageQuant TL ( GE Lifesciences ) . Quantified data was best described by a linear fit over the measured time scale , with the exception of HeDNA-stimulated H31-32K9me2 mono-ubiuitylation , which remained linear within the first 5 min of the reaction . ( C ) UHRF1 ubiquitylation assays on HeLa mononucleosomes in the presence of the indicated concentrations of HeDNA and/or an H31-15K9me2 peptide . ( D ) Ubiquitylation of an H31-43K9un peptide by UHRF1 and the indicated mutants ( see Figure 1A for mutant annotation ) in the absence or presence of HeDNA . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 00910 . 7554/eLife . 17101 . 010Figure 4—figure supplement 1 . UHRF1 ubiquitin ligase assays . ( A ) Time course assay of UHRF1 auto-ubiquitylation in the presence of UnDNA or HeDNA in the absence of a histone peptide substrate ( left ) . Rates of UHRF1 auto-ubiquitylation ( right ) were quantified from blots using ImageQuant TL ( GE Lifesciences ) . Quantified data was best described by a linear fit over the measured time scale . ( B ) UHRF1 ubiquitylation of HeLa mononucleosomes in the presence or absence of HeDNA reveals that mononucleosome ubiquitylation is significantly enhanced in the presence of HeDNA . ( C ) Ubiquitylation of H31-32K9me2 by MBP-tagged UHRF1 and the indicated mutants . Mutant UHRF1 proteins display significant defects in HeDNA stimulated ubiquitylation activity with the exception of the D469G . ( D ) FP binding assays quantifying the interaction of the MBP-UHRF1 D469G with FAM-labeled HeDNA , SyDNA , or UnDNA . Error is represented as ± s . e . m . for two independent experiments . While the literature suggests a HeDNA binding defect for the D469G mutation , these assays reveal this mutant binds DNA similarly to wild-type ( for comparison see Figure 1B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 010 To further characterize HeDNA-stimulated UHRF1 ubiquitylation , we compared the rate of UHRF1 enzymatic activity on itself ( auto-ubiquitylation measurements are often used as a proxy to monitor E3 ligase activity ) and an H31-32K9me2 peptide substrate . We measured a 2 . 5-fold rate enhancement of UHRF1 auto-ubiquitylation in the presence of HeDNA vs . UnDNA , both in the presence or absence of H31-32K9me2 peptide ( Figure 4B and Figure 4—figure supplement 1A ) . In sharp contrast , the rate of HeDNA-stimulated H31-32K9me2 mono-ubiquitylation was stimulated by more than 100-fold over the rate obtained with UnDNA ( Figure 4B ) . Comparing the rate of activity on UHRF1 substrates ( peptide vs . self ) , the rate of auto-ubiquitylation was 20-fold faster than the rate of peptide ubiquitylation in the presence of UnDNA . Conversely , the rate of peptide ubiquitylation was seven-fold faster than the rate of auto-ubiquitylation in the presence of HeDNA ( Figure 4B ) . Based on these observations , we propose that HeDNA-binding acts as an allosteric switch to enhance ubiquitylation of histone substrates . Similar to peptide substrates , UHRF1 mono- , di- , and tri-ubiquitylation of purified HeLa mononucleosomes was stimulated by HeDNA ( Figure 4—figure supplement 1B ) , confirming that the enhanced ubiquitylation activity of UHRF1 is relevant in the context of chromatin . In addition , when excess H31-15K9me2 peptide ( Supplementary file 1 ) ( which harbors the TTD-PHD binding site but not the published ubiquitin target lysines ) was added to UHRF1 mononucleosome ubiquitylation assays , H31-15K9me2 effectively inhibited enzymatic activity towards mononucleosome substrates ( Figure 4C ) . In contrast , HeDNA concentrations as high as 40 μM did not block mononucleosome ubiquitylation , consistent with its role as an activator of UHRF1 E3 ligase activity . These results suggest that the N-terminus of H3 is the primary binding site for substrate recognition through the TTD-PHD and that DNA interaction can occur in trans to the nucleosome being targeted for ubiquitylation . To further investigate the role of UHRF1 reader domain functions in ligase activity , we tested the previously described H3mut , Linkermut , DNAmut , HeDNAmut ( Figure 1A ) and D469G ( Avvakumov et al . , 2008 ) mutants in ubiquitylation assays using H31-32K9me2 ( Figure 4—figure supplement 1C ) and H31-43K9un peptides as substrate ( Figure 4D ) . Reacting UHRF1 with peptide substrates , we observed low ubiquitin ligase activity in the absence of HeDNA , while HeDNA binding permitted robust formation of mono- , di- , and tri-ubiquitylated H3 peptides ( Figure 4D and Figure 4—figure supplement 1C ) . Characterizing DNA- , HeDNA- , and histone-binding loss-of-function UHRF1 mutants in ubiquitylation assays revealed defects in HeDNA-dependent H3 ubiquitylation , with the exception of the previously reported SRA loss-of-function mutant ( D469G ) ( Avvakumov et al . , 2008 ) , that exhibited wild-type binding to HeDNA in our assays ( Figure 4—figure supplement 1D ) . Ubiquitylation defects observed for H3mut and Linkermut confirmed a critical role for the TTD-PHD as the substrate-binding domain for HeDNA-dependent H3 ubiquitylation and demonstrated that multivalent cis engagement of H3K9me3 by the UHRF1 TTD-PHD is required for proper ubiquitylation ( Figure 4D and Figure 4—figure supplement 1C ) . In addition , complete loss of HeDNA-dependent ubiquitylation for DNAmut UHRF1 and the absence of multi-ubiquitylated H3 for HeDNAmut UHRF1 further support the role of DNA binding and HeDNA recognition to fully activate UHRF1 ubiquitin ligase activity . The histone ubiquitylation defects observed for these loss-of-function mutants further highlights the interplay between UHRF1 functional domains to support proper UHRF1 ubiquitin ligase activity . There is a growing appreciation for the role of allosteric regulation of RING E3 ubiquitin ligase activity in the field of ubiquitin biology ( Vittal et al . , 2015 ) . Most often , the regulation of RING E3 ligases is accomplished through modulation of the E3 affinity for an E2-ub ( thioesterified E2-ubiquitin ) conjugate . Auto-inhibition release is the primary mechanism observed to date , in which steric occlusion of the RING domain prevents E3 interaction with the E2-ub until the E3 receives an appropriate release signal . Examples of RING auto-inhibition release include neddylation of Cullins ( Duda et al . , 2008; Saha and Deshaies , 2008 ) , phosphorylation of Cbl ( Dou et al . , 2012 ) , and substrate/peptide mimetic binding to inhibitor of apoptosis 1 ( Dueber et al . , 2011 ) . Taking into consideration the above-described allosteric regulatory mechanism of E3 ligase activity , we first sought to determine whether the interaction with HeDNA could affect association with E2-ub . Using isothermal titration calorimetry ( ITC ) and nuclear magnetic resonance ( NMR ) spectroscopy , we monitored the interaction between UHRF1 and E2-N-ub ( isopeptide-linked C85K E2-ubiquitin conjugate ) or 15N-E2-O-ub ( oxyesterified C85S E2-ubiquitin conjugate ) , respectively . Surprisingly , neither ITC nor NMR spectrum intensity loss measurements indicated a change in affinity for the E2 conjugates in the absence or presence of HeDNA ( Figure 5A–B and Figure 5—figure supplement 1 ) . In addition , we readily observed UHRF1 auto-ubiquitylation in the presence of HeDNA and UnDNA ( Figure 4B and Figure 4—figure supplement 1A ) , indicating the RING domain of UHRF1 could productively interact with E2-ub regardless of the methylation status of the bound DNA . Notably , upon E3 binding to the conjugated E2 , NMR resonances belonging to the conjugated ubiquitin did not suffer as great a loss in intensity ( Figure 5B and Figure 5—figure supplement 1B–C ) , indicating that ubiquitin retained its dynamics when bound to the E3 . This observation suggests that UHRF1 binding does not promote closed E2-ub states as strongly as other canonical RING domain E3’s ( Christensen et al . , 2007; Pruneda et al . , 2011a , 2012 ) . 10 . 7554/eLife . 17101 . 011Figure 5 . HeDNA binding directs ubiquitin to histone substrates . ( A ) ITC measuring the interaction of UHRF1 with E2-N-ub ( UbcH5c ( C85K ) -ub linked by isopeptide bond ) in the presence and absence of HeDNA ( see also Figure 5—figure supplement 1A ) . ( B ) Average peak intensities for 1H-15N HSQC-TROSY spectra of the 15N-E2-o-ub ( E2-o-Ub , UbcH5c ( S22R/C85S ) -ub esterified conjugate ) ( see also Figure 5—figure supplement 1B–C ) . Percentages indicate the reduction in intensity due to addition of UHRF1 or HeDNA . The addition of HeDNA to E2-ub ( comparing blue to red ) results in the same decrease in intensity as the addition of HeDNA to a sample containing E2-ub and UHRF1 ( comparing green to purple ) for both the E2 or ub within the conjugate . ( C ) Coomassie-stained gel of ubiquitin discharge assays in the presence of the indicated ligands and 20 mM free lysine ( left ) . Densitometry analysis of the indicated components of the reaction ( right ) . Line coloring corresponds to lane labels at the top of the gel . ( D ) Coomassie-stained gel of ubiquitin discharge assays in the presence HeDNA and either no peptide , H31-20 , H31-20K14aK18ac , H31-20K9acK14acK18ac and 20 mM free lysine ( left ) . Densitometry analysis of the indicated components of the reaction ( right ) . Line coloring corresponds to lane labels at the top of the gel . We conducted at least five ubiquitin discharge assays , and the trends observed for each condition in panels C and D were consistent across all experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 01110 . 7554/eLife . 17101 . 012Figure 5—figure supplement 1 . HeDNA binding does not modulate the interaction of UHRF1 with E2-ubiquitin conjugate . ( A ) Isotherm from ITC experiments monitoring E2-N-Ub binding to UHRF1 in the absence or presence of HeDNA . ( B ) , NMR analysis of UHRF1 interaction with E2-conjugated ubiquitin in the absence or presence of HeDNA . 1H-15N HSQC-TROSY spectra of the 15N-E2-o-ub conjugate ( UbcH5c ( Ser22Arg/Cys85Ser-O-ub; 200 μM ) in the presence or absence of HeDNA ( 0 . 11 Molar equivalents ) and/or MBP-UHRF1 ( 0 . 09 Molar equivalents ) . HeDNA does not appear to significantly perturb the 15N-E2-o-ub spectrum ( left ) . MBP-UHRF1 binds to the E2-ub conjugate as indicated by peak intensity loss ( middle ) , however HeDNA does not promote further binding ( right ) . ( C ) NMR peak intensities relative to the 15N-E2-o-ub conjugate spectrum for the E2 UbcH5c ( left ) and ub ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 012 Recently , another mechanism of allosteric regulation of RING activity has been described where a ligand , Poly-ADP-ribose ( PAR ) , induces a conformational change directly in the E3 RNF146 RING domain . This alternative RING conformation stabilizes the E2-ub/RING complex , thereby enhancing ubiquitin discharge from the conjugated E2 ( DaRosa et al . , 2015 ) . To test whether HeDNA-induced UHRF1 ubiquitin ligase activity enhanced ubiquitin discharge from the conjugated E2 , we performed single turnover ubiquitylation assays where purified E2-ub served as the ubiquitin donor and excess free lysine was present as a proxy ubiquitin substrate . The rate of ubiquitin discharge from E2 in these assays ( monitored by the loss of the E2-ub and the appearance of free E2 ) showed only a modest increase in the reactivity of the conjugate in the presence of HeDNA ( Figure 5C ) . These results suggest that HeDNA-dependent activation of UHRF1 RING activity does not occur through enhancement of the intrinsic rate of ubiquitin discharge from the E2 to non-specific lysine sidechains . Remarkably , when an H31-20 peptide ( Supplementary file 1 ) was added to ubiquitin discharge reactions , we observed rapid conversion of E2-ub to E2 ( Figure 5C ) . Additionally , we observed the appearance of a band corresponding to ubiquitylated H31-20 ( Figure 5C ) . Notably , we also observed a decrease in the amount of free ubiquitin and UHRF1 auto-ubiquitylation formed when peptide was present , presumably because more ubiquitin was being transferred to H3 ( Figure 5C ) . Thus , even under conditions where free lysine was in great excess , ubiquitin was transferred rapidly and preferentially to H3 substrate in the presence of HeDNA . To determine whether activation of UHRF1 occurs upon substrate binding ( i . e . , substrate-assisted activation ) , we performed single turnover assays in the presence of H31-20 , H31-20K14acK18ac , and H31-20K9acK14acK18ac ( Figure 5D , see also Supplementary file 1 ) . We previously demonstrated the interaction of the UHRF1 TTD-PHD with these potential substrates by peptide microarray ( Rothbart et al . , 2013 ) and reasoned that the acetylated peptides would maintain interaction with the TTD-PHD but would be unable to accept ubiquitin . Neither the H31-20K14acK18ac nor H31-20K9acK14acK18ac were capable of being modified with ubiquitin ( Figure 5D ) . Additionally , rapid E2-ub depletion was only observed in the sample containing H31-20 ( Figure 5D ) , indicating that substrate binding alone does not enhance the E3 ligase activity of UHRF1 in the presence of HeDNA . Consistent with our previous single turnover results ( Figure 5C ) , reactions that contained unblocked lysines on H31-20 accumulated less auto-ubiquitylated UHRF1 and free ubiquitin compared to assays with no peptide or with the acetylated H3 peptides ( Figure 5D ) , supporting a model where HeDNA alters the substrate specificity of UHRF1 ubiquitylation . It is worth noting that in all single and multiple turnover assays performed , virtually no ubiquitylated species of H3 were observed in assays that lacked HeDNA . Taken together , these data demonstrate that HeDNA stimulates UHRF1 ubiquitin ligase activity through a novel regulatory mechanism and suggest that HeDNA binding serves as an allosteric switch that directs the TTD-PHD bound H3 substrate to the E2-ub active site for transfer . We next used high-resolution mass spectrometry to further characterize the histone lysine specificity of HeDNA-stimulated UHRF1 ubiquitylation . HeLa mononucleosomes were reacted with UHRF1 in the presence of HeDNA or UnDNA for 2 hr using the enrichment strategy depicted in Figure 6A . Since histone proteins are highly basic , propionic anhydride was used to chemically modify free lysines and facilitate the identification of peptide fragments ( Garcia et al . , 2007 ) . We first identified peptides that were enriched in the HeDNA sample relative to the UnDNA sample , and the only histone peptides that were enriched greater than ten-fold were derived from H3 ( Supplementary file 2 ) ( see Materials and methods for information about normalization and quantification ) . Consistent with our findings that HeDNA stimulates UHRF1 ubiquitin ligase activity , ubiquitin remnants on H3 peptides were heavily enriched in samples reacted in the presence of HeDNA relative to UnDNA ( Figure 6B and Figure 6—figure supplement 1 , 2 ) . In addition to the previously identified sites H3K18 and H3K23 , we also found H3K14ub , H3K27ub , and H3K36ub heavily enriched in the presence of HeDNA ( Figure 6B , Figure 6—figure supplement 1 , 2 ) . We further identified abundant H3K18ub and H3K23ub ( multi-ubiquitylation ) on the same peptide . These results are consistent with laddering observed for H3ub in our experiments with synthetic peptides ( see Figure 4B ) and recombinant and native mononucleosomes ( see Figure 4C and Figure 4—figure supplement 1C ) , as well as immunoblots from HeLa cells in previous studies ( Nishiyama et al . , 2013; Qin et al . , 2015 ) . H3K18ub was the most abundant ubiquitylated peptide based upon spectral counts ( greater than 15-fold more than any other site ) , and H3K23ub was only observed in the context of H3K18ub ( Figure 6B ) . Collectively these results suggest that H3K18 is the preferred ubiquitylation site for UHRF1 , but that UHRF1 can target a number of lysines on the H3 tail . 10 . 7554/eLife . 17101 . 013Figure 6 . HeDNA stimulates UHRF1-directed ubiquitylation of multiple N-terminal lysines on histone H3 . ( A ) Schematic of the assay and sample preparation strategy to identify by LC-MC/MS products of UHRF1 ubiquitylation reactions with HeLa mononucleosomes in the presence of UnDNA or HeDNA . ( B ) Quantification of the area under the curve ( AUC ) from extracted-ion chromatograms for the indicated ubiquitylated H3 peptides enriched by immunoprecipitation of FLAG-ub . See Figure 6—figure supplement 1 for retention times and fragmentation for identified peptides . ( C ) Immunoblot analysis for Flag-ub and the indicated histone PTMs following UHRF1 ubiquitylation of HeLa mononucleosomes reacted in the presence of HeDNA or UnDNA , ( - ) indicates unreacted nucleosomes . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 01310 . 7554/eLife . 17101 . 014Figure 6—figure supplement 1 . Ion-extracted chromatograms ( left ) and fragmentation patterns ( right ) for each ubiquitylated peptide identified using the search procedures described in Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 01410 . 7554/eLife . 17101 . 015Figure 6—figure supplement 2 . Characterizing lysine prioritization of UHRF1 ubiquitylation on mononucleosomes . ( A ) Ubiquitylation of Hela mononucleosomes after 2 hr in the presence of HeDNA or UnDNA . ( B ) Ratio and normalized ratio ( HeDNA/UnDNA ) for H3 ubiquitylated peptides from propionylated samples . Samples were normalized to the mean ratio of free ubiquitin peptides . Quantification was performed with Skyline and values for each peptide charge state were summed . ( C ) Ratio and normalized ratio ( HeDNA/UnDNA ) for H3 ubiquitylated peptides from propionylated and unreacted samples . A ubiquitylated peptide from the PHD ( marked with an asterisk ) was enriched in the HeDNA sample . ( D ) Location of UHRF1 auto-ubiquitylation sites lining the TTD-PHD . TTD , pink; PHD , blue; modified lysines , red . The C-terminal atom on the H3 peptide ( S10 ) is shown as a sphere . ( E ) Auto-ubiquitylation of UHRF1 in the absence of H3 peptides with either HeDNA or UnDNA when run under conditions to resolve higher molecular weight species . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 01510 . 7554/eLife . 17101 . 016Figure 6—figure supplement 3 . UHRF1 targets H3K9me2 histones for ubiquitylation . HeDNA-stimulated UHRF1 ubiquitylation assays titrating recombinant histones H2A , H2B , H3 , or H3K9me2 ( 22 . 5 μM , 7 . 5 μM , 2 . 5 μM , 0 . 75 μM , 0 . 225 μM , 0 . 075 μM , 0 . 025 μM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 016 Consistent with our model of HeDNA altering UHRF1 substrate preferences , we also observed changes to the sites of UHRF1 auto-ubiquitylation in the presence of HeDNA ( Figure 6—figure supplement 2C–E ) . In particular , we identified a nine-fold enrichment of UHRF1 K303ub , a solvent exposed lysine in the PHD near the C-terminus of a bound H3 peptide ( Figure 6—figure supplement 2D ) . This region may represent the target zone for HeDNA-dependent ubiquitylation where the RING domain would be in proximity to this region . Accordingly , we observed an additional auto-ubiquitylated UHRF1 band in the presence of HeDNA compared to that observed with UnDNA ( Figure 6—figure supplement 2E ) , corroborating our mass spectrometry results . We also identified several histone PTMs that co-occurred with ubiquitylated H3 peptides , including all three states of H3K9 methylation ( Figure 6B ) . Additionally , several of the most enriched H3 peptides not containing ubiquitin remnants also contained H3K9me2 ( Supplementary file 2 ) . We confirmed this epigenetic link by immunoblotting HeLa mononucleosomes ubiquitylated by UHRF1 . Ubiquitylated H3 was detected in the presence of HeDNA ( but not UnDNA ) on nucleosomes marked with H3K9me3 , but not on nucleosomes marked with H3K9acK14ac ( Figure 6C ) . In addition , titrating recombinant human histones H3 , H2A , and H2B into ubiquitylation assays revealed that while UHRF1 could modify H2A and H2B at concentrations above 5 μM , H3 could be modified at sub-micromolar concentrations , and H3K9me2 protein ( synthesized by native chemical ligation ) could be modified at even lower concentrations ( Figure 6—figure supplement 3 ) . Taken together , these findings strongly support the role of H3K9 methylation in directing UHRF1 ubiquitylation to adjacent lysine residues in the presence of HeDNA . Other histone PTMs co-occurring on ubiquitylated peptides were: H3K23ac , also identified in another study ( Qin et al . , 2015 ) ; H3K27me2 , which often co-occurs with H3K9me2/me3 and is considered a hallmark of facultative heterochromatin ( Boros et al . , 2014 ) ; H3K36ac , and H3K37me3 ( Figure 6B and Figure 6—figure supplement 1 , 2 ) . However , future studies will be required to dissect the biological significance of these PTM combinations to ubiquitin ligase-dependent UHRF1 function . Our studies define an orchestrated sequence of histone- and DNA-binding events targeting UHRF1 to chromatin and identify a key regulatory mechanism controlling DNA methylation inheritance through UHRF1 E3 ligase activation following recognition of HeDNA . This mechanism is consistent with the observation that UHRF1-dependent H3 ubiquitylation accumulates in S-phase when HeDNA intermediates are generated behind replicating DNA polymerase ( Nishiyama et al . , 2013; Qin et al . , 2015 ) . Building on recent studies connecting H3 ubiquitylation to DNMT1 recruitment ( Nishiyama et al . , 2013; Qin et al . , 2015 ) , we propose a model where UHRF1 is targeted to chromatin through its coordinated histone and DNA reading activities ( Figure 7A ) . When UHRF1 encounters HeDNA , H3 ubiquitylation serves as a mechanism to facilitate the recruitment of DNMT1 to replicating regions of the genome to copy parental DNA methylation patterns ( Figure 7B ) . 10 . 7554/eLife . 17101 . 017Figure 7 . Proposed model for the contributions of DNA and histone recognition events to the DNA methylation regulatory function of UHRF1 . ( A ) UHRF1 is targeted to and retained on chromatin by the combined actions of H3K9me2/me3 recognition through the TTD-PHD and DNA interaction , independent of methylation status , through the SRA . ( B ) The interaction of the SRA with HeDNA , a DNA replication intermediate , directs the ubiquitin ligase activity of UHRF1 towards N-terminal lysines on the histone H3 tail . H3 ubiquitylation by UHRF1 contributes to the retention of DNMT1 in chromatin environments enriched for HeDNA and facilitates the epigenetic inheritance of DNA methylation patterns . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 01710 . 7554/eLife . 17101 . 018Figure 7—figure supplement 1 . Mouse UHRF1 ( Np95 ) SRA adopts different conformations bound to UnDNA ( left;PDB:2ZO2 ) and HeDNA ( right;PDB:3F8I ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17101 . 018 Structural characterization of the UHRF1 SRA bound to HeDNA ( Arita et al . , 2008; Avvakumov et al . , 2008; Hashimoto et al . , 2008 ) and cellular localization of UHRF1 with DNMT1 and PCNA ( proliferating cell nuclear antigen ) at replicating heterochromatic foci ( Bostick et al . , 2007; Sharif et al . , 2007 ) contribute to the model in which UHRF1 ubiquitylation of S-phase chromatin is mediated through HeDNA recognition . While it remains to be seen whether uncoupling UHRF1 from HeDNA recognition changes its residence genome-wide , our studies show HeDNA sensing is not required to target UHRF1 to bulk chromatin . Rather , coordinated recognition of H3 and DNA , independent of HeDNA discrimination , drives chromatin interaction . We propose that the avidity resulting from sub-μM affinities of UHRF1 for both DNA and H3 peptides through reciprocal positive allostery provides a biochemical basis by which UHRF1 is exclusively localized on chromatin . This may also explain why small perturbations to histone binding affinity through the TTD-PHD ( e . g . , Linkermut and TTD aromatic cage mutation [Rothbart et al . , 2013 , 2012] ) so dramatically affect chromatin targeting of this protein . Our studies define the UHRF1 TTD-PHD as the substrate-binding domain for HeDNA-stimulated ubiquitylation , further demonstrating a functional role for the coordinated recognition of H3K9me2/me3 and HeDNA in UHRF1 ubiquitylation . UHRF1 appears to be versatile in targeting lysines for ubiquitylation on the H3 tail ( Figure 6B ) : this ability may be related to the complexities of PTM patterning found on this region of the H3 tail ( Young et al . , 2009 ) and the necessity to promote efficient recruitment of DNMT1 to differentially modified chromatin environments . In addition , UHRF1 histone ubiquitylation may serve other roles in DNA related processes ( i . e . , DNA repair ) ( Liang et al . , 2015; Tian et al . , 2015; Zhang et al . , 2016 ) . Further studies are necessary to examine the biological consequence of different patterns of H3 ubiquitylation by UHRF1 and their relationship to pre-existing histone PTM signatures . Important to note are studies on the enzymology of DNMT1 activity that show the enzyme has an intrinsic preference for HeDNA substrates ( Goyal et al . , 2006 ) and methylates in a processive manner ( Bestor and Ingram , 1983; Hermann et al . , 2004 ) . Our studies define a major function for HeDNA , beyond direct stimulation of DNMT1 activity , in the regulation of UHRF1 histone ubiquitylation and DNA methylation inheritance . Considering DNMT1 behavior on oligonucleotide substrates , it is intriguing to speculate that UHRF1 functions to provide a nucleation event for DNMT1 recruitment to chromatin . Future studies mapping the genome-wide distribution of UHRF1-directed H3 ubiquitylation in relation to DNA methylation patterning will clarify the relationship between UHRF1 and DNMT1 activities . How might HeDNA binding alter the substrate preference of UHRF1 directed ubiquitylation ? We speculate that HeDNA is bound by the UHRF1 SRA in a manner that positions the RING in proximity to the H3 binding region of UHRF1 . RING activity towards H3 may be conformationally restricted when UHRF1 binds SyDNA or UnDNA ( Figure 7 ) . Consistent with this hypothesis , structures of the SRA domain from mouse UHRF1 ( Np95 ) show this domain can adopt different conformations bound to UnDNA and HeDNA ( Figure 7—figure supplement 1 ) ( Hashimoto et al . , 2008 ) . Also the NKR finger of the UHRF1 SRA , which harbors the HeDNAmut , adopts a highly ordered conformation upon HeDNA binding ( Avvakumov et al . , 2008; Hashimoto et al . , 2008 ) . This allows for pseudo-base pairing to the exposed guanosine nucleotide ( Figure 1—figure supplement 1 ) , and we hypothesize this stable finger conformation is critical for HeDNA-stimulated H3 ubiquitylation . Unfortunately , efforts to crystallize the enzymatically active conformation of UHRF1 proved unsuccessful . Determining the active conformation of UHRF1 will be an important step in further understanding the regulation imparted by HeDNA . Why might ubiquitin be an ideal PTM to accompany a temporally controlled process like replication-coupled DNA methylation ? Ubiquitin itself is a functional protein domain capable of participating a wide variety of protein-protein interactions ( Harrison et al . , 2016 ) and can sterically occlude surfaces , as has been proposed for H2BK120ub in the formation of a productive complex with DOT1L ( Zhou et al . , 2016 ) . Ubiquitin modifications are also dynamic and can be rapidly removed by deubiquitylases . In fact , recent analysis of ubiquitin turnover kinetics showed that the half-life of H2BK123ub in budding yeast is approximately one minute ( Yumerefendi et al . , 2016 ) . Thus , discovering the identity of the deubiquitylase that removes H3 ubiquitylation may provide key insight into the dynamics of DNA methylation regulation at the level of histone ubiquitylation . The initial study implicating H3 ubiquitylation in the inheritance of DNA methylation indirectly suggested that USP7 may be responsible for this function through interaction with DNMT1 ( Nishiyama et al . , 2013 ) . This is notable , as USP7 has also been shown to interact with UHRF1 ( Zhang et al . , 2015 ) . Additionally , recent studies have tied USP7 to DNA replication ( Lecona et al . , 2016 ) and the maintenance of heterochromatin ( Mungamuri et al . , 2016 ) , providing a biological link to replication-coupled inheritance of DNA methylation . However direct evidence of USP7 catalyzed deubiquitylation of H3 is lacking . In conclusion , our study defines the relationship between UHRF1 histone-binding , DNA-binding , and ubiquitylation activities and connects HeDNA recognition to UHRF1 enzymatic function . Additionally , we characterize HeDNA as an active epigenetic mark that allosterically regulates UHRF1 ubiquitylation towards histone H3 . More broadly , these finding provide a function for epigenetic patterning associated with UHRF1 beyond protein recruitment . We speculate that epigenetic mechanisms of multivalency and allostery are more widespread and add additional layers of complexity , specificity , and connectivity to chromatin recognition , modification patterning , and genome regulation . The cDNA that encodes amino acids 1–793 of human UHRF1 ( full length ) was cloned into a modified pGEX vector in frame with an N-terminal 6xHis-MBP tag that can be cleaved with TEV protease . E . coli were grown to O . D . 0 . 6 and induced with 600 mM IPTG overnight at 18°C . Cells were collected by centrifugation and resuspended in lysis buffer ( 50 mM Tris-HCl , pH 8 . 0 , 300 mM NaCl , 2 mM PMSF , 1 μM Bestatin , 1 μM Pepstain A , and 10 μM Leupeptin [Thermo Fisher Scientific , Waltham , MA] ) , lysed with sonication on ice , and cellular debris was pelleted at 15 , 000 x g for 30 min . The supernatant was passed over a HisTRAP nickel column ( GE Lifesciences , Pittsburgh , PA ) , washed ( 50 mM Tris-HCl , pH 8 . 0 , 1 M NaCl , and 15 mM imidazole ) and eluted ( 25 mM HEPES , pH 7 . 5 , 100 mM NaCl , and 250 mM imidazole ) . Eluted protein was concentrated to −2 mL using a 10 kDa spin concentrator ( Amicon Ultra ) and further purified by size-exclusion chromatography ( SEC ) over a Superdex S-200 ( 16/600 ) column ( GE Lifesciences ) in 25 mM HEPES , 100 mM NaCl , and 1 mM DTT . Monomeric fractions were pooled and concentrated to 100–200 μM . The purified protein was either used directly or was bound to MBP resin for overnight cleavage with TEV protease purified as previously described ( Tropea et al . , 2009 ) . Cleaved UHRF1 was less stable at higher concentrations than 6xHis-MBP-UHRF1 but behaved similarly in binding and ubiquitylation assays . To complete the study , we purified UHRF1 from bacteria more than 10 times , and all protein preparations were functional and behaved similarly . Mutations were introduced into cDNAs by Quick Change ( Agilent , Santa Clara , CA ) and purified mutant proteins behaved similarly to wild-type protein , but were generally less stable at higher concentration . To circumvent this issue , UHRF1 mutants were characterized as MBP fusions . Histone peptides N- and C-terminally labeled with 5-carboxyfluorescein ( FAM ) were synthesized as described ( Rothbart et al . , 2013 ) . 6-FAM-labeled double-stranded DNA was generated by annealing the following combinations of synthetic oligonucleotides ( Eurofins , Louisville , KY ) ; FAM-5’-CCATGXGCTGAC-3’ and 5’-GTCAGYGCATGG-3’ , where X and Y are both cytosine ( UnDNA ) , X is cytosine and Y is 5mC ( HeDNA ) , or X and Y are both 5mC ( SyDNA ) . Binding experiments were performed in 25 μL in black flat-bottom 384-well plates ( Corning , Tewskbury , MA ) . Protein was titrated with 10 nM FAM-labeled DNA or histone peptides in buffer containing 25 mM HEPES , pH 7 . 5 , 0 . 05% NP-40 , 100 mM NaCl ( unless otherwise indicated ) . Where indicated , 10 μM unlabeled DNA or histone peptide was included in the reaction mix . Following a 10 min incubation period , fluorescence polarization measurements were performed at 25°C with a PHERAstar fluorescence microplate reader ( BMG Labtech , Cary , NC ) using a 480-nm excitation filter and 520/530 ± 10-nm emissions filters . Gain settings in the parallel ( \|\| ) and perpendicular ( ⊥ ) channels were calibrated to a polarization measurement of 100 milli-polarization units ( mP ) for the FAM tracer in the absence of protein . Polarization ( P ) was determined from raw intensity values of the parallel and perpendicular channels using the equation P = \|\| – ⊥ / \|\| + 2 ( ⊥ ) and converted to anisotropy ( A ) units using the equation A = 2P / 3 – P . Equilibrium dissociation constants ( Kd ) were determined by non-linear regression analysis of anisotropy curves using a one-site binding model in GraphPad Prism . To control for variability in salt concentration , each experiment included a wild-type protein as a reference . Accordingly , the methyl preference for DNA binding ( HeDNA , SyDNA , and UnDNA ) and the positive allostery of histone and DNA binding of the wild-type protein was observed in greater than ten independent experiments in various buffers and salt concentrations with several batches of purified protein . Ubiquitylation assays were typically performed in 20 μL reactions containing 1 . 5 μM UHRF1 , 100 nM E1 activating enzyme ( Boston Biochem #E-304; Cambridge , MA ) , 200 nM E2 Ubc5c ( purified in house over HisTRAP column ) , 2 . 5 mM MgCl2 , 1 mM DTT , 5 μM FLAG-ubiquitin ( Boston Biochem ) , 10 mM ATP , 25 mM HEPES , pH 7 . 5 , and 100 mM NaCl . Unless otherwise indicated , peptide concentrations were 13 μM , and HeDNA , SyDNA , and UnDNA concentrations were 3 μM , 10 μM , and 40 μM , respectively . Assays were performed at 25°C and quenched after 20 min with SDS-PAGE loading buffer ( 2% SDS , 10% glycerol , 1% 2-Mercaptoethanol , 50 mM Tris-HCl pH 6 . 8 , 0 . 01% bromophenol blue ) . Reactions were ran on 16% SDS-PAGE gels , transferred to PVDF membranes , and visualized using fluorescent imaging of immunoblots probed for FLAG-ubiquitin with FLAG ( Sigma #F3165 , 1:5000; St . Louis , MO or BioLegends #637304 , 1:5000; San Diego , CA ) and Alexa Fluor 488 or 647 ( Life Technologies 1:5 , 000; Carlsbad , CA ) antibodies on a Typhoon Trio+ fluorescent scanner ( GE Lifesciences ) . Histone peptide substrates were synthesized as previously described ( Rothbart et al . , 2013 ) . Recombinant histone proteins and mononucleosomes were obtained commercially from Epicypher ( H2A , #15–0301; H2B , 15–0302; and H3 . 1 , #15–0303; mononucleosomes , #16–0002; Research Triangle Park , NC ) . Allosteric activation of UHRF1 ubiquitylation activity towards histone peptides and nucleosomes was observed in more than ten independent experiments , and DNA and UHRF1 titrations were repeated three times . The activities of the mutant proteins were tested in five independent experiments with at least two protein preparations for each mutant . Rate measurements for UHRF1 ubiquitylation activities in the presence of HeDNA vs UnDNA were conducted three times with similar results as Figure 4B . C-terminal H3 peptide ( amino acids 11–135; T11C ) was prepared as described ( Shogren-Knaak et al . , 2003 ) by cleavage of precursor with Factor 10X . Purification by reverse-phase HPLC followed by pooling of appropriate fractions and lyophilization afforded a white solid ( 6 . 2 mg ) . The theoretical mass of C622H1040N196O172S3 product is 14112 . 54 Da and the measured mass of the product was 14112 . 94 Da . N-termial peptide thioester ARTKQTARK ( me2 ) S-Mes-OH was synthesized as described ( Mahto et al . , 2011 ) and purified by reverse-phase HPLC to 70% purity . After purification peptide contained 30% of hydrolysis product ( ARTKQTARK ( me2 ) S-OH ) . A mixture of 1 mg of C-terminal peptide ( 70 . 86 nmoles ) and 0 . 52 mg of an N-terminal peptide thioester ( 70% pure; 280 . 3 nmoles; 4 molar equivalents ) in 0 . 5 mL of ligation buffer ( 3 M Guanidine-HCl , pH 7 . 9 , 100 mM potassium phosphate ) was treated with benzyl mercaptan ( 2 . 5 μL ) and thiophenol ( 2 . 5 μL ) , and the mixture shaken vigorously for 24 hr . The reaction mixture was diluted with ligation buffer ( 500 μL ) , treated with MeCN:water:trifluoroethanol ( 750 μL; 25:75:0 . 1 ) , and desalted by dialysis ( 2 x 30 min with water change ) . Analysis by reverse-phase HPLC and by gel electrophoresis on SDS-18% polyacrylamide gel followed by staining with coomassie blue indicated a complete ligation reaction . Purification by reverse-phase HPLC followed by pooling of appropriate fractions and lyophilization afforded H3K9me2 ( 1–135 ) T11C as a white solid ( 0 . 70 mg; 65% ) . The theoretical mass of C670H1129N215O186S3 is 15268 . 90 Da and the measured mass of the product was 15269 . 19 Da . Ligated peptide ( H3K9me2 T11C; 0 . 7 mg ) was dissolved in argon-degassed desulfurization buffer ( 200 mM phosphate , 6 M guanidine-HCl , pH 6 . 7; 0 . 15 mL ) and treated with ethanethiol ( 2 μL ) , TCEP ( 0 . 15 mL of 0 . 5 M in desulfurization buffer ) , t-butanethiol ( 10 μL ) , and VA-061 ( 2 , 2'-azobis[2- ( 2-imidazolin-2-yl ) propane] ) in methanol ( 2 μL of 0 . 2 M solution ) and incubated at 37°C for 24 hr . The resultant mixture was purified by reverse-phase HPLC followed by pooling of appropriate fractions and lyophilization to afford H3 1–135 T11A as a white solid ( 0 . 55 mg ) . The theoretical mass of C670H1129N215O186S2 is 15236 . 84 Da and the measured mass of product was 15237 . 16 Da . Lysine reactivity assays were performed as previously described ( DaRosa et al . , 2015; Wenzel et al . , 2011 ) . Briefly , the UbcH5c-Ub conjugate was generated in 25 mM sodium phosphate , pH 7 . 0 and 100 mM NaCl containing 1 . 5 μM human E1 , 250 μM Ub , 100 μM UbcH5c , 2 . 5 mM MgCl2 , and 2 mM ATP ( Sigma ) . Reactions were incubated for 40 min at 37°C , then purified by SEC to isolate E2-Ub . SEC-purified E2-Ub was added to UHRF1 E3 samples incubated with HeDNA or buffer for 30 min on ice to form a final concentration of 8 μM E3 , 25 μM E2-Ub , and , where indicated , 13 μM HeDNA and 12 μM peptide . After a zero min time point was taken , buffered L-lysine HCl ( Sigma ) was added to a final concentration of 20 mM and samples were incubated at 35°C , removing samples at indicated time points . Samples were quenched in non-reducing SDS sample loading buffer and analyzed by SDS-PAGE stained with either Coomassie or Oriole fluorescent gel stain ( Bio-Rad , Hercules , CA ) . Lysine reactivity assay performed in the presence of excess free lysine were performed in the following conditions: 32 μM E2-Ub ( UbcH5c and WT Ub ) , 8 μM UHRF1 , 13 μM HeDNA , 11 μM H3 ( 1–20 ) K9me3 , with 20 mM Lysine . Ubiquitin discharge assays were performed at least five times in the lab and yielded results consistent with Figures 5C–D . His-MBP-tagged UHRF1 SRA-RING ( amino acids 405–793 ) was produced in E . coli as described above . GST-tagged UHRF1 TTD-PHD ( amino acids 123–366 ) and BPTF PHD-Bromo ( gift from Dr . Alex Ruthenburg [Ruthenburg et al . , 2011] ) were produced as previously described ( Rothbart et al . , 2013 ) . Proteins ( each at 1 μM ) were incubated overnight at 4°C with MBP magnetic beads ( NEB , Ipswich , MA ) in binding buffer containing 50 mM Tris-HCl , pH 8 . 0 , 100 mM NaCl , 0 . 1% NP-40 , 0 . 5% BSA , and , where indicated , 25 μM DNA oligonucleotides or histone peptides . Pulldown experiments with the SRA-RING DNAmut were performed with 5 μM DNA . Samples were washed extensively with binding buffer , eluted in SDS sample buffer , resolved by SDS-PAGE , transferred to PVDF membrane ( Thermo ) , and probed with GST antibody ( Sigma #G7781 , 1:2 , 000 ) . Pull-down assays were performed in triplicate . Asynchronously growing HeLa cells were harvested by trypsinization 48 hr post transfection with the indicated FLAG-tagged human UHRF1 constructs . Pellets were washed once with cold 1x PBS , snap frozen in liquid N2 and either processed immediately or stored at −80°C . Cell pellets were resuspended in 1x volume CSK buffer ( 10 mM PIPES pH 7 . 0 , 300 mM sucrose , 100 mM NaCl , 3 mM MgCl2 , 0 . 1% Triton X-100 and 1x Complete EDTA-Free protease inhibitor cocktail from Roche ) and incubated on ice for 20 min . Total protein was quantified by Bradford Assay ( BioRad ) , and 10% of this total fraction was combined with an equivalent volume of CSK buffer supplemented with Universal Nuclease ( Thermo , 1:5 , 000 ) . Note that the concentration of the total fraction is now 0 . 5x . The remaining cell lysate was centrifuged at 1300 x g for 5 min at 4°C . The supernatant ( soluble fraction ) was collected . The chromatin pellet was resuspended in 1x volume CSK buffer and kept on ice for 10 min before being spun again at 1300 x g for 5 min at 4°C . The supernatant was discarded and the chromatin pellet was solubilized in CSK buffer supplemented with Universal Nuclease . 1–5 μg of protein from each fraction ( estimated from Bradford on total extract ) was resolved by SDS-PAGE , transferred to PVDF membrane ( Thermo ) , and probed with the indicated antibodies ( Flag , Sigma #F1804 , 1:5 , 000; β-tubulin , Millipore #05–661 , 1:5 , 000 , H3 , Epicypher #13–0001 , 1:25 , 000 ) . Immunofluorescence analysis of 5mC content was performed essentially as described with the following modifications ( Rothbart et al . , 2012 ) . HeLa cells grown in 4-well chamber slides ( Nunc Lab-Tek ) were fixed with ice-cold methanol at −20°C for 10 min . To denature the DNA , fixed cells were treated with 2 N HCl for 30 min at 37°C and washed twice with 0 . 1 M boric acid , pH 8 . 5 . Cells were blocked for 30 min in PBS containing 1% ( w/v ) BSA and labeled with an anti-5mC antibody ( Active Motif #39649 , 1:500; Carlsbad , CA ) in PBS containing 1% BSA for 1 hr at room temperature . Cells were washed with PBS and incubated with an Alexa Fluor 647-conjugated secondary antibody ( Life Technologies #A21236 , 1:1000 ) for 1 hr at room temperature protected from light . Cells were washed with PBS and mounted with SlowFade Gold Antifade with DAPI ( Thermo #S36942 ) . Images were acquired using a Nikon A1+ RSi confocal microscope using a 60x objective following excitation with 403-nm and 640-nm solid-state lasers . The 5mC signal from each image was quantified using the equation ∑i1[bi>t]1[ri>t] ( ri−t ) ∑i1[bi>t] , where bi is DAPI signal intensity for an individual pixel , ri is 5mC signal intensity for an individual pixel , and t defines the background signal threshold . The percent of control 5mC was calculated using the mean 5mC signal from at least four fields of view . 20 μg of Hela extracted mononuclesomes ( Epichyper #16–0002 ) were used as substrate in each ubiquitylation reaction supplied with either HeDNA or UnDNA ( described above ) for 2 hr . The reactions were placed on ice , treated with Universal Nuclease ( Thermo , 1:5 , 000 ) , and the ubiquitylated products were immunoprecipitated with FLAG M2 magnetic beads ( Sigma ) . The resin was washed 3x with 1 mL of wash buffer ( HEPES , pH 7 . 5 , 100 mM NaCl ) , split in half , and the beads were transferred to spin columns ( Vivacon ) and sequencing grade modified trypsin ( Promega , Madison , WI ) . Half of the sample was reacted with proprionic anhydride ( Alfa Aesar ) using a modified version of this procedure ( Lin and Garcia , 2012 ) . 100 μl of a 1:3 ratio of proprionic anhydride diluted in 100 mM NH4CO3 , pH 8 . 0 was added to each spin column followed by 50 μl NH4OH to adjust the pH to 8 . 0 . Each reaction was incubated for 30 min at 30°C before being spun through the column . This protocol was repeated to ensure complete proprionylation of free lysines in the sample . The proprionylated and unreacted samples were then digested on resin using sequencing grade modified trypsin ( Promega ) digested at 37°C for 2 hr . This mixture was analyzed with LC-MS/MS without proprionylation of the free amines exposed after trypsin digestion . The peptide mixture was analyzed in positive mode using a nanoAquity UPLC coupled LTQ Orbitrap Elite mass spectrometer ( Thermo ) . Chromatographic separation used a 2 cm trapping column ( Acclaim PepMap 100 ) and a 15 cm EASY-spray analytical column ( 75 μm ID , C18 beads of 3 . 0 μm particle size , 100 Å pore size ) . The HPLC flow rate was set to 350 nL/min over a gradient of 1% buffer B ( 0 . 1% formic acid in acetonitrile ) to 25% buffer B in 150 min . The full mass scan ( 300 to 2000 m/z ) was acquired at a resolution of 120 , 000 with a maximum injection time of 500 ms , and MS/MS was performed in a data-dependent manner for the top 15 intense ions in the linear ion trap by collision-induced dissociation . Raw data were converted to mzXML format using ProteoWizard ( Kessner et al . , 2008 ) and searched using the Crux pipeline ( McIlwain et al . , 2014 ) ( version 2 . 1 . 16867 ) against the human UniProtKB/Swiss-Prot sequence database ( downloaded on 2/20/15 ) ( Boutet et al . , 2007 ) . Search parameters were set as the following: peptides between 6 and 25 amino acids long with a precursor mass tolerance of 0 . 5 amu , no missed cleavages , fully-enzymatic Arg-C digestion , a static propionyl modification ( +56 . 026215 ) on lysines , and a maximum of 4 variable modifications consisting of up to 2 lysine ubiquitinations ( +58 . 016716 ) , 2 methylations ( +14 . 01565 ) , 2 dimethylations ( −27 . 994915 ) , 2 trimethylations ( −13 . 979264 ) , 2 acetylations ( −14 . 015644 ) , 1 methionine oxidation ( +15 . 99492 ) , and 1 STY phosphorylation ( +79 . 966331 ) . The mass of propionyl was subtracted from variable lysine modification masses ( except methylation ) due to the already applied static propionyl modification . For unpropionylated samples , the differing parameters were: up to 3 missed cleavages , fully-enzymatic trypsin digestion , no static modifications , and a maximum of 4 variable modifications consisting of up to 2 lysine ubiquitinations ( +114 . 042931 ) , 2 methylations ( +14 . 01565 ) , 2 dimethylations ( +28 . 0313 ) , 2 trimethylations ( +42 . 046951 ) , 2 acetylations ( +42 . 010571 ) , 1 methionine oxidation ( +15 . 99492 ) , and 1 STY phosphorylation ( +79 . 966331 ) . Prior to execution of the Percolator algorithm supplied by Crux , deltaCn scores were re-computed using an alternate definition: deltaCni = 1 – ( ( xcorr1-xcorri ) / xcorr1 ) . This adjustment was performed because the similar mass of trimethylation and acetylation results in identical xcorr values for the low mass accuracy MS/MS spectra from linear ion traps , which then led to invalid deltaCn values with the default equation used by Percolator . After application of a 5% FDR threshold , peptides were further filtered by ensuring they had the expected retention time relative to peptides having the identical unmodified sequence . We used the following procedure . First , peptides with the same unmodified sequence were sorted in ascending order by their Percolator PEP ( posterior error probability ) . Then , each peptide ( starting from lowest to highest PEP ) was accepted if at least one of its MS/MS scan’s retention time was consistent relative to all currently accepted peptides having the same unmodified sequence . The expected relative retention time constraints were: ubiquitin < dimethyl ≤ trimethyl < acetyl < propionyl < methyl , oxidation < unmodified , and phosphorylation ≤ unmodified . Peptides expected to have the same retention times were allowed to elute within 2 min of each other . Finally , peptide H3K9me3 + K14ub was accepted after manual inspection of its corresponding MS/MS spectra , isotopic distribution , and its consistent retention time despite being above the 5% FDR threshold . Quantification was performed within Skyline ( MacLean et al . , 2010 ) and the results were exported for further visualization and analysis using the R programming language . Proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD003983 . A 10 μM solution of apo-UHRF1 or 1:1:1 ratio of ligands ( HeDNA and H3 ( 1–15 ) K9me3 peptide ) was passed over a Superdex 200 ( 10/300 ) GL column using an AKTA purifier FPLC ( GE Lifesciences ) in size exclusion buffer ( 25 mM HEPES pH 7 . 4 , 100 mM NaCl , 1 mM DTT ) with a flow rate of 0 . 5 mL/min . Samples with ligand were allowed to equilibrate for 10 min prior to injection onto the column . The apparent molecular weight was calculated using a linear fit to the retention time for a set of molecular weight standards ( BioRad #1511901 ) . Analytical size exclusion experiments were repeated three times with identical results . Dynamic light scattering was measured using were a DynaPro Plate Reader ( Wyatt Technology , Goleta , CA ) . UHRF1 was at 5 μM and 10 μM H3 ( 1–15 ) K9me2 or DNA was added to a final volume of 50 μL in buffer ( HEPES pH 7 . 5 100 mM NaCl and 1 mM DTT ) . Samples were incubated for 10 min before monitoring light scattering for over 100 s for each sample . Light scatter for each ligand alone yielded a low intensity and poly-dispersed signal that did not significantly contribute to the scattering when UHRF1 was present . Addition of the ligand however likely accounts for the small increases to poly-dispersity observed upon addition of ligand . A 20 nM solution of UHRF1 ( 25 mM HEPES pH 7 . 4 , 100 mM NaOAc , 1 mM DTT ) was mixed with or without HeDNA ( 5 μM ) and deposited on freshly peeled mica , immediately rinsed with water ( Sigma #W4502 ) , and dried with nitrogen gas before imagining . All images were acquired on the same day as the deposition . Images were collected on an MFP3D Atomic Force Microscope ( Asylum Research Oxford Instruments using the following parameters: scan rate 1 Hertz , scan size 1 μM x 1 μM , image resolution 1024 x 512 . Images were collected in intermittent contact mode ( AC mode ) using AFM probes from NanoSensor ( PPP-FMR , force constant = 2 . 8 N/m ) . Images were analyzed using the Asylum Research AFM software package . The images were flattened to a second-degree polynomial to account for surface warping artifacts and volume analysis was performed using built-in particle analysis ( a more detailed review of this methodology can be found here Ratcliff and Erie , 2001 ) . Volume distributions were plotted to a peak fit model and visualized using Origin 6 . 1 ( origin labs ) . The fact we could only identify a single volume species indicates monomeric UHRF1; the kD ( data not shown ) we calculated from AFM volume is also in agreement with monomeric UHRF1 ) . Oxyester-linked 15N E2-O-Ub conjugate ( UbcH5c ( Ser22Arg/Cys85Ser ) -O-Ub was generated as previously described ( Pruneda et al . , 2011b ) . Two-dimensional 1H-15N HSQC-TROSY experiments were performed with 200 μM 15N E2-O-Ub conjugate in 25 mM sodium phosphate , pH 7 . 0 and 150 mM NaCl on a Bruker 500 MHz AVANCE II NMR spectrometer . MBP-UHRF1 and/or HeDNA was added to experiments to a final concentration of 18 μM and 22 μM , respectively . NMR data was processed with NMRPipe ( Delaglio et al . , 1995 ) and peak intensities were determined using NMRViewJ ( Johnson and Blevins , 1994 ) ( OneMoonScientific ) . Relative peak intensity changes were determined as the absolute peak intensity divided by the initial intensity of the E2-O-Ub conjugate in the absence of additives . E2-N-ub was generated as previously described ( Branigan et al . , 2015 ) . ITC experiments were performed at 25°C in a MicroCal iTC200 in 25 mM Hepes pH 7 . 4 and 100 mM NaCl . UHRF1 was at 12 μM and the E2-N-ub was at 218 μM . Data was fit to a single site binding model with Origin .	Cells are able to regulate the activity of their genes in response to different cues . Genetic information is encoded in DNA and one way to regulate gene activity is to modify the DNA by attaching chemical “epigenetic” markers to it . When a cell divides , these epigenetic markers can be inherited by the daughter cells so that they share the same patterns of gene activity as the parent cell . When the DNA of the parent cell is copied prior to cell division , the epigenetic markers are also copied onto the new DNA . Mistakes in this process are linked to a wide range of diseases in humans , such as cancer and neurological disorders . One type of epigenetic marker is known as a methyl tag and it is added to DNA by certain enzymes in a process called DNA methylation . A protein called UHRF1 is required for human cells to inherit patterns of DNA methylation through cell division . This protein binds to newly copied DNA that lacks some methyl tags as well as to another protein associated with DNA called histone H3 . UHRF1 modifies histone H3 by attaching a small protein molecule called ubiquitin to it . This helps to recruit a DNA methylation enzyme to place methyl tags on the newly copied DNA . However , it was not clear how the various properties of UHRF1 allow it to control how DNA methylation is inherited . Harrison et al . addressed this question by studying purified proteins and DNA fragments outside of living cells . The results show that UHRF1 binding to DNA and histone H3 work together to bring UHRF1 to the sites on DNA that require methylation . Further experiments revealed that the methylation pattern on newly copied DNA is able to activate the ability of UHRF1 to place ubiquitin on histone H3 . The findings of Harrison et al . reveal a new mechanism by which dividing cells control how DNA methylation is inherited by their daughter cells . A future challenge will be to find out how attaching ubiquitin to histone H3 activates DNA methylation .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2016	Hemi-methylated DNA regulates DNA methylation inheritance through allosteric activation of H3 ubiquitylation by UHRF1
A well-established cascade of transcription factor ( TF ) activity orchestrates adipogenesis in response to chemical cues , yet how cell-intrinsic determinants of differentiation such as cell shape and/or seeding density inform this transcriptional program remain enigmatic . Here , we uncover a novel mechanism licensing transcription in human mesenchymal stem cells ( hMSCs ) adipogenically primed by confluence . Prior to adipogenesis , confluency promotes heterodimer recruitment of the bZip TFs C/EBPβ and ATF4 to a non-canonical C/EBP DNA sequence . ATF4 depletion decreases both cell-density-dependent transcription and adipocyte differentiation . Global profiling in hMSCs and a novel cell-free assay reveals that ATF4 requires C/EBPβ for genomic binding at a motif distinct from that bound by the C/EBPβ homodimer . Our observations demonstrate that C/EBPβ bridges the transcriptional programs in naïve , confluent cells and early differentiating pre-adipocytes . Moreover , they suggest that homo- and heterodimer formation poise C/EBPβ to execute diverse and stage-specific transcriptional programs by exploiting an expanded motif repertoire . Post-mitotic adipocytes arise from the differentiation of MSCs in a process termed adipogenesis ( Cristancho and Lazar , 2011; Rosen and Spiegelman , 2014 ) . The cell-fate decisions occurring during the stages of adipogenesis are controlled by multiple sequence-specific transcription factors ( TFs ) ( Farmer , 2006; Rosen and MacDougald , 2006; Cristancho and Lazar , 2011 ) . Best described among these are PPARγ and C/EBP TFs , which drive a process of terminal differentiation that results in the expression of metabolic genes and adipokines important for the adipocyte phenotype ( Hwang et al . , 1997; Rosen and MacDougald , 2006; Lefterova and Lazar , 2009 ) . PPARγ is recognized as a master regulator because it is necessary ( Barak et al . , 1999; Kubota et al . , 1999; Rosen et al . , 1999 ) and sufficient ( Tontonoz et al . , 1994; Hu et al . , 1995; Shao and Lazar , 1997 ) for adipogenesis . C/EBP proteins can also induce adipocyte differentiation of fibroblasts , although none can induce differentiation in the absence of PPARγ ( Rosen et al . , 2002 ) . Cistromic analyses have revealed that PPARγ and C/EBPα occupy sites near most of the genes that are up-regulated during adipogenesis , including their own , suggesting that they coordinate expression of the majority of genes determining adipocyte function ( Lefterova et al . , 2008; Nielsen et al . , 2008; Mikkelsen et al . , 2010; Schmidt et al . , 2011; Soccio et al . , 2011 ) . Thus , the transcriptional circuitry formed by PPARγ and C/EBPα may explain how adipocyte cell identity is established and maintained . Genomics studies have provided important new insights into the TFs acting before PPARγ and C/EBPα . Treatment of pre-adipocyte fibroblasts with a cocktail including glucocorticoid , cyclic AMP agonists and insulin ( DMI ) and activates thousands of putative enhancers with increased sensitivity to DNase I , enrichment for activating histone modifications , and induced binding by TFs ( Mikkelsen et al . , 2010; Siersbæk et al . , 2014a ) . Acting transiently during the early stages of adipogenesis , these TFs appear to serve at least two important functions: they recruit co-activators such as p300 and MED1 to activate genes important for terminal differentiation including PPARγ ( Steger et al . , 2010; Siersbæk et al . , 2014b ) , and they remodel chromatin to facilitate later binding by PPARγ and C/EBPα in adipocytes ( Siersbæk et al . , 2011 ) . While DMI is a potent trigger for adipogenesis in both mouse 3T3-L1 fibroblasts and human mesenchymal stem cells ( hMSCs ) , it is effective only when cells are densely packed ( Green and Kehinde , 1976; Pittenger et al . , 1999; McBeath et al . , 2004; Cristancho et al . , 2011 ) . Here , we use this cell density ‘checkpoint’ as a means to identify transcriptional mechanisms that prime cells to a permissive pre-adipogenic state . We identify on a genome-wide scale putative enhancer and promoter regions with differential occupancy for RNA Polymerase II ( RNAPII ) in response to changes in cell density and treatment with DMI cocktail . The findings support roles for C/EBPβ and GR as primary drivers of DMI-induced gene expression . We also observe a surprising enrichment for C/EBPβ-binding sites at RNAPII enhancers responding to high-seeding density prior to the addition of DMI cocktail . While lacking a canonical , palindromic C/EBPβ motif , these enhancers exhibit dramatic enrichment for an asymmetric , composite motif that juxtaposes half-sites for canonical C/EBP and AP-1 motifs . We demonstrate that these hybrid motifs recruit C/EBPβ as a heterodimer with another bZIP family member , ATF4 . Genome-wide binding by ATF4 demonstrates its exclusive co-localization with C/EBPβ , and depletion of ATF4 decreases adipogenesis . Together , these observations suggest that a program of C/EBPβ-ATF4-dependent gene expression triggered by high-seeding density plays an important role in priming hMSCs for adipogenesis . The observation that C/EBPβ hetero- and homodimeric complexes exhibit different sequence specificities provides novel mechanistic insights into how C/EBPβ can be differentially targeted to control distinct programs of gene expression at distinct phases of adipocyte differentiation , and revises the prevailing view that C/EBPβ is transcriptionally inactive in the absence of exogenous adipogenic stimuli ( Wiper-Bergeron et al . , 2003; Raghav et al . , 2012 ) . RNAPII is recruited to active enhancers on a global scale ( Szutorisz et al . , 2005; Koch et al . , 2008; Kim et al . , 2010 ) , and we performed RNAPII ChIP-seq in primary hMSCs to annotate putative cis-acting regulatory elements during human adipocyte differentiation . Using a feature detection algorithm that robustly identifies enriched regions , we captured a progression of differentiated states when cells were cultured either at non-permissive , low density ( LD ) or permissive , high density ( HD ) in the presence or absence of DMI adipogenic cocktail ( Figure 1A ) . K-means clustering uncovered differential RNAPII occupancy that fell broadly into three categories: associated with LD hMSCs ( clusters 1–8 , uncommitted ) , preferentially associated with HD hMSCs ( clusters 9–12 , primed ) , and associated with DMI induction ( clusters 13–16 ) . Clusters 17–19 displayed an ambiguous relationship to adipocyte differentiation and were excluded from subsequent analyses . 10 . 7554/eLife . 06821 . 003Figure 1 . RNAPII-annotated enhancers reveal stage-specific transcriptional programs during adipogenic commitment . ( A ) Heat map of RNA Polymerase II ( RNAPII ) peaks clustered as a function of differential enrichment in response to changes in cell density ( low , LD vs high , HD ) and differentiation cocktail ( + or − DMI , 24 hr ) in human mesenchymal stem cells ( hMSCs ) ( tracks 1–4 ) . Density heat maps ( red ) of C/EBPβ ( 0 and 6 hr DMI ) and GR ( 6 hr DMI ) binding in hMSCs within a 4 kb window of differential RNAPII peaks ( tracks 5–7 ) . Color bar specifically refers to scaling for C/EBPβ tracks . ( B ) Gene ontologies ( GOs ) resulting from mapping RNAPII enhancers to genes with correlated changes in gene body RNAPII . ( C ) Identification of de novo motifs in differential RNAPII clusters 9–11; reported sequences met detection thresholds of at least 5% of targets and p-value ≤ 1e-14 . ( D ) Comparison of the AP-1 , hybrid and C/EBP motifs , with the AP-1- and C/EBP-half sites boxed . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 00310 . 7554/eLife . 06821 . 004Figure 1—figure supplement 1 . Characterization of RNAPII-annotated enhancers . ( A ) Identification of de novo motifs in differential RNAPII clusters; reported sequences met detection thresholds of at least 5% of targets and p-value ≤ 1e-14 . Note the striking enrichment in clusters 1 and 3 of E2F4 and NF-YA , transcription factors ( TFs ) with well-established roles in controlling cell cycle progression ( Lukas et al . , 1996; Benatti et al . , 2011; Lee et al . , 2011; Fleming et al . , 2013 ) , as well as the identification of C/EBP and GR motifs in the DMI-induced clusters . ( B ) Density heat maps of ENCODE marks at differential RNAPII clusters implying occupancy of motifs by their cognate factors such as AP-1 , E2F and NF-YA within the uncommitted clusters . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 00410 . 7554/eLife . 06821 . 005Figure 1—figure supplement 2 . C/EBPβ and GR ChIP-seq in hMSCs . ( A ) Summary of ChIP-Seq peak calls and de novo motif analyses ( HOMER ) for C/EBPβ and GR from hMSCs cultured at HD either in the absence or presence of DMI cocktail for 6 hr . ( B ) Bar graph of differential RNAPII peaks co-localized with C/EBPβ in the presence or absence of DMI , as a function of RNAPII cluster . Absolute number of RNAPII/CEBPβ co-occupied peaks in the absence of DMI treatment is indicated above each bar . * , indicates hypergeometric p-value ≤ 5 × 10−6 comparing cluster-specific C/EBPβ co-occupancy to the average co-occupancy across all clusters for the DMI-negative condition . Enrichment for C/EBPβ was also statistically significant in the presence of DMI for clusters 8 , 10 , and 13 . ( C ) Venn diagram showing overlap between the cistromes of GR and C/EBPβ in DMI-treated hMSCs . ( D ) Scatterplot comparing sequence tags from C/EBPβ ChIP-seq peaks identified in low ( LD ) and high ( HD ) density hMSCs treated with DMI for 6 hr . C/EBPβ-binding strength is highly correlated for the two conditions . ( E ) GR ChIP in DMI-treated hMSCs comparing LD to HD cells at 7 genomic sites and a non-specific control ( INS ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 005 The uncommitted , HD-primed , and DMI-induced RNAPII clusters mapped to distinct gene ontologies ( GOs ) ( Figure 1B , Supplementary file 1 ) . Cytoskeleton and cell proliferation genes were enriched in the uncommitted clusters , and genes associated with biological processes characteristic of terminally differentiated adipocytes , namely PPAR signaling , pyruvate and lipid metabolism , and adipocytokine signaling were only enriched when the cells were cultured at high-seeding density and treated with DMI cocktail . Of particular note , genes involved in amino acid transport and metabolism were associated with the HD-primed clusters , where high-seeding density promoted RNAPII recruitment prior to DMI induction . Intriguingly , the HD-primed , but not the uncommitted and DMI-induced , RNAPII sites associated with a CHOP/ATF4/hybrid motif ( Figure 1C , Figure 1—figure supplement 1 ) . It is a hybrid of AP-1 ( TGA ) and C/EBP ( TTGC ) half-sites ( Figure 1D ) , and binds heterodimers between different C/EBP proteins and AP-1 or ATF factors ( Vinson et al . , 1993; Shuman et al . , 1997; Wolfgang et al . , 1997; Cai et al . , 2008 ) . ChIP-seq for C/EBPβ , which is known to control early phases of adipogenesis in murine cells along with GR ( Steger et al . , 2010; Siersbæk et al . , 2011 ) , revealed extensive binding in HD hMSCs that selectively mapped to the HD-primed RNAPII sites ( Figure 1A , track 5 , Figure 1—figure supplement 2A , B ) . Treatment of HD cultures with DMI cocktail led to a massive up-regulation of C/EBPβ genomic occupancy ( Figure 1—figure supplement 2A ) . It also induced binding by GR that mapped to regions with DMI-induced RNAPII ( Figure 1A and Figure 1—figure supplement 2B ) and co-localized with C/EBPβ ( Figure 1—figure supplement 2C ) in a density-independent manner ( Figure 1—figure supplement 1D , E ) . These data indicate that C/EBPβ targets HD-primed enhancers before , and DMI-induced enhancers during , human adipogenesis . To test whether C/EBPβ drives the activity of HD-primed enhancers via a non-canonical binding site as predicted by the motif analysis of RNAPII sites , we screened C/EBPβ-binding regions at all RNAPII-annotated regions for the presence of a canonical CEBP sequence ( TTGCnnAA ) ( Jolma et al . , 2013 ) or a hybrid motif ( TTKCATCA ) ( Figure 2A ) . Whereas the hybrid motif is present at 27% of C/EBPβ peaks in the HD-primed clusters , only 7% and 4% of peaks have this sequence in the uncommitted and DMI-induced clusters , respectively , indicating its enrichment within the C/EBPβ-binding sites of HD-primed enhancers . Conversely , 23% of C/EBPβ peaks in the DMI-induced clusters are associated with the canonical C/EBP motif , compared to 13% and 8% of peaks in the uncommitted and HD-primed clusters , respectively . These results demonstrate that C/EBPβ preferentially targets the hybrid motif in the HD-primed enhancers , whereas it binds the canonical C/EBP motif in the DMI-induced enhancers . 10 . 7554/eLife . 06821 . 006Figure 2 . A non-canonical C/EBP motif mediates transcription at HD-primed enhancers . ( A ) Pie charts comparing the frequency of hybrid ( red ) vs canonical C/EBP ( blue ) motifs present at RNAPII-C/EBPβ co-bound sites at stage-specific enhancers . ( B ) C/EBPβ enhancers harboring hybrid ( green ) or canonical C/EBP ( red ) motifs were assayed by a luciferase reporter in the absence ( day 0 ) or presence ( day 1 ) of DMI cocktail in C3H10T1/2 cells . ( C ) Mutations disrupting the hybrid motif specifically ( cyan ) or C/EBP motif generally ( blue ) were assayed by a luciferase reporter in basal medium in C3H10T1/2 cells . ( D ) Gene expression was assayed for genes associated with C/EBPβ enhancers containing hybrid ( green ) or canonical C/EBP motifs ( red ) , and in a panel of endogenous control genes ( gray ) , in DMI-treated ( day 1 ) vs untreated ( day 0 ) hMSCs . * , denotes p < 0 . 05 , Student's t-test comparing stimulated vs unstimulated or mutant vs WT; † , denotes p < 0 . 002 , comparison of fold changes between stage-specific enhancers vs endogenous controls , Mann–Whitney test . Error bars depict SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 006 The HD-primed and DMI-induced enhancers are active during distinct stages of differentiation , and employ different motifs to recruit C/EBPβ , suggesting that distinct classes of C/EBPβ-recognition sequences confer temporal regulation of gene transcription during adipocyte development . To address this idea , we placed several RNAPII-annotated regions that are targeted by C/EBPβ and harbor either a hybrid or canonical motif into luciferase reporters , and assayed their activity in response to DMI cocktail ( Figure 2B ) . Interestingly , reporters carrying the hybrid motif had reduced in activity in the presence of DMI , while those with the canonical C/EBP sequence were increased by DMI . Moreover , the hybrid motif mediates transcriptional activity since targeted mutation of either the hybrid sequence specifically ( Hyb- ) or the C/EBP motif generally ( C/EBP- ) lowered luciferase activity ( Figure 2C ) . To further test the functional significance of C/EBP motifs in regulating gene transcription , we quantified mRNAs for genes with nearby RNAPII-annotated regions bound by C/EBPβ ( Figure 2D ) . Consistent with the changes in both RNAPII occupancy and the luciferase reporters , genes associated with the hybrid motif showed decreased expression upon DMI stimulation . In contrast , genes associated with the canonical C/EBP motif ( AKR1C1 , ALOX15b , FKBP5 , HSD11b1 , and ZBTB16 ) were activated by DMI treatment along with a control adipocyte marker ( PDK4 ) . As a whole , these findings suggest that C/EBPβ targets a non-canonical C/EBP motif to activate HD-primed enhancers . Furthermore , although the addition of DMI cocktail increases C/EBPβ occupancy globally , including sites associated with the HD-primed enhancers , it reduces the activity of the HD-primed enhancers while inducing enhancers harboring canonical C/EBP motifs . Hybrid motifs recruit C/EBP TFs as heterodimers with bZip proteins from the AP-1 or ATF families ( Vinson et al . , 1993; Shuman et al . , 1997; Wolfgang et al . , 1997; Cai et al . , 2008 ) . Therefore , we performed ChIP to measure the occupancy of several candidate partners for C/EBPβ , and found robust binding of ATF4 at hybrid motifs ( Figure 3A and Figure 3—figure supplement 1A ) . To map ATF4-binding sites on a genome-wide scale , we performed ChIP-seq in pre-adipocyte hMSCs ( HD , without DMI ) , and identified 1451 discrete ATF4 peaks . Remarkably , ATF4 binds almost exclusively with C/EBPβ , with more than 90% of its sites co-localized with C/EBPβ-binding sites ( Figure 3B ) . Moreover , motif analysis shows an extraordinary level of sequence specificity at ATF4 binding sites , with strong nucleotide selection at nearly every position of the core hybrid motif , and extremely high concordance between the presence of ATF4 and the hybrid motif , with 85% of peaks containing the recognition sequence ( Figure 3C ) . In marked contrast , the C/EBPβ cistrome yields a motif with a higher degree of sequence degeneracy and a lower detection frequency ( 65% ) at binding sites ( Figure 1—figure supplement 2A ) . 10 . 7554/eLife . 06821 . 007Figure 3 . ATF4 is a density-regulated factor that binds to HD-primed enhancers as a heterodimer with C/EBPβ . ( A ) ATF4 ChIP in high density ( HD ) hMSCs interrogating genomic loci that contain the hybrid motif vs non-specific control ( INS ) . ( B ) Density heat map anchored on ATF4 ChIP-seq peaks to examine co-distribution with C/EBPβ on a genome-wide scale in HD hMSCs . ( C ) De novo motif analysis of ATF4 ChIP-seq peaks . ( D ) Changes in ATF4 and C/EBPβ protein levels assessed by Western blotting during adipogenic differentiation . MW markers , 50 and 37 kDa . ( E ) Bar graphs of ATF4 ( top ) and C/EBPβ ( bottom ) occupancy at hybrid or canonical motifs in hMSCs cultured at low ( green bars ) or high ( blue bars ) seeding density in the absence of DMI or following 24 hr of treatment ( red bars ) . * , denotes p < 0 . 05 , Student's t-test comparison of HD samples with vs without DMI treatment; † , denotes p < 0 . 05 , Student's t-test comparison of LD vs HD samples without DMI treatment . ( F ) Sequential ChIP of C/EBPβ followed by ATF4 or IgG to interrogate simultaneous occupancy of both factors at hybrid motifs . * , denotes p < 0 . 05 for a one-tailed Student's t-test . Error bars depict SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 00710 . 7554/eLife . 06821 . 008Figure 3—figure supplement 1 . Characterization of ATF4-binding sites . ( A ) bZip TF ChIP in HD hMSCs interrogating the same genomic loci from Figure 3A . With the exception of insulin ( INS ) , each target carries a hybrid motif and is occupied by ATF4 . ( B ) Top-ranked GOs from mapping ATF4-binding sites to nearby genes . ( C ) Percent of ATF4 or C/EBPβ sites co-localized with or without RNAPII in HD hMSCs . Co-bound sites were further subdivided into those with or without changed RNAPII occupancy during adipogenesis . ( D ) Density heat map of ATF4 occupancy in HD hMSCs juxtaposed against the differential RNAPII clusters and C/EBPβ and GR datasets from Figure 1A . Heat map color bar refers C/EBPβ tracks . Maximum signal intensity set to 0 . 5 for GR and 0 . 25 for ATF4 to facilitate visualization . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 008 Genomic occupancy of ATF4 is associated predominantly with genes involved in tRNA aminoacylation for protein synthesis , response to endoplasmic reticulum ( ER ) stress , and the unfolded protein response ( Figure 3—figure supplement 1B ) . ATF4 occupies sites near all of the genes observed in the HD-primed clusters that function in amino acid transport and metabolism and tRNA aminoacylation ( Supplementary file 1 ) . Relative to C/EBPβ , ATF4 occupies sites that are highly enriched for RNAPII , with the majority of these falling in the HD-primed clusters ( Figure 3—figure supplement 1C , D ) . Thus , ATF4 is intimately tied to HD-primed RNAPII , implicating it in the activation of gene transcription in response to high-cell-density seeding of hMSCs . These data also imply that ATF4 abundance may be regulated by cell density . Indeed , western analysis revealed that while ATF4 levels in uncommitted LD cells were negligible , ATF4 was up-regulated in response to high-cell-seeding density ( Figure 3D ) . Furthermore , upon addition of DMI cocktail , ATF4 expression diminished modestly at 3 hr and prominently by 24 hr . As expected , C/EBPβ showed a rapid and sustained induction by DMI in hMSCs . To determine if ATF4 genomic occupancy was correlated with its abundance , we performed ChIP in cells cultured at LD or HD with or without DMI treatment . ATF4 occupancy at hybrid sites was strongly reduced in either untreated LD hMSCs or DMI-treated HD hMSCs ( Figure 3E ) . C/EBPβ , in contrast , bound to the hybrid sites irrespective of the cell seeding conditions . Concomitant with C/EBPβ up-regulation by differentiation cocktail , a trend of increased binding was apparent in the DMI condition . These observations indicate that ATF4 protein level is modulated by signals derived from both cell density and DMI cocktail , leading to functional interaction with C/EBPβ in primed hMSCs as well as early in adipogenesis . Essentially no ATF4 peaks were observed that lack co-bound C/EBPβ , suggesting heterodimer binding . To interrogate whether ATF4 and C/EBPβ can simultaneously occupy the same genomic loci , we performed sequential ChIP experiments ( Figure 3F ) . After an initial ChIP for C/EBPβ , ATF4 binding was highly enriched relative to the IgG control at all sites containing a hybrid motif , while no such enrichment occurred at control sites where C/EBPβ binds independently of ATF4 ( PPARγ ) or is absent ( INS ) . Together , the data demonstrate ATF4 is a density-dependent TF that co-localizes with HD-primed enhancers in hMSCs by targeting the hybrid motif as a heterodimer with C/EBPβ . To investigate whether ATF4 regulates adipocyte differentiation , we depleted its expression in primary hMSCs , and in the well-characterized mouse pre-adipocyte cell line 3T3-L1 ( Figure 4A ) . Transient transfection with ATF4 siRNAs decreased the level of Oil Red O staining relative to controls , indicating lower lipid accumulation in the ATF4 knockdown cells ( Figure 4B ) . In parallel , early markers of adipogenesis were decreased threefold–fourfold in ATF4 knockdown cells ( Figure 4C ) , including the mRNAs encoding the lineage-determining TFs PPARγ2 and C/EBPα , as well as several of their downstream transcriptional targets . This deficit in adipogenic gene expression became more pronounced over time ( Figure 4—figure supplement 1 ) , and together , these data suggest that ATF4 expression is important for adipogenesis . To better define the role of ATF4 , we measured genomic binding in ATF4 knockdown cells . As expected , ATF4 occupancy was significantly reduced at all sites ( Figure 4D ) . Notably , RNAPII occupancy was diminished at all but one of the ATF4-binding sites , and most of these showed a significant reduction , indicating that ATF4 functions to recruit RNAPII . C/EBPβ binding was decreased at all sites , but reached significance at a minority , which is not surprising given that C/EBPβ can occupy these sites under conditions with little or no ATF4 binding . These data establish ATF4 as a novel density-dependent TF that transduces signals from multiple pro-adipogenic stimuli including cell density and DMI cocktail to program the activity of HD-primed enhancers and promote adipocyte differentiation . 10 . 7554/eLife . 06821 . 009Figure 4 . ATF4 promotes adipocyte differentiation and density-dependent transcription . ( A ) ATF4 knockdown in hMSCs and 3T3-L1 pre-adipocytes assessed by western blotting . RAN , loading control . ( B ) hMSCs ( top ) and 3T3-L1 cells ( bottom ) were transfected with siATF4 RNA duplexes or control sequences , and assessed for lipid droplet formation following 1 week of adipogenic differentiation . Top images , Oil Red O staining at the level of individual cells ( hMSCs ) or for an entire well of a 6-well plate ( 3T3-L1s ) . Bottom panel , phase contrast images . ( C ) HD hMSCs were transfected with control ( PPIB ) or ATF4 siRNAs and assayed for mRNA expression of early adipogenic markers at days 0 ( HD ) and 3 ( HD+DMI ) of differentiation . Expression was normalized to the maximum level for the control siRNA . ( D ) Changes in ATF4 ( top ) , RNAPII ( middle ) and C/EBPβ ( bottom ) occupancy as a function of ATF4 knockdown assessed by ChIP in undifferentiated hMSCs cultured at high seeding density . * , denotes p < 0 . 05 . Error bars depict SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 00910 . 7554/eLife . 06821 . 010Figure 4—figure supplement 1 . Gene expression in ATF4 knockdown cells . Adipogenic gene expression assayed following DMI treatment for 5 or 7 days of hMSCs transfected with siRNAs targeting PPIB ( control ) or ATF4 . The average of two biological replicates is shown . Error bars denote the range of the data . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 010 Our data suggest a model for density-dependent gene expression whereby C/EBPβ-ATF4 heterodimers activate hybrid-motif-bearing enhancers in response to increased cell density by recruiting RNAPII . This stage-specific gene expression program is driven by robust sequence-specificity of the C/EBPβ-ATF4 heterodimer , such that it preferentially occupies hybrid sequences relative to the C/EBPβ homodimer , with ATF4's genomic function tied exclusively to C/EBPβ . Whether these findings are emergent properties of the intrinsic DNA-binding specificities of ATF4 , C/EBPβ and the C/EBPβ-ATF4 heterodimer or result from other mechanisms is unknown . To gain insight into this , we performed EMSA with DNA probes representing the different motifs ( Figure 5A ) , and recombinant ATF4 and C/EBPβ purified from Escherichia coli to ensure homogenous preparations that lacked other potential bZip binding partners typically present in mammalian cell extracts ( Figure 5B ) . ATF4 exhibited no detectable binding to either sequence by itself , while C/EBPβ bound both , but preferred the palindromic C/EBP motif ( Figure 5C , D ) . Addition of recombinant ATF4 to C/EBPβ increased binding to the ATF4 motif at the lower protein concentrations ( Figure 5C , lanes 4–9 ) , indicating that ATF4 and C/EBPβ readily form heterodimers in solution as shown previously for other bZip proteins ( Cao et al . , 1991 ) , and revealing that the heterodimer recognizes this sequence with higher affinity than the C/EBPβ homodimer . In contrast , ATF4 decreased C/EBPβ occupancy of the palindromic C/EBP motif ( Figure 5D , lanes 1–6 ) , suggesting that the C/EBPβ-ATF4 heterodimer has lower affinity relative to the C/EBPβ homodimer for this sequence . Similar results were obtained with a reverse titration scheme ( Figure 5—figure supplement 1 ) , and antibody controls demonstrated that ATF4 and/or C/EBPβ produce the protein-DNA complexes , further supporting these conclusions . 10 . 7554/eLife . 06821 . 011Figure 5 . Distinct DNA-binding activities for ATF4 , C/EBPβ and the C/EBPβ-ATF4 heterodimer . ( A ) EMSA probe design for the hybrid and C/EBP sequences . Motifs are capitalized . ( B ) Silver-stained SDS-PAGE of recombinant his-tagged ATF4 and C/EBPβ . Arrowheads indicate the expected MWs for the tagged constructs . ( C and D ) EMSA titration of ATF4 ( 10 , 30 and 90 μM ) and C/EBPβ ( 100 , 300 and 900 nM ) with 0 . 25 nM of either the hybrid ( C ) or C/EBP ( D ) radiolabeled probe . Bar graphs show phosphoimager-based quantification of bound complexes relative to free probe . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 01110 . 7554/eLife . 06821 . 012Figure 5—figure supplement 1 . DNA-binding activities for ATF4 , C/EBPβ and the C/EBPβ-ATF4 heterodimer . ( A ) EMSA titration of ATF4 ( 2 , 6 and 18 μM ) and C/EBPβ ( 50 , 150 and 450 nM ) with 0 . 25 nM of either the hybrid ( upper ) or C/EBP ( lower ) radiolabeled probe . Antibody controls show a supershift for the C/EBPβ homodimer , however are inhibitory for DNA binding by the C/EBPβ-ATF4 heterodimer . Antibodies were added to recombinant proteins prior to the addition of probe . Arrowheads indicate gel–shift complexes . ( B ) Bar graphs quantifying probe binding from two independent EMSA experiments; error bars depict the data range . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 012 To extend our findings beyond two synthetic DNA templates , we tested the DNA-binding specificities of C/EBPβ and ATF4 to native sequences comprising the entire human genome . Genomic DNA was purified from hMSCs , stripped of protein , and sonicated to a relatively homogenous population of 250 bp fragments . Akin to EMSA , recombinant TFs and isolated DNA were interrogated via solution phase binding . Samples were processed subsequently in a manner similar to ChIP-seq using immunoprecipitation to isolate non-crosslinked protein-DNA complexes followed by analysis of bound DNA by next generation sequencing . A similar in vitro cistromics assay ( IVC ) has successfully examined TF binding with Drosophila components ( Guertin et al . , 2012 ) . Here , we interrogated how TF concentration and dimerization state affect genomic occupancy in the absence of the confounding effects encountered in cells . Titration of C/EBPβ over two orders of magnitude produced cistromes with robust peaks ( Figure 6A ) that scaled with protein amount ( Figure 6B ) . The total number of bound sites exceeding a 1 RPM threshold plateaued near the low end of the C/EBPβ titration , and gained sites at the upper end of the titration had lower binding strength ( Figure 6C ) , indicating that the addition of more protein drives binding to low-affinity DNA sites . In support of this , de novo motif analyses revealed stronger enrichment of the C/EBP motif at the top-ranked sites ( Figure 6D ) . Yet , the C/EBP motif was the top-ranked sequence for both the strongest and weakest sites , indicating appropriate specificity for the C/EBPβ homodimer even at the higher protein levels . We anticipated that the number of C/EBPβ-binding sites observed in vitro would exceed that found in cells due to the removal of accessibility barriers imposed by chromatin . To test this , we examined the level of histone H3 lysine 27 acetylation ( H3K27ac ) , a marker of active transcription and open chromatin ( Rada-Iglesias et al . , 2011; Stasevich et al . , 2014 ) , in hMSCs at regions bound by C/EBPβ ( Figure 6—figure supplement 1A ) . Unlike sites occupied in hMSCs , the in-vitro-specific sites have little or no H3K27ac in cells , suggesting that they reside in inactive chromatin . Taken together , these results demonstrate that vitro cistromics interrogates protein-DNA specificity effectively . 10 . 7554/eLife . 06821 . 013Figure 6 . Heterodimer formation with C/EBPβ is necessary for ATF4 binding and is sufficient to alter C/EBPβ-sequence specificity . ( A ) Browser shot comparing C/EBPβ peaks from the in vitro cistromics assay ( IVC ) and hMSC ChIP-seq . 0 . 1 and 1 μl C/EBPβ molarity equivalents in the initial binding reaction are ∼60 and 600 nM , respectively . Tracks are RPM normalized , y-axes scaled from 0–3 ( in vitro ) and 0–5 ( hMSC ) . ( B ) In vitro ChIP-seq peaks total vs a 1 RPM threshold , plotted as a function of C/EBPβ titration ( 0 . 1 , 0 . 25 , 1 and 10 μl ) . Total C/EBPβ peaks found in hMSCs ( red line ) is included for point of comparison . ( C ) Box plot of peak strength in the 10 μl C/EBPβ cistrome comparing gained vs co-bound sites . The co-bound fraction met a 0 . 5 RPM threshold in all C/EBPβ homodimer cistromes . ( D ) Motif enrichment for the strongest- and weakest-1000 sites of the C/EBPβ cistromes . ( E ) Browser shots of C/EBPβ peaks upon titration of ATF ( 0 . 1 , 1 and 10 μM ) with C/EBPβ ( 60 nM ) . Tracks are RPM normalized , 0–3 RPM scale on the left panel , 0–20 RPM on the right . ( F ) K-means clustered density heat maps of C/EBPβ occupancy as a function of ATF4 titration . Titration of ATF4 ( 0 . 1 , 1 and 10 μM , C/EBPβ IP; 1 and 10 μM , ATF4 IP ) with C/EBPβ ( 60 nM ) is indicated . ATF4 homodimer binding was examined with 250 μM protein . Peaks thresholded to meet 1 RPM in at least two C/EBPβ cistromes . Three discrete clusters were identified and the de novo motif enrichment was based on all sites in each cluster . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 01310 . 7554/eLife . 06821 . 014Figure 6—figure supplement 1 . Characterization of binding associated with either in vitro-selective C/EBPβ sites or ATF4 homodimers . ( A ) Average profile plot of H3K27ac in hMSCs at sites bound by C/EBPβ in hMSCs or in vitro only ( 0 . 1 μl cistrome ) . ( B ) De novo motif enrichment at 111 peaks identified in the ATF4 homodimer cistrome . Top two motifs shown . ( C ) Blocks diagram showing multimerization of motifs at ATF4 homodimer peaks . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 014 To investigate the effects of heterodimerization on the global genomic distribution of C/EBPβ , we titrated ATF4 into the IVC . Upon addition of ATF4 , C/EBPβ exhibited reduced occupancy of C/EBPβ-homodimer sites and de novo binding at sites not represented in the homodimer cistromes generated with higher C/EBPβ concentrations ( Figure 6E and Figure 6F , tracks 1–4 ) . Induced C/EBPβ-binding sites displayed parallel enrichment for ATF4 occupancy ( tracks 5 and 6 ) . Motif analysis showed robust enrichment of the hybrid motif at these sites , reproducing the exquisite specificity mapped in the hMSC ATF4 cistrome . Remarkably , ATF4 by itself exhibited no detectable binding to these regions ( track 7 ) , demonstrating that its interaction with the genome is exclusively heterodimeric . In total , we identified 111 peaks in the ATF4 homodimer cistrome , and these were enriched for multimers of a c-JUN motif ( Figure 6—figure supplement 1B , C ) . These rare sites are not occupied by ATF4 in hMSCs , revealing that homodimeric binding by ATF4 may not be physiological . Together , the EMSA and in vitro cistromics experiments demonstrate that the DNA-binding properties of ATF4 and C/EBPβ are sufficient to reconstitute their sequence specificities observed in cells . Our data reveal that C/EBPβ controls distinct gene expression programs in hMSCs at different stages of adipogenesis ( Figure 7 ) . Prior to treatment with DMI cocktail , high seeding density induces ATF4 , which heterodimerizes with C/EBPβ , driving a unique set of genes that may prime hMSCs into a pre-adipocyte state . In contrast , in response to exogenous adipogenic signals , C/EBPβ activates a different set of genes , in part by co-localizing with GR . GR promotes adipocyte differentiation in mouse cells ( Steger et al . , 2010; Asada et al . , 2011 ) , and its association with induced genes during early hMSC adipogenesis suggests that it functions similarly in human cells . 10 . 7554/eLife . 06821 . 015Figure 7 . C/EBPβ controls distinct gene expression programs in hMSCs . Cell density regulates ATF4 expression to control C/EBPβ-ATF4 heterodimer formation . Prior to DMI treatment , heterodimer binding to hybrid motifs promotes a gene expression program priming hMSCs for adipogenesis . In this basal state , C/EBPβ is a poor transcriptional activator . In response to DMI , ATF4 level decreases , C/EBPβ level and GR activity increase . C/EBPβ initiates a program of transcription at new genes to orchestrate adipocyte differentiation with cooperating factors such as GR . DOI: http://dx . doi . org/10 . 7554/eLife . 06821 . 015 This study identifies ATF4 as a cell density sensor that links the physical and chemical requirements for adipogenesis . High cell density has long been known to be an important condition for adipocyte differentiation in vitro ( Green and Kehinde , 1975; Pittenger et al . , 1999; McBeath et al . , 2004; Cristancho et al . , 2011 ) . It is likely that a combination of cues is required in tissue microenvironments supporting adipogenesis in vivo . ATF4 joins a small , yet growing list of TFs that confer adipogenic competency . A major block in identifying TFs controlling the formation of pre-adipocyte cells is the lack of molecular markers distinguishing naïve hMSCs from committed pre-adipocytes . Nevertheless , several candidate TFs have been identified in murine cell culture models . ZFP423 is expressed almost exclusively in adipogenic fibroblasts , and is necessary for adipocyte differentiation in vitro and for the development of adipose tissue in mice ( Gupta et al . , 2010 ) . TCF7L1 promotes commitment of 3T3-L1 fibroblasts to a pre-adipocyte state in response to cell confluency ( Cristancho et al . , 2011 ) . KAISO , in contrast , represses pre-adipocyte differentiation by recruiting the co-repressor SMRT to promoters ( Raghav et al . , 2012 ) . Each of these studies marks an important advance in understanding pre-adipocyte biology , yet it remains unclear as to how , or if , these pre-adipocyte TFs functionally interact with each other or with the TFs responding to adipogenic signals . Here , the identification of the C/EBPβ-ATF4 heterodimer points to a mechanism that connects adipogenic competency and early differentiation through the sharing of C/EBPβ . Mice lacking ATF4 are lean and resist diet-induced obesity , although they have adipose tissue ( Masuoka and Townes , 2002; Yang et al . , 2004; Seo et al . , 2009 ) . Our demonstration of a role for ATF4 in adult stem cell adipocyte differentiation suggest that this regulatory mechanism could be especially relevant in adult adipose tissue , consistent with the recent demonstration that adipogenesis occurs via distinct cell lineages during development vs adulthood ( Jiang et al . , 2014 ) . Depletion of ATF4 in 3T3-L1 cells blocks adipocyte differentiation ( Yu et al . , 2014 ) . ATF4 is also required for the development of blood and bone cell lineages ( Masuoka and Townes , 2002; Yang et al . , 2004; Seo et al . , 2009 ) . It is of interest to note that ATF4 plays a central role in handling ER stress induced by amino acid imbalance ( Harding et al . , 2003; Kilberg et al . , 2009 ) , and accumulating evidence indicates that ER stress pathways couple obesity to other metabolic dysfunctions such as diabetes ( Ozcan et al . , 2004; Scheuner and Kaufman , 2008 ) , suggesting that ATF4 may regulate lipid and carbohydrate metabolism . Our observation that ATF4 regulates similar biological processes in hMSCs primed for differentiation may suggest an anticipatory response to meet an increased demand for protein synthesis during adipogenesis . The binding profiles of ATF4 and C/EBPβ in hMSCs can be explained by the inherent sequence specificities of the homo- and heterodimeric complexes . First , genomic binding of ATF4 requires heterodimerization with C/EBPβ . While elegant biochemical studies have demonstrated the propensity of ATF4 and C/EBPβ to heterodimerize ( Vinson et al . , 1993 ) , early studies also reported DNA-binding activity for ATF4 homodimers at palindromic cyclic AMP response elements ( CREs ) ( Vinson et al . , 1993; Podust et al . , 2001 ) . In this light , our finding of a paucity of ATF4 homodimeric sites is unexpected , and indicates that while such interactions may be detectable when the relative concentration of the motif is high , the affinity of ATF4 homodimers for the CRE sequence is insufficient to produce binding in cells , or in vitro when a large amount of competitor DNA is present . In support of these coclusions , PBM-based assays failed to identify a robust motif for ATF4 homodimers ( Weirauch et al . , 2014 ) , whereas HT-SELEX assays showed robust sequence preference of ATF4 for hybrid motifs ( Jolma et al . , 2013 ) . The IVC demonstrates that C/EBPβ and ATF4 target distinct DNA half sites , TTRS and TGAT , respectively . The relative plasticity of C/EBPβ homodimer in accommodating multiple DNA sequence variants is borne out in vivo , as the top motif associated with the cellular C/EBPβ cistrome shows considerable degeneracy at the −1 and −2 positions . In contrast , the C/EBPβ-ATF4 heterodimer is restricted to hybrid motifs of two inverted half-sites , TTRSATCA . Thus the biochemical properties of the heterodimer are sufficient to explain the nearly 1:1 correspondence between ATF4 binding and the presence of hybrid motifs in vivo . Interestingly , ATF4 heterodimerizes with the C/EBP related protein , CHOP , in the context of ER stress to bind hybrid motifs ( Han et al . , 2013 ) , suggesting that while heterodimerization is required for DNA binding , the specific C/EBP partner can vary in different biological contexts . C/EBPβ homodimers bind to hybrid sequences with low affinity relative to the C/EBPβ-ATF4 heterodimer , and this interaction requires elevated protein concentrations . DMI-treated hMSCs show persistent occupancy of the hybrid sites by C/EBPβ despite low ATF4 levels . An interesting possibility is that persistent binding is attributable to the elevated C/EBPβ concentrations in differentiating cells , however the effects of DMI-induced post-translational modification of C/EBPβ on homodimer binding cannot be excluded . Notably , robust RNAPII recruitment at hybrid sites was observed in cells only in the presence of ATF4 . Thus , hetero vs homodimeric binding may have important consequences for licensing transcription activity , especially under conditions where the basal activity of C/EBPβ is limiting , possibly due to associated repressor complexes ( Wiper-Bergeron et al . , 2003; Raghav et al . , 2012 ) . Our findings may have broad impact on the understanding of bZip TFs that readily form different dimer pairs ( Vinson et al . , 2002; Newman and Keating , 2003; Reinke et al . , 2013 ) , as well as TF families using other structural domains for the formation of homo- and heterodimeric protein complexes . By analogy with the C/EBPβ-ATF4 heterodimer , other C/EBPβ-bZip heterodimers may selectively recognize non-degenerate DNA sequences , which could illuminate fundamental rules governing their operation on a genome scale , and help to explain discrepancies between the palindromic C/EBP motif identified in vitro ( Jolma et al . , 2013; Weirauch et al . , 2014 ) and the ChIP-seq motif with substantial variation at the central positions . Consistent with this paradigm , heterodimers displaying greatest affinity for TGACGCAA , which differs from the C/EBPβ-ATF4 motif by single T/C substitution at the fourth position , were observed through forced expression of C/EBPα with either c-JUN or c-FOS ( Hong et al . , 2011 ) and inferred from DNase I footprinting in 3T3-L1 cells ( Siersbæk et al . , 2014a ) . We propose that the degenerate C/EBP motif observed in cells results from the binding of different homo- and heterodimeric protein complexes with distinct sequence specificities . Because C/EBP proteins are widely expressed with additional bZip family members , determining how they target the genome is critical to understanding their control of basic processes including cell growth and development .	Human body fat consists mostly of fat-storing cells called adipocytes . These cells develop in two main steps . First , mesenchymal stem cells—which can potentially become one of many different types of cell—commit to becoming pre-adipocyte cells . These pre-adipocytes then develop into mature adipocytes . Proteins called transcription factors control both steps of this process by binding to particular sites in the DNA of the cell and activating certain genes that control the cell's identity and activity . Various different transcription factors are known to stimulate the development of mesenchymal stem cells into adipocytes . Experiments performed on cells that have been grown in the laboratory suggest that the cells must also be tightly packed together to become adipocytes . Cohen et al . have now investigated the role a protein called C/EBPβ plays in the development of adipocytes , and have found that it plays different roles at different stages of development . When mesenchymal stem cells become tightly packed , more of another protein called ATF4 is produced . This protein binds to C/EBPβ , and the resulting two-protein complex then binds to sites on the DNA of the mesenchymal stem cell to activate genes that turn the stem cell into a pre-adipocyte . Reducing the amount of ATF4 in mesenchymal stem cells reduces the number of pre-adipocytes that develop . When not bound to ATF4 , and in response to certain cell signals , C/EBPβ binds to different DNA sites and helps pre-adipocytes develop into mature adipocytes . Whether transcription factors other than ATF4 partner with C/EBPβ to change DNA-binding-site preferences remains unknown . Future studies will search for these , with the aim of providing a clearer molecular understanding for how C/EBPβ acts as a ‘bridge’ that links together the two stages of adipocyte development .	[ "Abstract", "Introduction", "Results", "Discussion" ]	[ "chromosomes", "and", "gene", "expression", "developmental", "biology" ]	2015	ATF4 licenses C/EBPβ activity in human mesenchymal stem cells primed for adipogenesis
Anchoring proteins sequester kinases with their substrates to locally disseminate intracellular signals and avert indiscriminate transmission of these responses throughout the cell . Mechanistic understanding of this process is hampered by limited structural information on these macromolecular complexes . A-kinase anchoring proteins ( AKAPs ) spatially constrain phosphorylation by cAMP-dependent protein kinases ( PKA ) . Electron microscopy and three-dimensional reconstructions of type-II PKA-AKAP18γ complexes reveal hetero-pentameric assemblies that adopt a range of flexible tripartite configurations . Intrinsically disordered regions within each PKA regulatory subunit impart the molecular plasticity that affords an ∼16 nanometer radius of motion to the associated catalytic subunits . Manipulating flexibility within the PKA holoenzyme augmented basal and cAMP responsive phosphorylation of AKAP-associated substrates . Cell-based analyses suggest that the catalytic subunit remains within type-II PKA-AKAP18γ complexes upon cAMP elevation . We propose that the dynamic movement of kinase sub-structures , in concert with the static AKAP-regulatory subunit interface , generates a solid-state signaling microenvironment for substrate phosphorylation . Intrinsically disordered regions of proteins are widespread in nature , yet the mechanistic roles they play in biology are underappreciated . Such disordered segments can act simply to link functionally coupled structural domains or they can orchestrate enzymatic reactions through a variety of allosteric mechanisms ( Dyson and Wright , 2005 ) . The regulatory subunits of protein kinase A provide an example of this important phenomenon where functionally defined and structurally conserved domains are connected by intrinsically disordered regions of defined length with limited sequence identity ( Scott et al . , 1987 ) . In this study , we show that this seemingly paradoxical amalgam of order and disorder permits fine-tuning of local protein phosphorylation events . Phosphorylation of proteins is a universal means of intracellular communication that is tightly controlled within the spatial context of the cell . A variety of stimuli trigger these events , which are catalyzed by numerous protein kinases and reversed by phosphoprotein phosphatases ( Hunter , 1995 ) . A classic example is production of the second messenger cyclic AMP ( cAMP ) , which stimulates a cAMP-dependent protein kinase ( PKA ) to phosphorylate a range of cellular targets ( Taylor et al . , 2012 ) . The PKA holoenzyme is a tetramer composed of two regulatory subunits ( R ) and two autoinhibited catalytic subunits ( PKAc ) . Binding of cAMP to each R subunit is believed to liberate active kinase and phosphorylation ensues . The local action of PKA is dictated by A-kinase anchoring proteins ( AKAPs ) that impose spatial constraint by tethering this kinase in proximity to substrates ( Wong and Scott , 2004 ) . AKAPs also organize higher-order macromolecular signaling complexes through their association with G-protein coupled receptors , GTPases and additional protein kinases . Likewise , AKAP-associated phosphatases and phosphodiesterases act to locally terminate these signals . While physiological roles for AKAPs that sequester enzymatic activity with ion channels , cytoskeletal components and regulatory enzymes have been well established , the structural mechanisms involved in these protein–protein interactions have been difficult to characterize . Currently , structural details on PKA anchoring are limited because most AKAPs are large , intrinsically disordered macromolecules that lack recognizable structural domains . An exception is the crystal structure of the central domain of AKAP18γ that bears homology to bacterial 2H phosphoesterase domains ( Gold et al . , 2008 ) . Likewise , high-resolution crystallographic structures of the catalytic subunit ( PKAc ) when free and in complex with the C-terminal autoinhibitory and cAMP binding domains of the type I or type II regulatory subunits of PKA ( RI and RII ) have provided details on the mechanisms of catalysis and autoinhibition ( Knighton et al . , 1991 , 1992; Gold et al . , 2006 , 2008; Wu et al . , 2007 ) . Yet , despite decades of effort , a complete structural picture of the PKA holoenzyme is lacking . This is presumably due to the presence of long flexible intrinsically disordered linker regions within the R subunit that tether this complex . NMR spectroscopy and X-ray crystallographic studies show that the N-terminal domains of RI and RII homodimerize through a four-helix bundle docking and dimerization interface ( D/D ) ( Newlon et al . , 2001; Gold et al . , 2006; Kinderman et al . , 2006; Sarma et al . , 2010 ) . The D/D creates a high-affinity binding groove for a canonical amphipathic helix on each AKAP that forms the reciprocal binding surface ( Gold et al . , 2006 ) . Although the physiological consequences of anchored PKA phosphorylation events have been established in a variety of cellular contexts , we have yet to discern how the individual protein components are assembled and operate within AKAP complexes ( Scott and Pawson , 2009 ) . Even more elusive is a mechanistic role for intrinsically disordered domains within these macromolecular assemblies . For example , does internal flexibility within these signaling assemblies modulate other aspects of enzyme action in addition to sequestering PKA at specific regions of the cell ? Here we have used electron microscopy ( EM ) to evaluate the topological arrangement of the fully assembled type IIα PKA holoenzyme when anchored to AKAP18γ . This analysis unveils a remarkable level of conformational plasticity that resides within the AKAP–PKA complex . In vitro and cell-based structure-function approaches reveal an unexpected functional requirement for intrinsically disordered regions within RIIα . These regions not only define a radius of action of the anchored catalytic subunit but also modulate the enzymatic efficiency of the kinase . Insights from these studies may be broadly applicable to the understanding of other intrinsically disordered proteins and anchoring complexes that organize enzymatic action in the cell . Macromolecular assemblies were formed when purified human γ isoform of AKAP18 ( AKAP18γ ) was incubated with the PKA holoenzyme ( PKAholo ) , formed by the RIIα and PKAc subunits ( Figure 1A , ‘Materials and methods’ ) . Following separation by size-exclusion chromatography , fractions from the front of the elution peak ( indicated by arrow in Figure 1B ) were analyzed by SDS-PAGE ( Figure 1C , left ) . The subunit composition was confirmed by immunoblotting ( Figure 1C , mid ) . Native electrophoresis established that a majority of this material migrated as a single species with an apparent molecular weight in excess of 240 kD ( Figure 1C , right ) . This molecular mass is consistent with a hetero-pentameric complex composed of a single AKAP18γ molecule anchored to an RIIα subunit dimer and two PKAc subunits . 10 . 7554/eLife . 01319 . 003Figure 1 . Purification and electron microscopy of the AKAP18γ–PKAholo complex . ( A ) SDS-PAGE and Coomassie staining of purified individual complex components . ( B ) Size-exclusion chromatography ( SEC ) trace for purification of the assembled AKAP18γ–PKAholo complex . Fractions at the leading edge of the peak ( indicated by gray bar ) were chosen for further analysis . ( C ) SDS-PAGE ( left ) , western blot ( middle ) and native gel electrophoresis ( right ) obtained from the SEC peak elution fraction ( arrow in B ) . ( D ) Electron micrograph of the negatively stained AKAP18γ–PKAholo complexes ( circles ) . Triangles indicate the three major densities of the AKAP18γ–PKAholo complex . Inset , shows labeling with a gold nanoparticle ( arrow ) conjugated to an AKAP18γ-streptavidin moiety ( arrow ) . ( E ) Left , enlarged images of individual AKAP18γ–PKAholo complexes ( denoted by asterisks in D ) . ( E ) Right , highlighted outline ( yellow ) of particle shapes . ( F ) Projection averages of the AKAP18γ–PKAholo complex classified into distinct conformations using ISAC ( Yang et al . , 2012 ) . Unlabeled scale bars represent 25 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 003 The structure of the AKAP18γ–PKAholo complex was resolved by electron microscopy of negatively stained particles since these complexes are too small to be imaged in vitrified ice ( Figure 1D ) . Single particle analysis revealed clusters of three densities , each approximately 60–100 Å in size resembling beads on a string ( Figure 1D , E ) . Closer inspection of individual particles established that these tripartite structures adopted a range of conformations ( Figure 1E ) . In solution , these complexes may adopt many more configurations and flattening on the carbon support of the EM grid likely captured only a subset of the possible topologies . Approximately 7 , 000 particles were selected from electron micrographs for structural analysis . Projection averages were derived by classifying particles of similar orientation and structural conformation using reference-free multivariate statistical analysis ( van Heel et al . , 1996 ) , and iterative stable alignment and clustering procedures ( Yang et al . , 2012 ) ( Figure 1F , ‘Materials and methods’ ) . Class averages revealed a remarkable variety of configurations ( Figure 1F; Video 1 ) . These ranged from a tightly packed pseudo-symmetric triangular configuration ( Figure 1F , left panel ) to a fully extended linear configuration of the three densities ( Figure 1F , right panel ) . In either case , the central density was consistently smaller ( ∼60 Å ) than the two peripheral densities ( 85 × 100 Å ) . Similar size differences between the central and peripheral lobes were observed at the single particle level in raw micrographs ( Figure 1D , E ) . Affinity-labeling of strep-tagged AKAP18γ with a 5 nm gold particle targeted the smaller central density , indicating that this element includes the anchoring protein ( Figure 1D , inset ) . 10 . 7554/eLife . 01319 . 004Video 1 . Conformational dynamics of the wild-type AKAP18γ–PKA holoenzyme complex . A montage of projection averages obtained for the wild-type AKAP18γ–PKA holoenzyme complex displays the variety of topological configurations sampled by the dynamic signaling particle . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 004 Three-dimensional ( 3D ) reconstructions for the triangular and linear configurations of the AKAP18γ–PKAholo complex were determined at 35 Å resolution from a tilted-series dataset ( Figure 2A , Figure 2—figure supplements 1 and 2 ) . Negative staining in EM can cause flattening of particles , which might lead to apparent structural distortions . However , inspection of particles at various tilt angles showed that particles of linear and triangular conformations remained clearly distinguishable , even at high tilt angles ( Figure 2—figure supplement 1 ) . In both 3D reconstructions , a central mass with dimensions of 60 × 60 × 80 Å ( corresponding to the site of AKAP18γ as demonstrated by affinity gold labeling , Figure 2A , black triangle and Figure 1D , inset ) was flanked on either side by larger densities of 100 × 100 × 85 Å ( Figure 2A ) . These flanking densities can each accommodate a sub-complex of RIIα and PKAc . In the triangular conformation , the two peripheral densities are oriented at a 100° angle with respect to the central density and exhibit an end-to-end length of ∼300 Å ( Figure 2A , top ) . The end-to-end length increases to ∼385 Å in the extended linear configuration ( Figure 2A , bottom ) . Back-projections calculated from the final 3D maps compare well with the experimental class averages ( Figure 2—figure supplement 2 ) . Moreover , this back-projection analysis demonstrated that when the triangular model is tilted completely on its edge ( where it may appear more linear in projection ) its dimensions ( max length = 300 Å ) are significantly smaller than the maximum end-to-end length obtained for the linear reconstruction ( max length = 385 Å ) . Hence , we conclude that the linear and triangular conformations are structurally distinct . 10 . 7554/eLife . 01319 . 005Figure 2 . 3D reconstructions and pseudo-atomic structure of the AKAP18γ–PKAholo complex . ( A ) Three-dimensional ( 3D ) EM reconstructions and 90° rotated views of the fitted molecular models for the AKAP18γ–PKAholo complex . High-resolution structures for regions of AKAP18γ ( gold ) , RIIα ( green ) and C subunit of PKA ( blue ) that fit within the EM densities are indicated . Models are presented in the ( top ) compact triangular and ( bottom ) extended linear conformation . ( B ) Pseudo-atomic model of the AKAP18γ–PKAholo complex . ( C ) ( top ) Projection averages of the AKAP18γ–PKAholo complex with structural domains fitted into the EM densities and connected by lines representing the RIIα flexible linker regions . ( bottom ) Projection averages were removed for clarity . Scale bar represents 25 nm . ( D ) Statistical analysis of individual particle lengths in angstroms ( Å ) displays a Gaussian distribution ( green line ) with a mean value of 275 Å and a standard deviation ( σ ) ±65 Å . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 00510 . 7554/eLife . 01319 . 006Figure 2—figure supplement 1 . Tilted-series electron microscopy data . ( A ) Example electron micrographs collected using a series of image tilts ( α = 0° , 20° , 40° , and 50° ) . Colored circles indicate positions of the numbered particles in the tilt series . ( B and C ) Left , zoom-views of circled particles in ( A ) . At all tilt angles particles are readily classified as either linear ( B , yellow ) or as non-linear ‘triangular’ ( C , blue ) conformations . Right , particle boundaries are outlined for clarity . The tilted-series data was used to provide unique particle ‘views’ used for angular reconstitution methods applied in IMAGIC ( van Heel et al . , 1996 ) for initial 3D reconstruction of the AKAP18–PKAholo models . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 00610 . 7554/eLife . 01319 . 007Figure 2—figure supplement 2 . Three-dimensional EM maps of the AKAP18γ–PKAholo complex . Rotated views of the 3D reconstructions for the AKAP18γ–PKAholo complex determined to 35 Å resolution . ( A ) In the triangular conformation , the three spherical densities are arranged in a pseudo-symmetric fashion , with the two peripheral densities extending 300 Å apart and at an angle of 100° with respect to the central density . The smaller central density is 60 × 60 × 80 Å and the two larger peripheral densities are 100 × 100 × 85 Å . ( B ) In the linear conformation , the three spherical densities are arranged in a linear fashion , with the two peripheral domains extending 385 Å apart . ( C and D ) Back-projections ( right ) of the calculated 3D maps are compared to class averages ( left ) used for generating the initial reconstructions in IMAGIC ( van Heel et al . , 1996 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 007 Pseudo-atomic models of the pentameric protein assembly in both configurations were constructed by fitting Protein Data Bank coordinates for regions of AKAP18γ and PKAholo subunits ( Knighton et al . , 1991 , 1992; Gold et al . , 2006 , 2008; Wu et al . , 2007 ) ( Figure 2A , ‘Materials and methods’ ) . A model for AKAP18γ ( residues 88–317 ) was derived by connecting the central domain ( residues 88–290 ) ( Gold et al . , 2008 ) via a short linker to the PKA anchoring helix ( AKAP ( helix ) , residues 301–317 ) ( Gold et al . , 2006 ) ( Figure 2A , yellow , ‘Materials and methods’ ) . When AKAP18γ is docked to residues 1–43 of RIIα ( RII ( D/D ) ) , the resulting sub-structure fits within the central density of maps for both the triangular and extended configurations ( Figure 2A , yellow ) . This tallies with the single particle affinity-labeling studies that map AKAP18γ to this region ( Figure 1D , inset ) . In a similar manner , the peripheral densities each accommodate a PKA sub-structure consisting of one catalytic subunit ( Figure 2A , blue ) in complex with the cAMP-binding domain of RIIα , residues 91–392 ( Figure 2A , green ) ( Knighton et al . , 1991; Wu et al . , 2007 ) . Finally these models were completed by connecting the central AKAP18γ–RII ( D/D ) sub-complex to each RIIα-PKAc unit through a flexible chain corresponding to residues 44–90 of RIIα ( Figure 2A , ‘Materials and methods’ ) . The complete pseudo-atomic model of AKAP18γ–PKAholo complex is presented in Figure 2B . Our model of the anchored PKA complex implies that an intrinsically disordered flexible linker region within RIIα supports the array of conformations that were observed in the raw micrographs and the projection averages ( Figures 1 and 2C ) . To quantitatively assess the distribution of conformations assumed by this complex , we measured the end-to-end distance between the two large peripheral densities of 223 individual particles ( Figure 2D ) . This population of structures followed a Gaussian distribution ( Figure 2D , green trace ) with a mean particle length of 275 ± 65 Å ( n = 223 ) . We propose that conformational plasticity observed in these analyses is facilitated by this intrinsically disordered region between residues 44 and 90 of RIIα , a linker that connects the AKAP docking site ( D/D domain ) to the cAMP-responsive transduction domains . This notion is further substantiated by a primary sequence analysis of RIIα orthologs , showing that the linker regions are of similar length but exhibit low amino acid identity ( Figure 3A ) . 10 . 7554/eLife . 01319 . 008Figure 3 . Flexibility within RIIα constrains the configuration of the anchored kinase assembly . ( A ) Amino acid sequence alignment of the linker region in RIIα that connects the conserved N-terminal D/D domain to the C-terminal autoinhibitor and cAMP binding domains . This region shows low sequence homology and is likely to be structurally disordered . ( B ) Schematic representations of modified AKAP18γ–PKAholo complexes with AKAP18γ depicted in yellow , RIIα in green and PKAc in blue . ( C ) Biochemical analysis of the purified AKAP18γ–RIIα-PKAc complexes assembled with the wild-type RIIα subunit , the RIIα Δ44–86 mutant where the linker region was deleted , and the RIIα ZeChimera mutant where the mouse RIIα linker region was replaced with the corresponding and extended sequence from zebrafish . Top panel shows SDS-PAGE and Coomassie blue staining of the protein components . The next three panels show western blotting for RIIα , PKAc subunit and AKAP18γ , respectively . ( D ) Electron micrograph of negatively stained AKAP18γ–PKAholo complexes ( circles ) assembled with the RIIα Δ44–86 construct . ( E ) Electron micrograph of negatively stained AKAP18γ–PKAholo complexes ( circles ) assembled with the RIIα ZeChimera construct . Insets in ( D and E ) show enlarged views of individual particles outlined in gold for clarity . ( F ) Projection averages of the AKAP18γ–PKAholo complexes assembled with a truncated RIIα Δ44–86 construct using ISAC ( Yang et al . , 2012 ) . ( G ) Projection averages of AKAP18γ–PKAholo complex assembled an RIIα ZeChimera construct . Scale bars in ( F ) and ( G ) represent 25 nm . ( H ) Statistical analysis of particle radius in angstroms ( Å ) for each AKAP18γ–PKAholo complexes . Box plot displays second and third quartile values , tails corresponding to minimum and maximum distances , ( ) indicates p<0 . 01; ( ** ) indicates p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 008 We reasoned that if the linker region in RIIα contributes to conformational flexibility of the holoenzyme , altering its length could affect the structure and function of the anchored kinase . This structural postulate was tested by producing modified AKAP18γ–PKAholo complexes in which the linker region of RIIα was deleted ( RIIα Δ44–86 ) or replaced with an extended sequence of 60 residues found in zebrafish ( RIIα ZeChimera; Figure 3A , B ) . Formation of modified AKAP–PKA complexes ( assembled as described above ) was monitored by Coomassie blue staining and immunoblot detection of the component proteins ( Figure 3C ) . The conformations of these modified AKAP18γ–PKAholo complexes were analyzed by electron microscopy as previously described ( Figure 3D–G ) . Single-particle EM and class averages of the assemblies formed with RIIα Δ44–86 yielded uniform complexes exclusively in a compact triangular configuration ( Figure 3D , F ) . In contrast , complexes formed with RIIα ZeChimera resembled the array of conformations observed for the wild-type AKAP18γ–PKAholo assembly ( Figure 3E , G ) . Quantitative analysis of the differing linker lengths was assessed by measuring the radius of each particle , defined by the center of the AKAP18γ subunit to the distal end of each PKA subunit . Particle radii for the RIIα Δ44–86 complexes ranged in length from 55 to 125 Å , ( mean value of 87 ± 13 Å , n = 142; Figure 3H ) whereas the assemblies formed with the RIIα ZeChimera were extended with lengths ranging from 100 to 265 Å ( mean value of 168 ± 33 Å , n = 296; Figure 3H ) . These latter measurements are similar to the parameters of the native complex , ( 160 ± 29 Å , n = 216; Figures 2D and 3D ) . Thus we conclude that a flexible linker in RIIα is responsible for the conformational plasticity of the AKAP18γ–PKAholo assemblies . One mechanistic ramification of our structural analyses is that flexibility within PKAholo complex could permit precise orientation of the anchored catalytic subunit toward substrates . This would be particularly true for substrates that are physically associated with AKAPs . For example , the type 4 phosphodiesterase isoforms PDE4D3 and PDE4D5 associate with the central domain of AKAP18γ and are phosphorylated on two sites by PKA ( Sette and Conti , 1996; Carlisle Michel et al . , 2004; Stefan et al . , 2007 ) . We confirmed this protein–protein interaction upon co-expression of the components in HEK293 cells ( Figure 4A ) . PDE4Ds co-precipitate with AKAP18γ and the RII subunits of PKA ( Figure 4A , lane 1 ) but not with GFP controls ( Figure 4A , lane 2 ) . Accordingly , cAMP phosphodiesterase activity was enriched 3 . 27 ± 0 . 48-fold ( n = 5 ) in AKAP18γ–GFP immune complexes as compared to GFP controls or samples treated with the PDE4 selective inhibitor rolipram ( Figure 4B ) . These data allowed us to move to an in vitro system to study phosphorylation of PDE4D by AKAP18–PKA complexes . 10 . 7554/eLife . 01319 . 009Figure 4 . RIIα linker length influences basal PKA phosphorylation of associated substrates . ( A ) PDE4D isoforms associate with AKAP18 in cells . HEK293 cells expressing AKAP18–GFP or GFP alone along with PDE4D and RIIα were subjected to immunoprecipitation with anti-GFP antibodies . Immunocomplexes were separated by SDS-PAGE and immunoblotted for PDE4D and RIIα . AKAP18 was detected by RII overlay analysis . ( B ) Similar immunocomplexes as in ( A ) were used for phosphodiesterase activity assays . Inclusion of the small molecule rolipram ( 10 μM , bar 3 ) inhibited associated PDE4D activity . ( C ) Schematic of AKAP18–PKA holoenzyme-PDE4D3/5 complexes used in subsequent in vitro substrate phosphorylation assays . Based on our structural data , the PKA catalytic subunit is expected to be positioned within ∼100 Å of its substrate PDE4D . ( D ) Anchoring of PKA and PDE4D stimulates cAMP-independent phosphorylation of the phosphodiesterase . ( Top panel ) Basal 32P incorporation into PDE4D ( detected by autoradiograph ) is shown in the absence or presence of AKAP18γ . ( Bottom panels ) Levels of PDE4D , RIIα , PKAc and AKAP18γ were assessed by immunoblot . ( E ) Densitometeric quantification of phospho-PDE4D in panel ( D ) , n = 4 ( p<0 . 05 ) . ( F–G ) Deletion of the flexible linker augments cAMP-independent phosphorylation of PDE4D by 1 . 97 ± 0 . 18-fold ( p<0 . 05 ) . ( F ) ( Top panel ) Autoradiograph showing incorporation of 32P into PDE4D in each complex . ( Bottom panel ) Coomassie blue staining of the SDS-PAGE gel showing components of the assay . The PDE4D is at the top , while the different complexes are shown at the bottom . ( G ) Densitometric quantification of phospho-PDE4D levels in ( F ) , n = 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 00910 . 7554/eLife . 01319 . 010Figure 4—figure supplement 1 . Phosphorylation of PDE4D is time-dependent . AKAP18 complexes formed with either RIIα or each variant were incubated with PDE4D3 and γ-32P-ATP for the times indicated . ( Top panel ) Autoradiograph showing incorporation of 32P into PDE4D in each complex . ( Middle and bottom panels ) India ink staining of the membrane , showing components of the assay . Total PDE4D3 is in the middle , and RII , PKAc and AKAP18γ are shown in the bottom panel . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 010 In vitro phosphorylation studies on AKAP18-associated PDE4D were performed in two phases . Our structural data predict that the C subunit of anchored PKA resides within ∼100 Å of its substrate PDE4D ( shown schematically in Figure 4C ) . We reasoned that this tight configuration could permit cAMP-independent phosphorylation of the phosphodiesterase . Therefore , in the first phase , ‘basal’ phosphorylation of PDE4D was measured in the context of the AKAP18γ signaling complex ( Figure 4D ) . Phosphate incorporation into PDE4D was increased 1 . 87 ± 0 . 14-fold ( p<0 . 05 ) upon its tethering to AKAP18γ as assessed by autoradiography ( Figure 4D , E ) . This suggests that the proximity to a substrate afforded by AKAP18γ augments the cAMP-independent action of PKA . This could explain why physiologically relevant PKA targets such as aquaporins or phospholamban , which require continual or instantaneous phosphorylation to fulfill their biological roles , have their own AKAP-associated pool of kinase ( Henn et al . , 2004; Lygren et al . , 2007; Gold et al . , 2012 ) . Accordingly , a primary function of AKAPs may be to ensure basal phosphorylation of such tethered substrates . In the second phase , experiments were conducted using higher-order complexes formed with wild-type RIIα , RIIα Δ44–86 , or the RIIα ZeChimera ( assembled as described above ) to investigate whether manipulating the intrinsic flexibility of PKA altered phosphorylation of anchored PDE4D ( Figure 4F–G ) . We measured cAMP-independent phosphorylation of PDE4D at 5 min , a time point that showed sub-maximal substrate phosphorylation ( Figure 4F , G , Figure 4—figure supplement 1 ) . Basal PDE4D phosphorylation was enhanced 1 . 97 ± 0 . 18-fold ( n = 6 , p<0 . 05 ) in complexes formed with RIIα Δ44–86 when compared to a wild-type complex ( Figure 4G , bars 1 and 4 ) . In contrast , extension of the linker region in the context of AKAP18γ–RIIα ZeChimera PKAholo assembly had no effect as compared to wild type ( Figure 4G , bars 1 and 7 ) . Control experiments confirmed that addition of cAMP further augmented phosphorylation of PDE4D in all cases ( Figure 4F , lanes 2 , 5 and 8 ) and pretreatment with PKI inhibitor peptide abolished anchored kinase activity ( Figure 4F , lanes 3 , 6 and 9 ) . These data show that the AKAP can be thought of as a catalyst that physically brings the reactants together , and the flexibility within the anchored PKA holoenzyme allows for the precise orientation of the enzyme and substrate . This mechanism may be particularly relevant for cAMP-independent phosphorylation events that are believed to represent approximately 30% of PKA action ( Taylor et al . , 2012 ) . Therefore , this hitherto unexplained but critical component of cellular PKA activity may be accomplished by the persistent phosphorylation of substrates embedded in higher-order AKAP signaling assemblies . To follow these in vitro studies , we tested our hypothesis that flexibility within the anchored PKA holoenzyme influences cAMP signaling in living cells . We generated a modified fluorescence resonance energy transfer ( FRET ) based PKA activity sensor using the A-kinase activity reporter ( AKAR2 ) backbone ( Zhang et al . , 2005 ) . Our modified sensor ( AKAR-18RBS ) was constructed by fusing the PKA binding helix of AKAP18 ( 18RBS ) ( Fraser et al . , 1998; Gray et al . , 1998 ) to the amino terminus of AKAR2 ( Figure 5A ) . This genetically encoded reporter detects PKA phosphorylation in real-time by monitoring changes in the YFP/CFP emission ratio inside cells ( Figure 5A ) . As a prelude to these studies AKAR-18RBS association with wild-type RIIα or either of the modified RIIα constructs was confirmed by co-immunoprecipitation of each complex from HEK293 cells ( Figure 5B ) . In parallel , immunoblot and confocal fluorescent imaging analyses confirmed that mCherry tagged versions of each RIIα form were expressed to equivalent levels ( Figure 5C ) and uniformly distributed in HEK293 cells ( Figure 5—figure supplement 1A ) . 10 . 7554/eLife . 01319 . 011Figure 5 . Flexibility within the anchored PKA holoenzyme impacts cAMP responsive signaling inside cells . ( A ) Schematic of the modified AKAR-18RBS FRET reporter used in these studies . The PKA RII binding site common to all AKAP18 isoforms was fused to the N-terminus of AKAR2 to create a FRET-based kinase activity sensor that anchors type II PKA holoenzyme . Phosphorylation of a consensus site Thr by PKA induces recruitment of the adjacent Forkhead homology-associated ( FHA ) phospho-Thr binding domain . Subsequent rearrangement brings together the ECFP and YFP ( citrine ) moieties to produce an increase in FRET signal as readout of kinase activity . ( B ) Co-immunoprecipitation of AKAR-18RBS-PKA complexes with holoenzymes composed of three different RII forms . AKAR-18RBS was immunoprecipitated with anti-GFP antibodies and bound PKA subunits were detected by western blotting . ( C ) Lysates from cells transfected with the AKAR-18RBS reporter and each of the three RIIαforms were immunoblotted to confirm similar expression levels in each cohort . ( D ) Basal ( unstimulated ) raw FRET signals from cells expressing RIIα ( gray ) , RIIα Δ44–86 ( green ) and RIIα ZeChimera ( orange ) . ( E and G ) Time course of AKAR-18RBS activation in response to the β-adrenergic agonist isoproterenol in cells co-expressing the FRET reporter with ( E ) RIIα wild type , ( F ) RIIα Δ44–86 , or ( G ) RIIα ZeChimera . Isoproterenol ( Iso , 10 μM ) was added at t = 100 s and FRET was recorded for 5 min post-stimulation . Warmer colors indicate increasing phosphorylation as shown in the pseudo-color scale . ( H ) Amalgamated traces of the Iso stimulated changes in FRET in each cohort ( 0–400 s ) . Data are normalized to unstimulated basal FRET level for each respective RIIα form . Changes in the AKAR-18RBS normalized FRET ratio are shown from cells expressing RIIα wild type ( black ) , RIIα Δ44–86 ( green ) and RIIα ZeChimera ( orange ) . Mutation of the phosphoacceptor threonine in the FRET reporter to create AKAR-18RBS–T/A blocks phosphorylation and abolishes FRET in response to Iso treatment ( Inset ) . ( I ) Scatter plot representation of peak FRET signals from all cells at 200 s . This plot indicates that AKAR-18RBS-PKA complexes formed with RIIα wild type ( black ) or the RIIα ZeChimera ( orange ) had similar peak responses , while complexes formed with RIIα Δ44–86 displayed significantly greater peak FRET responses ( p<0 . 001 ) . This plot also shows that in all cases , some cells fail to respond entirely; these non-responders are included in the amalgamated data analysis presented in ( H ) . ( J ) C subunit of PKA association with AKAR-18RBS , AKAR-18RBS–pro or AKAR-18RBS–T/A complexes . AKAR-18RBS and AKAR-18RBS–T/A co-precipitate RIIα and endogenous PKA catalytic subunit; AKAR-18RBS–pro , a mutant that cannot bind RII subunits , fails to co-precipitate PKA complexes ( top panels ) . Iso treatment ( 1 μM , 5 min ) does not cause dissociation of PKAc from the AKAR-18RBS or AKAR-18RBS–T/A complexes ( top panel , lanes 1–2 and 5–6 ) . Control immunoblots show equivalent levels of AKAR-18RBS , AKAR-18RBS–pro and AKAR-18RBS–T/A in immunoprecipitates as well as RIIα and PKAc expression in cell lysates ( middle panels ) . Immunoblotting for phospho-PKA substrates ( R-X-X-pS/T motif ) confirms that Iso treatment activates endogenous β-ARs and initiates downstream phosphorylation events ( bottom panel ) . ( K ) Immunoprecipitation of full-length AKAP18 complexes following Iso treatment . Cells expressing AKAP18γ and RIIα variants were treated with vehicle or Iso ( 1 μM , 5 min ) and AKAP18γ complexes were immunoprecipitated . Immunoblotting shows that Iso has no effect on the amount of PKA catalytic subunit in complexes formed with RIIα wild type , RIIα Δ44–86 , or RIIα ZeChimera . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 01110 . 7554/eLife . 01319 . 012Figure 5—figure supplement 1 . Comparison of AKAR-18RBS and AKAR-18RBS–pro anchoring and FRET controls . ( A ) Representative confocal micrographs of live cells show comparable expression levels and the intracellular localization of AKAR-18RBS ( green ) and mCherry-RIIα ( red ) . Composite images are also shown . HEK293 cells were transiently transfected with AKAR-18RBS and either PKA RIIα wild type ( upper panels ) , RIIα Δ 44–86 ( middle panels ) or RIIα ZeChimera ( lower panels ) . Scale bars represent 5 μm . ( B ) Schematic of AKAR-18RBS–pro that contains two helix-breaking proline substitutions in the PKA anchoring site that abolish RII binding . ( C ) AKAR-18RBS co-precipitates all three RIIα variants as well as endogenous PKA catalytic subunit from transiently transfected HEK293 cells ( top panels ) . AKAR-18RBS–pro no longer anchors RII and does not co-precipitate RIIα or PKAc ( top panels ) . Control immunoblots show equivalent levels of AKAR-18RBS and AKAR-18RBS–pro in immunoprecipitates as well as RIIα and PKAc expression in cell lysates . ( D ) FRET recordings using AKAR-18RBS–pro . Traces show Iso stimulated changes in FRET in each cohort ( 0–400 s ) . Data are normalized to unstimulated basal FRET level for each respective RIIα form . The increase in the AKAR-18RBS normalized FRET ratio in cells co-expressing RIIα Δ44–86 ( Figure 5H ) is no longer present when the reporter is unable to bind PKA . DOI: http://dx . doi . org/10 . 7554/eLife . 01319 . 012 Initial experiments evaluated basal FRET of the AKAR-18RBS reporter using a modified Leica DMI 6000B microscope . The raw YFP/CFP emission ratio of AKAR-18RBS was elevated in unstimulated cells expressing RIIα Δ44–86 as compared to RIIα wild type or the RIIα ZeChimera ( Figure 5D ) . These data further develop the concept introduced in Figure 4 that the RIIα linker region influences basal phosphorylation of AKAP-associated substrates . Next , real-time changes in the YFP/CFP emission ratio of the AKAR-18RBS reporter were monitored in cells expressing equivalent levels of the individual RIIα forms ( Figure 5E–G ) . The β-adrenergic ( β-AR ) agonist isoproterenol ( Iso ) was administered after 100 s to initiate the cAMP response ( Figure 5E–G ) . Notable increases in AKAR-18RBS FRET were evident in cells expressing RIIα wild type or the RIIα ZeChimera ( Figure 5E , G ) . In both the cases the normalized FRET ratio was maximal at 200 s and gradually declined over the remainder of the time course ( Figure 5H , black and orange traces ) . This latter phenomenon may be attributed to either desensitization of the β-AR system or dephosphorylation of the reporter by phosphoprotein phosphatases ( Bouvier et al . , 1987; Pitcher et al . , 1992; Violin et al . , 2003 ) . Importantly , cells expressing the more compact RIIα Δ44–86 form responded more rapidly and robustly to isoproterenol with a maximal FRET response at 200 s ( Figure 5F , H , green trace ) . The RIIα Δ44–86 effect was abolished when control experiments were performed with a proline modified AKAR-18RBS derivative that is unable to anchor PKA ( Fraser et al . , 1998; Gray et al . , 1998; Figure 5—figure supplement 1 ) . Other control experiments established that application of isoproterenol did not stimulate an AKAR-18RBS derivative ( T/A ) where the threonine phospho-acceptor was replaced with alanine ( Figure 5H , inset , red trace ) . Scatter plot representation of the Iso stimulated FRET responses at 200 s reveal that AKAR-18RBS-PKA complexes formed with RIIα or the RIIα ZeChimera responded with similar magnitudes ( Figure 5I , black and orange ) . In contrast , the normalized FRET responses of the more compact and less flexible AKAR-18RBS-PKA complexes formed with RIIα Δ44–86 were significantly higher ( Figure 5I , green , p<0 . 001 ) . Taken together , these cell-based studies indicate that removal of the flexible linker region in RIIα augments cAMP-responsive PKA phosphorylation of AKAR-18RBS inside the cells . These findings complement the interpretation of our EM reconstructions and in vitro phosphorylation studies . These observations argue that compact and rigid AKAP–PKA assemblies enhance kinase action within the immediate vicinity of the substrate . Local activation of PKA is believed to involve dissociation of the C subunits from the R subunit dimer . Yet , cAMP binds to the R subunits avidly ( KD 6–8 nM ) , and its rate of release is so slow that it is not readily apparent how the cAMP-binding sites turnover within the cell ( Poppe et al . , 2008 ) . Additionally , there is a report that cAMP can activate PKA without C subunit release ( Yang et al . , 1995 ) . Therefore , we evaluated whether cAMP stimulation altered the composition of PKA holoenzymes associated with the AKAR-18RBS reporter . Interestingly , both the RII and C subunits of PKA were present in AKAR-18RBS immune complexes , even after stimulation of cAMP production by isoproterenol ( Figure 5J , lanes 1 and 2 ) . The isoproterenol-stimulated activation of PKA was validated by immunoblot detection of phosphorylated substrates with a phospho-PKA site antibody ( Figure 5J , bottom panel ) . Control experiments confirmed that PKA subunits did not associate with the proline-modified AKAR-18RBS derivative , whereas the AKAR-18RBS-T/A reporter retained the ability to anchor PKA ( Figure 5J , lanes 3–6 ) . This AKAR-18RBS reporter data raised the question of whether anchoring can stabilize the PKA holoenzyme even in the presence of cAMP . To further test this hypothesis , co-precipitation experiments were repeated with full-length AKAP18γ . Again the C subunit was retained within the AKAP complex upon isoproterenol stimulation ( Figure 5K , lanes 1 and 2 ) . This occurred regardless of whether the PKA holoenzymes were formed with wild-type RIIα , RIIα Δ44–86 , or RIIα ZeChimera ( Figure 5K , top panel ) . These results confirm that the increased agonist-stimulated activity of the AKAR-18RBS–RIIα Δ44–86 complex shown in Figure 5H , I is not due to enhanced dissociation of the PKA catalytic subunit . A broader implication is that release of the C subunit from anchored PKA holoenyzmes may not be required for the phosphorylation of nearby substrates . Under this scenario , association with AKAPs limits and thereby defines the range of kinase action within the cell . This underscores the central role of AKAPs in the governance of cAMP responsive phosphorylation events . A combination of steric effects and environmental factors may explain the enhanced catalytic efficiency of this condensed AKAP–PKA assembly inside cells . For example , the less flexible RIIα Δ44–86 dimer may constrain the PKAc subunits in a manner that minimizes the distance to their target substrate . Moreover , the more compact AKAP-PKAΔ44-86-substrate assembly may augment kinase activity by inhibiting the actions of signal termination enzymes . This could involve the steric exclusion of cellular protein phosphatases that catalyze the removal of phosphate or protection from phosphodiesterase activity that metabolizes cAMP . Deletion of the flexible linker between residues 44–86 of RIIα not only enhances basal phosphorylation of these anchored substrates , but also supplements the catalytic efficiency of the anchored PKA holoenzyme in this cellular context . However , this enhanced anchored kinase action may be an undesirable feature in vivo . Hence , we propose that the evolution of flexible linkers in RII may be a means to ensure bi-directional regulation of phosphorylation events by signal termination enzymes . It is noteworthy that extending the number of residues in the RIIα ZeChimera linker had little effect on the overall structure of the AKAP-PKA complex , or on the activity of PKA monitored by our in vitro and cell-based assays . According to our structural analysis , this construct samples a range of radial conformations that are very similar to wild-type particles ( Figure 3H ) . Moreover , this observation is consistent with theoretical analyses of random coiled chain models that demonstrate the average end-to-end length of random coil scales by the square root of the number of amino acids in the chain ( average length ∼ N0 . 5 , where N is the number of residues ) ( Flory , 1975 ) . This statement implies that the average length separating the ends of a random coil reaches a plateau , or maximum , that is only marginally altered upon the inclusion of more residues . Consequently extending the linker in the context of the RIIα ZeChimera may explain why this affords minimal benefit to the overall structure and dynamics of this seemingly more extended and malleable anchored enzyme complex . Hence , evolutionary constraint of the RIIα linker domain to 45–47 residues may ensure appropriate separation between the anchoring and transduction domains of anchored mammalian PKA holoenzymes . Another key element of our model is that the array of conformations adopted by the anchored RII subunit dimer defines a ‘radius of action’ for the tethered catalytic subunit . Not only does this structural plasticity explain why an intact PKA holoenzyme structure has eluded X-ray crystallographers for over 20 years but also suggests that the ‘built in’ flexibility within this protein complex sets a dynamic range for basal and cAMP-dependent phosphorylation of nearby substrates . This latter concept could have profound implications for PKA and other kinases , where enzyme activity is constrained within anchored complexes or enzyme scaffolds . The conformational variability of the tethered complex could enhance kinase access to multiple phosphorylation sites on the same substrate , or , alternatively , orient each catalytic subunit toward distinct substrate proteins within the vicinity of the scaffolding site . Our structural and cellular characterization of higher-order AKAP assemblies emphasizes how static protein–protein interactions act in concert with the dynamic movement of sub-structures to orchestrate local phosphorylation events . We believe that these combined molecular approaches are broadly applicable for the investigation of related enzyme complexes that are spatially organized within the cell . RIIα and PKAc were expressed in BL21 ( DE3 ) pLysS . Escherichia coli was transformed with pET28b-RIIα-6xHis ( encoding full-length RIIα , accession NM_008924 ) or pET15a-6xHis-mPKAc ( Addgene plasmid 14921; www . addgene . org ) , encoding full-length mouse PKA catalytic subunit ( Narayana et al . , 1997 ) and grown to an OD600 of 0 . 6 . Protein expression was induced with 1 mM IPTG , and cells were grown with shaking for 16 hr at 28°C ( for RIIα-His ) or 24 hr at 16°C ( for PKAc ) . The cells were harvested by centrifugation at 5 , 000×g for 10 min at 4°C . Pellets containing induced protein were stored at −20°C until use . For purification , the cells were thawed in 50 mM NaPhosphate , pH 7 . 5 , 100 mM NaCl , 10 mM Imidazole , 2 mM MgCl2 containing 4 mg/ml lysozyme , 1 mM AEBSF , 50 μg/ml soybean trypsin inhibitor , 2 μg/ml leupeptin , 16 μg/ml benzamidine , and 25 U/ml benzonase . After resuspension , Triton X-100 was added to 0 . 5% and lysates were rocked at 4°C for 30–60 min . When the lysate was no longer viscous , debris was pelleted by centrifugation at 37 , 000×g for 30 min at 4°C . Cleared lysates were added to 1/25 vol Ni+–NTA beads ( bed volume; GE Healthcare , Pittsburg , PA ) and rocked for 1 hr at 4°C . The bead slurry was then added to a column and allowed to settle . Beads were washed with >10 bed volumes of 50 mM NaPhosphate , pH 7 . 5 , 500 mM NaCl , 20 mM Imidazole . Purified protein was eluted by collecting 2 × 2 bed volume fractions of 50 mM NaPhosphate , pH 7 . 5 , 500 mM NaCl , 50 mM Imidazole followed by 4 × 2 bed volumes 50 mM NaPhosphate , pH 7 . 5 , 500 mM NaCl , 300 mM Imidazole . Fractions were analyzed by SDS-PAGE and Coomassie staining . The fractions containing the protein of interest were pooled , concentrated and applied to a Superdex 16/600 S200 gel filtration ( GF ) column . The column was run in 25 mM HEPES , pH 7 . 5 , 200 mM NaCl , 1 mM EDTA , 1 mM TCEP . Peak fractions were pooled , concentrated , brought to 10% glycerol and frozen in liquid N2 for storage at −80°C . AKAP18γ ( accession NM_016377 [Note: This sequence was updated during the course of these studies and corresponds to the longer AKAP18δ isoform; the construct used in these studies encodes the γ isoform starting at Met23 of the referenced ORF] ) was expressed with an N-terminal strep-II tag and a C-terminal 6x-His tag in High Five insect cells ( Trichopulsia ni; Invitrogen ) . Sf-9 cells were transfected with a Bacmid containing the expression cassette from pFastBac-C-His-strep-AKAP18γ . After two rounds of viral amplification , High Five cells were infected with baculovirus ( at a cell density of ∼2 × 106 cells/ml ) and grown for 72 hr at 26°C shaking at 105 rpm . Cells were harvested by centrifugation ( 10 min , 700×g ) and frozen at −20°C . Protein was purified as above using Ni+–NTA Sepharose , except lysozyme was omitted from the lysis buffer . Gel filtration chromatography was performed as described above . MBP-PDE4D3-8xHis was expressed in BL21 ( DE3 ) pLysS transformed with the vector pMAL-C5P2-PDE4D3-8His , which was constructed by PCR and InFusion ( Clontech ) cloning into a modified pMAL vector ( NEB ) containing an 8xHis tag prior to MBP and a PreScission protease cleavage site after MBP . Protein was purified using Ni+–NTA Sepharose , and gel filtration chromatography was performed as described above . Constructs for deleted and extended linker forms of mRIIα were created using standard cloning methods . Briefly , to create the linker deletion , AscI restriction sites were introduced into the cDNA by site-directed mutagenesis . The resulting plasmid was digested with AscI and then re-ligated closed . To create the Zebrafish linker chimera , a minigene encoding the Zebrafish linker from NCBI Reference sequence NM_212958 . 1 was synthesized with flanking AscI sites ( Integrated DNA Technologies ) . This minigene was digested out of the carrier vector and inserted into the AscI sites in the modified RIIα cDNA . Clones were screened for orientation and verified by sequencing ( Genewiz ) . Mammalian expression constructs encoding RII variants were created by PCR cloning into Gateway DONR vectors ( Invitrogen ) followed by LR cloning into pcDNA-DEST40 . Vectors that express AKAR-18RBS , the AKAR-18RBS–T/A and the AKAR-18RBS–pro ( L9P , A13P in the AKAP18 anchoring helix ) mutants were created upon insertion of a cDNA encoding the PKA binding site from AKAP18 into pcDNA3-AKAR2 . 2 ( Zhang et al . , 2005 ) followed by site directed mutagenesis ( QuikChange XL II , Stratagene ) . AKAP18γ–PKA complexes were assembled using purified proteins . RIIα and PKAc were mixed at a 1:1 . 2 molar ratio , and AKAP18γ was added in molar excess to favor the isolation of stoichiometric complexes by gel filtration . Complexes were dialyzed against 15 mM MOPS , pH 6 . 5 , 100 mM NaCl , 1 mM MgCl2 , 1 mM TCEP , 2% glycerol for >4 hr at 4°C in 0 . 5 ml Slide-A-Lyzer MINI dialysis cups ( 88401; Thermo Scientific ) . The complexes were then loaded onto a Superdex 16/600 S200 gel filtration column on an AKTA Purifier ( GE Healthcare ) and separated at a flow rate of 0 . 5 ml/min in 25 mM HEPES , pH 7 . 5 , 200 mM NaCl , 1 mM EDTA , 1 mM TCEP . Peak fractions were analyzed by SDS-PAGE and Coomassie staining . The fractions containing approximately stoichiometric complex components were pooled , concentrated , brought to 10% glycerol and frozen in liquid N2 for storage at −80°C . RIIα-PKAc holoenzyme complexes were assembled as above , omitting AKAP18γ . Duplicate samples were separated on 10% Tris-Glycine-SDS gels . Proteins were transferred to nitrocellulose membranes and blocked with 5% NFDM , 1% BSA in TBST . Membranes were probed with either anti-AKAP18 rabbit polyclonal antiserum ( VO57 ) , mouse anti-PKAc or anti-RIIα monoclonal antibodies ( BD Biosciences ) overnight at 4°C . The membranes were washed , incubated with appropriate secondary antibodies , washed again and protein was detected using ECL ( Super-signal Pico , Thermo Pierce ) on an AlphaInnotech Multiimager III . AKAP18γ–PKA complexes were analyzed by native PAGE using precast 3–8% Tris-Acetate gels ( Invitrogen ) . The samples were mixed with Invitrogen’s 2X native sample buffer and run in 50 mM Tris pH 8 . 3 , 50 mM Tricine , along with native MW markers ( Invitrogen ) . Protein was detected by silver staining ( SilverQuest , Invitrogen ) . PDE assays were performed as described ( Hill et al . , 2005 ) . Briefly , after washing , immunocomplexes were suspended in 50 μl KHEM ( 50 mM KCl , 50 mM HEPES-KOH , pH 7 . 2 , 10 mM EGTA , 1 . 92 mM MgCl2 ) . Next , 50 μl of PDE assay buffer ( 20 mM Tris-HCl , pH 7 . 4 , 10 mM MgCl2 , 2 μM cAMP , 100 nM okadaic acid ) containing ∼50 , 000 cpm [3H]-cAMP was added , and the samples were incubated at 30°C for 20 min with shaking . The samples were boiled for 3 min and incubated on ice for 15 min . Snake venom ( 25 μl of a 1 mg/ml solution ) was added , and the samples were incubated at 30°C for 10 min with shaking . Dowex AG 1×8 ( 200–400 mesh , CL form , washed and used as a 1:1:1 slurry of resin:water:ethanol ) ion exchange resin ( 400 μl ) was added , and the samples were incubated on ice for 15 min . The resin was centrifuged for 3 min at 14 , 000×g , and 150 μl of the supernatant was mixed with 1 ml scintillation fluid for counting . All assays were performed in duplicate . PKA holoenzyme was incubated in the absence or presence of molar equivalent of purified AKAP18γ . These pre-formed complexes ( 1 μg ) were mixed with excess substrate MBP-PDE4D3 ( 2 μg ) in 40 mM Tris , pH 7 . 5 , 10 mM MgOAc , 0 . 1 mM EGTA , 100 μM IBMX . Mg2+-ATP containing 1 μCi 32P-ATP was added to a final concentration of 10 μM . The samples were incubated at 30°C with rapid shaking for 5 min . Reactions were stopped by the addition of 5X Laemmli sample buffer and boiling for 5 min . The samples were separated by SDS-PAGE , and gels were either stained with Coomassie blue or transferred to nitrocellulose for India ink staining . Phosphorylation of PDE4D3 was detected by autoradiography . Films were scanned for quantification by densitometry , followed by analysis using GraphPad Prism . Analysis of AKAP18γ–PKA complexes containing different RIIα variants was performed and analyzed similarly . In these assays , PKI was used at a final concentration of 10 μM and added to appropriate samples 10 min prior to the start of the assay . Prior to initiation of the assay , cAMP was added to some samples to a final concentration of 30 μM . HEK293 cells were transiently transfected with cDNAs encoding AKAR-18RBS and individual RII subunits , and cultured on 35-mm glass coverslip dishes ( MatTek Corporation ) . 48 hr after transfection , the cells were imaged in HEPES-Buffered Saline Solution ( HBSS; 116 mM NaCl , 20 mM HEPES , 11 mM Glucose , 5 mM NaHCO3 , 4 . 7 mM KCl , 2 . 5 mM CaCl2 , 1 . 2 mM MgSO4 , 1 . 2 mM KH2PO4 , pH 7 . 4 ) . The cells were stimulated by isoproterenol ( 10 μM ) . Fluorescence emission was acquired using a DMI6000B inverted microscope ( Leica ) , an EL6000 component ( fluorescent light source , filter wheel , ultra fast shutter , Leica ) and a CoolSnap HQ camera ( Photometrics ) , all controlled by MetaMorph 7 . 6 . 4 ( Molecular Devices ) . Dual-emission images were obtained simultaneously through a Dual-View image splitter ( Photometrics ) with S470/30 and S535/30 emission filters and 505 dcxr dichroic mirror ( Chroma ) . Exposure time was 200 ms with an image interval of 10 s . FRET changes within a region of interest were calculated as the ratio of measured fluorescent intensities from two channels ( Mdonor , MindirectAcceptor ) after subtraction of background signal . FRET ratio ( YFP/CFP ) changes were normalized to the average FRET ratio value before stimulation . HEK293 cells were transiently transfected ( TransIT LT1; Mirus ) with vectors encoding various proteins according to figure legends . After 40–48 hr and prior to harvesting , the cells were serum starved in DMEM for 1 hr at 37°C . The cells were then treated with vehicle or Isoproterenol for 5 min at 37°C and harvested in lysis buffer ( 25 mM HEPES , pH 7 . 4 , 150 mM NaCl , 1 mM EDTA , 1 mM EGTA , 20 mM NaF , 2% glycerol , 0 . 3% Triton X-100 ) containing protease inhibitors . AKAP18γ or AKAR-18RBS complexes were immunoprecipitated with anti-GFP rabbit IgG ( Invitrogen ) and protein A agarose for 2 hr at 4°C . Beads were washed 1 × 1 ml in lysis buffer containing 350 mM NaCl , 1 × 1 ml wash buffer ( lysis buffer with no detergent ) containing 350 mM NaCl and 2 × 1 ml wash buffer . Proteins were separated on 4–12% gradient gels ( Invitrogen ) and transferred to nitrocellulose membranes . Primary antibodies ( PKA catalytic mAb [BD Biosciences] , 1:1000; RIIα mAb [BD Biosciences] , 1:2000; GFP mAb [Santa Cruz Biotechnology] , 1:2000; phospho-PKA substrates rabbit mAb [Cell Signaling Technology] , 1:1000 ) were incubated with membranes overnight at 4°C in TBST/Blotto . The membranes were washed extensively in TBST , incubated with secondary antibodies , washed again and developed using ECL ( Thermo Pierce ) on an Alpha Innotech MultiImage III with FluoroChem Q software . Confocal microscopy was performed using a Zeiss LSM-510 META laser scanning confocal microscope equipped with a Zeiss 63 × g , 1 . 4 numerical aperture , oil immersion lens . For live cell imaging , HEK293 cells expressing AKAR-18RBS and individual RII-mCherry constructs were imaged in HBSS . Colocalization studies were performed using dual excitation ( 488 , 543 nm ) and emission ( bandpass 505–530 nm and long-pass 560 nm for YFP and mCherry , respectively ) filter sets . The wild-type and mutant AKAP18γ–PKA holoenzyme complexes were prepared similarly for electron microcopy studies . Freshly purified samples were diluted 1:25 times to a final concentration of 5 μg/ml with EM buffer containing 25 mM HEPES ( pH—7 . 4 ) , 200 mM NaCl , 0 . 5 mM EDTA and 1 mM dithiothreitol ( DTT ) . For gold-labeling studies , a 0 . 05% solution of streptavidin-conjugated gold particles ( 5 nm , Nanocs ) was diluted 1:100 with EM buffer and mixed 1:1 ( vol/vol ) with the wild-type AKAP18γ–PKA complex and allowed to incubate for 1 hr at 4°C . 2 μl of specimen samples were applied to carbon coated EM grids and negatively stained with 0 . 75% ( wt/vol ) uranylformate . Wild-type particles were visualized on a 120 kV TEM ( FEI ) , and images were recorded at a nominal magnification of 52 , 000 × at the specimen level on a 4 k × 4 k CCD ( Gatan ) . Serial tomographic images were obtained using automated data collection software ( Xplore3D , FEI ) collected at tilt angles of 0° , 20° , 40° and 50° ( Figure 2—figure supplement 1 ) . The mutant complexes and gold-labeled particles were visualized on a 120 kV TEM ( FEI ) and recorded at a nominal magnification of 68 , 000 × at the specimen level on a 4 k × 4 k CMOS-based camera ( Tietz ) . For image processing , micrographs were binned two times yielding final pixel sizes of 4 . 2 Å per pixel and 3 . 0 Å per pixel for the wild-type dataset and mutant datasets , respectively . Thon rings in the power spectra were used to select only those micrographs free of drift or significant astigmatism . The contrast transfer function ( CTF ) parameters were determined for each micrograph using the program CTFTILT ( Mindell and Grigorieff , 2003 ) . ∼7 , 000 wild-type particles and 1 , 000 mutant particles were hand selected from micrographs using Ximdisp ( Smith , 1999 ) . Two-dimensional ( 2D ) projection averages were generated using reference-free multivariate statistical analysis methods in IMAGIC ( van Heel et al . , 1996 ) and with the Iterative Stable Alignment and Clustering routines implemented using the ISAC program ( Yang et al . , 2012 ) . Initial three-dimensional ( 3D ) reconstructions were generated from 0° , 20° , 40° and 50° tilted particle class-average datasets displaying either linear or bent conformations using angular reconstitution routines in IMAGIC ( van Heel et al . , 1996 ) . Refinement of the initial 3D density maps was performed using individual particle images in FREALIGN ( Grigorieff , 2007 ) without symmetry constraints . The final 3D density maps were filtered to 35 Å resolution as suggested by Fourier shell correlation ( FSC ) analysis . Back-projection analysis was carried out in SPIDER ( Frank et al . , 1996 ) . Statistical analysis of individual particle dimensions was obtained by measuring particle lengths on unbinned micrographs using ImageJ ( Schneider et al . , 2012 ) and evaluated by the two-tailed t test . Structural models of the AKAP18γ–PKA holoenzyme complex were constructed to represent the triangular and extended conformations determined by the 3D reconstructions . A model of AKAP18γ ( residues 88–317 ) was initially constructed as follows . The atomic coordinates of the AKAP–PKA binding helix in complex with the RIIα docking and dimerization domain ( RIID/D-AKAPhelix; PDB 2IZX ) ( Gold et al . , 2006 ) was used as a template to model the corresponding region in AKAP18γ , residues 301–317 , by computational mutation and minimization in COOT ( Emsley and Cowtan , 2004 ) . The AKAPhelix domain was oriented with the N-terminus ( residue 301 ) placed in close proximity to the C-terminus of the AKAP18γ central domain ( PDB 2VFL ) ( Gold et al . , 2008 ) , residues 88–290 . The relative orientations were guided by their corresponding fit into the 3D density map . The two AKAP18γ domains were connected by a 9-residue linker corresponding to the AKAP18γ sequence 291–299 using COOT ( Emsley and Cowtan , 2004 ) . This model of the linked AKAP18γ structure ( residues 88–317 ) bound to the RIID/D domain was placed within the central density of the density map , and the fit was refined by computational minimization in Chimera ( Pettersen et al . , 2004 ) . The central density was determined to be the site of AKAP18γ localization by affinity gold labeling studies ( see above ) . The atomic coordinates corresponding to the mouse PKA catalytic domain ( PKAc ) bound to the RIIα cAMP binding domain ( PDB 2QVS ) ( Wu et al . , 2007 ) were placed into each of the peripheral EM densities , and their fits were independently minimized in Chimera ( Pettersen et al . , 2004 ) . Each RIIα regulatory domain ( residues 91–392 ) was connected to a corresponding RIID/D domain ( residues 3–43 ) by a 46-residue linker region ( corresponding to residues 44–90 from mouse ) using the modeling program COOT ( Emsley and Cowtan , 2004 ) . The RIIα linker regions were modeled as unstructured polypeptides connecting the sub-domains of the AKAP18γ-PKA holoenzyme complex oriented in the triangular and linear conformations determined by EM . Figures were prepared in Adobe Photoshop . Protein structures were visualized and images captured in UCSF chimera version 1 . 6 .	It was once thought that proteins needed to have structures that were both ordered and stable , but this view was changed by the discovery that certain proteins contain regions that are disordered and flexible . In some cases these regions of intrinsic disorder help the protein to function by linking more stable regions that are active . However , in other proteins the disordered regions are themselves biologically active and can , for example , function as enzymes . Protein kinase A is a family of enzymes that contains both ordered and disordered regions , with the ordered sections being involved in phosphorylation , a chemical process that is widely used for communication within cells . However , in order to initiate phosphorylation , these kinases must be anchored to a rigid substrate nearby , so a second group of proteins called AKAPs–which is short for A-kinase anchoring proteins–hold the kinases in place by binding to their disordered regions . These AKAPs also help the kinases to dock with other molecules involved in phosphorylation . A full structural picture of how the kinases induce phosphorylation has yet to be obtained , partly because it is extremely difficult to determine the structure of the disordered regions within the kinases . Moreover , the AKAPs are also disordered , which makes it difficult to work out how the kinases are held in position . Smith , Reichow et al . have used electron microscopy to reveal that the disordered region has two important roles: it determines how far away from the anchoring protein that the active region of the kinase can operate , and it influences how efficiently the kinase can bind to its target molecule in order to induce phosphorylation . Future challenges include investigating how the inherent flexibility of AKAP complexes contribute to the efficient phosphorylation of physiological targets .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "biochemistry", "and", "chemical", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2013	Intrinsic disorder within an AKAP-protein kinase A complex guides local substrate phosphorylation
The NR5A-family nuclear receptors are highly conserved and function within the somatic follicle cells of the ovary to regulate folliculogenesis and ovulation in mammals; however , their roles in Drosophila ovaries are largely unknown . Here , we discover that Ftz-f1 , one of the NR5A nuclear receptors in Drosophila , is transiently induced in follicle cells in late stages of oogenesis via ecdysteroid signaling . Genetic disruption of Ftz-f1 expression prevents follicle cell differentiation into the final maturation stage , which leads to anovulation . In addition , we demonstrate that the bHLH/PAS transcription factor Single-minded ( Sim ) acts as a direct target of Ftz-f1 to promote follicle cell differentiation/maturation and that Ftz-f1’s role in regulating Sim expression and follicle cell differentiation can be replaced by its mouse homolog steroidogenic factor 1 ( mSF-1 ) . Our work provides new insight into the regulation of follicle maturation in Drosophila and the conserved role of NR5A nuclear receptors in regulating folliculogenesis and ovulation . Female fertility , an essential half of the reproductive equation , requires proper follicle maturation and ovulation . The NR5A family of nuclear receptors are critical for the success of these complex ovarian processes across species ( Jeyasuria et al . , 2004; Meinsohn et al . , 2019; Mlynarczuk et al . , 2013; Sun and Spradling , 2013; Suresh and Medhamurthy , 2012 ) . The majority of what is known concerning these NR5A receptors in female fertility stems from studies performed over the past two decades in rodent models . These investigations have shown that both members of this family , NR5A1 ( steroidogenic factor-1 or SF-1 ) and NR5A2 ( liver receptor homolog-1 or LRH-1 ) , are expressed in the follicle cells that encapsulate the oocyte throughout oogenesis ( Falender et al . , 2003; Hinshelwood et al . , 2003 ) . Follicle-cell-specific loss of either receptor leads to drastically impaired fertility . LRH-1 knockout in granulosa cells in either primary or more developed antral follicles results in severe anovulation , which is attributed to the inhibition of ––cumulus expansion , expression of steroidal biosynthetic genes , and granulosa cell proliferation/differentiation ( Bertolin et al . , 2014; Bertolin et al . , 2017; Bianco et al . , 2019; Duggavathi et al . , 2008; Meinsohn et al . , 2018 ) . Targeted depletion of SF-1 in granulosa cells of primary follicles has shown to result in hypoplastic ovaries and a dramatically reduced number of developing follicles ( Pelusi et al . , 2008 ) . Much less is known about the molecular mechanism of SF-1 in these ovarian follicle cells . SF-1 was initially recognized as the mammalian homolog of the Drosophila fushi tarazu-factor 1 ( ftz-f1 ) , which was first identified as a transcription factor binding to the promoter of the pair-rule segmentation gene fushi tarazu ( ftz ) during early embryogenesis ( Lala et al . , 1992; Ueda et al . , 1990 ) . Drosophila ftz-f1 gene encodes two protein isoforms ( αFtz-f1 and βFtz-f1 ) , each comprised of unique N-terminal sequences and common C-terminal sequences ( Lavorgna et al . , 1991; Lavorgna et al . , 1993 ) . αFtz-f1 is maternally supplied and functions as a cofactor for Ftz during early embryogenesis ( Guichet et al . , 1997; Yu et al . , 1997 ) . On the other hand , βFtz-f1 is only transiently induced after each ecdysone pulse in the late embryo , larvae , and pupae , and functions as a competency factor for stage-specific responses to ecdysone pulses and progression into the next developmental stages ( Broadus et al . , 1999; Cho et al . , 2014; Lavorgna et al . , 1993; Woodard et al . , 1994 ) . In addition , βFtz-f1 precisely controls the timing of ecdysone pulses through regulating ecdysteroid synthesis enzymes ( Akagi et al . , 2016; Parvy et al . , 2005; Talamillo et al . , 2013 ) . Therefore , βFtz-f1 is essential for late embryogenesis , larval molting , metamorphosis , and pupal development ( Bond et al . , 2011; Boulanger et al . , 2011; Sultan et al . , 2014; Yamada et al . , 2000 ) . Ftz-f1 has also been found to function as an oncogene and promote tumorigenesis and tumor invasiveness in Drosophila imaginal discs ( Atkins et al . , 2016; Külshammer et al . , 2015; Song et al . , 2019 ) . Even though initial studies demonstrated the potential for Ftz-f1 in adult tissues ( Ueda et al . , 1990 ) , little has been done to study what roles Ftz-f1 plays in adult flies , particularly in oogenesis . Drosophila oogenesis is an excellent model for studying many cell biology questions in the last few decades . Drosophila oogenesis occurs in the ovariole , ~16 of which bundle together to form an ovary . At the anterior tip of the ovariole , germline and follicle stem cells proliferate to produce daughter cells to form a stage-1 egg chamber ( also named follicle in this paper ) , which develop through 14 morphologically distinct stages into a stage-14 egg chamber [also named mature follicle; ( Spradling , 1993 ) . Each follicle contains a layer of somatic follicle cells encasing 16 interconnected germ cells , one of which differentiates into an oocyte , while the rest become nurse cells to support oocyte growth and are eventually degraded in mature follicles . Somatic follicle cells proliferate at stages 1–6 and transition into endoreplication at stages 7-10A induced by Notch signaling ( Klusza and Deng , 2011 ) . At stage 10B , a pulse of ecdysone signaling induces follicle cell transition from endoreplication to synchronized gene amplification via zinc-finger transcription factor Ttk69 ( Sun et al . , 2008 ) . This is also accompanied by the downregulation of the zinc-finger transcription factor Hindsight ( Hnt ) and the upregulation of the homeodomain transcription factor Cut in stage-10B follicle cells . As follicles develop from stage 10B onwards , Ttk69 and Cut are diminished . By stage 14 , another critical follicle cell transition occurs , accompanied by re-upregulation of Hnt and complete loss of Cut and Ttk69 ( Knapp et al . , 2019 ) . This transition is critical for the follicle to gain ovulatory competency via upregulation of Octopamine receptor in mushroom body ( Oamb ) and Matrix metalloproteinase 2 ( Mmp2 ) ( Deady and Sun , 2015; Deady et al . , 2015; Deady et al . , 2017; Knapp et al . , 2019 ) . In addition , stage-14 follicle cells upregulate NADPH oxidase ( Nox ) expression , downregulate EcR . B1 and EcR . A , and receive another ecdysteroid signaling via EcR . B2 to become ovulatory competent ( Knapp and Sun , 2017; Li et al . , 2018 ) . However , it is largely unknown how follicle cells differentiate from stage 10B to stage 14 . In this study , we demonstrate that Ftz-f1 is transiently expressed in Drosophila follicle cells at stages 10B-12 and this expression is induced by ecdysteroid signaling in stage-10B follicle cells , independent of Ttk69 . Loss of ftz-f1 in follicle cells after stage 10B severely inhibits follicle cell differentiation into the final maturation stage , resulting in follicles incompetent for follicle rupture and ovulation . In addition , we identify the basic helix-loop-helix/PAS ( bHLH/PAS ) transcription factor Single-minded ( Sim ) , whose functions are known in the central nervous system development ( Crews et al . , 1988; Muralidhar et al . , 1993; Nambu et al . , 1990; Thomas et al . , 1988 ) , functioning downstream of Ftz-f1 for follicle cell differentiation/maturation . RNA-seq and CUT&RUN analyses ( Meers et al . , 2019; Zhu et al . , 2019; Skene and Henikoff , 2017 ) suggest that Sim is a direct target of Ftz-f1 in follicle cells . Furthermore , we demonstrate the role of Ftz-f1 in follicle cell maturation is functionally conserved as ectopic expression of mouse SF-1 is able to rescue Ftz-f1’s function in this process . These findings demonstrate a more conserved role of NR5A nuclear receptors in Drosophila and mammalian reproduction and help elucidate potential mechanisms downstream of NR5A nuclear receptor signaling required for female fertility across species . To investigate the role of Ftz-f1 in female fertility , we first analyzed the expression of Ftz-f1 throughout oogenesis using anti–Ftz-f1 antibody . Ftz-f1 protein is not detected in germline cells and ovarian follicle cells from stage 1 to stage 10A ( Figure 1A ) ; however , it is drastically upregulated in all follicle cells at stage 10B ( Figure 1B ) , when follicle cells transition into synchronized gene amplification . Following stage 10B , Ftz-f1 begins to progressively decrease in follicle cells ( except anterior stretch follicle cells ) and is no longer detectable in stage-13/14 follicle cells ( Figure 1C–F ) . A ftz-f1::GFP . FLAG transgene showed that the expression of Ftz-f1::GFP . FLAG tagged protein completely matches Ftz-f1 antibody staining ( Figure 1—figure supplement 1A–E ) . In addition , we also examined the ftz-f1 transcription using the enhancer trap line ftz-f1 fs ( 3 ) 2877 , which has a P-element containing lacZ gene inserted in the ftz-f1 gene ( Karpen and Spradling , 1992 ) . Expression of βGal is also induced in stage-10B follicle cells and stays high in stage-13/14 follicle cells ( Figure 1—figure supplement 1F–J ) , which is likely a result of βGal not being subjected to endogenous protein regulation . Together , our data suggest that both ftz-f1 mRNA and protein are transiently induced in stage-10B to 12 follicle cells during Drosophila oogenesis . The drastic upregulation of Ftz-f1 at stage 10B is concurrent to the ecdysone-induced transition from endoreplication to gene amplification at stages 10A/10B , which is mediated by the upregulation of the zinc-finger transcription factor Ttk69 ( Sun et al . , 2008 ) . Therefore , we examined whether ecdysone signaling induces ftz-f1 expression in follicle cells . Using the flip-out Gal4 system ( Pignoni and Zipursky , 1997 ) , we disrupted the ecdysone signaling via misexpressing a dominant-negative ( DN ) form of ecdysone receptor ( EcRDN ) ( Cherbas et al . , 2003 ) . EcRDN-overexpressing follicle cells showed a complete loss of Ftz-f1 in stage-10B egg chambers ( Figure 1G ) , indicating that Ftz-f1 expression is induced by ecdysone signaling . We also investigated whether premature activation of ecdysone signaling in follicle cells was sufficient to induce premature Ftz-f1 expression . Treating egg chambers with exogenous 20-hydroxyecdysone ( 20E ) is able to prematurely activate the EcR ligand sensor in follicle cells prior to stage 10 ( Sun et al . , 2008; Figure 1—figure supplement 2A ) but is not sufficient to induce premature expression of Ftz-f1 ( Figure 1—figure supplement 2B ) . Previous work also showed that Ftz-f1 is only induced during low ecdysone titer . Manipulation of Cyp18a1 , encoding a cytochrome P450 enzyme involved in lowering 20E titer , influences Ftz-f1 expression during the prepupa-to-pupa transition ( Rewitz et al . , 2010 ) . In contrast , neither ectopic expression nor knockdown of Cyp18a1 in follicle cells was able to affect Ftz-f1 expression ( Figure 1—figure supplement 2C–F ) . Altogether , our data suggest that Ftz-f1 expression in stage-10B follicle cells is induced by ecdysone signaling and seems insensitive to the ecdysone level . To determine whether Ftz-f1 is induced by Ttk69 , the downstream target of ecdysone signaling , we knocked down Ttk69 expression by overexpressing ttkRNAi in the flip-out Gal4 clones . Follicle-cell clones with ttkRNAi overexpression showed no detectable Ttk69 ( Figure 1—figure supplement 3A ) but normal Ftz-f1 expression in stage-10B egg chambers ( Figure 1H ) . To determine whether Ftz-f1 regulates Ttk69 expression , we generated ftz-f1ex7 mutant clones using the MARCM system ( Wu and Luo , 2006 ) . ftz-f1 mutant follicle cells exhibited normal expression of Ttk69 ( Figure 1I ) . In addition , ftz-f1 mutant follicle cells properly transitioned into the gene amplification stage according to punctate EDU staining ( Figure 1J ) . Our results indicate that ecdysone signaling induces both Ftz-f1 and Ttk69 upregulation in stage-10B follicle cells; the latter one leads to the endoreplication/gene amplification transition , while the former one does not . To determine the function of Ftz-f1 in follicle cells , we knocked down ftz-f1 expression in follicle cells using Vm26Aa-Gal4 , which starts to express in all follicle cells ( except anterior stretch follicle cells ) at stage 10 ( Peters et al . , 2013 ) . Both ftz-f1RNAi1 and ftz-f1RNAi2 showed efficient knockdown of ftz-f1 in stage-10B and stage-12 follicle cells when driven by Vm26Aa-Gal4 , although ftz-f1RNAi1 is more efficient than ftz-f1RNAi2 ( Figure 2A–C , Figure 2—figure supplement 1A–C ) . Females with such genetic manipulation ( named ftz-f1RNAi females ) laid significantly fewer eggs than control females ( Figure 2D and Figure 2—figure supplement 1D ) . In addition , ftz-f1RNAi1 females showed a severe retention of stage-14 follicles inside their ovaries ( Figure 2—figure supplement 1E ) , indicating an ovulation defect . To support this observation , we examined whether stage-14 follicles from ftz-f1RNAi females are competent to Octopamine ( OA ) -induced follicle rupture ( Deady and Sun , 2015; Knapp et al . , 2018 ) . Using the 47A04-LexA driving LexAop2-6XGFP as a reporter for isolating mature follicles , we found that mature follicles from control females had ~83% of follicles ruptured after OA stimulation , consistent with our previous result ( Deady and Sun , 2015 ) . In contrast , mature follicles from ftz-f1RNAi1 and ftz-f1RNAi2 females showed 6% and 17% follicle rupture , respectively ( Figure 2—figure supplement 1F ) . Since hexameric GFP showed punctate GFP signal in mature follicle cells ( Figure 2—figure supplement 1J–L ) , we also used Oamb-RFP as a reporter for isolating mature follicles from both control and ftz-f1RNAi females to perform OA-induced follicle rupture . We observed 67% follicle rupture from control females , but 2% and 18% follicle rupture from ftz-f1RNAi1 and ftz-f1RNAi2 females , respectively ( Figure 2E and Figure 2—figure supplement 1G–I ) . All the data suggest that expression of Ftz-f1 in follicle cells from stage 10B to stage 12 is required for follicle rupture and ovulation . Our recent work has demonstrated that OA/Oamb signaling leads to calcium influx , which activates both Mmp2 and Nox to regulate follicle rupture ( Deady and Sun , 2015; Li et al . , 2018 ) . To determine what is defective in follicles from ftz-f1RNAi females , we first examined whether ionomycin , a Ca2+ ionophore , is sufficient to induce these follicles to rupture . Mature follicles from control females showed 75% follicle rupture with ionomycin stimulation; however , mature follicles from ftz-f1RNAi females only showed ~3% follicle rupture ( Figure 2E ) . Similar results were also found when mature follicles were isolated according to LexAop2-6XGFP ( Figure 2—figure supplement 1F ) . The incompetency of ionomycin to induce follicle rupture in follicles isolated from ftz-f1RNAi females suggests that either components downstream of the calcium rise or ovulatory genes parallel to the calcium pathway are defective in these follicles . Consistent with this , we found that Mmp2 expression in posterior follicle cells was completely disrupted in stage-14 follicles from ftz-f1RNAi females ( Figure 2F–G ) . In addition , we found that these follicles were defective in OA-induced and ionomycin-induced superoxide production ( Figure 2J–K ) , indicating that Nox expression might also be disrupted in mature follicles of ftz-f1RNAi females . Furthermore , we noticed that Oamb-RFP expression became patchy in mature follicles of ftz-f1RNAi females when examined in higher magnification , indicating that Oamb expression is also disrupted ( Figure 2H–I ) . Follicles from ftz-f1RNAi females also exhibited morphological defects in overall shape and dorsal appendage formation ( Figure 2—figure supplement 1M–O ) . Altogether , these results indicate that expression of Ftz-f1 in stage-10B–12 follicle cells is essential for follicles to mature and become competent to OA-induced follicle rupture and ovulation . We have recently demonstrated that follicle cells experience a novel transition from stage 13 to 14 by downregulation of Cut and Ttk69 and upregulation of Hnt , which promotes Oamb and Mmp2 expression and follicle maturation ( Deady et al . , 2017; Knapp et al . , 2019 ) . Analysis of Hnt expression in stage-14 follicles from ftz-f1RNAi females revealed a patchy expression of Hnt that overlaps with Oamb-RFP expression ( Figure 3—figure supplement 1A–B ) . In addition , Cut and Ttk69 were still detected in follicle cells without Oamb-RFP ( Figure 3—figure supplement 1C–F ) , consistent with the fact that Cut antagonizes Hnt expression in stage-14 follicle cells ( Knapp et al . , 2019 ) . The patchy nature of follicle cell markers is likely due to the incomplete knockdown of ftz-f1 using RNAi . All these data support the hypothesis that ftz-f1 is required for follicle cells to transition into the final maturation stage . To determine whether Ftz-f1 functions cell-autonomously in follicle cell differentiation , we generated ftz-f1 mutant follicle-cell clones . Consistent with our hypothesis , ftz-f1 mutant clones did not upregulate Hnt expression and continued to express Cut and Ttk69 in stage-14 follicles in a cell-autonomous fashion ( Figure 3A–C ) . In addition , EcR . A and EcR . B1 , two isoforms downregulated in wild-type stage-14 follicle cells , were still detected at the high level in ftz-f1 mutant follicle cells ( Figure 3D–E ) . Furthermore , we also found that another zinc-finger transcription factor Broad-Complex ( Br-C; DiBello et al . , 1991 ) was downregulated in wild-type follicle cells but remained high in ftz-f1 mutant follicle cells ( Figure 3F ) . Finally , ftz-f1 mutant follicle cells continue to have punctate EDU staining , while neighboring wild-type follicle cells have already ceased gene amplification in stage 14 ( Figure 3G ) . To determine which stages ftz-f1 mutant follicle cells were arrested in , we carefully examined Hnt and Cut expression in ftz-f1 mutant clones from stage 10B to stage 13 . Previous work showed that Hnt is undetectable at the end of stage 10B , while Cut is fully upregulated ( Sun et al . , 2008 ) . Indeed , we found that Hnt was downregulated in ftz-f1 mutant clones at stage 10B; however , Hnt expression was not fully diminished in ftz-f1 mutant clones at stage 10B or stage 12 ( Figure 3H–I ) . In addition , Cut expression was upregulated in ftz-f1 mutant clones at stage 10B , but it was not upregulated as high as that in neighboring wild-type follicle cells ( Figure 3J and Figure 3—figure supplement 2A ) . This difference was undetectable at stage 12 when Cut is reduced in wild-type follicle cells ( Figure 3K and Figure 3—figure supplement 2B–D ) . Altogether , these data suggest that ftz-f1 mutant follicle cells were arrested at the end of stage 10B . Therefore , ecdysone-induced Ftz-f1 functions cell-autonomously to promote follicle cell differentiation and progression into the final stages of maturation . To understand how Ftz-f1 promotes follicle cell differentiation in late oogenesis , we tried to identify the direct targets of Ftz-f1 . We first performed RNA-seq analysis using hand-dissected stage-10B–12 follicles from control and ftz-f1RNAi1 females with Vm26Aa-Gal4 . Principle component analysis clearly showed separation of control samples from ftz-f1RNAi1 samples ( Figure 4A ) . DE-seq analysis identified 197 downregulated genes and 192 upregulated genes that had more than two-fold change in expression level and adjusted p-value less than 0 . 01 ( Figure 4B and Supplementary file 1 ) . It is worth noting that neither hnt nor cut and ttk are among the differentially expressed genes . To profile the Ftz-f1-binding sites throughout the genome in follicle cells , we carried out CUT&RUN experiment , an assay utilizing transcription factor-specific antibody to bring micrococcal nuclease ( MNase ) to release transcription factor-bound short fragments in intact cells followed by next-generation sequencing ( Meers et al . , 2019; Skene and Henikoff , 2017 ) . We implemented the CUT&RUNTools workflow developed by Yuan’s group with minor modification ( see materials and methods; Zhu et al . , 2019 ) . With highly stringent criteria , we identified 520 , 943 , and 550 narrow peaks in three biological replicates , respectively . All three biological replicates showed similar peak distribution throughout the genome ( Figure 4—figure supplement 1A–C ) . Majority of the peaks are located within 3 kb of transcriptional start site ( TSS; Figure 4—figure supplement 1C ) , consistent with the idea that Ftz-f1 is a transcriptional regulator . Using MEME-chip ( Machanick and Bailey , 2011 ) , de novo motif search with sequences flanking the peak summit identified similar motifs ( CAAGGTCARV for replicate 1 , CAAGGTCR for replicate 2 , and DBTCAAGGTCA for replicate 3; Figure 4C ) , which are also similar to the canonical Ftz-f1 binding motif YCAAGGYCR ( Murata et al . , 1996; Ueda et al . , 1990 ) . Footprinting analysis for all three motifs showed the typical pattern of a high posterior probability of cut ( or cut-frequency ) in the motif flanking region and a low posterior probability of cut in the motif core ( Figure 4D ) , presumably due to the protection of transcription factor-bound DNA . In addition , all three motifs showed a symmetric motif footprint profile ( Figure 4D ) . Altogether , these data suggest that de novo-identified motifs are the true Ftz-f1-binding motifs . In total , we identified 166 , 505 , and 389 motif sites within the narrow peaks in each replicate , respectively ( Supplementary file 2 ) . The nearest gene/transcript associated with each motif site were also identified using ChiPseeker ( Yu et al . , 2015 ) and listed in Supplementary file 2 . To identify the direct target genes of Ftz-f1 in follicle cells , we set the following criteria: ( 1 ) the gene must be differentially expressed according to the RNA-seq analysis; and ( 2 ) the gene must contain a direct Ftz-f1 binding site , which is defined as a site containing overlapping Ftz-f1-binding motifs appeared in at least two of the three biological replicates and with a binding log-odds score >5 . The log-odds score is a binding probability score that quantifies the similarity between the cuts at each motif occurrence and the aggregate footprint pattern ( Zhu et al . , 2019 ) . With these criteria , we identified 15 genes/transcripts that were likely direct targets of Ftz-f1 ( Supplementary file 3 ) . Among these genes , 13 were downregulated genes and 2 were upregulated genes . Only two of the genes ( Eip74EF and sim ) encode transcription factors . Eip74EF ( Ecdysone-induced protein 74EF ) encodes a transcription factor that responds to different concentration of 20E during puparium formation ( Burtis et al . , 1990 ) , while sim ( single-minded ) encodes a bHLH/PAS-domain transcription factor in embryonic neuronal development ( Crews et al . , 1988; Nambu et al . , 1990; Thomas et al . , 1988 ) . To understand how Ftz-f1 promotes follicle cell differentiation in late oogenesis , we focused on the bHLH/PAS transcription factor Sim for the following reasons: 1 ) transcription factors will make profound changes during cell differentiation; 2 ) sim was identified in an ongoing genetic screen for Drosophila ovulatory genes; and 3 ) only one single peak containing Ftz-f1-binding site was clearly identified at the proximal promoter region ( −200 bp ) of one of sim’s transcripts ( FBtr0334613; Figure 4E ) . Most strikingly , FBtr0334613 was the only sim transcript expressed in stage-10B–12 follicles and was downregulated in ftz-f1–knockdown follicles , through reanalyzing the RNA-seq data using the HISAT-Stringtie ( Figure 4F ) . To test whether sim is indeed a downstream target of Ftz-f1 , we performed Sim antibody staining in wild-type follicles and follicles with ftz-f1 mutant clones . Sim was not expressed in follicle cells before stage 10B ( except in stalk follicle cells connecting two egg chambers; Figure 5A and Figure 5—figure supplement 1 ) . Sim was drastically upregulated in stage-10B/11 follicle cells ( except anterior stretch follicle cells ) and progressively downregulated to the lowest point at stage 13 ( Figure 5B–E ) . Sim was re-upregulated at stage 14 and its function at this stage will be reported in another manuscript ( Figure 5F ) . Consistent with the idea that sim is a downstream target of Ftz-f1 , ftz-f1 mutant follicle cells completely lack Sim expression at stage 10B and 12 ( Figure 5G–H ) . In contrast , ttk-knockdown follicle cells have normal expression of Sim ( Figure 5I ) . In addition , misexpression of ftz-f1 is sufficient to induce premature Sim expression in stage-10A follicle cells ( Figure 5—figure supplement 2A–D ) , which seemed to disrupt the follicle cell transition from stage 10A to stage 10B manifested by the continuous expression of Hnt and no upregulation of Cut at/after stage 10B ( Figure 5—figure supplement 2E–H ) . Altogether , these data suggest that Sim is a direct target of Ftz-f1 but not Ttk69 . To determine whether Sim is required for follicle cell differentiation , we generated flip-out Gal4 clones with overexpression of simRNAi . Follicle cells with simRNAi overexpression have no detectable Sim expression at stage 10B , 12 , or 14 ( Figure 1—figure supplement 3B–D ) , indicating efficient knockdown . Similar to the ftz-f1 mutant follicle cells , simRNAi-overexpressing follicle cells also failed to fully upregulate Hnt expression at stage 14 ( Figure 6A ) , as well as downregulate Cut , Ttk69 , EcR . A , EcR . B1 , and Br-C ( Figure 6B–F ) . In addition , occasional faint expression of Hnt was detected in sim-knockdown follicle cells at stage 10B and 12 ( Figure 6G–H ) , while the different level of Cut expression in sim-knockdown and adjacent wild-type follicle cells was detected at stage 10B but not at stage 12 ( Figure 6I–J ) , similar to ftz-f1 mutant follicle cells . The similarity between ftz-f1 mutant and sim-knockdown follicle cells is not due to Sim regulating Ftz-f1 expression , as Ftz-f1 is properly upregulated in sim-knockdown follicle cells at stage 10B ( Figure 6K ) . Our data suggest that Sim is essential for follicle cell differentiation in late oogenesis , like Ftz-f1 . We aimed to rescue differentiation defects of ftz-f1-knockdown follicle cells with misexpression of sim in the flip-out Gal4 system . Unfortunately , ectopic sim expression led to early follicle cell defects manifested by the smaller nuclei starting at stage 9 , continuous expression of Hnt , and no expression of Cut from 7 to stage 14 ( Figure 6—figure supplement 1A–E ) . Alternatively , we tested the ability of ectopic sim to rescue ftz-f1 knockdown defects when driven by Vm26Aa-Gal4 . However , ectopic expression of sim alone or in the ftz-f1-knockdown background led to disrupted Hnt and Cut expression patterns at stage 10B/11 ( Figure 6—figure supplement 2A–H ) . These follicles showed mild rescue ( if any ) of Hnt , Cut , and Oamb-RFP expression at stage-14 , but had abnormal morphology and no dorsal appendage formation as ftz-f1-knockdown follicles ( Figure 6—figure supplement 2I–P ) . These data likely suggest that the level and temporal expression of Sim is essential for proper follicle cell differetiation . Nonetheless , the phenotypic similarity between ftz-f1 and sim mutant follicle cells and the induction of sim expression by Ftz-f1 support the idea that Sim acts downstream of Ftz-f1 to promote follicle cell differentiation . Next , we examined whether ectopic expression of ftz-f1 is sufficient to rescue ftz-f1RNAi defects . As expected , flip-out Gal4 clones with both ftz-f1 and ftz-f1RNAi2 showed rescue of Ftz-f1 expression in stage-10B follicle cells , despite it is slightly weaker than that in wild-type follicle cells ( Figure 7—figure supplement 1A–B ) . This is likely due to ftz-f1RNAi targeting not only endogenous ftz-f1 gene but also ectopically expressed ftz-f1 mRNA . We also observed complete rescue of Hnt and Cut expression ( Figure 7—figure supplement 1E–H ) . Unlike ftz-f1 overexpression alone ( Figure 5—figure supplement 2C ) , we did not observe premature induction of Sim ( Figure 7—figure supplement 1C–D ) , since Ftz-f1 was not overexpressed in early stages ( Figure 7—figure supplement 1A ) . To determine whether Ftz-f1’s role in follicle cell differentiation is conserved , we investigated the potential of mouse SF-1 ( mSF-1 ) , the mouse homolog of Ftz-f1 , to substitute for Ftz-f1 in follicle cell maturation . We generated flip-out Gal4 clones that express either ftz-f1RNAi2 , mSF-1 , or both and examined follicle cell maturation markers . Consistent with ftz-f1 mutant follicle cells ( Figure 3 ) , ftz-f1RNAi2-overexpressing follicle cells could not upregulate Hnt expression at stage 14 ( Figure 7A ) . In contrast , follicle cells with both ftz-f1RNAi2 and mSF-1 had normal Hnt upregulation at stage 14 , the same as follicle cells with mSF-1 alone ( Figure 7B–D ) . In addition , follicle cells with ftz-f1RNAi2 showed strong Cut expression at stage 14 , while follicle cells with both ftz-f1RNAi2 and mSF-1 had no Cut expression , similar to follicle cells with mSF-1 alone ( Figure 7E–H ) . These data suggest that mSF-1 can replace Ftz-f1’s role in promoting follicle cell differentiation and maturation . Strikingly , we also noticed that ectopic mSF-1 was sufficient to promote premature differentiation of follicle cells . In wild-type follicle cells , Hnt expression was not downregulated until stage 10B; however , Hnt was prematurely downregulated in follicle cells with both mSF-1 and ftz-f1RNAi2 at stage 10A but not in earlier stages ( Figure 7I–J ) . In addition , Hnt was not re-upregulated until stage 14 in wild-type follicle cells but was prematurely upregulated in follicle cells with both mSF-1 and ftz-f1RNAi2 at stages 12/13 ( Figure 7K–L ) . In accordance with Hnt , Cut was prematurely upregulated in follicle cells with both mSF-1 and ftz-f1RNAi2 at stage 10A and prematurely downregulated at stage 12/13 ( Figure 7M–P ) . We consistently observed cytoplasmic staining of Cut in the clone cells , indicating that Cut was evicted from follicle cell nuclei for degradation ( Figure 7O–P ) . These data indicate that overexpression of mSF-1 is sufficient to promote follicle cell differentiation prematurely . The rescue of follicle cell maturation by mSF-1 prompted us to examine whether mSF-1 is also sufficient to restore Sim expression in ftz-f1–knockdown follicle cells . Like ftz-f1 mutant clones ( Figure 5G–H ) , Sim is barely detected in follicle cells with ftz-f1RNAi2 overexpression at stage 14; however , it is readily detected in follicle cells with both ftz-f1RNAi2 and mSF-1 or mSF-1 alone ( Figure 8A–D ) . Most strikingly , ectopic mSF-1 was able to prematurely induce Sim expression in follicle cells with ftz-f1RNAi2 at stage 10A but not earlier stages ( Figure 8E–F ) . In addition , Sim was also prematurely downregulated in these follicle cells at stage 12 ( Figure 8G–H ) . All these data are consistent with the idea that ectopic mSF-1 promotes the premature differentiation of follicle cells via Sim . In conclusion , our data suggest that ecdysone-induced Ftz-f1 promotes follicle cell differentiation and maturation partly via bHLH/PAS transcription factor Sim , and this role is likely conserved ( Figure 8I ) . Since the identification of the ftz-f1 gene almost three decades ago ( Lavorgna et al . , 1991; Ueda et al . , 1990 ) , previous work has primarily focused on Ftz-f1’s role in embryogenesis , larval development , pupation , and metamorphosis . The expression and function of Ftz-f1 in adult flies , particularly in oogenesis , is largely lacking . Work in this study demonstrated for the first time that Ftz-f1 is transiently expressed in the adult ovarian follicle cells from stage 10B to stage 12 according to three different reporters . It is worth noting that we were unable to detect Ftz-f1 expression in follicle cells before stage 10B , unlike the work reported previously ( Talamillo et al . , 2013 ) . In addition , we didn’t observe any morphological and molecular defects in ftz-f1 mutant follicle cells before stage 10 ( data not shown ) . Ftz-f1 antibody used in this study is raised against βFtz-f1 protein; however , it can potentially detect αFtz-f1 since αFtz-f1 and βFtz-f1 share common C-terminal regions ( personal communication with Dr . Ueda ) . Therefore , it is unknown whether follicular Ftz-f1 is αFtz-f1 or βFtz-f1 . Since αFtz-f1 is maternally supplied and only detected in early embryos , we favor the idea that it is βFtz-f1 expressed in follicle cells . This is consistent with the fact that follicular Ftz-f1 is regulated by ecdysteroid signaling , similar to the transient expression of βFtz-f1 after each ecdysone pulses during larval and pupal development ( Yamada et al . , 2000 ) . It is striking that follicular Ftz-f1 is so transiently expressed , similar to transient expression of βFtz-f1 in development . Our data showed that ecdysteroid signaling is essential for Ftz-f1 expression at stage 10B . It seems contradictory to the fact that βFtz-f1 is inhibited by high ecdysone titer and only induced when ecdysone titer is low during development ( Broadus et al . , 1999; Woodard et al . , 1994; Yamada et al . , 2000 ) . However , there’s no precise measurement of ecdysone titer at each stage of oogenesis . It is plausible that ecdysone signaling at stage 10A leads to sequential activation of genes that are responsible for Ftz-f1 expression at stage 10B . Consistent with this idea , Cyp18a1 , encoding a cytochrome P450 enzyme involved in inactivating 20-hydroxyecdysone and inducing Ftz-f1 expression during the prepupa-to-pupa transition ( Rewitz et al . , 2010 ) , is significantly enriched in stage-10B follicles and likely required for follicle cell differentiation ( Tootle et al . , 2011 ) . Unfortunately , either overexpression or knockdown of Cyp18a1 did not affect Ftz-f1 expression in follicle cells . In addition , exogenous 20E is also not sufficient to induce Ftz-f1 expression in earlier stages . Thus , Ftz-f1 expression is precisely regulated in follicle cells and is not sensitive to the 20E level . It will be interesting to investigate whether other ecdysone-induced genes that regulate βFtz-f1 expression during the larva-to-pupa transition , such as Blimp-1 , DHR3 , E75 , and Nos ( Akagi et al . , 2016; Cáceres et al . , 2011; Yamanaka and O'Connor , 2011 ) , contribute to precise upregulation of Ftz-f1 in stage-10B follicle cells . It is unknown what factors contribute to downregulation of Ftz-f1 at stage 12 . It is worth noting that several Ftz-f1-binding sites were identified in the ftz-f1 gene ( Supplementary file 2 ) and that βFtz-f1 can negatively regulates its own expression during prepupa-to-pupa transition ( Woodard et al . , 1994 ) . A similar negative-feedback mechanism could occur in follicle cells . Previous work regarding follicular epithelium mostly focused on egg chambers ––before stage 10 , at the stage 10A/10B transition , or at the stage 13/14 transition ( Duhart et al . , 2017; Klusza and Deng , 2011; Knapp et al . , 2019; Osterfield et al . , 2017 ) . Little is known about how stage-10B follicle cells differentiate into final maturation . With both global knockdown and mutant clone analyses , our work clearly demonstrated that Ftz-f1 is a key factor required for promoting stage-10B follicle cells to differentiate into final maturation , which is essential for releasing fertilizable oocytes at the end of oogenesis . Molecular marker analysis showed that all known stage-14 follicle cell markers , including upregulated Hnt , Oamb , Mmp2 expression and downregulated Cut , Ttk69 , Br-C , EcR . A/B1 expression ( Figure 8I ) , are disrupted in ftz-f1 mutant follicle cells . In fact , ftz-f1 mutant follicle cells seem to be arrested at the end of stage 10B . All these data suggest that Ftz-f1 is a master regulator for the final differentiation of follicle cells after stage 10B . Consistent with this idea , loss of ftz-f1 also led to disrupted dorsal appendage formation and chorion gene amplification , and likely eggshell formation . It is not completely understood how Ftz-f1 can have such profound influence on cell differentiation . During the larva-to-pupa transition , Ftz-f1 seems to regulate ecdysteroid synthesis enzymes and thus influence the next ecdysone pulse ( Akagi et al . , 2016; Parvy et al . , 2005 ) . Could the same mechanism apply in follicle cells ? Indeed , we have previously demonstrated that another pulse of ecdysteroid signaling occurs in stage-14 follicle cells in addition to the ecdysteroid signaling at the stage 10A/10B transition ( Knapp and Sun , 2017 ) . This is controlled by the upregulation of Shade ( Shd ) , the enzyme converting ecdysone to active 20-hydroxyecdysone . However , preliminary analysis showed that Shd is continuously expressed in ftz-f1 mutant follicle cells ( data not shown ) . Therefore , Ftz-f1 is unlikely to regulate follicle cell differentiation through regulating the next pulse of ecdysteroid signaling . This is also supported by the fact that Ftz-f1 promotes follicle cell differentiation in a cell-autonomous fashion and that Sim functions as a downstream target to promote follicle cell differentiation ( see below ) . Few direct targets of Ftz-f1 have been identified . Among those , ftz , Edg84A , and Adh are best characterized , and all of them have Ftz-f1 binding motif ( YCAAGGYCR ) in the promoter region within 500 bp upstream of TSS ( Ayer and Benyajati , 1992; Murata et al . , 1996; Ueda et al . , 1990 ) . With RNA-seq , we identified 389 differentially expressed genes in follicles with ftz-f1 knockdown . GO term analysis showed that genes related to intrinsic and integral components of membrane are most enriched among downregulated genes , while genes related to secondary active transmembrane transporter activity , developmental process , and chorion are most enriched among upregulated genes ( Supplementary file 1 ) . Using CUT&RUN experiment , we identified Ftz-f1 binding motifs in follicle cells that were similar to the canonical Ftz-f1 binding motif ( YCAAGGYCR ) . More than 250 sites could be potential Ftz-f1 direct binding sites ( Supplementary file 2 ) . Combining both experiments , we were able to identify 15 genes/transcripts that could be potential direct targets of Ftz-f1 in follicle cells . Among these , our data illustrated that one of sim’s transcript ( FBtr0334613 ) is the only transcript expressed in follicle cells and is the direct target of Ftz-f1 ( Figure 4 ) . This is also supported by our finding that sim3 . 7-Gal4 , which utilizes a 3 . 7 kb promoter region of sim’s longest transcript ( FBtr0082711 ) that does not contain Ftz-f1-binding site ( Xiao et al . , 1996 ) , was not detected in follicle cells ( data not shown ) . In the future , it will be interesting to isolate the entire promoter region that is required for sim expression in follicle cells and identify all the factors regulating its expression . In addition , future work will be focused on the other direct targets of Ftz-f1 to better understand the molecular network of Ftz-f1 regulated follicle cell differentiation and maturation . Sim is a master regulator of central nervous system ( CNS ) midline cell development and has been extensively studied in the development of the CNS midline , the central complex , and optic ganglia in the last two decades ( Nambu et al . , 1991; Pielage et al . , 2002; Umetsu et al . , 2006 ) . Its role outside the nervous system is sparse . Our findings here also illustrated for the first time that Sim is upregulated in stage-10B follicle cells and is essential for follicle cell differentiation . This is consistent with a previous report that sim mutant flies are sterile ( Pielage et al . , 2002 ) . We also demonstrated that Sim upregulation depends on Ftz-f1 , not vice versa , which places Sim downstream of Ftz-f1 . In addition , phenotypic defects of sim-knockdown follicle cells are strikingly similar to those of ftz-f1 mutant follicle cells . Furthermore , mSF-1 overexpression leads to premature Sim upregulation at stage 10A as well as premature follicle cell differentiation . All these data support the conclusion that Sim function as the downstream effector of Ftz-f1 to promote follicle cell differentiation . Sim belongs to the bHLH/PAS transcription factor family and dimerizes with another bHLH-PAS transcription factor Tango to activate downstream gene expression ( Ohshiro and Saigo , 1997; Sonnenfeld et al . , 1997 ) . It will be interesting to investigate whether Tango is a cofactor for Sim in follicle cells and what are the direct targets of Sim in follicle cell differentiation in the future . It will be also interesting to know whether Sim also acts downstream of Ftz-f1 during larval and pupal development . Our work also illustrated the importance of precise control of Sim expression in follicle cells . Ectopic sim expression in early-stage follicle cells seemed to disrupt the organization of the follicle-cell monolayer ( Figure 6—figure supplement 1 ) . It also disrupts the endoreplication as follicle cell nuclei are smaller than the adjacent wild-type cells . This is not due to the disruption of Notch signaling and mitotic/endocycle transition ( Sun and Deng , 2005; Sun and Deng , 2007 ) , because Cut is properly downregulated in these cells and the nuclei size defect is only manifested after stage 8 . Therefore , premature upregulation of Sim may also disrupt the cell differentiation program . In addition , Sim is also expressed in stalk follicle cells and its role in stalk follicle cells is completely unknown . The mammalian NR5A homolog SF-1 , is expressed in somatic follicle cells of the ovary in both rodents and humans ( Hinshelwood et al . , 2003; Tajima et al . , 2003 ) , and loss of this SF-1 expression in murine granulosa cells leads to a severe depletion of developing follicles and infertility ( Pelusi et al . , 2008 ) . Despite the critical role for SF-1 in female fertility , it still remains unknown how SF-1 within these follicle cells regulates folliculogenesis . Drosophila poses as a valuable model for the study of the function of NR5A receptors , considering the DNA binding sequence of NR5A receptors is highly conserved from Drosophila to humans , with over 80% in sequence similarity ( Fayard et al . , 2004 ) . Furthermore , studies have already begun to show the functional conservation of these NR5A receptors in both the embryo and female reproductive tract of Drosophila ( Lu et al . , 2013; Sun and Spradling , 2012; Splinter et al . , 2018 ) . In this work , we demonstrated that Ftz-f1 is also expressed in the somatic follicle cells of the ovary and plays a crucial role in female fertility , akin to SF-1 . Furthermore , our work demonstrated that Ftz-f1’s function in follicle cell differentiation is functionally conserved , as mSF-1 is sufficient to rescue defects in follicle cell maturation caused by loss of Ftz-f1 . Our results also showed that mSF-1 is sufficient to induce expression of the Ftz-f1 target Sim . The mammalian homologs of Sim are encoded by sim1 and sim2 ( Yamaki et al . , 1996 ) . The role of Sim1 and Sim2 in female fertility have never been studied; however , Sim2 seems to be expressed in human ovarian follicle cells ( according to Human Protein Atlas ) . Thus , it would be interesting to probe if Sim1 or Sim2 is expressed in ovarian follicle cells and whether they function downstream of SF-1 for follicle development . Overall , our findings could help to further elucidate the genetic and molecular mechanisms of NR5A signaling and how it regulates follicle development and female fertility . Flies were reared on standard cornmeal and molasses food at 25°C , unless noted otherwise . ftz-f1ex7 is a P-element excision line and is considered as a null allele ( Yamada et al . , 2000 ) . For ftz-f1 expression analysis , ftz-f1::GFP . FLAG [Bloomington Drosophila Stock Center ( BDSC ) , stock #38645] and ftz-f1fs ( 3 ) 2877 ( Karpen and Spradling , 1992 ) were utilized . The protein trap line Mmp2::GFP/Cyo ( Deady et al . , 2015 ) was used for Mmp2 expression . The Vm26Aa-Gal4 ( Peters et al . , 2013 ) was used to drive expression in follicle cells starting at stage 10 . Isolation and identification of stage-14 follicles for follicle rupture assay were performed using Oamb-RFP ( Knapp et al . , 2019 ) or 47A04-LexA ( BDSC , stock #54873 ) driving lexAop2-6XGFP ( BDSC , stock #52265 ) . sim3 . 7-Gal4 ( Xiao et al . , 1996 ) was also from BDSC ( stock #26784 ) . The following transgenic lines were used to knock down or overexpress genes in experiments: UAS-EcRDN ( BDSC , stock #6872 ) , UAS-ttkRNAi [Vienna Drosophila Resource Center ( VDRC ) , stock #101980] , UAS-Cyp18a1 ( Rewitz et al . , 2010 ) , UAS-Cyp18a1RNAi ( VDRC , stock #5602 ) , UAS-ftz-f1RNAi1 ( BDSC , stock #33625 ) , UAS- ftz-f1RNAi2 ( VDRC , stock #104463 ) , UAS-ftz-f1 ( Yussa et al . , 2001 ) , UAS-simRNAi ( VDRC , stock #26888 ) , UAS-sim-3xHA ( Fly-ORF , stock #000719 ) and UAS-mSF1 ( Yussa et al . , 2001 ) . Ecdysone sensor hsGal4DBD-EcRLBD , UAS-nlacZ was a gift by Wu-Min Deng ( Kozlova and Thummel , 2002 ) . All experiments involving RNAi lines are performed at 29°C and contain UAS-dcr2 in order to enhance the RNAi efficiency . Control flies for all experiments were prepared by crossing Gal4 driver to Oregon-R flies . Mosaic analysis with repressible cell marker ( MARCM ) was used to generate follicle cell clones homozygous for the ftz-f1ex7 allele , via crossing hsFLP , tub-Gal4 , UAS-GFP; tub-Gal80 , FRT2A/TM6B to ftz-f1ex7 , FRT 79D/TM3 , Ser . Flip-out Gal4 clones were generated using either the hsFLP; act <CD2<Gal4 , UAS-GFP/TM3 or hsFLP; act <CD2<Gal4 , UAS-RFP/TM3 stock to cross to indicated transgenes of interest . For clone induction , adult female progeny with correct genotypes were heat shocked for 45 min at 37°C to induce FLP/FRT mediated recombination and incubated at 25°C for 2–4 days before dissection . For analysis of EcR ligand sensor , flies were heat shocked for 45 min at 37°C and allowed to recover at 29°C for 16 hr before dissection . Dissected ovaries were treated with 100 nM of 20E ( Cayman Chemical ) in Grace’s medium for five hours before fixation and antibody staining . Egg-laying experiments were performed as previously described ( Deady and Sun , 2015 ) . Five-day-old females ( fed with wet yeast for 1 day ) were housed with Oregon-R males ( five females: 10 males ) in one bottle to lay eggs on molasses plates over two days at 29°C ( with removal and replacement of plates every 22 hr ) . After egg laying , the ovary pairs for each female were dissected out and the number of mature follicles within the ovary pair were quantified . The ex vivo follicle rupture assays were performed as described previously ( Knapp et al . , 2018 ) . Ovaries from 5- to 6-day-old virgin females fed with wet yeast for 3 days were dissected out and stage-14 follicles were isolated in Grace’s insect medium ( Caisson Labs , Smithfield , UT ) . After isolation , follicles were separated into groups ~ 30 follicles and cultured at 29°C for 3 hr in culture medium ( Grace’s insect medium +10% fetal bovine serum + 1% penicillin-streptomycin ) containing 20 μM OA ( Sigma-Aldrich ) or 2 μM ionomycin ( Cayman Chemical , Ann Arbor , MI ) . Each data point represents the percentage ( mean ± standard deviation ( SD ) ) of ruptured follicles per experimental group . Measurement of superoxide production was performed as previously described ( Li et al . , 2018 ) , with slight modifications . Five mature follicles were isolated and placed in each well of a 96-well plate with 100 μl of Grace’s insect medium containing either 20 μM OA or 2 μM ionomycin and 200 μM of L-012 ( Wako Chemicals ) . Plates were placed in a CLARIOstar microplate reader ( BMG Labtech ) for luminescence reading for 60 min . Eight to ten wells ( technical repeats ) were used in each experiment for each genotype , and the mean ± standard error of the mean ( SEM ) of the technical repeats was calculated . Each experiment was performed at least twice . Immunostaining was performed following a standard procedure ( Sun and Spradling , 2012 ) . The following primary antibodies were used: mouse anti-Hnt ( 1G9 , 1:75 ) , anti-Cut ( 2B10 , 1:15 ) , anti-Br-C ( 25E9 . D7 , 1:15 ) , anti-EcR . A ( 15G1a , 1:30 ) , and anti-EcR . B1 ( AD4 . 4 , 1:30 ) from the Developmental Study Hybridoma Bank; rabbit anti-Ftz-f1 ( 1:50000; a gift from Dr . Hitoshi Ueda , Okayama University , Japan ) , rabbit anti-Ttk69 ( 1:100; a gift from Dr . Wanzhong Ge , Zhejiang University , China ) , rabbit anti-GFP ( 1:4000; Invitrogen ) , mouse anti-GFP ( 1:1000; Invitrogen ) , rabbit anti-RFP ( 1:2000 , MBL international ) , Chicken anti-β-Gal ( ab9361 , 1:500; Abcam ) , and guinea pig anti-Sim ( 1:1000; a gift from Dr . Stephen Crews , University of North Carolina at Chapel Hill School of Medicine , Chapel Hill , USA ) . Alexa Fluor 488 and Alexa Flour 568 goat secondary antibody ( 1:1000; Invitrogen ) were used as secondary antibodies . EdU detection was performed as previously described ( Alexander et al . , 2015 ) . Ovaries were dissected out in room temperature Grace’s insect medium and incubated in 50 μM EdU for 30 min . Ovaries were then fixed in 4% EM-grade paraformaldehyde for 13 min and permeabilized in PBX ( 0 . 1% TritonX in PBS ) for 30 min . For detection of EdU , the Invitrogen’s Click-iT EdU Alexa Fluor 555 Imaging Kit was utilized following the manufacturer’s instructions . Images were acquired using a Leica TCS SP8 confocal microscope or Leica MZ10F fluorescent stereoscope with a sCOMS camera ( PCO . Edge ) and assembled using Photoshop software ( Adobe ) and ImageJ . Around 60 stage-10B–12 egg chambers from 7 to 10 flies were isolated in Grace’s medium ( Caisson labs ) and grounded in 300 µl of TRIzol ( Life Technologies , 15596018 ) directly . Total RNAs were extracted using Direct-zol RNA MicroPrep Kit ( Zymo Research , Irvine , CA ) . mRNA libraries were prepared using Illumina TruSeq Stranded mRNA Sample Preparation kit following the manufacturer’s protocol ( Illumina , San Diego , CA ) and were then sequenced on an Illumina NextSeq 550 sequencer to achieve single-end 75 bp reads in UConn’s Center for Genome Innovation . Three biological replicates were prepared for each genotype . Raw reads from RNA-seq were trimmed with Sickle ( -q 30 l 50 ) . Trimmed reads were mapped to Drosophila melanogaster genome ( dm6 ) with HISAT2 ( Kim et al . , 2015 ) . The counts were generated against the features with HTSeq-count ( Anders et al . , 2015 ) . Principal component analysis ( PCA ) was used to test the reproducibility between the replicates . One ftz-f1-knockdown sample was an outlier due to unknown reason and was dropped from the analysis . The differential expression of genes between conditions were evaluated using DESeq2 ( Love et al . , 2014 ) . In DESeq2 , genes showing less than 10 cumulative counts across the compared samples were dropped from the analysis . Genes with ( a ) base mean counts >10 , ( b ) a False discovery Rate ( FDR ) < 0 . 01 , and ( c ) absolute value of log2FoldChange > 1 were considered to be significant and used in the downstream analysis . For transcript level expression , HISAT , Stringtie and Ballgown method was used ( Pertea et al . , 2016 ) . Stringtie was used to estimate FPKM for each transcript . The sample preparation for CUT&RUN followed the previous protocol with slight modification ( Skene et al . , 2018 ) . In short , approximately 200 stage-10B–13 egg chambers from ~15 ftz-f1::GFP . FLAG females were isolated in 1xPBS . These egg chambers were equally separated into two halves , quickly spun and washed three times with wash buffer ( 20 mM HEPES-NaOH pH 7 . 5 , 150 mM NaCl , 0 . 5 mM Spermidine , with 1x protease inhibitor EDTA free ) , and incubated in primary antibody at 4°C overnight . Samples were washed twice in dig-wash buffer and incubated for 1 hr at 4°C with protein-AG MNase ( 1:800 ) expressed and purified in house with the plasmid from Addgene ( #123461 ) . For chromatin digestion and release , high Ca2+/low salt option was chosen and performed as in Meers et al . , 2019 . For library preparation , NEBNext Ultra II DNA Library Prep Kit ( NEB ) was performed as described in Liu et al . , 2018 . For amplification , after the addition of indexes , 14 cycles of 98°C , 20 s; 65°C , 10 s were run . A 1 . 2x SPRI bead cleanup was performed ( Agencourt Ampure XP , Beckman ) . Libraries were sequenced on an Illumina NextSeq 500 sequencer to achieve pair-end 75 bp reads . The following primary antibody were used: mouse anti-FLAG M2 ( 1:250; Sigma F1804; experimental antibody ) and mouse IgG1 ( 1:125; Sigma MABC002 , control antibody ) . Three biological replicates were performed for each experimental antibody and control antibody . For the data analysis , we followed the CUT&RUNTools workflow with the following modification ( Zhu et al . , 2019 ) . In short , trimmed pair-end reads were mapped to Drosophila melanogaster genome ( dm6 ) using Bowtie2 ( option --dovetail --local --very-sensitive-local --no-unal --no-mixed --no-discordant ) ( Langmead and Salzberg , 2012 ) . Fragments < 120 bp from experimental and control samples were used in MACS2 ( Zhang et al . , 2008 ) for identifying the narrow peaks ( macs2 callpeak -t experiment . bam -c control . bam -g dm -f BAMPE -n outprefix --outdir outdir -q 0 . 01 -B --SPMR --keep-dup all ) . de novo motif search and motif footprint analysis were exactly followed in Zhu et al . , 2019 . Chipseeker was used to analyze the peak distribution and motif sites relevant to nearest genes ( Yu et al . , 2015 ) . All sequencing data are deposited in NCBI Sequence Read Archive ( SRA ) with BioProject ID PRJNA624186 . Statistical tests were performed using Prism 7 ( GraphPad , San Diego , CA ) . Quantification results are presented as mean ± SD or mean ± SEM as indicated . Statistical analysis was conducted using Student’s t-test .	When animals reproduce , females release eggs from their ovaries which then get fertilized by sperm from males . Each egg needs to properly mature within a collection of cells known as follicle cells before it can be let go . As the egg matures , so do the follicle cells surrounding it , until both are primed and ready to discharge the egg from the ovary . Mammals rely on a protein called SF-1 to mature their follicle cells , but it is unclear how this process works . Most animals – from humans to fruit flies – release their eggs in a very similar way , using many of the same proteins and genes . For example , the gene for SF-1 in mammals is similar to a gene in fruit flies which codes for another protein called Ftz-f1 . Since it is more straightforward to study ovaries in fruit flies than in humans and other mammals , investigating this protein could shed light on how follicle cells mature . However , it remained unclear whether Ftz-f1 plays a similar role to its mammalian counterpart . Here , Knapp et al . show that Ftz-f1 is present in the follicle cells of fruit flies and is required for them to properly mature . Ftz-f1 controlled this process by regulating the activity of another protein called Sim . Further experiments found that the gene that codes for the SF-1 protein in mice was able to compensate for the loss of Ftz-f1 and drive follicle cells to mature . Studying how ovaries release eggs is an essential part of understanding female fertility . This work highlights the similarities between these processes in mammals and fruit flies and may help us understand how ovaries work in humans and other mammals . In the future , the findings of Knapp et al . may lead to new therapies for infertility in females and other disorders that affect ovaries .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "genetics", "and", "genomics" ]	2020	Nuclear receptor Ftz-f1 promotes follicle maturation and ovulation partly via bHLH/PAS transcription factor Sim
Innate immune recognition is critical for the induction of adaptive immune responses; however the underlying mechanisms remain incompletely understood . In this study , we demonstrate that T cell-specific deletion of the IL-6 receptor α chain ( IL-6Rα ) results in impaired Th1 and Th17 T cell responses in vivo , and a defect in Tfh function . Depletion of Tregs in these mice rescued the Th1 but not the Th17 response . Our data suggest that IL-6 signaling in effector T cells is required to overcome Treg-mediated suppression in vivo . We show that IL-6 cooperates with IL-1β to block the suppressive effect of Tregs on CD4+ T cells , at least in part by controlling their responsiveness to IL-2 . In addition , although IL-6Rα-deficient T cells mount normal primary Th1 responses in the absence of Tregs , they fail to mature into functional memory cells , demonstrating a key role for IL-6 in CD4+ T cell memory formation . Innate immune recognition is mediated by a variety of pattern recognition receptors ( PRRs ) , including Toll-like receptors ( TLRs ) ( Iwasaki and Medzhitov , 2004 ) , NOD-like receptors ( NLRs ) ( Martinon et al . , 2009 ) , C-type lectin receptors ( CLRs ) ( Geijtenbeek and Gringhuis , 2009 ) , and retinoic acid inducible gene-I ( RIG-I ) -like receptors ( RLRs ) ( Pichlmair and Reis e Sousa , 2007 ) , all of which recognize conserved molecular structures specific to microbes ( pathogen-associated molecular patterns , or PAMPs ) and activate adaptive immune responses through the induction of DC maturation . The DC maturation process involves a redistribution of major histocompatibility complex ( MHC ) molecules from intracellular endocytic compartments to the cell surface , increased expression of costimulatory molecules , and secretion of inflammatory cytokines and chemokines , which is essential for the activation of naive T cells ( Banchereau and Steinman , 1998 ) . Previously , we demonstrated that TLR-induced DC maturation and migration to the draining lymph nodes , in the absence of TLR-induced inflammatory cytokines , is insufficient for the induction of T cell activation ( Pasare and Medzhitov , 2004 ) . T cell responses are controlled by naturally occurring CD4+ CD25+ regulatory T cells ( Tregs ) , which play an important role in maintaining immune tolerance and homeostasis . Accordingly , the absence of Foxp3 , the lineage-defining transcription factor that is critical for the generation and function of Tregs results in the development of a fatal autoimmune disease ( Fontenot and Rudensky , 2005; Sakaguchi , 2005 ) . We previously found that cytokines produced by DCs following TLR activation are critical for releasing responder CD4+ T cells from suppression by Tregs ( Pasare and Medzhitov , 2003 ) . Based on in vitro experiments as well as the analysis of complete IL-6 KO mice , we implicated IL-6 as a mediator for this block of suppressor activity ( Pasare and Medzhitov , 2003 ) . However , the pleiotropic nature of IL-6 has made it difficult to assess the T cell-specifc function of this cytokine in vivo . IL-6 has been previously described as a B cell growth factor , an initiator of acute phase responses , and a mediator of T cell survival ( Kamimura et al . , 2003 ) . Recently , IL-6 has also been implicated in the differentiation of specific CD4+ T cell subsets . IL-6 has been demonstrated to be an important cytokine that governs the differentiation of Th17 cells . In particular , it is thought to regulate the switch between the induction of Foxp3+ Tregs and Th17 cells ( Bettelli et al . , 2006; Veldhoen et al . , 2006 ) . Stimulating T cells with TGF-β results in the induction of Foxp3 , whereas combining TGF-β with IL-6 represses Foxp3 expression and induces retinoid-related orphan receptor ( ROR ) -γt , resulting in the differentiation of Th17 cells ( Ivanov et al . , 2006 ) . IL-6 has also been suggested to be essential for the differentiation of the T follicular helper ( Tfh ) cell lineage , which is defined by expression of the markers CXCR5 and programmed death receptor-1 ( PD-1 ) . Treating T cells with IL-6 leads to the upregulation of the transcriptional repressor B cell lymphoma 6 ( Bcl-6 ) , a process that is thought to drive Tfh cell generation ( Johnston et al . , 2009; Nurieva et al . , 2009 ) . Tfh cells specialize in providing help to germinal center ( GC ) B cells ( Vinuesa et al . , 2005; King et al . , 2008 ) . The IL-6 receptor complex consists of the ligand-binding subunit , the IL-6Rα chain , and the signal-transducing subunit , gp130 ( Taga et al . , 1989 ) , which is a common signaling transducer for several cytokines including IL-6 , IL-11 , leukemia inhibitory factor ( LIF ) , oncostatin M ( OSM ) , cardiotrophin-1 ( CT-1 ) , and IL-35 . Homodimerization of gp130 upon IL-6 binding results in the activation of the gp130-associated tyrosine kinases JAK1 , JAK2 and TYK2 and the subsequent phosphorylation of the transcription factors STAT1 and STAT3 ( Kamimura et al . , 2003 ) . In addition to the membrane-bound receptor , a soluble form of the IL-6Rα chain can be generated through either alternative splicing or proteolytic cleavage and is capable of binding to IL-6 and activating cells through association with gp130 ( Kamimura et al . , 2003 ) . Although both IL-6-deficient mice and mice bearing a T cell-specific deletion of gp130 have been used to examine the in vivo function of IL-6 in T cell responses , the pleiotropic nature of IL-6 , as well as the promiscuous use of the gp130 signaling subunit , has complicated these analyses ( Kamimura et al . , 2003 ) . Thus , the T cell-specific role of IL-6 signaling in vivo has remained poorly understood . In the present study , we therefore used mice carrying a T cell-specific ablation of the IL-6Rα chain , in order to investigate the role of IL-6 in the induction of CD4+ T cell responses following TLR activation . Our data not only confirm an important function for IL-6 in Th17 cell differentiation , but also reveal a critical role for IL-6 in the induction of the Th1 cell response in vivo . IL-6 enables T cell activation by acting on responder CD4+ T cells to make them less sensitive to the suppressive activity of Tregs , in part by blocking Treg-mediated inhibition of IL-2Rα expression in responder CD4+ T cells in cooperation with IL-1β . Although not absolutely required for the generation of Tfh cells , we also found that IL-6 signaling is important for the ability of these cells to provide help to B cells . In addition , we reveal a role for IL-6 in the generation of functional memory CD4+ T cells . To examine the function of IL-6 in CD4+ T cell responses in vivo , we generated mice in which T cells were specifically deficient of the IL-6Rα by crossing a floxed allele of Il6ra with mice expressing the Cre recombinase under the control of the CD4 promoter ( hereafter called IL-6RαT-KO mice ) . Since the Cre-encoding transgene is expressed at the double positive stage in thymic development , both CD4+ and CD8+ T cells in the periphery of IL-6RαT-KO mice failed to express the IL-6Rα ( Figure 1A ) . Importantly , both CD4+ and CD8+ T cells from IL-6RαT-KO mice remained deficient of the IL-6Rα after immunization with Ovalbumin ( OVA ) and LPS emulsified in Incomplete Freund's Adjuvant ( IFA ) as a carrier , suggesting that the release of the soluble form of the IL-6Rα during the immune response does not restore IL-6 signaling in these cells ( Figure 1A ) . Furthermore , IL-6-induced STAT3 phosphorylation was blocked in IL-6Rα-deficient CD4+ and CD8+ T cells compared to control wild-type ( WT ) T cells ( Figure 1B ) . To evaluate whether deficiency of the IL-6Rα on CD4+ T cells compromised the gp130-dependent signaling axis , we stimulated CD4+ T cells in vitro with α-CD3e and α-CD28 mAbs in the presence of gp130-dependent cytokines and measured the phosphorylation of STAT3 1 hr later by Western blot . Addition of IL-6 to the cells phosphorylated STAT3 very effectively in WT cells but not in IL-6Rα-deficient cells , thus confirming the results obtained by flow cytometry ( Figure 1—figure supplement 1 ) . Importantly , the addition of the soluble form of the IL-6Rα ( sIL6Rα ) together with IL-6 rescued the phosphorylation of STAT3 in IL-6Rα-deficient CD4+ T cells whereas IL-11 , OSM , or CNTF did not phosphorylate STAT3 in either wild-type or IL-6Rα-deficient CD4+ T cells ( Figure 1—figure supplement 1 ) . These results suggest that the STAT3-dependent signaling pathway remains intact in IL-6Rα-deficient CD4+ T cells and that other tested cytokines of the IL-6 family do not play a major role in the activation of naive CD4+ T cells . We therefore demonstrate efficient deletion of the IL-6Rα and abrogation of IL-6 signaling in T cells from IL-6RαT-KO mice . 10 . 7554/eLife . 01949 . 003Figure 1 . Impairment of both Th1 and Th17 responses in IL-6RαT-KO mice . ( A ) Expression of the IL-6Rα chain by CD4+ and CD8+ T cells from WT and IL-6RαT-KO mice was examined by flow cytometry in naive mice ( upper panels ) and in mice immunized with OVA plus LPS in IFA ( lower panels ) . ( B ) CD4+ and CD8+ T cells purified from WT and IL-6RαT-KO mice were either left untreated ( shaded histogram ) or stimulated with recombinant IL-6 for 20 min ( open histogram ) and expression of phosphorylated STAT3 ( Y705 ) was assessed by flow cytometry . ( C ) CD4+ T cells were purified from the popliteal and inguinal lymph nodes of WT and IL-6RαT-KO mice 7 days following immunization in the footpads with OVA and LPS emulsified in IFA . Proliferation was assessed by [3H]-thymidine incorporation following coculture of purified CD4+ T cells with irradiated splenocytes presenting titrating doses of OVA for approximately 72–84 hr . ( D ) Supernatants of CD4+ T cells from immunized mice were collected approximately 84 hr after restimulation with antigen in vitro . The production of IFN-γ and IL-17 by CD4+ T cells was examined by ELISA . ( E ) Proliferation and cytokine expression were measured by CFSE-labeling and intracellular cytokine staining , respectively , 72 hr after in vitro restimulation . Stimulations were performed as described in ( C ) . ( F ) Day 7 following immunization with 2W peptide and LPS emulsified in IFA , the percentages of antigen-specific T cells were determined by 2W:I-Ab tetramer staining . Gated on total CD4+ cells . ( G ) Total cell numbers and absolute numbers of 2W:I-Ab tetramer positive CD4+ T cells in the draining lymph nodes of WT and IL-6RαT-KO mice after the immunization . Data are representative of three independent experiments . Line graphs and bar graphs represent mean ± SEM; for all panels: p≤0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 00310 . 7554/eLife . 01949 . 004Figure 1—figure supplement 1 . Phosphorylation of STAT3 in IL6Rα-deficient CD4+ T cells after stimulation through gp130-containing receptors . Isolated CD4+ T cells from IL-6Rα-deficient mice and wild-type controls were stimulated in vitro with α-CD3e and α-CD28 mAbs and the indicated cytokines . Phosphorylation of STAT3 was measured 1 hr later by Western blot analysis . Detection of β-actin was used as loading control . sIL-6Rα , soluble IL-6R α chain; OSM , oncostatin M; CNTF , ciliary neurotrophic factor . Shown is a representative experiment reflecting the results from one mouse per genotype out of two independent experiments using a total of four mice per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 00410 . 7554/eLife . 01949 . 005Figure 1—figure supplement 2 . T cell homeostasis is unaffected when IL-6 signaling is abrogated specifically in T cells . ( A ) Total numbers of CD4+ and CD8+ T cells in the thymus and lymph nodes of WT , IL-6RαT-KO and complete IL-6-deficient mice were assayed by flow cytometry . ( B ) CD4+ T cells from the thymus , spleen , and lymph nodes of WT and IL-6RαT-KO mice were stained for annexin V . ( C ) CD4+ T cells purified from the spleen of WT and IL-6RαT-KO mice were stimulated with plate-bound α-CD3 and α-CD28 plus or minus recombinant IL-6 for 72 hr and analyzed for expression of CD44 , CD62L , and active Caspase-3 . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 00510 . 7554/eLife . 01949 . 006Figure 1—figure supplement 3 . IL-6Rα-deficient CD4+ CD45RBhi T cells fail to induce Colitis and IL-6RαT-KO mice are resistant to EAE . ( A ) RAG2 KO mice were injected intraperitoneally with 5 × 105 CD4+ CD45RBhi cells purified from WT or IL-6RαT-KO mice . Mice were monitored for weight loss . ( B ) Histological sections of the middle colon of RAG2 KO mice that received WT or IL-6Rα-deficient CD4+ CD45RBhi cells 8 weeks before . ( C ) Mesenteric lymph node and lamina propria cells isolated from RAG2 KO recipient mice were stimulated with α-CD3 in vitro and supernatants were assayed for IL-17 production by ELISA . ( D ) WT and IL-6RαT-KO mice were immunized with MOG35-55 peptide in CFA plus pertussis toxin to induce EAE . Mice were monitored weekly for disease progression . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 00610 . 7554/eLife . 01949 . 007Figure 1—figure supplement 4 . IL-6RαT-KO mice show impaired CD4+ T cell responses . ( A–C ) WT and IL-6RαT-KO mice were immunized with OVA and PGN ( A ) or CpG DNA ( B ) emulsified in IFA , or OVA emulsified in CFA ( C ) . 7 days later , purified CD4+ T cells were cultured with irradiated splenocytes and titrating doses of OVA for 72 hr . Proliferation and cytokine production were measured by [3H]-thymidine incorporation and ELISA , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 00710 . 7554/eLife . 01949 . 008Figure 1—figure supplement 5 . Defective CD4+ T cell response in IL-6RαT-KO mice immunized with 2W peptide plus LPS in IFA . CD4+ T cells from immunized IL-6RαT-KO and control mice were isolated and restimulated in vitro . Proliferation ( A ) , IFN-γ production ( B ) , and frequency of 2W:I-Ab+ CD4+ T cells ( C ) were determined by [3H]-thymidine incorporation , ELISA , and flow cytometry , respectively . Representative experiments are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 00810 . 7554/eLife . 01949 . 009Figure 1—figure supplement 6 . BrdU and active Caspase-3 staining in the presence or absence of T cell-intrinsic IL-6 signaling . ( A ) WT and IL-6RαT-KO mice were immunized with 2W peptide and LPS emulsified in IFA . Mice were injected with 1–2 mg/mouse of BrdU i . p . daily from day 4 to day 7 . Draining lymph nodes were harvested on day 7 and single cell suspensions were stained for antibodies against BrdU and active Caspase-3 , gated on CD4+ T cells ( B ) . Data are representative of two independent experiments showing mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 009 Prior studies suggested that IL-6 is a mediator of T cell survival . Specifically , IL-6 has been shown to protect CD4+ T cells from α−CD3-induced apoptosis and Fas-mediated cell death in vitro ( Takeda et al . , 1998 ) . Moreover , complete IL-6-deficient mice were reported to have reduced T cell numbers in the thymus and peripheral lymphoid organs ( Kamimura et al . , 2003 ) . We therefore examined how IL-6Rα deficiency affects T cell homeostasis . We verified that complete IL-6-deficient mice had decreased numbers of T cells ( Figure 1—figure supplement 2A ) . In contrast , we found that the absolute numbers of CD4+ and CD8+ T cells in the thymus and lymph nodes of IL-6RαT-KO mice were similar to that of WT mice ( Figure 1—figure supplement 2A ) . Consistent with WT levels of CD4+ and CD8+ T cells in IL-6RαT-KO mice , we did not observe an increased tendency of IL-6Rα-deficient CD4+ T cells to undergo apoptosis . These cells became activated ( CD44hi , CD62Llo ) and cleaved caspase-3 to the same extent as control cells upon stimulation with α-CD3 and α-CD28 in vitro , irrespective of the presence of IL-6 in the culture medium ( Figure 1—figure supplement 2C ) . Likewise , CD4+ T cells from the thymus or peripheral lymphoid organs of IL-6RαT-KO mice stained positive for annexin V in similar proportions as WT CD4+ T cells ( Figure 1—figure supplement 2B ) . Taken together , these findings demonstrate that IL-6 signaling in T cells is dispensable for T cell homeostasis . Thus , the reduced T cell numbers in complete IL-6-deficient mice are likely a consequence of IL-6 regulating T cell homeostasis indirectly through other cell types . Next , we determined whether IL-6 signaling in T cells is required for the initiation of CD4+ T cell responses by immunizing IL-6RαT-KO and WT mice with OVA plus LPS in IFA . 7 days following immunization , CD4+ T cells purified from the popliteal and inguinal lymph nodes were co-cultured with irradiated WT splenocytes from naïve mice , as antigen-presenting cells , in order to examine the recall response to OVA . We found that T cells isolated from IL-6RαT-KO mice showed impaired proliferation upon restimulation with the antigen as measured by [3H]-thymidine incorporation ( Figure 1C , Figure 2—figure supplement 2A ) . Furthermore , in contrast to WT CD4+ T cells , CD4+ T cells from IL-6RαT-KO mice failed to produce IL-17 ( Figure 1D , Figure 2—figure supplement 2B ) . In order to confirm this phenotype and characterize the IL-6RαT-KO mice further , we also tested for the induction of Th17-dependent autoimmune diseases in these mice . IL-6Rα-deficient CD4+ CD45RBhi T cells failed to induce colitis when transferred into RAG2 KO mice ( Figure 1—figure supplement 3A–C ) and compared to WT mice , IL-6RαT-KO mice were also resistant to the induction of EAE following immunization with MOG35-55/CFA ( Figure 1—figure supplement 3D ) . Thus , the Th17 response of IL-6RαT-KO mice was defective , which was expected given the role of IL-6 in Th17 differentiation . However , immunization of the IL-6RαT-KO mice with OVA and LPS in IFA also revealed a profound defect in the Th1 response as IL-6Rα-deficient CD4+ T cells failed to secrete robust amounts of IFN-γ after restimulation with antigen ( Figure 1D ) . We obtained similar results following immunization with OVA and other TLR ligands , such as peptidoglycan ( PGN ) , CpG DNA , or the more complex Complete Freund's Adjuvant ( CFA ) ( Figure 1—figure supplement 4A–C ) . To address whether the defective cytokine production in in vitro restimulation assays were a consequence of defective proliferation and/or T cell differentiation , we also analyzed CD4+ T cell proliferation and cytokine production using CFSE labeling and intracellular cytokine staining , respectively . We confirmed that T cells from IL-6RαT-KO mice were impaired in their ability to proliferate following restimulation ( Figure 1E ) . Although we observed an overall reduction in T cell proliferation , a small fraction of T cells from IL-6RαT-KO mice proliferated normally as assessed by CFSE dilution ( Figure 1E ) and were able to produce IFN-γ at similar levels as WT cells . The cells remained deficient in IL-17 , regardless of proliferation ( Figure 1E ) . These results suggest that the defect in the IFN-γ response is not due to a requirement for IL-6 in the differentiation of IFN-γ-producing cells . Instead , our results imply that the IFN-γ response is deficient in IL-6RαT-KO mice as a consequence of impaired T cell proliferation . Conversely , the deficient IL-17 response , regardless of proliferative status , reflects a dual requirement of IL-6 for both CD4+ T cell proliferation and differentiation of IL-17 producing T cells . The experiments so far assessed the CD4+ T cell response after restimulation with OVA in vitro . Staining CD4+ T cells with MHC class II tetramers is a way to track antigen-specific CD4+ T cells more directly . In contrast to class II tetramers using OVA-derived peptides , 2W:I-Ab tetramers , which stain CD4+ T cells that are specific for the 2W peptide , offer a better staining performance . Furthermore , naive 2W-specific CD4+ T cells are more frequent than OVA-specific CD4+ T cells , which facilitates the in vivo analysis of the CD4+ T cell response ( Moon et al . , 2007 ) . We therefore opted to track antigen-specific T cells with 2W:I-Ab tetramers following immunization with 2W peptide and LPS in IFA ( Moon et al . , 2007 ) . To ensure that the choice of a different antigen did not affect the phenotype of IL-6RαT-KO mice , we first immunized the mice with the 2W peptide and measured the proliferation and IFN-γ secretion following restimulation in vitro . IL-6Rα-deficient CD4+ T cells did not expand and secrete IFN-γ efficiently ( Figure 1—figure supplement 5A , B ) . Antigen-specific 2W:I-Ab+CD4+ T cells from IL-6RαT-KO mice were also less frequent in these cultures compared to WT controls ( Figure 1—figure supplement 5C ) . These results resembled those obtained with OVA-based immunization and demonstrated that the choice of antigen did not affect the outcome of the CD4+ T cell response . We thus tracked antigen-specific 2W:I-Ab+ CD4+ T cells directly in vivo . We found that the percentage of 2W:I-Ab+ CD44+ CD4+ T cells was reduced in IL-6RαT-KO mice compared with the WT controls ( Figure 1F ) . Moreover , relative to WT mice , IL-6RαT-KO mice had fewer numbers of cells in their lymph nodes on day 7 after immunization ( Figure 1G ) , which resulted in lower absolute numbers of antigen-specific T cells in IL-6RαT-KO mice ( Figure 1G ) . Although the percentage and absolute number of 2W:I-Ab+CD4+ T cells were significantly reduced in IL-6RαT-KO mice on day 7 following immunization , we found that the small proportion of remaining IL-6Rα-deficient CD4+ T cells in these mice proliferated to a similar extent as the WT , if not slightly better , as assessed by BrdU incorporation ( Figure 1—figure supplement 6A ) . Consistently , we failed to see a difference in cell death between WT and IL-6Rα-deficient CD4+ T cells on day 7 following immunization as determined by active Caspase-3 staining ( Figure 1—figure supplement 6B ) . Collectively , these results confirm the expected role of IL-6 signaling in Th17 differentiation , but they also reveal a role for IL-6 signaling in regulating T cell expansion and the subsequent induction of Th1 responses in vivo . We previously showed that IL-6 produced by DCs following TLR activation is required to overcome Treg-mediated suppression in vitro ( Pasare and Medzhitov , 2003 ) . Therefore , we asked whether impaired T cell responses in IL-6RαT-KO mice are due to unopposed suppression by Tregs in the absence of IL-6 signaling . To test this possibility , IL-6RαT-KO mice were either left untreated or injected with a monoclonal antibody ( mAb ) against CD25 , which resulted in the depletion of approximately 80% of Foxp3+ Tregs ( Figure 2—figure supplement 1A , B ) prior to immunization with OVA and LPS in IFA . While CD4+ T cells from IL-6RαT-KO mice with an intact Treg compartment showed impaired proliferation and production of IFN-γ upon restimulation , these responses were largely rescued upon depletion of Tregs ( Figure 2A , B , Figure 2—figure supplement 2A , B ) . This result suggests that IL-6 may control T cell responses by overcoming the suppressive effect of Tregs . Although the IFN-γ response was restored in the absence of Tregs , the IL-17 response remained defective in these mice , confirming the nonredundant role of IL-6 in Th17 differentiation in vivo , at least under the experimental conditions used in this study . 10 . 7554/eLife . 01949 . 010Figure 2 . IL-6 signaling in responder T cells is required to overcome suppression by Tregs . ( A and B ) WT and IL-6RαT-KO mice received a single intravenous injection of α-CD25 monoclonal antibody 3 days prior to immunization to transiently deplete Tregs . Mice were immunized in the footpads with OVA and LPS emulsified in IFA and 7 days following immunization , purified CD4+ T cells were restimulated as in Figure 1C . Proliferation ( A ) and cytokine production ( B ) were measured by [3H]-thymidine incorporation and ELISA , respectively , 72–84 hr following restimulation . ( C and D ) WT and IL-6RαTREG-KO mice were immunized with OVA and LPS emulsified in IFA and purified CD4+ T cells were restimulated 7 days following immunization as in Figure 1C . Proliferation ( C ) and production of IFN-γ and IL-17 ( D ) were measured as before . ( E ) Percentages of Foxp3+ CD4+ T cells in the draining lymph nodes of WT and IL-6RαT-KO mice 7 days following immunization with OVA and LPS in IFA . ( F ) WT and IL-6RαT-KO mice were immunized with 2W peptide and LPS emulsified in IFA and 7 days after immunization cells isolated from the draining lymph nodes were stained with 2W:I-Ab tetramer to determine percentages of CD4+ Foxp3+ 2W:I-Ab+ cells . Gated on total CD4+ cells . Data are representative of at least three independent experiments . Line graphs and bar graphs represent mean ± STD . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 01010 . 7554/eLife . 01949 . 011Figure 2—figure supplement 1 . Treg depletion efficiency with α-CD25 treatment . WT and IL-6RαT-KO mice were either left untreated or injected with α-CD25 monoclonal antibody ( PC61 ) by the intravenous route . To confirm Treg depletion , peripheral blood leukocytes were examined for expression of CD4 , CD25 and Foxp3 3 days later using an α-CD25 monoclonal antibody directed against a different epitope ( A ) . We also verified the successful depletion of Tregs after PC61 treatment in FoxP3-GFP mice ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 01110 . 7554/eLife . 01949 . 012Figure 2—figure supplement 2 . Statistical representation of the CD4+ T cell response in the presence or absence of Tregs . ( A and B ) Statistical representation of the CD4+ T cell response following immunization with OVA + LPS in IFA . When indicated , Tregs were transiently depleted by administrating an α-CD25 antibody 3 days prior to the immunization . CD4+ T cells were isolated from the draining lymph nodes 7 days after the immunization and re-stimulated with 900 µg/ml OVA in vitro in the presence of irradiated splenocytes . Proliferation was measured 3 days later by [3H]-thymidine incorporation ( A ) and cytokine secretion of CD4+ T cells was measured by ELISA ( B ) . Shown are the combined data of at least three independent experiments . Each dot represents one mouse and the experiments were normalized to the first replicate of triplicates of the wild-type mice . p≤0 . 05; p≤0 . 005; ns , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 01210 . 7554/eLife . 01949 . 013Figure 2—figure supplement 3 . Suppression of T cell proliferation by WT and IL-6RαT-KO regulatory cells . Purified CD4+CD25- responder T cells from WT or IL-6RαT-KO mice were incubated at a 1:1 ratio with WT or IL-6RαT-KO CD4+CD25+ regulatory T cells for 3 days with soluble anti-CD3 and anti-CD28 . [3H]-thymidine was added for the last 15–17 hr of culture to assess the proliferative capacity of the cells . The data are representative of two independent experiments . Bar graph represents mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 01310 . 7554/eLife . 01949 . 014Figure 2—figure supplement 4 . Absolute cell numbers of antigen-specific Foxp3+ Tregs in IL-6RαT-KO and control mice . Mice were immunized with 2W peptide plus LPS in IFA . 7 days later the absolute numbers of Tregs were calculated within the draining lymph nodes using flow cytometry; p≤0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 014 Because both conventional CD4+ T cells and Tregs are unresponsive to IL-6 in IL-6RαT-KO mice , we wanted to determine which T cell subset was targeted by IL-6 to mediate its effects . One possibility is that IL-6 targets conventional CD4+ T cells in order to make them refractory to Treg-mediated suppression . Therefore , in the absence of IL-6 signaling , CD4+ T cells would remain susceptible to suppression by Tregs . Alternatively , IL-6 might target Tregs directly to block their suppressive activity . To test this latter possibility , we generated IL-6RαFL/FL; Foxp3Cre/+ mice ( hereafter called IL-6RαTREG-KO mice ) to specifically delete the IL-6Rα in Tregs . We found that CD4+ T cells from IL-6RαTREG-KO mice proliferated normally and produced both IFN-γ and IL-17 at similar levels as WT CD4+ T cells following immunization with OVA and LPS in IFA ( Figure 2C , D ) . These findings demonstrate that , at least under the experimental conditions used here , IL-6 signaling is not required in Tregs , and instead IL-6 acts on conventional CD4+ T cells to make them less sensitive to the suppressive effect of Tregs . To further rule out the possibility that the difference that we see in the response of CD4+ T cells in IL-6RαT-KO mice after immunization is due to altered Treg function in the absence of IL-6 signaling , we sorted CD4+CD25+ IL-6Rα-deficient Tregs to examine their ability to suppress proliferation of purified conventional C4+CD25− T cells . We found that IL-6Rα-deficient Tregs were able to inhibit the proliferative response of WT responder CD4+ T cells to the same degree as WT Tregs ( Figure 2—figure supplement 3 ) . Several studies have shown reciprocal regulation of Tregs and Th17 cells and demonstrated that IL-6 inhibits TGF-β-driven induction of Tregs and induces the production of IL-17 ( Bettelli et al . , 2006; Veldhoen et al . , 2006 ) . Moreover , mice bearing a T cell-specific deletion of gp130 as well as IL-6-deficient mice have been shown to have elevated numbers of Foxp3+ Tregs in their secondary lymphoid organs , although this has not been universally observed for the latter mouse strain ( Korn et al . , 2007 , 2008 ) . Therefore , it remained possible that in the absence of the IL-6Rα , Tregs were abnormally expanded in vivo , resulting in the defective T cell proliferation observed in Figure 1 . To address this , we immunized IL-6RαT-KO mice with OVA and LPS in IFA and analyzed the percentages of Foxp3+ T cells . We found no significant increase in the proportion of these cells among CD4+ T cells , compared to the WT ( Figure 2E ) . In support of this notion , we also did not detect significant expansion of antigen-specific Foxp3+ T cells following immunization with 2W peptide and LPS in IFA and the absolute numbers of Tregs were lower in IL-6RαT-KO mice compared with WT controls ( Figure 2F , Figure 2—figure supplement 4 ) . Altogether , these results suggest that expansion of induced Tregs or impairment of their function in the absence of IL-6 signaling does not account for the diminished CD4+ T cell response , at least under the experimental conditions used here . Instead , IL-6 signaling is required in CD4+ T cells to render them refractory to Treg-mediated suppression . Induction of CD4+ T cell responses is essential for mounting productive T cell-dependent B cell responses . Since Tfh cells play a critical role in providing help for GC B cells , we wished to characterize the role of IL-6 in regulating this aspect of CD4+ T cell biology . It was previously shown that relative to WT controls , complete IL-6-deficient mice have a strong decrease in the frequency of Tfh cells following protein immunization ( Nurieva et al . , 2008 ) . However , when we immunized IL-6RαT-KO mice with OVA and LPS in IFA and examined Tfh cell percentages 7 days later by staining for CXCR5 and PD-1 , we found only a modest reduction of Tfh cells in IL-6RαT-KO mice . In these mice , the frequency of Tfh cells , which were also ICOShi and PSGL-1low , was reduced by approximately 25% compared to WT controls ( Figure 3A , B ) , suggesting that IL-6 signaling in T cells does not play an essential role in Tfh cell development , and that IL-6 may affect Tfh cell development indirectly . We then examined Tfh cell function by several approaches . First , we determined the structure of the GCs and the location of CD4+ T cells within the GCs in IL-6RαT-KO and WT controls by immunofluorescence . However , we did not observe obvious changes in the GC structure in IL-6RαT-KO mice 14 days after immunization with OVA plus LPS in IFA ( Figure 3C ) . We also did not observe a strong reduction in the number of CD4+ T cells within the GCs of IL-6RαT-KO mice ( Figure 3D ) . Next , we analyzed the expression of Bcl-6 , the lineage-defining transcription factor for Tfh cells , and IL-21 , the Tfh cell-associated cytokine essential for GC formation and antibody production . Indeed , Tfh cells from both WT and IL-6RαT-KO mice upregulated both Bcl6 and IL-21 when compared to non-TFH cells ( Figure 3E , F ) . However , IL-6Rα-deficient Tfh cells failed to reach WT levels of expression for either of the two genes , suggesting a defect in the function of these cells in IL-6RαT-KO mice . 10 . 7554/eLife . 01949 . 015Figure 3 . Tfh cells generated in the absence of IL-6 signaling have reduced expression of Bcl-6 and IL-21 . ( A ) Frequency of CXCR5hi PD-1hi CD4+ Tfh cells in draining lymph nodes of WT and IL-6RαT-KO mice 7 days following immunization with OVA and LPS in IFA ( left panels ) . Additional surface markers expressed by Tfh cells from IL-6RαT-KO and control mice ( right panels ) . ( B ) Statistical representation of the frequency of Tfh cells in IL-6RαT-KO and control mice . Shown are the combined results of multiple experiments , each closed circle represents one mouse , p≤0 . 005 . ( C ) GC structure in the lymph nodes of IL-6RαT-KO and control mice of immunized mice . Adjacent tissue sections were stained for PNA ( red ) and B220 ( green ) , or CD4 ( red ) and B220 ( green ) . The dashed line demarcates the approximate border of the GC and was used to identify the GC location in the stainings with mAbs again B220 and CD4 ( in addition to a marked decrease of B220 signal at the site of the GCs ) . Representative images are shown . ( D ) Number of CD4+ T cells within the GCs per 75 µm2 . ( E and F ) Quantitative PCR measuring the expression of Bcl-6 ( E ) and IL-21 ( F ) in sorted CXCR5hi PD-1hi CD4+ Tfh and CXCR5low PD-1low CD4+ non-Tfh cells from immunized IL-6RαT-KO and control mice . Data show fold-induction over non-Tfh cells . A representative out of three independent experiments is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 015 To directly test the possibility that reduced expression of Bcl6 and IL-21 in IL-6Rα-deficient Tfh cells has functional consequences for GC formation and the resulting antibody response , we tested these responses in IL-6RαT-KO mice after immunization with OVA and LPS in IFA . Interestingly , although significant numbers of Tfh cells were generated in the absence of T cell-specific IL-6 signaling , the fraction of PNA+Fas+CD19+ GC cells was reduced by approximately 50% ( Figure 4A , B ) . Moreover , IL-6RαT-KO mice also exhibited decreased percentages of CD138+B220− plasma cells following immunization ( Figure 4C , D ) . We then examined the antigen-specific antibody production by ELISA and detected a moderate reduction in the amount of antigen-specific IgG1 in the absence of IL-6 signaling in T cells ( Figure 4E ) . However , the titers of OVA-specific IgG2c , the isotype regulated by IFN-γ , were significantly reduced in these mice compared to WT controls ( Figure 4E ) . 10 . 7554/eLife . 01949 . 016Figure 4 . B cell responses are impaired in IL-6RαT-KO mice . ( A ) Percentage of PNA+ Fas+ CD19+ GC B cells in the draining lymph nodes of WT and IL-6RαT-KO mice 7 days after immunization with OVA and LPS in IFA . ( B ) Statistical representation of the results shown in ( A ) . Shown are the combined results of multiple experiments , each closed circle represents one mouse , p≤0 . 005 . ( C ) Frequencies of CD138hiB220neg plasma cells in the draining lymph nodes of WT and IL-6RαT-KO mice on day 7 following immunization with OVA and LPS in IFA . ( D ) Statistical representation of the results shown in ( C ) . Shown are the combined results of multiple experiments , each closed circle represents one mouse , p≤0 . 005 . ( E ) Antibody response in WT and IL-6RαT-KO mice in the presence of an intact Treg compartment ( E ) or following the transient depletion of Tregs ( F ) . Mice were immunized with OVA and LPS in IFA and the antigen-specific antibody titers in the serum were measured by ELISA . Tregs were depleted 3 days prior to the immunization . Representative experiments of at least three are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 01610 . 7554/eLife . 01949 . 017Figure 4—figure supplement 1 . Size of the GC compartment in IL-6RαT-KO and control mice in the presence or absence of Tregs . Tregs were transiently depleted 3 days prior to immunization with OVA plus LPS in IFA . The frequency of PNA+ Fas+ CD19+ GC B cells in the draining lymph nodes was determined 14 days later by flow cytometry . Shown are the combined data of three independent experiments , each dot represents one mouse; p≤0 . 05 , **p≤0 . 0005 , ns = not significant ( p>0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 017 Since our results suggested that IL-6 is required for T cells to overcome Treg-mediated suppression , we asked whether the effects of deficient IL-6 signaling in CD4+ T cells on the B cell response are conditional on the presence of Tregs . To test this possibility , we depleted IL-6RαT-KO and control mice of Tregs prior to immunization with OVA and LPS in IFA and examined GC formation and antigen-specific antibody production . Treg depletion indeed restored the size of the GC compartment and OVA-specific IgG2c responses in IL-6RαT-KO mice ( Figure 4F , Figure 4—figure supplement 1 ) . Collectively , these results suggest that while not essential for Tfh cell development , IL-6 signaling in T cells is important for their ability to provide help to B cells , presumably in part by overcoming suppression by Tregs . The mechanism by which CD4+ T cells differentiate into long-lived memory cells remains poorly understood ( Kaech et al . , 2002 ) . Previously , we demonstrated that while depletion of CD4+ CD25+ Tregs in MyD88-deficient mice restored the primary Th1 cell response , the memory response remained defective in these mice after OVA and LPS in IFA immunization ( Pasare and Medzhitov , 2004 ) . This result suggested that a MyD88-dependent signal ( s ) was required for the generation of memory CD4+ T cell responses . Because IL-6 is produced in a MyD88-dependent manner following TLR activation , we investigated whether memory CD4+ T cell responses were dependent on IL-6 . First , we determined whether IL-6 signaling is required for the generation or maintenance of memory CD4+ T cells . We therefore immunized IL-6RαT-KO mice and WT controls with 2W peptide and LPS in IFA in the presence or absence of Tregs and tracked antigen-specific T cells 30–60 days later by 2W:I-Ab tetramer staining . IL-6RαT-KO mice generated significant amounts of 2W:I-Ab-specific memory CD4+ T cells that in some experiments reached WT levels , regardless of whether or not Tregs were depleted ( Figure 5A ) . Memory CD4+ T cells can be divided into different subsets based on expression of CXCR5 and PD-1 , which are used to identify CXCR5− PD-1− Th1 effector memory cells , CXCR5int PD-1− Th1 central memory cells and CXCR5hi PD-1+ Tfh cells ( Pepper et al . , 2011 ) . Therefore , we also stained for different populations of memory CD4+ T cells using these markers . We found no major differences in the percentages of these subsets , when we compared IL-6RαT-KO with WT mice ( Figure 5B ) . Thus , the memory CD4+ T cell compartment is largely intact in the absence of IL-6 signaling in T cells , regardless of the presence of Tregs . 10 . 7554/eLife . 01949 . 018Figure 5 . IL-6 signaling in T cells is dispensible for phenotypic , but required for functional differentiation of memory CD4+ T cells . ( A and B ) Generation of memory CD4+ T cells . WT and IL-6RαT-KO mice , with or without Treg depletion , were immunized once with 2W peptide and LPS in IFA . The frequency of all antigen-specific 2W:I-Ab+ CD4+ memory T cells ( A ) or individual subsets of 2W:I-Ab+ CD4+ memory T cells ( B ) in the draining lymph nodes was measured 60 days later by flow cytometry . 2W:I-Ab+ CXCR5−PD-1− represent Th1 effector memory cells , CXCR5intPD-1− represent Th1 central memory cells , and CXCR5hi PD-1+ represent Tfh cells . Data are representative of at least three independent experiments . ( C and D ) Expansion and cytokine secretion of memory CD4+ T cells . WT and IL-6RαT-KO mice , with or without Treg depletion , were immunized with OVA and LPS in IFA . 60 days later , mice that were previously depleted of Tregs were depleted for a second time prior to re-immunization with OVA and LPS in IFA . 7 days after the second immunization , purified CD4+ T cells were restimulated and proliferation ( C ) and cytokine production ( D ) were measured by [3H]-thymidine incorporation and ELISA , respectively . Line graph and bar graph represent mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 01810 . 7554/eLife . 01949 . 019Figure 5—figure supplement 1 . Statistical representation of the CD4+ T cell memory response following immunization with OVA + LPS in IFA . ( A and B ) Statistical representation of the CD4+ T cell response following immunization with OVA + LPS in IFA . Mice were immunized again at least 30 days after the primary immunization . When indicated , Tregs were transiently depleted by administrating an α-CD25 antibody 3 days prior to both the primary and secondary immunization . CD4+ T cells were isolated from the draining lymph nodes 7 days after the secondary immunization and re-stimulated with 900 µg/ml OVA in vitro in the presence of irradiated splenocytes . Proliferation was measured 3 days later by [3H]-thymidine incorporation ( A ) and cytokine secretion of CD4+ T cells was measured by ELISA ( B ) . Shown are the combined data of three independent experiments . Each dot represents one mouse and the experiments were normalized to the first replicate of triplicates of the wild-type mice . p≤0 . 05; **p≤0 . 005; ns , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 01910 . 7554/eLife . 01949 . 020Figure 5—figure supplement 2 . Frequency of antigen-specific 2W:I-Ab+ CD4+ memory T cells in IL-6RαT-KO and control mice after the secondary immune response in the presence or absence of Tregs . IL-6RαT-KO and control mice were immunized with 2W peptide plus LPS in IFA twice , 30–60 days apart . In some mice , Tregs were transiently depleted using an α-CD25 antibody 3 days prior to the primary immunization and again 3 days prior to secondary immunization ( lower panels ) . The frequency of 2W:I-Ab+ CD44+ CD4+ T ells was determined 7 days after the secondary immunization by flow cytometry . Numbers represent the mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 020 To examine whether memory CD4+ T cells generated in the absence of the IL-6Rα were functionally intact , we employed OVA or 2W-based immunization , which we used interchangeably since they behave similarly in our assays ( Figure 1—figure supplement 5 ) . Thus , we depleted Tregs twice , prior to the first immunization with either OVA or 2W and again prior to secondary challenge with OVA or 2W 60 days later . This was done because Tregs recover approximately 2 weeks after α-CD25 treatment . We then monitored the memory CD4+ T cell response after in vitro restimulation . Although depletion of Tregs in IL-6RαT-KO mice rescued primary CD4+ T cell proliferation and the induction of IFN-γ-producing cells following OVA and LPS in IFA immunization ( Figure 2A , B ) , this was not sufficient for the generation of memory Th1 responses . Proliferation of antigen-specific CD4+ T cells and induction of the IFN-γ response was still defective in IL-6RαT-KO mice compared with their WT counterparts following secondary immunization ( Figure 5C , D , Figure 5—figure supplement 1 ) . Consistent with this finding , we also noticed a trend towards reduced frequencies of antigen-specific 2W:I-Ab+CD4+ T cells in mice 7 days after the second immunization with 2W peptide , even under conditions where Tregs were transiently absent ( Figure 5—figure supplement 2 ) . Collectively , these findings demonstrate that induction of a robust primary CD4+ T cell response is insufficient for the formation of the memory response and suggests that although IL-6 signaling is not critical for the generation or maintenance of memory CD4+ T cells , it is essential for their function . Our results demonstrate that IL-6 is required to overcome the suppressive effect of Tregs on T cell proliferation in vivo . To gain insights into the mechanism by which IL-6 performs this function , we cultured CD4+ T cells with Foxp3+ Tregs at a 1:1 ratio and stimulated these cells with α−CD3ε/CD28 mAbs in the presence or absence of IL-6 . Since T cell responsiveness to IL-2 is important for T cell proliferation , we examined the expression of the IL-2 receptor α chain ( CD25 ) by CD4+ Foxp3− responder T cells in the culture on day 3 . We found that expression of CD25 by these cells was inhibited in the presence of Foxp3+ CD4+ T cells ( Figure 6A ) . However , addition of exogenous IL-6 to these cultures resulted in partial recovery of CD25 expression ( Figure 6A ) . Because the effect of IL-6 was partial , we investigated whether other cytokines could fulfill a similar function as IL-6 . Separate work in our laboratory indicated that IL-1 may also operate in this process ( Schenten et al . , 2014 ) . We therefore wondered whether IL-6 and IL-1 cooperate in mediating this function . Indeed , IL-1β also partially rescued expression of CD25 by responder CD4+ T cells , although to a lesser extent than IL-6 , and the combined addition of both IL-6 and IL-1β resulted in greater recovery of CD25 expression even in the presence of Tregs ( Figure 6A ) . Since IL-15 is another TLR-induced cytokine that had been previously implicated in overcoming Treg-mediated suppression ( Shevach , 2009 ) , we also tested whether IL-15 had a similar effect on CD25 expression of CD4+ responder T cells in the presence of Tregs . Indeed , addition of IL-15 to the culture medium also resulted in the maintenance of CD25 expression ( Figure 6—figure supplement 1 ) . 10 . 7554/eLife . 01949 . 021Figure 6 . IL-6 , together with IL-1β , overcomes Treg-mediated suppression by maintaining T cell responsiveness to IL-2 and other cytokines . ( A ) Purified CD4+ Foxp3- responder T cells ( TN ) were either cultured alone or at a 1:1 ratio with CD4+ Foxp3+ cells and α-CD3 and α-CD28 , with or without the indicated cytokines . CD25 expression by CD4+ Foxp3− cells was examined on day 3 by flow cytometry . ( B ) Proliferation of responder T cells in the presence of Tregs and indicated cytokines following stimulation with soluble α-CD3 and α-CD28 . Data are representative of at least three independent experiments . ( C ) Purified CD4+ Foxp3− responder T cells ( TN ) were either cultured alone or at a 1:1 ratio with CD4+ Foxp3+ cells and stimulated as in ( A ) . The expression of the indicated cytokine receptors was measured by flow cytometry . ( D ) Purified CD4+ T cells were stimulated with α-CD3 and α-CD28 in the presence or absence of the indicated cytokines . Expression of CD25 by CD4+ T cells was assessed on day 4 of culture . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 02110 . 7554/eLife . 01949 . 022Figure 6—figure supplement 1 . IL-15 overcomes Treg-mediated inhibition of CD25 in responder CD4+ T cells . Purified CD4+ Foxp3− cells were either cultured alone or at a 1:1 ratio with CD4+ Foxp3+ cells and α-CD3 and α-CD28 with or without recombinant IL-15 . CD25 expression by CD4+ Foxp3− cells was examined on day 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 02210 . 7554/eLife . 01949 . 023Figure 6—figure supplement 2 . Pathway analysis of the differentially expressed genes in antigen-specific CD4+ T cells of IL-6RαT-KO mice . Differences in the gene expression of antigen-specific CD4+ T cells from immunized IL-6RαT-KO mice and wild-type controls were dissected by gene-set enrichment analysis ( GSEA ) . Relative differences in the gene expression in IL-6Rα-deficient CD4+ T cells and wild-type controls were used to rank the genes against known gene-sets . Shown are the enrichment plots for gene sets reflecting STAT3-dependent genes ( A ) , cell cycle genes ( B ) , and ribosomal genes ( C ) . The enrichment profile is depicted as a green graph and the distribution of the ranked genes is shown as black lines below the graph . Genes that are more highly expressed in IL-6Rα-deficient CD4+ T cells cluster towards the left side of the graphs while genes that are more highly expressed in wild-type cells cluster towards the right side of the graph . The top 10 genes of the leading edge are listed to the right of each panel . Data are derived from the pooled samples of three independent experiments representing a total of 15–20 mice per genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 01949 . 023 The rescue of CD25 expression by IL-6 and/or IL-1β was accompanied by a recovery of T cell proliferation in the presence of Tregs . As expected , CD4+ responder T cells failed to proliferate following stimulation with α-CD3ε/CD28 mAbs , when equal numbers of Tregs were added to the culture ( Figure 6B ) . However , the addition of either IL-6 or IL-1β resulted in a partial recovery of proliferation and the presence of both IL-6 and IL-1β in the culture medium allowed CD4+ T cells to proliferate at similar levels as CD4+ responder T cells stimulated in the absence of Tregs . These data therefore implied that the IL-6 and IL-1-induced maintenance of CD25 expression in the presence of Tregs enabled CD4+ responder T cells to proliferate . Next , we asked whether Tregs also have an affect on the expression levels of other cytokine receptors of CD4+ responder T cells and , if so , whether IL-6 and IL-1 can reverse this effect . Indeed , Tregs suppressed the expression of the receptors for IL-12 and IFN-γ , two cytokine receptors that are essential for the generation of Th1 responses , and the suppression of these two cytokine receptors was rescued by the addition of IL-6 and IL-1 to the culture medium ( Figure 6C ) . We obtained similar data for the expression of the IL-21 receptor and the common γ-chain , although the expression levels for these receptors changed more moderately ( data not shown ) . The regulation of all these receptors , including CD25 , occurred at the transcriptional level as the RNA expression of the corresponding genes largely mirrored the protein expression on the cellular surface ( data not shown ) . Tregs suppress T cell responses by multiple mechanisms , including via the release of inhibitory cytokines such as TGF-β and IL-10 ( Shevach , 2009 ) . To determine whether inhibitory cytokines associated with Tregs can suppress expression of CD25 and whether their activity is countered by IL-6 , purified CD4+ T cells were stimulated in the presence of either TGF-β alone or TGF-β with IL-6 . We found that treating T cells with TGF-β also resulted in impaired expression of CD25 in responder T cells ( Figure 6D ) . However , the negative effect of TGF-β on CD25 expression was hindered when exogenous IL-6 was added to the culture . Additionally , we found that IL-1β also functioned similarly to IL-6 in countering the negative effect of TGF-β on expression of CD25 . Moreover , adding both IL-6 and IL-1β resulted in greater recovery of CD25 . Collectively , our results suggest that IL-6 cooperates with IL-1β to counter Treg suppression by maintaining T cell responsiveness to IL-2 and other cytokines important for the proliferation and differentiation of CD4+ T cells . This mode of action for IL-6 and IL-1β may therefore indeed be a part of the mechanism that renders CD4+ T cells refractory to Treg-mediated suppression in vivo . The gene expression profile of antigen-specific CD4+ T cells of IL-6RαT-KO mice after immunization with protein and LPS is different from the profile of wild-type T cells and T cells from MyD88T-KO mice , whose phenotype is similar to that of IL6-RαT-KO mice ( Schenten et al . , 2014 ) . To gain further insights into the IL-6-induced changes in antigen-specific CD4+ T cells , we performed a gene set enrichment analysis ( GSEA ) of the differentially expressed genes of the cells from IL-6RαT-KO mice and wild-type controls on day 7 after immunization ( Figure 6—figure supplement 2 ) . Genes whose induction has been associated with the STAT3 signaling pathway were repressed in IL-6RαT-KO mice , thus validating the overall strategy for the analysis . Interestingly , IL-6Rα-deficient CD4+ T cells also upregulated transcripts for cell cycle genes . However , we found the most striking difference between IL-6Rα-deficient CD4+ T cells and wild-type controls in ribosomal genes . A large number of these genes were down-regulated in IL-6Rα-deficient CD4+ T cells , suggesting that protein synthesis is impaired in IL-6Rα-deficient CD4+ T cells , which presumably negatively affects proliferation and cytokine secretion in these cells . IL-6 is one of the most pleiotropic cytokines that acts on multiple cell types and regulates many aspects of innate and adaptive immunity ( Kamimura et al . , 2003 ) . Consequently , the role of IL-6 signaling specifically in T cells has been very difficult to evaluate . In this study , we focused on a specific question about IL-6 signaling , namely its role in the connection between TLR activation and the generation of a CD4+ T cell response . Using mice bearing a T cell-specific deletion of the IL-6Rα , we found that the absence of IL-6 signaling in T cells resulted in abrogated expansion of CD4+ T cell and impaired development of Th1 and , as expected , Th17 cells following immunization with TLR agonists as adjuvants . In the steady state , IL-6RαT-KO mice had similar numbers of CD4+ and CD8+ T cells as their WT counterparts and we did not observe an increase in the fraction of apoptotic T cells in the thymus or periphery of these mice . These results indicate that the impairment of CD4+ T cell responses in IL-6RαT-KO mice was not due to a T cell-intrinsic defect in survival in the absence of IL-6 signaling . Because we had previously demonstrated that cytokines produced by DCs in response to TLR activation , in particular IL-6 , render CD4+ T cells refractory to Treg-mediated suppression in vitro , we hypothesized that the defect in T cell responses in these mice might be caused by unopposed suppression by Tregs . Indeed , depletion of CD25+ Tregs restored the development of the Th1 response in IL-6RαT-KO mice , while the Th17 response was not rescued , which is consistent with the role of IL-6 in the differentiation of Th17 cells . Moreover , deleting the IL-6Rα specifically in Foxp3+ Tregs revealed that IL-6 signaling is not required in these cells for the generation of CD4+ T cell responses and instead suggested that IL-6 signaling in conventional CD4+ T cells was critical for the generation of a CD4+ T cell response . Collectively , these data therefore support a model in which IL-6 signaling in naive T cells makes them refractory to the suppressive activity of Tregs in vivo . We found that Tfh cells were generated following immunization of IL-6Rα T-KO mice , albeit at moderately reduced frequencies , suggesting that IL-6 is not essential as an instructive signal for the induction of this T cell lineage . STAT3 has been demonstrated to be important for Tfh cell differentiation after protein immunization ( Nurieva et al . , 2008 ) . Thus , in the absence of IL-6 signaling , another STAT3-dependent cytokine such as IL-21 might compensate for IL-6 in inducing the development of Tfh cells . Indeed , IL-21 has been demonstrated to play an important role in Tfh cell generation ( Nurieva et al . , 2008 ) . Tfh cells play a critical role in providing help for GC B cells . Therefore , to examine the effects of IL-6 on Tfh cell function , we analyzed GC formation , plasma cell differentiation and antibody production in IL-6RαT-KO mice . Although Tfh cells could develop in these mice , there was a reduction in the GC compartment size the percentage of plasma cells and , importantly , the amount of antigen-specific antibodies following immunization . Defective IL-6 signaling in CD4+ T cells affected predominantly the titers of antigen-specific IgG2c , while the titers of antigen-specific IgG1 were only modestly reduced . In this context , it is interesting to note that IFN-γ , the signature cytokine of Th1 cells , is important for class-switch recombination ( CSR ) to IgG2c . The role of Tfh cells in the initiation of CSR has not been fully understood . Under conditions of Th2 immunity , Tfh cells have been identified as the major source of the IL-4 and IFN-γ required for the induction of CSR to IgG1 and IgG2c , respectively ( Veldhoen et al . , 2006 ) . In analogy , defective IFN-γ secretion by Tfh cells in IL-6RαT-KO mice may cause the impairment of the IgG2c response in these mice . Indeed , we found substantial amounts of Tfh cells in the T cell assays with which we measured IFN-γ secretion , which would be consistent with this view ( data not shown ) . Alternatively , it is possible that under the conditions used in our study , CSR is already initiated outside of the GCs and the absence of a functional Th1 response negatively influences CSR to IgG2c . Regardless of these possibilities , however , our data collectively demonstrate that IL-6 , while dispensable for the generation of Tfh cells , plays a nonredundant role in the function of these cells . Interestingly , depletion of Tregs recovered the antigen-specific IgG2c titers in IL-6αT-KO mice , suggesting that the activity of Tfh cells is negatively regulated by Tregs and this negative control is opposed by T cell-intrinsic IL-6 signaling . Tregs have been demonstrated to suppress T cell responses by a variety of mechanisms including the sequestration of IL-2 by Tregs or the production of inhibitory cytokines such as TGF-β , IL-10 , and IL-35 . In search of the mechanism by which IL-6 counteracts the suppressive activity of Tregs , we found that both Tregs and TGF-β inhibited expression of CD25 on CD4+ T cells and this inhibition could be partially prevented by IL-6 and completely prevented by the combined effect of IL-6 and IL-1 . Studies have shown that IL-1 is important for Th17 cell differentiation ( Acosta-Rodriguez et al . , 2007; Hu et al . , 2011; Sutton et al . , 2006; Volpe et al . , 2008 ) . IL-1 has been demonstrated to augment Th17 differentiation induced by IL-6 in combination with TGF-β ( Veldhoen et al . , 2006 ) . Recently , it has been suggested that IL-6 induces the upregulation of the IL-1R and that this effect is a pre-requisite for the differentiation of Th17 cells ( Chung et al . , 2009 ) . Consistent with this view , our own gene expression studies on antigen-specific CD4+ T cells from immunized IL-6RαT-KO and MyD88T-KO mice also revealed a downregulation of the IL-1R in antigen-specific CD4+ T cells in the absence of IL-6 signaling ( Schenten et al . , 2014 ) . While further studies are required to determine whether this particular mechanism also applies to other T cell lineages , it appears that IL-6 and IL-1 cooperate not only in the induction of Th17 differentiation but also in the generation of a Th1 response . In the context of the latter response , our work demonstrates that IL-6 and IL-1 are both important to overcome Treg-mediated suppression , perhaps in part by cooperating in the maintenance of the sensitivity of CD4+ responder T cells to IL-2 and other relevant instructive cytokines such as IL-12 ( Schenten et al . , 2014 ) . It is likely that other cytokines or cytokine combinations may play a similar role in other types of T cell responses , depending on the pathogen and PRRs involved . It should be noted in this regard that several cytokines , including IL-15 , have been shown to render CD4+ T cells resistant to the suppressive activity of Tregs in vitro ( Ben Ahmed et al . , 2009 ) . Interestingly , we found that similar to IL-6 and IL-1 , IL-15 is also able to counter Treg-mediated downregulation of CD25 in CD4+ T cells ( Figure 6—figure supplement 1 ) . This raises the possibility that control of CD25 expression on naive CD4+ T cells may be a common mechanism utilized by different cytokines induced by different pathways of innate immune recognition . It was previously demonstrated that IL-15 overcomes Treg-mediated suppression by activating phosphatidylinositol 3-kinases ( PI-3 kinases ) , which are key regulators of cell growth , survival , and proliferation ( Ben Ahmed et al . , 2009 ) . Interestingly , both IL-6 and IL-1 have also been implicated in the initiation of PI-3 kinase signaling ( Hideshima et al . , 2001; Madge and Pober , 2000 ) . This raises the possibility that the similarity in the ability of these cytokines to maintain the sensitivity of responder T cells to IL-2 signals in the presence of Tregs may be due to their shared function in inducing this important signaling pathway . Further studies are needed to determine whether inhibiting the PI-3 kinase pathway downstream of IL-6 and IL-1 signaling results in a failure to rescue the down-regulation of CD25 expression and susceptibility to Treg-mediated suppression . It should also be noted that while control of CD25 expression may be applicable to naive CD4+ T cells and the Th1 differentiation pathway , other stages of T cell responses ( e . g . , memory response ) and other differentiation pathways ( e . g . , Th17 and Tfh ) are likely to be counter-regulated by IL-6 and Tregs through a different control point . The counter-regulation by IL-6 and Tregs in these cases may still rely on the same principle , that is , regulated expression of a receptor for an autocrine cytokine that controls cell expansion and maintenance . Possible candidates include IL-21 and the IL-21R ( for Tfh and Th17 cells ) , and IL-4 and the IL-4 receptor for Th2 cells . Indeed , our in vitro studies indicate that IL-6 and Tregs may also counter-regulate the IL-4Rα , IL-21R , and the common γ-chain ( data not shown ) . Future studies will need to address this possibility in vivo . We investigated the changes in the cellular pathways in IL-6Rα-deficient CD4+ T cells by GSEA of antigen-specific CD4+ T cells on day 7 after immunization . The late time point allowed us to obtain enough cells for the analysis but likely missed important changes in the gene expression at early stages of the immune response . Thus , the observed changes in the cellular pathways in IL-6Rα-deficient CD4+ T might be biased by the presence of IL-6Rα-deficient CD4+ T cells that have escaped the Treg-mediated control mechanisms . This caveat of the analysis aside , we did observe a repression of STAT3-induced genes , thereby confirming the general approach of the analysis . IL-6Rα-deficient CD4+ T cells up-regulated cell cycle genes during the late stage of the immune response . At this point , the cell may attempt to proliferate , particularly in light of the lower absolute numbers of antigen-specific CD4+ T cells from IL-6RαT-KO mice , but are unable to do so because of a failure to induce ribosome production . The latter observation , namely a defect in the up-regulation of ribosome-associated genes in IL-6Rα-deficient CD4+ T cells , appeared to be one of the most significant changes in these cells . Interestingly , CD4+ T cells heterozygous for the ribosomal protein S6 , which is also down-regulated in IL-6Rα-deficient CD4+ T cells , increase their size after TCR stimulation but fail to proliferate ( Sulic et al . , 2005 ) . Indeed , increased ribosome production accompanies increased cytokine production following TCR stimulation ( Asmal et al . , 2003 ) . Thus , decreased ribosome generation may be a consequence of Treg-mediated suppression in the absence of IL-6 signal . Further studies will be required to elucidate this aspect . After microbial infection , a population of memory precursor cells are generated that progressively differentiate into functionally mature , long-lived memory T cells ( Cui and Kaech , 2010 ) . It is currently unknown whether particular cytokines produced during infection are required for the differentiation and maintenance of memory CD4+ T cells . Previously , using MyD88-deficient mice , we showed that a TLR-induced signal was necessary for the formation of memory CD4+ T cell responses ( Pasare and Medzhitov , 2004 ) . However , the nature of the signal was unclear at the time . In the present study , we found that although depleting Tregs rescued the primary Th1 response in the absence of IL-6 signaling in T cells , the memory response was still defective . Furthermore , we found that antigen-specific memory CD4+ T cells were still generated , even though the IFN-γ response was impaired in IL-6RαT-KO mice . This result suggests that IL-6 signaling is required for the generation of functionally competent memory CD4+ T cells . Importantly , in a parallel study , we found that the functionally competent memory CD4+ T cells also depend on T cell-specific IL-1 signaling , suggesting that IL-6 and IL-1 cooperate in this aspect of the CD4+ T cell response as well ( Schenten et al . , 2014 ) . Finally , it is interesting to note that STAT3-dependent cytokines , such as IL-6 , IL-10 , and IL-21 , as well as MyD88-dependent IL-1 have been suggested to play a role in memory CD8+ T cell differentiation and functional maturation following immunization and infection ( Foulds et al . , 2006; Hinrichs et al . , 2008; Castellino and Germain , 2007; Cui et al . , 2011; Yi et al . , 2010 ) . The fact that both IL-6Rα- and MyD88-deficient memory cells persist for several months following the primary immunization suggests that other signals control the maintenance of these cells . In summary , this study demonstrates that T cell-intrinsic IL-6 signaling plays a critical role in CD4+ T cell expansion , differentiation , and memory formation . Moreover , we have shown that IL-6 in cooperation with IL-1β counteract the suppressive activity of Tregs by maintaining CD25 expression in CD4+ T cells . Thus , cytokines induced upon innate immune recognition , including IL-6 and IL-1 , can control activation of adaptive immune responses by rendering antigen-specific T cells refractory to suppression by Treg cells . Disregulation of this mechanism may also contribute to the development of autoimmune diseases . The generation of the conditional Il6ra allele has been published previously ( Wunderlich et al . , 2010 ) . Briefly , exon 2 and 3 were flanked by loxP sites . Deletion of these exons upon Cre-mediated recombination leads to a frameshift mutation and a premature stop codon in exon 4 . As a result , only exon 1 and an additional non-sense 20 aa are encoded by the deleted allele , resulting in a non-functional protein without a binding site to IL-6 . Cd4-Cre ( Lee et al . , 2001 ) , Foxp3-Cre ( Rubtsov et al . , 2008 ) , and Foxp3-GFPKI mice ( Fontenot et al . , 2005 ) have been previously described . All animals were kept on a C57BL/6 background . Animals were housed in a conventional , specific pathogen-free facility at Yale University and all animal experiments were performed in accordance with the guidelines set by the Institutional Animal Care and Use Committee of Yale University . Ovalbumin , LPS , Complete Freund's Adjuvant ( CFA ) and Incomplete Freund's Adjuvant ( IFA ) were purchased from Sigma Aldrich ( St . Louis , MO ) . Endotoxin-free OVA was obtained from Biovendor , LLC ( Candler , NC ) . Low endotoxin CpG and peptidoglycan ( PGN ) were purchased from Invivogen ( San Diego CA ) . MOG35-55 peptide was purchased from Anaspec and 2WS1 peptide ( EAWGALANWAVDSA ) was from Genscript ( Piscataway , NJ ) . Pertussis toxin was purchased from Calbiochem/EMD Millipore ( Billerica , MA ) . Antibodies used included: CD4 , CD8 , CD126 , CD25 , CD44 , CD19 , B220 , IFN-γ , IL-17 , CXCR5 , PD-1 , CD138 , CD95 , CD45Rb , Active Caspase-3 , CD62L , pSTAT3 ( Y705 ) , ICOS , PSGL1 , CD3ε , CD28 , IgG1 and IgG2c ( all purchased from BD biosciences , San Diego , CA ) and PNA ( Vector Laboratories , Burlingame , CA ) . Recombinant IL-6 , IL-1β , and IL-15 were purchased from R&D Systems ( Minneapolis , MN ) . Annexin-V and the Foxp3 staining kit were purchased from Ebioscience ( San Diego , CA ) . Anti-CD4 microbeads were purchased from Miltenyi Biotec ( Auburn , CA ) . Mouse cells were cultured in complete RPMI-1640 supplemented with 10% FCS , 2 mM L-glutamine , 1 mM Sodium pyruvate , 50 μM β-mercaptoethanol , 10 mM Hepes , 100 U/ml Penicillin , 100 μg/ml Streptomycin , all from Gibco/Life Technologies ( Grand Island , NY ) . 2W:I-Ab tetramers were a generous gift from Marc Jenkins ( University of Minnesota , Minneapolis , Minnesota ) . Mice were injected subcutaneously in both hind footpads with either 50 μg/footpad of OVA or 50 μg/footpad of 2WS1 peptide ( EAWGALANWAVDSA; Genscript ) along with 5 μg/footpad LPS emulsified in IFA . For certain experiments , mice were immunized with OVA and CpG or PGN emulsified in IFA or OVA emulsified in CFA . For memory experiments , mice were immunized in only one hind footpad for the primary immunization and 60 days later a secondary immunization was done in the opposite hind footpad . For in vivo depletion of CD4+ CD25+ Tregs , mice received an intravenous injection of 100 μg of monoclonal α-CD25 antibody ( clone PC61 ) . 3 days later , depletion of Tregs was confirmed by staining peripheral blood lymphocytes for CD4 and CD25 markers . Cells were stained with relevant antibodies for 30 min on ice for cell surface staining . For 2W:IAb tetramer staining , cells were stained for 1 hr at room temperature . For intracellular staining of cytokines , cells were stimulated in the presence of Golgiplug ( BD Biosciences ) for 5 hr , stained for cell surface molecules , fixed , permeabilized , and stained for intracellular cytokines using the BD Biosciences intracellular cytokine staining kit . For detecting Foxp3 expression , cells were stained using the BD Biosciences Foxp3 staining kit . For pSTAT3 staining , cells were stimulated with IL-6 in vitro for 20 min , fixed and permeabilized with 90% methanol , then stained with anti-pSTAT3 ( Y705 ) antibody . Cells were analyzed on a FACSCalibur flow cytometer or LSRII ( BD Biosciences ) with FlowJo software ( Tree Star , Ashland , OR ) . Popliteal and inguinal lymph nodes were isolated from mice and single-cell suspensions were incubated with anti-CD4 microbeads ( Miltenyi Biotec ) and subsequently MACS-purified . T cell purity was confirmed by flow cytometry . Purified CD4+ T cells ( 1 × 105 ) were cultured in U-bottom 96-well tissue culture plates with 3 × 105 irradiated splenocytes as antigen-presenting cells ( APCs ) and titrating doses of antigen for 72–84 hr . To assess proliferation of T cells , [3H]-thymidine was added for the last 12–16 hr of the culture . Supernatants were collected at approximately 84 hr to determine cytokine production by ELISA . CD4+ T cells were MACS-purified using anti-CD4 beads and stimulated with mAbs against CD3e and CD28 for 15 min . When indicated , the cells were stimulated in the presence of 20 ng/ml IL-6 ( R&D Systems ) , 20 ng/ml IL-6 + 400 ng/ml sIL-6Ra ( Santa Cruz , Germany ) , 20 ng/ml IL-11 ( Peprotec , Germany ) , 20 ng/ml OSM ( Sigma ) , or 20 ng/ml CNTF ( R&D Systems ) . Phosphorylation of STAT3 was detected by Western blot using an anti-pSTAT3 mAb ( Cell Signaling , Germany ) . Equal loading was ensured by staining for β-actin ( Santa Cruz ) . Following immunization , RNA was isolated from PD-1+ CXCR5+ CD4+ Tfh and PD1− CXCR5− CD4+ non-Tfh cells sorted by flow cytometry , three mice per genotype , using the microRNeasy Kit ( Qiagen , Valencia , CA ) . Quantitative PCR was performed using the following primers: bcl6 , 5′-CACACTCGAATTCACTCTG-3′ ( forward ) and 5′-TATTGCACCTTGGTGTTGG-3′ ( reverse ) and il21 , 5′-AGGGCCAGATCGCCTCCTGATT-3′ ( forward ) and 5′-GAGCTGGGCCACGAGGTCAATG-3′ ( reverse ) . Popliteal and inguinal lymph nodes isolated from mice 14 days following immunization were immediately frozen in OCT tissue-freezing medium . Sections were cut to 6-μm thickness on a cryostat and fixed in acetone . Adjacent sections were stained with B220-FITC and PNA-biotin followed by Streptavidin-APC or B220-FITC and CD4-APC , respectively . Paired antibodies against IFN-γ , IL-17A , IgG1 , and IgG2c were purchased from BD biosciences to perform ELISAs . CD4+ Foxp3− and CD4+ Foxp3+ cells were sorted from the spleens of Foxp3-GFP mice by flow cytometry . Purified CD4+ Foxp3− T cells ( 5 × 104 ) were cultured in U-bottom 96-well tissue culture plates with CD4+ Foxp3+ T cells ( 5 × 104 ) and soluble α−CD3ε ( 4 μg/ml ) and α−CD28 ( 4 μg/ml ) plus or minus stimulation with recombinant cytokines: IL-6 ( 100 ng/ml ) , IL-1β ( 100 ng/ml ) and IL-15 ( 100 ng/ml ) . Alternatively , CD4+Foxp3− T cells were stimulated in flat-bottom 96-well tissue culture plates with plate-bound α-CD3ε ( 1 μg/ml ) and α-CD28 ( 1 μg/ml ) and TGF-β ( 10 ng/ml ) was added to the cultures instead of CD4+Foxp3+ cells . While we kept the concentration of IL-6 constant , we noticed a loss of IL-6 bioactivity upon reconstitution in PBS containing BSA and storage at −80°C . Fresh IL-6 was largly sufficient to maintain CD25 expression on CD4+ T cell in the presence of Tregs , whereas limiting amounts of IL-6 necessitated the addition of IL-1β to reach the same effect . Gene set enrichment analysis ( GSEA ) is a computational tool for determining the enrichment of previously characterized gene sets at the top or bottom of a rank ordered list . The generation of the original data describing the general gene expression profile of antigen-specific CD4+ T cells from IL-6RαT-KO mice and wild-type control has been published elsewhere ( Schenten et al . , 2014 ) . For the GSEA , the rank-ordered-list was created by calculating the fold change between the FPKM of antigen-specific CD4+ T cells from IL-6RαT-KO and wild-type mice and subsequently ordering the list by the fold change . FPKM values of the CD4+ T cells from two mouse strains were determined using the Cuff-diff program . A pseudocount of 0 . 01 was added to all FPKM values to prevent dividing by zero and only genes with an FPKM ≥1 in either group were included in the analysis . The GseaPreranked tool , part of the javaGSEA Desktop Application v2 . 0 . 14 , was used in conjunction with the Molecular Signatures Database v4 . 0 to run the analysis . EAE was induced by subcutaneous immunization of mice in the rear flank region with 200 μl of an emulsion of 300 μg of MOG35-55 peptide and 250 μg of M . tuberculosis H37RA in CFA on days 0 and 7 . Mice also received two injections of 500 ng of pertussis toxin in 200 μl total volume intraperitoneally on days 0 and 2 . Clinical signs of EAE were assessed according to the following score: 0 , no sign of disease; 1 , loss of tone in tail; 2 , paraparesis; 3 , hind limb paralysis; 4 , quadraplegia; 5 , moribund . Mice were euthanized before they reached stage 5 . CD4+ CD45RBhi cells were sorted from the spleen of WT mice by flow cytometry and 5 × 105 cells were adoptively transferred intraperitoneally into recipient RAG2 KO mice . Mice were monitored for disease progression by weighing each mouse weekly using a top-loading balance . Mice typically started developing signs of disease 4–5 weeks after the transfer of CD4+ CD45RBhi T cells . Where indicated , p values for statistical significance were determined by either two-tailed unpaired Student's t test or Mann–Whitney test . Number of asterisks represents the extent of significance with respect to p value .	The human body's ability to defend itself against pathogens relies on two distinct but connected systems: the innate and the adaptive immune systems . Innate immune cells survey their environment and use receptors located on their surface to distinguish between molecules that are harmless and molecules that stem from pathogens . When the cells of the innate immune system detect a pathogen , they secrete signaling molecules to alert adaptive immune cells to the invaders . Both sets of immune cells then mount a coordinated attack that usually kills the pathogen . The adaptive immune system also produces memory cells that retain information about the pathogen: this allows the organism to mount a fast and efficient immune response the next time the same type of pathogen strikes . However , it is not completely understood how the innate immune system communicates with the adaptive immune system to allow these processes to take place . One of the signaling molecules involved in the communication between different types of immune cells is a protein called Interleukin 6 ( IL-6 ) . This protein must be produced in order to trigger the immune response: however , many immune cells are able to recognize and respond to IL-6 , so it has been difficult to study its impact on specific cell types . Nish et al . have now investigated the effects of IL-6 on T cells , one of the main types of adaptive immune cell , by creating mice with T cells that are not able to recognize IL-6 . The detection of pathogens by innate immune cells normally has several effects: the population of T cells increases; the T cells produce daughter cells—T helper cells—that support innate immune cells in killing pathogens; and memory cells are formed . Nish et al . find that these responses are impaired in the mutant mice . To understand why , Nish et al . turn to T regulatory cells; these are adaptive immune cells that control the strength of the immune response . These experiments show that when T cells are ‘blind’ to IL-6 , they are more sensitive to the action of T regulatory cells , and this disturbs the delicate balance between the stimulation and inhibition of the immune system . Nish et al . go on to show that IL-6 works together with another signaling molecule , Interleukin 1 , to regulate how the T cells respond . The work helps to explain how the adaptive immune system mounts an immune response against pathogens but not against the host's own tissues .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2014	T cell-intrinsic role of IL-6 signaling in primary and memory responses
The biological properties of pancreatic cancer stem cells ( PCSCs ) remain incompletely defined and the central regulators are unknown . By bioinformatic analysis of a human PCSC-enriched gene signature , we identified the transcription factor HNF1A as a putative central regulator of PCSC function . Levels of HNF1A and its target genes were found to be elevated in PCSCs and tumorspheres , and depletion of HNF1A resulted in growth inhibition , apoptosis , impaired tumorsphere formation , decreased PCSC marker expression , and downregulation of POU5F1/OCT4 expression . Conversely , HNF1A overexpression increased PCSC marker expression and tumorsphere formation in pancreatic cancer cells and drove pancreatic ductal adenocarcinoma ( PDA ) cell growth . Importantly , depletion of HNF1A in xenografts impaired tumor growth and depleted PCSC marker-positive cells in vivo . Finally , we established an HNF1A-dependent gene signature in PDA cells that significantly correlated with reduced survivability in patients . These findings identify HNF1A as a central transcriptional regulator of PCSC properties and novel oncogene in PDA . Pancreatic ductal adenocarcinoma ( PDA ) is projected to be the second leading cause of cancer deaths in the U . S . by 2020 ( Rahib et al . , 2014 ) . The exceeding lethality of PDA is attributed to a complex of qualities frequent to the disease including early and aggressive metastasis and limited responsiveness to current standards of care . While both aspects are in-and-of-themselves multifaceted and can be partially attributed to factors such as the tumor microenvironment ( Olive et al . , 2009; Provenzano et al . , 2012; Waghray et al . , 2016 ) and the mutational profile of the tumor cells ( Yachida et al . , 2012 ) , cancer stem cells ( CSCs ) have also been identified to contribute to promoting early metastasis and resistance to therapeutics ( Hermann et al . , 2007; Li et al . , 2011 ) . CSCs , which were originally identified in leukemias ( Bonnet and Dick , 1997; Graham et al . , 2002 ) , have been identified in a number of solid tumors including glioblastoma ( Singh et al . , 2003 ) , pancreas ( Li et al . , 2007; Hermann et al . , 2007 ) and colon ( O'Brien et al . , 2007 ) . In these cases , CSCs have been characterized by the ability to establish disease in immunocompromised mice , to resist chemotherapeutics , the capability of both self-renewal and differentiation into the full complement of heterogeneous neoplastic cells that comprise the tumor , and the propensity to metastasize . In each case , CSCs are distinguished from other tumor cell types by the expression of various , sometimes divergent cell surface markers . Our lab was the first to identify pancreatic cancer stem cells ( PCSCs ) , which were found to express the markers EPCAM ( ESA ) , CD44 , and CD24 ( Li et al . , 2007 ) . In addition to these markers , CD133 ( Hermann et al . , 2007 ) , CXCR4 ( Hermann et al . , 2007 ) , c-MET ( Li et al . , 2011 ) , aldehyde dehydrogenase 1 ( ALDH1 ) ( Kim et al . , 2011 ) , and autofluorescence ( Miranda-Lorenzo et al . , 2014 ) have all been proposed markers of PCSCs . In all cases , the identified cells are characterized by being able to form spheres of cells ( tumorspheres ) under non-adherent , serum-free conditions , as well as an increased ability to form tumors in mice compared to bulk tumor cells . While a number of markers have been identified for PCSCs , relatively little is known about the transcriptional platforms that govern their function and set them apart from the majority of bulk PDA cells . Transcriptional regulators such as NOTCH ( Wang et al . , 2009; Abel et al . , 2014 ) , BMI1 ( Proctor et al . , 2013 ) , and SOX2 ( Herreros-Villanueva et al . , 2013 ) have been demonstrated to play roles in PCSCs , although these proteins are also critical for normal stem cell function in many tissues . In this study , we sought to better understand the biological heterogeneity of PCSCs and their bulk cell counterparts in an effort to identify novel regulators of PCSCs in the context of low-passage , primary patient-derived PDA cells . Using microarray analysis and comparing primary PDA cell subpopulations with different levels tumorigenic potential and stem-cell-like function , we identified hepatocyte nuclear factor 1-alpha ( HNF1A ) , an endoderm-restricted transcription factor , as a key regulator of the PCSC state . Supporting this hypothesis , depletion of HNF1A resulted in a loss of PCSC marker expression and functionality both in vitro and in vivo . Additionally , ectopic expression of HNF1A augmented PCSC properties in PDA cells and enhanced growth and anchorage-independence in normal pancreatic cell lines . Mechanistically , we found that HNF1A directly regulates transcription of the stem cell transcription factor POU5F1/OCT4 , which is necessary for stemness in PCSCs . Based on these data , we postulate a novel pro-oncogenic function for HNF1A through its maintenance of the pancreatic cancer stem cell properties . A transcriptional profile of PCSCs has yet to be established , and we hypothesized that such a profile would contain key regulators of the PCSC state . To pursue this hypothesis , we utilized a series of low-passage , patient-derived PDA cell lines to isolate PCSC-enriching and non-enriching subpopulations for comparative analysis . Using two of our previously described PCSC surface markers , CD44 and EPCAM ( Li et al . , 2007 ) , we found that low-passage PDA cells generally formed three subpopulations ( abbreviated P herein ) based on surface staining: CD44High/EPCAMLow ( P1 ) , CD44High/EPCAMHigh ( P2 ) , or CD44Low/EPCAMHigh ( P3 ) ( Figure 1A ) . Similar expression patterns were observed in 10 primary tumor samples ( data not shown ) . Additionally , a CD44Low/EPCAMLow subpopulation was observed in five samples ( data not shown ) , consistent with our previous data ( Li et al . , 2007 ) . Using previously described measures of PCSC function ( Li et al . , 2007; Li et al . , 2011 ) , including co-expression of the PCSC marker CD24 ( Figure 1—figure supplement 1A ) , the abilities for isolated subpopulations to reestablish heterogeneous CD44 and EPCAM surface expression ( Figure 1B ) , to form tumorspheres under non-adherent/serum-free culture conditions ( Figure 1C , D ) , and to initiate tumors in immune-deficient mice ( Supplementary file 1 ) , we found that P2 cells showed greater enrichment for cells with PCSC properties than their P1 and P3 counterparts . Using two primary PDA lines ( NY8 and NY15 ) , P1 , P2 , and P3 PDA cells were sorted by flow cytometry , prepped immediately for mRNA , and analyzed by Affymetrix GeneChip microarray and validated by quantitative RT-PCR . We found that P2 cells from both lines exhibited a signature of 50 genes that was upregulated ( >1 . 5 fold ) relative to both P1 and P3 cell counterparts ( Figure 1E ) . To further refine this gene cohort , we utilized oPOSSUM ( Kwon et al . , 2012 ) , a web-based system to detect overrepresented transcription-factor-binding sites in gene sets . Interestingly , HNF1A , a P2 cohort gene itself ( Figure 1E , F ) , had predicted binding sites in the ±5000 regions ( from start of transcription ) of 17/50 of the enriched genes , and due to its stringent consensus sequence ( DGTTAATNATTAAC ) was the most highly ranked common transcription factor by Z-score ( 17 . 895 ) . Of these 50 genes , HNF1A is known to positively regulate cohort genes HNF4A ( Boj et al . , 2001 ) , NR5A2 ( Molero et al . , 2012 ) , CDH17 ( Zhu et al . , 2010 ) , IGFBP1 ( Babajko et al . , 1993; Powell and Suwanichkul , 1993 ) , and DPP4 ( Gu et al . , 2008 ) . Interestingly , genome-wide association ( GWA ) studies have recently identified certain single nucleotide polymorphisms ( SNPs ) in the HNF1A locus as risk factors for developing PDA ( Pierce and Ahsan , 2011; Li et al . , 2012; Wei et al . , 2012 ) , although the mechanism by which these SNPs exert their influence is currently unknown . Similarly , SNPs in the HNF1A target NR5A2 are also associated with the development of PDA ( Petersen et al . , 2010; Rizzato et al . , 2011 ) , further implicating a role for the HNF1A-transcriptional network in PDA . To further support the enrichment of HNF1A in PCSCs , sorted cells were western blotted for HNF1A and HNF1A-target proteins , CDH17 and DPP4 . These proteins were found to be elevated in P2 cell lysates compared to other subpopulations ( Figure 1—figure supplement 1B ) , in agreement with their transcript levels . CSCs are enriched in cancer cell populations grown under low-attachment tumorsphere ( S ) conditions compared to cells grown in adherent ( A ) conditions . In keeping with this observation , we found protein levels of HNF1A and CDH17 elevated in multiple PDA lines cultured under tumorsphere conditions ( Figure 1—figure supplement 2A and C ) . Using a GFP-based reporter driven by eight tandem copies of the HNF1A consensus sequence GGTTAATGATTAACC ( Figure 1—figure supplement 2B ) , we found GFP expression was elevated in NY5 , NY8 , and NY15 cells grown under tumorsphere ( S ) compared to adherent conditions ( A ) ( Figure 1—figure supplement 2C ) . This construct showed excellent dependence on HNF1A expression as targeting HNF1A with an HNF1A-specific siRNA ablated expression of both the ectopic GFP and endogenous CDH17 ( Figure 1—figure supplement 2D ) . Lastly , we found the frequency of GFP-positive cells increased in cells grown in suspension ( Figure 1—figure supplement 2E ) , with GFP expression being highest in the P2 subpopulation of NY15 cells ( Figure 1—figure supplement 2F ) . Based on our gene expression and tumorsphere data , we hypothesized that HNF1A is a central regulator of CSC function . Consistent with our hypothesis that HNF1A may be an integral component of PDA biology we observed higher levels of HNF1A protein and transcripts in PDA cells compared to non-transformed immortalized pancreatic cell lines HPNE ( N ) and HPDE ( D ) ( Figure 2A; Figure 2—figure supplement 1A ) . Immunostaining of a PDA tissue microarray showed HNF1A expression to be significantly elevated ( p<0 . 0001 ) in PDA neoplastic ducts ( n = 41 ) compared to normal pancreatic ducts ( n = 18 ) ( Figure 2—figure supplement 1B , C ) . To examine the role of HNF1A in PDA cells , we depleted the protein with two distinct siRNAs ( Figure 2B ) . Knockdown of HNF1A resulted in reduced cell numbers in multiple primary PDA lines ( Figure 2C ) . To determine whether the apparent loss in cell number was due to apoptotic cell death , we performed annexin V/DAPI staining on control and HNF1A-depleted NY5 , NY8 , and NY15 cells . In all cases , knockdown of HNF1A resulted in a significant ( p<0 . 05 ) increase in apoptotic cells , while not affecting necrotic cell numbers ( Figure 2D , data not shown ) . Furthermore , increased cleavage of caspases 3 , 6 , 7 , and 9 was observed in cells depleted of HNF1A ( Figure 2E ) , indicating apoptotic cell death . These data indicate that HNF1A is important for PDA cell growth and survival . Next , we pursued whether depletion of HNF1A impacted PDA subpopulation distribution . Consistent with a central role in maintaining heterogeneous EPCAM and CD44 expression , we observed a change in P2 in all cell lines ( Figure 3A , Figure 3—figure supplement 1A ) with a concomitant increase in the P3 population ( Figure 3—figure supplement 1A , C ) . NY8 cells showed a loss in the P1 population as well ( Figure 3—figure supplement 1A , B ) . Collectively , these results support a role for HNF1A in maintaining cellular heterogeneity , with the most dramatic change being the consistent loss of the PCSC population . In addition to changes in CD44 and EPCAM surface expression , we also observed a marked decrease in CD24 surface expression ( Figure 3B , Figure 3—figure supplement 1D ) and mRNA levels ( data not shown ) in multiple PDA lines; suggesting that loss of HNF1A depletes the CSC compartment . To assess functional consequences of HNF1A-depletion on the PCSC compartment , cells ( NY5 , NY8 , NY15 ) expressing HNF1A shRNAs were grown under tumorsphere-promoting conditions . These shRNAs effectively depleted HNF1A as well as CDH17 ( Figure 3C ) , indicating downstream signaling inhibition . Consistent with a role in PCSC function , HNF1A knockdown showed a marked reduction in tumorsphere formation ( p<0 . 05 ) ( Figure 3D , E; Figure 3—figure supplement 1E ) . We next sought to determine whether CSC properties could be augmented by ectopic expression of HNF1A in PDA cells . For these studies , we selected PDA lines with high ( NY15 ) , medium ( NY8 ) , and low ( NY53 ) expression of HNF1A ( Figure 2A ) to determine if additional HNF1A expression could bolster PCSC properties under different cellular contexts . Using doxycycline-inducible expression of HNF1A ( Figure 4A , B ) , we noted increased expression of CD24 , CD44 , and EPCAM in multiple primary PDA lines ( Figure 4B–D , data not shown ) , indicating that ectopic HNF1A can increase PCSC marker expression in PDA cells . Additionally , we found that HNF1A-expressing cells formed ~2 . 5 fold more tumorspheres than their counterparts ( Figure 4E ) in all PDA cells tested . Taken together , these data indicate that ectopic HNF1A can promote PCSC properties , even in the presence of higher endogenous expression ( i . e . NY15 ) . We next examined the effects of ectopic HNF1A expression in the non-tumorigenic pancreatic ductal cell lines HPDE and HPNE , which were devoid of endogenous HNF1A expression ( Figure 2A ) . Doxycycline-inducible ectopic expression of HNF1A alone or in concert with ectopic KRASG12D was readily achieved in HPDE cells ( Figure 4—figure supplement 1A ) . Consistent with previous reports , KRASG12D-induced phosphorylation of both ERK1/2 and AKT in HPDE cells . Similar effects were seen in HPNE cells constitutively expressing HNF1A and KRASG12D alone or in combination ( Figure 4—figure supplement 1A ) . We then tested the impact the of HNF1A and/or KRASG12D expression , either alone or in combination , on HPDE cell growth . Under normal growth conditions with serum , ( LacZ ) HPDE cells grew to confluency but did not form colonies , presumably due to contact-inhibition ( Figure 4—figure supplement 1B ) . Expression of KRASG12D , however , resulted in colony formation , indicating a bypass of contact inhibition . HNF1A alone resulted in significantly increased colony formation , which was further enhanced by the additional expression of KRASG12D . Similar effects were seen in HPNE cells ( data not shown ) . In clonogenicity assays , HNF1A-expressing HPNE cells formed similar numbers of colonies to control and KRASG12D-expressing cells ( Figure 4F , G ) ; however , HNF1A alone promoted enhanced colony size . HPDE cells failed to form colonies at clonal densities in the presence of serum . In addition to foci formation , anchorage-independent growth can indicate cellular transformation in vitro . When suspended in soft agar , control HPDE cells failed to grow over a 21-day period ( Figure 4H , I ) . The addition of KRASG12D alone did not significantly promote colony formation , consistent with its relatively weak transforming ability in HPDE cells . Interestingly , HNF1A alone resulted in numerous small colonies which in turn synergized with the expression of KRASG12D in the form of numerous large colonies . Neither HNF1A nor KRASG12D alone resulted in anchorage-independent growth in HPNE cells ( data not shown ) . Lastly , we examined the effects of both transgenes on PCSC marker expression . Expression of HNF1A increased expression of EPCAM , CD44 , and CD24 in HPDE cells ( Figure 4—figure supplement 1A , C ) . Control HPNE cells lacked expression of both EPCAM and CD24 , but expressed high levels of CD44 . Expression of HNF1A was able to increase CD44 surface expression , while not changing EPCAM status ( Figure 4—figure supplement 1C , data not shown ) . Remarkably , CD24 was potently induced upon HNF1A expression , with nearly 83% of HPNE cells expressing CD24 compared to 0 . 5% of LacZ-expressing control cells . These data would suggest that HNF1A possesses properties of an oncogene capable of cooperation with oncogenic KRAS . To determine whether HNF1A was necessary for tumorigenesis , we implanted two HNF1A-high primary lines ( NY5 and NY15 ) expressing control or two HNF1A-targeting shRNAs orthotopically in the pancreas of NOD/SCID mice . HNF1A-depleted cells showed significantly reduced tumor growth compared to their control cohorts ( p<0 . 05 ) , ( Figure 5A , B ) . Similar results were observed with HNF1A knockdown in subcutaneous xenografts of NY5 and NY15 cells ( Figure 5C , Figure 5—figure supplement 1A ) . To determine whether inhibition of tumor growth was due to effects on the PCSC compartment , NY5 tumors were dissociated and analyzed by flow cytometry . Consistent with our in vitro findings , the EPCAM+/CD44+/CD24 +cell population was significantly reduced in HNF1A-depleted tumors ( p<0 . 05 ) ( Figure 5D , E ) . Importantly , western blot analysis of resultant tumor lysates confirmed that shRNAs remained effective at depleting HNF1A during the course of the experiment ( Figure 5—figure supplement 1B ) . As a direct relationship between HNF1A and stem cell function has not been reported , we examined mRNA expression of central stemness regulators MYC , SOX2 , KLF4 , NANOG , and POU5F1/OCT4 in HNF1A-depleted cells . Of these transcription factors , only POU5F1/OCT4 mRNA showed consistent downregulation in multiple PDA cell lines in response to HNF1A knockdown ( Figure 6A , data not shown ) . Similarly , we found that POU5F1/OCT4 mRNA was upregulated in response to overexpression of HNF1A in both PDA cells and HPDE cells ( Figure 6B ) , indicating regulation of POU5F1/OCT4 expression by HNF1A in pancreatic-lineage cells . To determine whether POU5F1/OCT4 mRNA was correlated with HNF1A expression , qRT-PCR was performed in 22 primary PDA lines as well as HPNE and HPDE cells . The Pearson correlation coefficient of POU5F1/OCT4 mRNA was found to be significantly correlated ( p=0 . 0094 ) with HNF1A mRNA levels ( Figure 6C ) . Additionally , POU5F1/OCT4 and HNF1A mRNA levels were correlated ( Pearson’s r = 0 . 406 , p=8 . 9×10−8 ) in patient tumors samples from The Cancer Genome Atlas ( TCGA ) dataset for PDA ( PAAD cohort , data not shown ) , further supporting relationship between the two genes . Despite a strong association between POU5F1/OCT4 and HNF1A mRNA levels , we did not observe a significant association between POU5F1/OCT4 mRNA and any of the PDA subpopulations , indicating that factors other than HNF1A modulate the levels of POU5F1/OCT4 mRNA in different PDA subpopulations ( data not shown ) . Previously published HNF1A chromatin immunoprecipitation sequencing ( ChIP-seq ) data performed in HepG2 cells by The Encyclopedia of DNA Elements ( ENCODE ) project ( Consortium and ENCODE Project Consortium , 2012 ) identified a region of enrichment of HNF1A upstream of the POU5F1/PSORS1C3 gene loci proximal to recently identified retrotransposon long terminal repeat ( LTR ) -rich region that can serve as a promoter for both genes ( Malakootian et al . , 2017 ) . Additionally , enrichment of this LTR region by TATA-binding protein ( TBP ) and acetylated lysine 27 histone H3 supports the involvement of this region in the transcription of POU5F1/OCT4 . Interestingly , this LTR promoter region contains three consensus half-sites for HNF1A ( Figure 6—figure supplement 1A ) . To test whether HNF1A binds directly to these half-sites , ChIP-PCR was performed in NY5 , NY8 , and NY15 cells . Consistent with the ENCODE data we observed significant enrichment of two half-sites in NY5 and all three half-sites in NY8 and NY15 by HNF1A . By contrast , the canonical distal enhancer of POU5F1/OCT4 ( Yeom et al . , 1996 ) , located 14-kbp downstream of the LTR promoter , and HNF1A non-target gene MYOD showed no significant enrichment by HNF1A ( Figure 6—figure supplement 1B ) , demonstrating the specificity of enrichment observed . To validate the LTR promoter region as an HNF1A-responsive promoter region , a reporter construct was generated encompassing the three putative HNF1A half-sites ( Figure 6—figure supplement 1C ) . Co-transfection of 293FT cells ( which lack endogenous HNF1A ) with the LTR reporter and an HNF1A-expression plasmid resulted in a 4 . 5-fold induction of Cypridina luciferase expression over LacZ-expression plasmid co-transfected cells ( Figure 6—figure supplement 1D ) . Additionally , neither the cloning vector nor the canonical downstream promoter region of POU5F1/OCT4 showed responsiveness to HNF1A expression , supporting the POU5F1/OCT4 LTR promoter as the HNF1A-responsive promoter for the gene . POU5F1/OCT4 has previously been shown to be elevated in PCSCs ( Miranda-Lorenzo et al . , 2014; Luo et al . , 2017 ) , although a functional role for the protein has not been demonstrated in this context . To determine if POU5F1/OCT4 regulation was a key event in HNF1A-dependent stemness , we targeted POU5F1/OCT4 with multiple siRNAs , either in combination or as single sequences . Depletion of POU5F1/OCT4 resulted in a pronounced inhibition of tumorsphere formation , comparable to HNF1A knockdown ( Figure 6D–H ) . To determine whether changes in apoptosis or cell cycle were responsible for the loss of tumorsphere formation in response to POU5F1/OCT4 knockdown , we performed annexin V/DAPI staining and propidium iodide staining in NY8 cells following transfection with POU5F1/OCT4 siRNA . Consistent with its role as a regulator of stemness in normal and cancer stem cells ( Okita et al . , 2007; Takahashi and Yamanaka , 2006; Lu et al . , 2013; Kumar et al . , 2012; Nishi et al . , 2014 ) , we did not observe changes in either apoptosis or cell cycle in response to POU5F1/OCT4 knockdown ( Figure 6—figure supplement 2A , B ) . Importantly , knockdown of either HNF1A or POU5F1/OCT4 had comparable effects on the protein levels of OCT4A ( Figure 6D ) , the isoform responsible for imparting stemness ( Lee et al . , 2006 ) . To determine whether expression of OCT4A was sufficient to rescue stemness of PDA cells depleted of HNF1A , NY8 , and NY15 cells were transduced with OCT4A-expressing lentiviruses or vector controls and transfected with HNF1A siRNA . Consistent with our previous results , loss of HNF1A impaired tumorsphere formation in both lines expressing the vector control , however , this effect was rescued by the expression of OCT4A ( Figure 6I , Figure 6—figure supplement 2C , D ) . These data indicate that HNF1A mediates stemness of PCSCs through direct transcriptional regulation of POU5F1/OCT4 . Lastly , we sought to gain insight into the transcriptional activity and genomic binding of HNF1A in PDA and determine whether its targets held prognostic information similar to other signatures in PDA ( Bailey et al . , 2016; Collisson et al . , 2011 ) . In order to identify transcriptional targets of HNF1A , we performed Bru-seq with control and HNF1A-depleted NY8 and NY15 cells ( two replicates each of control shRNA and 2 HNF1A-targeting shRNAs per cell line ) . Bru-seq is a variation of RNA-seq which measures changes in nascent RNA levels ( bona fide transcription rate ) as opposed to steady-state mRNA changes measured by conventional RNA-seq and microarray ( Paulsen et al . , 2013 ) . Differentially expressed genes were defined by adjusted p value<0 . 1 for at least one HNF1A-targeting shRNA and a mean expression level across samples ( in RPKM ) greater than 0 . 25 . Of these differentially expressed genes , 243 HNF1A upregulated and 46 HNF1A downregulated were found to be in common between NY8 and NY15 ( Figure 7A ) . To assess genomic binding of HNF1A , we performed ChIP-seq using an HNF1A-specific antibody from NY8 and NY15 ( two replicates each ) . ChIP-seq peaks were called using MACS ( Feng et al . , 2012 ) with the a priori assumption of narrow , transcription factor-like peaks . Input DNA was used to discern peaks from the background . Peaks were assigned to genes based on proximity and minimum mean expression level ( 0 . 25 RPKM ) obtained from the Bru-seq data . Common peaks between NY8 and NY15 cells were defined as those peaks with overlap in at least one replicate of both cell lines . Genes were then classified as proximal , distal or neither , given the distance of the closest common peak to the transcription start site ( proximal:±5 kb , distal:±100 kb , neither:>100 kb or no peak ) . The closest peak to a gene must also identify that gene as its closest gene , to discern among genes with nearby TSSs . 139/239 ( 57 . 2% ) and 11/46 ( 23 . 9% ) HNF1A upregulated/downregulated genes had HNF1A binding based on this criteria ( Figure 7B ) , and supports the role of HNF1A as a transcriptional activator . To further understand the regulatory role of HNF1A , we asked whether the HNF1A peaks overlapped with enhancer regions . The ENCODE combined segmentation model ( a model for regulatory regions based on the ChromHMM and Segway models ) was selected for this purpose ( Hoffman et al . , 2013; Ernst and Kellis , 2012; Hoffman et al . , 2012 ) . Of the six cell lines represented in this data set , it is not clear if any one best represents our PDA cell lines . We therefore extracted regions designated ‘strong enhancer’ ( E ) from all the cell lines and merged them into one set of enhancer regions . 72 . 7% of HNF1A-bound genes had peaks overlapping in at least one of these putative enhancer regions ( Consortium and ENCODE Project Consortium , 2012 ) ( Figure 7B ) , suggesting that HNF1A has significant interaction with regulatory regions . A number of known HNF1A target genes exhibited HNF1A promoter-proximal binding and transcriptional responsiveness via Bru-seq/ChIP-seq , including CDH17 ( Figure 7C ) . Additionally , the PCSC marker EPCAM also showed HNF1A distal binding and transcriptional responsiveness , implicating HNF1A as a direct regulator of this gene . CD24 , which showed decreased transcription in response to HNF1A loss , did not show direct binding , indicating an indirect mechanism of regulation ( data not shown ) . POU5F1/OCT4 transcription was found to decrease in both NY8 ( 34 . 3% ) and NY15 ( 41 . 5% ) cells , with weak enrichment of the LTR promoter region ( data not shown ) , further supporting direct regulation of POU5F1/OCT4 transcription by HNF1A . To determine whether POU5F1/OCT4 contributes to the deregulation of genes by HNF1A knockdown , we tested for overrepresentation of TF-binding motifs in the proximal promoter regions ( ±5 kbp from TSS ) of HNF1A upregulated and downregulated genes using oPOSSUM . The POU5F1/OCT4 motif was the most significantly over-represented transcription factor motif in HNF1A downregulated genes ( z-score = 18 . 381; 13/45 genes ) . The POU5F1/OCT4 motif was enriched in the HNF1A upregulated genes , though less highly ranked ( rank #60; z-score = 2 . 104; 47/231 genes; Supplementary file 2 ) . Of the predicted POU5F1/OCT4 targets , four have previously been identified ( CACNA2D1 , GATA2 , SNAI1 , and ZEB2 ) ( Marsboom et al . , Li et al . , 2010; Ben-Porath et al . , 2008 ) . Additionally , other reported POU5F1/OCT4 targets ( Ben-Porath et al . , 2008 ) were identified among non-predicted targets , including the HNF1A upregulated genes KLF5 and ZHX2 and the HNF1A downregulated gene GJA1 . These data demonstrate an overlap between HNF1A and POU5F1/OCT4 transcriptional networks . Because CSC and oncogene gene signatures have been linked to prognosis in a variety of cancer types ( Bartholdy et al . , 2014; Eppert et al . , 2011; Glinsky et al . , 2005; Merlos-Suárez et al . , 2011; Rosenwald et al . , 2003 ) , we asked if expression of HNF1-regulated genes was related to survival as a clinical outcome . The TCGA dataset for PDA ( PAAD ) consists of 178 tumor samples from different patients where both gene expression ( RNA-seq ) and clinical survival data was collected . Of these , we selected those tumors ( n = 169 ) not identified as histologically neuroendocrine . For each gene , we generated a Cox proportional hazards survival model . We asked what fraction of genes in the HNF1A-responsive genes exhibited significance via Cox regression and whether they were associated with increased or reduced survival ( hazard ratio <1 or>1 , respectively ) . p Values were FDR-adjusted for multiple testing and two thresholds were explored . 13/237 ( 5 . 5% ) of HNF1A upregulated genes were associated with reduced survival at FDR < 0 . 1 and 57/237 ( 24 . 1% ) at FDR < 0 . 25 , with only one gene associated with increased survival passing the FDR < 0 . 25 threshold ( Figure 7—figure supplement 1A ) . For HNF1A upregulated and bound , we found a similar pattern; 11/137 ( 8 . 0% ) genes associated with reduced survival and 37/137 ( 27 . 0% ) genes at FDR < 0 . 25 and 0 genes passing the FDR < 0 . 25 threshold ( Figure 7E ) . For HNF1A downregulated genes , 1/45 ( 2 . 2% ) genes were significant at FDR < 0 . 25 only ( Figure 7—figure supplement 1B ) . A background set of genes , defined as those genes expressed above a minimal threshold in the Bru-seq data and mappable to gene identifiers in the TCGA data ( see Materials and methods , was selected for permutations testing ) . The permutation tests showed that HNF1A upregulated genes were significantly associated with poorer outcomes versus randomly selected genes ( insets , Figure 7E and Figure 7—figure supplement 1A; see Materials and methods for details ) . These findings further support the oncogenic role for HNF1A in PDA as a direct regulator of a set of genes associated with poor patient survival . In this study , we identified the transcription factor HNF1A as putative regulator of a PCSC gene signature . Functional studies revealed that HNF1A was not only central to the regulation of this gene signature , but also PCSC function . Depletion of HNF1A effectively inhibited PDA cell growth , tumorsphere formation , and tumor growth , with a loss of PCSC marker expression observed both in vitro and in vivo . Mechanistically , HNF1A appears to promote stemness through positive regulation of pluripotency factor POU5F1/OCT4 . Finally , we found that expression of HNF1A upregulated genes significantly predicted poor survival outcomes in patients with PDA . These data point to a novel oncogenic role for HNF1A in pancreatic cancer , particularly in promoting PCSC properties . A clear role for HNF1A in PDA has not previously been established . An early study of the putative oncogene FGFR4 , frequently expressed in PDA ( Ohta et al . , 1995 ) , is directly regulated by HNF1A through intronic binding sites ( Shah et al . , 2002 ) . More recently , 73% of PDA samples were found to stain positive for HNF1A ( Kong et al . , 2015 ) . A more direct role for HNF1A in PDA has been suggested by multiple GWA studies implicating certain SNPs in HNF1A as risk factors for the development of PDA ( Pierce and Ahsan , 2011; Wei et al . , 2012; Li et al . , 2012 ) . Nearly all the identified HNF1A SNPs are non-coding and relatively common ( minor allele frequencies between 30 and 40% ) , suggesting these SNPs may serve as potential contributing rather than driving factors in pancreatic tumorigenesis . Interestingly , PDA-associated HNF1A SNPs rs7310409 , rs1169300 , and rs2464196 are also associated with both an elevated risk ( 1 . 5–2 fold ) of developing lung cancer and elevated circulating C-reactive protein ( CRP ) . A well-established direct target of HNF1A ( Toniatti et al . , 1990 ) , CRP is downregulated in patients with inactivating mutations in HNF1A ( Thanabalasingham et al . , 2011 ) . As several PDA-associated SNPs are associated with elevated CRP , it is therefore possible that these SNPs augment the activity/expression of HNF1A rather than diminish it , as in the case of maturity-onset diabetes of the young 3 ( MODY3 ) variants which reduce or abolish HNF1A expression or function . Still , a tumor suppressive role for HNF1A in PDA has also been proposed ( Hoskins et al . , 2014; Luo et al . , 2015 ) . In these studies , HNF1A was found to possess pro-apoptotic/anti-proliferative properties contrary to the data in this study . Differences in these results may be technical in nature ( control cells in Luo et al . exhibited unusually high baseline apoptosis approaching 50% ) ; however , it is also possible that the role of HNF1A may differ between different molecular subtypes of PDA ( Bailey et al . , 2016 ) or in a dynamic manner like fellow transcription factor PDX1 ( Roy et al . , 2016 ) . Supporting the former , HNF1A expression has been proposed as a biomarker to distinguish between the exocrine/ADEX subtype ( HNF1A high/KRT81 low ) and the quasi-mesenchymal/squamous/basal-like subtype ( HNF1A low/KRT81 high ) ( Muckenhuber et al . , 2018; Noll et al . , 2016 ) , and supports previous observations that the quasi-mesenchymal/squamous/basal-like subtype is associated with poorer prognosis and drug resistance ( Bailey et al . , 2016; Collisson et al . , 2011; Moffitt et al . , 2015 ) . Although these studies suggest that HNF1A expression may be highest in the exocrine/ADEX subtype of PDA , HNF1A function was not specifically examined . It is possible that like other pancreas-lineage transcription factors , such as PDX1 ( Roy et al . , 2016 ) and FOXA1 ( Roe et al . , 2017 ) , HNF1A is associated with subtypes of PDA that retain elements of pancreatic identity ( classical and exocrine/ADEX ) , but are nonetheless important maintenance of the disease . Interestingly , Noll et al . demonstrated that high expression of CYP3A5 in the exocrine/ADEX subtype mediates resistance to tyrosine kinase inhibitors and paclitaxel . Our work identifies CYP3A5 as a direct target of HNF1A , suggesting that HNF1A may play a direct role in drug resistance in PDA , and future studies should explore this possibility . While we found an association between HNF1A upregulated genes and poor patient survival , we did not observe a significant association between HNF1A mRNA expression and survival ( p=0 . 7017 ) . As the promoters of HNF1A upregulated genes were enriched for transcription factor known to play roles in PDA including GATA ( likely GATA5 or GATA6 ) ( Roe et al . , 2017; Martinelli et al . , 2017; Zhong et al . , 2011 ) , PDX1 ( Roy et al . , 2016 ) , and SOX9 ( Camaj et al . , 2014; Kopp et al . , 2012; Tsuda et al . , 2018 ) , it is possible that HNF1A may work in concert with other transcription factors to elicit its full oncogenic function in PDA . A similar interaction between the transcription factors Foxa1 and Gata5 was recently described in driving metastasis in murine models of PDA ( Roe et al . , 2017 ) . Our data on HPDE and HPNE cells support a partially transforming capacity for HNF1A , wherein it overcomes contact-inhibition and anchorage-dependent growth . As cooperation with oncogenic KRAS was observed in these cells , it is feasible that HNF1A provides additional oncogenic input , possibly by altering the differentiation state of KRAS-mutant , precancerous pancreatic cells or by expanding the resident stem cell/cancer stem cell population . Indeed , expression of HNF1A alone was sufficient to increase CD24 expression/positivity in both HPDE and HPNE cells . Typically a marker of endodermal differentiation , HNF1A has not previously been reported as necessary for normal or cancer stem cells . HNF1A plays a critical role in the normal functionality of the endocrine pancreas , with hereditary inactivating mutations in the gene and promoter region resulting in MODY3 , an autosomal dominant form of diabetes resulting from β cell insufficiency . Additionally , murine knockout models recapitulate the diabetic phenotype seen in humans ( Lee et al . , 1998 ) , with elegant transcriptomic work demonstrating a requirement for murine Hnf1a in β cell proliferation ( Servitja et al . , 2009 ) . The role for HNF1A in the exocrine pancreas is less clear , and compared to islet and liver cells in the latter study , we only identified 11 overlapping HNF1A upregulated genes ( ANXA4 , CEACAM1 , CHKA , DPP4 , HNF4A , HSD17B2 , LGALS3 , MTMR11 , NR0B2 , SLC16A5 , TM4SF4 ) , suggesting distinct activity for HNF1A in PDA compared to either β cells or the liver . Regulation of POU5F1/OCT4 transcription by HNF1A is an especially exciting finding , connecting HNF1A with a previously unidentified role in regulating stemness . Our study identifies a recently described LTR promoter region ( Malakootian et al . , 2017 ) , upstream from the canonical POU5F1/OCT4 promoter , as a likely region of direct transcriptional regulation of POU5F1/OCT4 by HNF1A , supported by both ChIP and reporter assays ( Figure 6—figure supplement 1B , D ) . As this promoter region , is not conserved between humans and rodents , it is possible the interaction between HNF1A and POU5F1/OCT4 is an acquisition of human evolution and may explain why POU5F1/OCT4 has not previously been identified as an HNF1A target . Interestingly , a recent study SPINK1-positive castrate resistant prostate cancer identified POU5F1/OCT4 as part of a gastrointestinal gene signature present in SPINK1-positive prostate cancer and regulated by HNF1A and its target gene HNF4G ( Shukla et al . , 2017 ) . Consistent with our findings , this study showed downregulation/upregulation of POU5F1/OCT4 mRNA in response to HNF1A knockdown/overexpression , respectively . While direct regulation of POU5F1/OCT4 and HNF1A was not explored in this study , it does support an association between these two transcription factors , not only in gastrointestinal cells , but other cancers as well . This could indicate a more general role for HNF1A in regulating stem cell properties in human cells in which it is normally expressed . Given that HNF1A upregulated genes were found to be significantly associated with poor survival in patients with PDA , it is likely that multiple target genes contribute to HNF1A’s oncogenic influence , and future studies should be done to assess the functions of these genes in PDA to ascertain their value as either potential biomarkers or therapeutic targets . Further studies are also needed in regards to HNF1A’s role in the exocrine pancreas and whether its function is redirected during the development of PDA , particularly under the influence of oncogenic KRAS . Overall , this study further validates the importance of HNF1A to PDA while providing a novel and critical role for HNF1A in driving pancreatic cancer stem cell properties . Eight- to 10-week-old , evenly sex-mixed NOD/SCID mice were used for all experiments . Orthotopic implantation of PDA cells to the pancreas has previously been described ( Abel et al . , 2014 ) . Briefly , mice were anesthetized with an intraperitoneal injection of 100 mg/kg ketamine/5 mg/kg xylazine , and a small left subcostal incision was performed . 10 , 000 PatGFP-Luc2-labeled tumor cells in a volume of 50 µl ( 1:1 vol of cell suspension in growth media and Matrigel ) were injected into the tail of the pancreas using a 30-gauge needle . Weekly bioluminescent imaging of implanted orthotopic tumors in mice was performed using a Xenogen IVIS 200 Imaging System ( Xenogen Biosciences , Cranbury , NJ ) . For subcutaneous implantation of tumor cells , 10 , 000 tumor cells in a volume of 50 µl ( 1:1 vol of cell suspension in growth media and Matrigel ) was injected subcutaneously into both the left and right midflank regions of mice . Tumor growth was monitored weekly by digital caliper and tumor volumes calculated by the ( length x width2 ) /2 method . All mice were sacrificed once any tumors reached 20 mm3 in volume . Formalin-fixed , paraffin-embedded tumor samples were sectioned and processed for immunofluorescent staining by the University of Michigan ULAM Pathology Cores for Animal Research . Immunohistochemistry was performed using a Ventana BenchMark Ultra autostainer . HNF1A antibody ( GT4110 ) was used for immunohistochemistry at a 1:100 dilution . A PDA/normal pancreas tissue microarray was generated by the University of Michigan Department of Pathology . All microscopies were performed on an Olympus IX83 motorized inverted microscope with cellSens Dimension software ( Olympus Corporation , Waltham , MA ) . Lentiviral destination vectors were generously provided by Dr . Andrew Aplin ( Thomas Jefferson University ) . For construction of HNF1A , KRASG12D , GFP and LacZ cDNA lentiviruses , pLentipuro3/TO/V5-DEST , pLentineo3/TO/V5-DEST , pLentihygro3/TO/V5-DEST were used . For OCT4A , pLenti6 . 3/UbC/V5-DEST was used . An EcoRV digested/re-ligated pLenti6 . 3/UbC/V5-DEST ( removing the Gateway cloning element ) was used as an empty vector control . For construction of shRNA lentiviruses , pLentipuro3/BLOCK-iT-DEST was used . Human HNF1A and KRASG12D were cloned from primary PDA cDNA into pENTR/D-TOPO ( Invitrogen ) . Human OCT4A was cloned from pCR4-TOPO clone BC117435 ( Transomic Technologies ) into pENTR/D-TOPO . LacZ and PatGFP ( a variant of EGFP containing the following mutations: S31R , Y40N , S73A , F100S , N106T , Y146F , N150K , M154T , V164A , I168T , I172V , A207V ) were also cloned into pENTR/D-TOPO as control proteins . For labeling cells with firefly luciferase , PatGFP was fused to the N-terminus of firefly luciferase Luc2 ( subcloned from pGL4 . 10 ) and cloned into pENTR/D-TOPO using Gibson Assembly ( New England Biolabs ) . PatGFP-Luc2 was recombined into pLenti0 . 3/EF/V5-DEST , a modified version of pLenti6 . 3/UbC/V5-DEST with the human EF-1α promoter instead of the human UbC promoter and no downstream promoter/selective marker cassette , to generate pLenti0 . 3/EF/GW/PatGFP-Luc2 . To generate doxycycline-inducible cell lines , a cassette containing the IVS-TetR region from pLenti6/TR ( Invitrogen ) was subcloned into pLenti0 . 3/EF/V5-DEST , along with a C-terminal P2A peptide-blasticidin resistance gene ( Bsd ) reading frame to generate pLenti0 . 3/EF/GW/IVS-Kozak-TetR-P2A-Bsd . The resultant lentiviruses were used to transduce NY8 , NY15 , NY53 , and HPDE to generate doxycycline-inducible ‘TR’ lines . To generate the HNF1A-responsive reporter , the multiple cloning site and minimal promoter from pTA-Luc ( Takara , Mountain View , CA ) was subcloned upstream of PatGFP . Eight tandem repeats of the HNF1A-binding site with spacer nucleotides ( CTTGGTTAATGATTAACCAGA ) was cloned between the MluI and BglII sites of the multiple cloning site . LacZ2 . 1 ( CACCAAATCGCTGATTTGTGTAGTCGTTCAAGAGACGACTACACAAATCAGCGA ) , HNF1A shRNA#1 ( CACCGCTAGTGGAGGAGTGCAATTTCAAGAGAATTGCACTCCTCCACTAGC ) , and HNF1A shRNA#2 ( CACCGTCCCTTAGTGACAGTGTCTATTCAAGAGATAGACACTGTCACTAAGGGAC ) were cloned into pENTR/H1/TO ( Invitrogen ) . cDNA and shRNA constructs were recombined into their respective lentiviral plasmids using LR Clonase II ( Invitrogen ) . The resulting constructs were packaged in 293FT cells as previously described . Non-targeting control ( Cat#D-001810–01 ) HNF1A-targeting siRNA#1 ( GGAGGAACCGTTTCAAGTG ) HNF1A-targeting siRNA#2 ( GCAAAGAGGCACTGATCCA ) POU5F1/OCT4-targeting siRNA#5 ( CATCAAAGCTCTGCAGAAA ) POU5F1/OCT4-targeting siRNA#6 ( GATATACACAGGCCGATGT ) POU5F1/OCT4-targeting siRNA#9 ( GCGATCAAGCAGCGACTAT ) POU5F1/OCT4-targeting siRNA#10 ( TCCCATGCATTCAAACTGA ) HPDE cells were a generous gift from Dr . Craig Logsdon ( MD Anderson ) . HPNE , Capan-2 , HPAF-II , BxPC-3 , AsPC-1 , Panc-1 , and MiaPaCa-2 cells were purchased from ATCC ( Manassas , VA ) . For all low-passage human primary PDA cells , primary PDA xenograft tumors were cut into small pieces with scissors and then minced completely using sterile scalpel blades . Single cells were obtained described previously ( Li et al . , 2007 ) . The cells used in this article are passaged less than 10 times in vitro . All cells were authenticated by STR profiling ( University of Michigan DNA Sequencing Core ) . Cells were routinely tested for mycoplasma contamination using the MycoScope PCR Detection kit ( Genlantis , San Diego , CA ) and only mycoplasma-free cells were used for experimentation . ATCC and primary PDA cells were cultured in RPMI-1640 with GlutaMAX-I supplemented with 10% FBS ( Gibco ) , 1% antibiotic-antimycotic ( Gibco ) , and 100 µg/ml gentamicin ( Gibco ) . HPDE cells were maintained in keratinocyte SFM supplemented ( Invitrogen ) with included EGF and bovine pituitary extract as well as 1% antibiotic-antimycotic and 100 µg/ml gentamicin . Low-melting agarose ( Invitrogen ) was dissolved in serum-free RPMI-1640 with GlutaMAX-I to a final concentration of 2% at 60°C and cooled to 42°C . 200 µL per well 2% agarose was evenly spread at the bottom of a 24-well dish , followed by 250 µL of 0 . 6% agarose ( diluted with complete keratinocyte SFM and supplemented with FBS to 2 . 5% ) , a 250 µL of 0 . 4% agarose/cell suspension , and a 250 µL of acellular 0 . 4% agarose . Each layer was allowed to solidify a 4°C for 10 min and then heated to 37°C prior to adding the next layer . 500 µl of complete keratinocyte SFM and supplemented with 2 . 5% FBS was added atop each gel and replenished every 3 days . Flow cytometry was performed as described previously ( Li et al . , 2007 ) . Cells were dissociated with 2 . 5% trypsin/EDTA solution , counted and transferred to 5 mL tubes , washed with HBSS supplemented with FBS twice and resuspended in HBSS/2% FBS at a concentration of 1 million cells/100 µL . Primary antibodies were diluted 1:40 in cell suspensions and incubated for 30 min on ice with occasional vortexing . Cells were washed twice with HBSS/2% FBS and incubated for 20 min on ice with APC-Cy7 Streptavidin diluted 1:200 . Cells were washed twice with HBSS/2% FBS and resuspended in HBSS/2%FBS containing 3 µM 4' , 6-diamidino-2-phenylindole ( DAPI ) ( Invitrogen , Carlsbad , CA ) . Flow cytometry and sorting was done using a FACSAria ( BD Biosciences , Franklin Lakes , NJ ) . Side scatter and forward scatter profiles were used to eliminate cell doublets , APC-Cy7 was used to exclude mouse cells . For PatGFP-Luc2 labeling , GFP+/DAPI- cells were isolated by sorting and expanded for one passage prior to implantation . For analysis of apoptosis , APC-conjugated Annexin V and Annexin V binding buffer ( BD Biosciences ) was used following manufacturer’s recommendations with 3 µM DAPI added immediately before analysis to stain permeable cells/necrotic debris . Cells were trypsinized , washed in PBS and fixed in 70% ethanol for 4 hr . Cells were then permeabilized with PBS containing 0 . 1% Triton X100 and 200 µg/ml RNase A for 2 hr at 37°C and stained with 167 µg/ml propidium iodide for 30 min . DNA content was measured by flow cytometry on a CytoFLEX flow cytometer ( Beckman Coulter ) and analyzed Summit v6 . 2 software ( Beckman Coulter ) . Flow sorted NY8 and NY15 P1 , P2 , and P3 cells were immediately used for RNA isolation using the RNeasy Plus Mini Kit coupled with RNase-free DNase set ( Qiagen ) . Microarrays and analyses were performed by the University of Michigan DNA Sequencing Core . RNA labeling and hybridization was conducted using the Human Genome U133 Plus 2 . 0 microarray ( Affymetrix , Santa Clara , CA ) . Probe signals were normalized and corrected according to background signal . Adjusted signal strength was used to generate quantitative raw values , which were log-transformed for all subsequent analyses . For both the PCSC-enriched genes ( related to Figure 1 ) and the HNF1A target genes ( related to Figure 7 ) , oPOSSUM 3 . 0 ( http://opossum . cisreg . ca/oPOSSUM3/ ) ( Kwon et al . , 2012 ) was used to detect over-represented conserved transcription factor binding sites . The program was run using the following options: conservation cutoff of 0 . 4 , matrix score threshold of 85% , and search region of 5 kbp , upstream and downstream of the transcription start site . The query was entered against a background of 24 , 752 genes in the oPOSSUM database . Total RNA was extracted using RNeasy Plus Mini Kit coupled with RNase-free DNase set ( Qiagen ) and reverse transcribed with High Capacity RNA-to-cDNA Master Mix ( Applied Biosystem ) . The resulting cDNAs were used for PCR using Power SYBR Green PCR Master Mix ( Applied Biosystem ) in triplicates . qPCR and data collection were performed on a ViiA7 Real-Time PCR system ( Invitrogen ) . Conditions used for qPCR were 95°C hold for 10 min , 40 cycles of 95°C for 10 s , 60°C for 15 s , and 72°C for 20 s . All quantitations were normalized to an endogenous control ACTB . The relative quantitation value for each target gene compared to the calibrator for that target is expressed as 2- ( Ct-Cc ) ( Ct and Cc are the mean threshold cycle differences after normalizing to ACTB ) . Single cells were suspended in tumorsphere culture media containing 1% N2 supplement , 2% B27 supplement , 1% antibiotic-antimycotic , 20 ng/mL epidermal growth factor ( Gibco , Carlsbad , CA ) , 20 ng/mL human bFGF-2 ( Invitrogen ) , 10 ng/mL BMP4 ( Sigma-Aldrich , St . Louis , MO ) , 10 ng/mL LIF ( Sigma-Aldrich ) and plated in six-well Ultra-Low Attachment Plates ( Corning , Corning , NY ) . siRNA transfection siRNA were purchased from Dharmacon ( Lafayette , CO ) and were transfected at 25 nM using Lipofectamine RNAiMAX Reagent ( Invitrogen ) . siRNA sequences can be found in the Supplementary Material and methods . All lysates were boiled in 1x Laemmli sample buffer with β-mercaptoethanol for 5 min followed by electrophoresis on 4–20% Mini-PROTEAN TGX precast Tris-Glycine-SDS gels ( Bio-Rad , Hercules , CA ) . Proteins were transferred to low-fluorescent PVDF ( Bio-Rad ) and incubated overnight in primary antibody at 1:1000 dilution . Blots were incubated in IRDye-conjugated secondary antibodies at room temperature for 1 hr and imaged/quantitated by an Odyssey CLx imaging system ( Li-Cor , Lincoln , NE ) . For western blotting , HNF1A ( clone GT4110 ) and KRAS ( ab55391 ) from Abcam ( Cambridge , MA ) , β-Actin ( clone AC-74 ) from Sigma-Aldrich , Cadherin-17 ( CDH17 ) from Proteintech ( Rosemont , IL ) , β-Galactosidase from Promega ( Madison , WI ) and RASG12D , CD44 , EPCAM , DPP4 , Cleaved Caspase-3 ( D175 ) , Cleaved Caspase-6 ( D162 ) , Cleaved Caspase-7 ( D198 ) , Cleaved Caspase-9 ( D315 ) , Cleaved Caspase-9 ( D330 ) , phospho-ERK1/2 , phospho-AKT S473 , OCT4A and GFP from Cell Signaling Technology ( Danvers , MA ) . For flow cytometry , mouse anti-human EPCAM ( CD326 ) clone HEA-125 was purchased from Miltenyi Biotec ( San Diego , CA ) . Mouse anti-human CD44 clone G44-26 , CD24 clone ML5 and APC-Cy7 Streptavidin were purchased from BD Biosciences ( San Jose , CA ) . Biotinylated mouse anti-mouse H-2Kd/H-2Dd clone 34-1-2S was purchased from SouthernBiotech ( Birmingham , AL ) . For the Cypridina luciferase construct containing the full-length canonical OCT4 promoter , a 3 . 9-kbp insert was excised from phOct4-EGFP ( Gerrard et al . , 2005 ) by XhoI and BamHI digestion , followed by ligation into pCLuc-Basic2 ( New England Biolabs ) . phOct4-EGFP was a gift from Wei Cui ( Addgene plasmid # 38776 ) . For the POU5F1/OCT4 LTR construct , a 1 . 7-kbp insert was amplified from NY5 genomic DNA with the following primers: 5’-ATCTTGGAATTCTGGGCACTCAGTTTATTGTTAGG-3’ and 5’-GGTGGCGGATCCTGTGTTAATCCTCCTCGGGG-3’ . The insert was digested with EcoRI and BamHI and cloned into pCLuc-Basic2 . Cypridina luciferase constructs were co-transfected with Lipofectamine 2000 ( Invitrogen ) into 293FT cells with either LacZ or HNF1A lentiviral expression plasmids and the internal control plasmid pTK-GDLuc , a variant of pTK-GLuc ( New England Biolabs ) in which the Gaussia luciferase coding region was replaced with the coding region for Gaussia-Dura ( Millipore ) in order to provide a more stable luciferase signal . Cypridina and Gaussia-Dura luciferase activities were measured in conditioned media 48 hr post-transfection with the BioLux Cypridina Luciferase and BioLux Gaussia Luciferase Assay Kits ( New England Biolabs ) , respectively . A confluent 15 cm culture plate of cells was used per immunoprecipitation . Cells were fixed with 1% formaldehyde for 10 min . Nuclei were collected and chromatin sheared to 1–10 nucleosomes using the SimpleChIP Plus Enzymatic Chromatin IP kit and protocol ( Cell Signaling ) . HNF1A was immunoprecipitated with goat polyclonal antibody C-19 ( Santa Cruz ) . Libraries from HNF1A-immunoprecipitated chromatin and input chromatin was prepared by the University of Michigan Sequencing Core and sequenced on the Illumina HiSeq 4000 . Chromatin was prepared as indicated for ChIP-seq and immunoprecitated with either normal goat IgG ( R and D Systems ) or anti-HNF1A ( C-19 , Santa Cruz Biotechnology ) overnight using the SimpleChIP Plus Enzymatic Chromatin IP kit and protocol . Quantitative PCR was performed using immunoprecipitated DNA and 2% chromatin input DNA as described earlier for qRT-PCR using modified thermocycling conditions: 95°C hold for 10 min , 45 cycles of 95°C for 15 s and 60°C for 60 s . Percent Input for immunoprecipitated DNA was calculated using the formula 2% x 2 ( Ct 2% Input Sample - Ct IP Sample ) . Primers for POU5F1/OCT4 regulatory regions were as follows: half-site #1 ( HS1 ) ( 5’-GTGAAATCTTTAGTGTTGTGAG-3’ and 5’-CCAAGAAATGTAGCAGGACGAGCCCC-3’ ) , half-site #2 ( HS2 ) ( 5’-AACCTTTTACATGAGCAGGTTTG-3’ and 5’-AATGGTGGAAAGAATTACATGG-3’ ) , half-site #3 ( HS3 ) ( 5’-GGGCACTCAGTTTATTGTTAGG-3’ and 5’-TTTCCTGTCACAGGGGTTTAGTG-3’ ) , and distal enhancer ( DE ) ( 5’-GAGAGGCCGTCTTCTTGGCAGAC-3’ and 5’-GTTCACTTCTCGGCCTTTAACTGCCC-3’ ) . MYOD ( primers 5’-AGACTGCCAGCACTTTGCTATC-3’ and 5’-ATAGAAGTCGTCCGTTGTGGC-3’ ) was used as a non-HNF1A target gene control . Nascent RNA labeling and sequencing ( Bru-seq ) was performed as previously described ( Paulsen et al . , 2013 ) . For each shRNA ( LacZ2 . 1 , HNF1A shRNA#1 , and #2 ) , two replicates were performed in each cell line ( NY8 and NY15 ) . Cells were incubated in media containing 2 mM bromouridine ( Bru ) ( Aldrich ) for 30 min at 37°C . Total RNA was isolated after lysis in Trizol and Bru-RNA was isolated using anti-BrdU antibodies conjugated to magnetic beads . Strand-specific libraries were made using the Illumina TruSeq kit and sequenced on the Illumina HiSeq 4000 platform at the University of Michigan Sequencing Core ( Ann Arbor , MI ) . Genes were recognized as differentially expressed in both cell lines if the fold change after knockdown was greater than 1 . 5 ( and FDR < 0 . 1 in NY15 ) and the mean RPKM for a given comparison was greater than 0 . 25 in either HNF1A shRNA#1 or shRNA#2 per cell line . The HNF1A ChIP-seq experiment consisted of 2 replicates each of input and ChIP libraries from both NY8 and NY15 cells ( eight libraries altogether ) . 52-base , single end reads were aligned to the human reference genome ( hg19 ) using Bowtie v1 . 1 . 1 ( with options: -n 3 k 1 m 1 ) . Peaks were called using MACS v1 . 4 . 2 using the default options and input samples as controls . MACS peaks overlapping ENCODE blacklist regions ( https://www . encodeproject . org/annotations/ENCSR636HFF ) were removed . Peak counts were 5057 ( NY15 rep1 ) , 8616 *NY15 rep2 ) , 64603 ( NY8 rep1 ) , and 13169 ( NY8 rep2 ) . Each peak was assigned to the closest expressed gene’s transcription start site ( TSS ) . Then , for each TSS , the distance to the nearest peak was measured . If the nearest associated peak was within ±5 kb of the TSS , it was considered proximal . In the absence of a proximal peak , the nearest associated peak within ±100 kb of the TSS was considered distal . A gene was recognized as having a proximal or distal peak if at least one replicate in both cell lines identified a proximal or distal peak . If a gene was found to have both proximal and distal peaks ( usually due to differences between replicates ) , the gene was identified as distal if it had distal peaks in both replicates of both cell lines , otherwise it was identified as neither . Manipulation of genomic regions was performed using bedtools2 ( v2 . 26 . 0 ) . The HNF1A knockdown experiment used for Bru-seq consisted of a control shRNA and two different HNF1A-targeting shRNAs for each of NY8 and NY15 cells , and 2 replicates of each ( 12 samples altogether ) . 52-base , stranded , single end reads were aligned first to ribosomal DNA ( U13369 . 1 ) using Bowtie v0 . 12 . 8 and the remaining reads aligned to the human reference genome ( hg19 ) using TopHat v1 . 4 . 1 . Differential gene expression analysis was performed using DESeq v1 . 24 . 0 ( R v3 . 3 . 1 ) . Gene annotation and counting was performed as previously described ( Paulsen et al . , 2014 ) . Differentially expressed genes were selected based on the following criteria: mean RPKM >0 . 25 across samples , minimum gene length 300 , absolute value of log2 fold-change >0 . 58 ( 1 . 5 fold-change ) , adjusted p value<0 . 1 , and these requirements met for at least one HNF1A shRNA in both cell lines . All ChIP-seq and Bru-seq data from this study are available at the NCBI Gene Expression Omnibus ( GEO; accession # GSE108151 ) . Enhancer regions used in this study were taken from the ENCODE Combined Segmentation annotation ( http://hgdownload . soe . ucsc . edu/goldenPath/hg19/encodeDCC/wgEncodeAwgSegmentation/ ) ( Hoffman et al . , 2013; Ernst and Kellis , 2012; Hoffman et al . , 2012 ) . Regions labeled ‘E’ ( strong enhancers ) were extracted from all six cell lines used in the Combined Segmentation analysis , then merged to create a set of general putative enhancer regions . The enhancer regions were then queried against peak coordinated from each list of ChIP-seq peaks ( see Supplemental file 2 ) . All genomic region manipulations were performed using bedtools2 ( v . 2 . 26 . 0 ) . Gene expression and patient survival data for pancreatic adenocarcinoma were obtained through the Broad Institute TCGA Genome Data Analysis Center ( PAAD cohort; 2016; Firehose stddata__2016_01_28 run; Broad Institute of MIT and Harvard; doi:10 . 7908/C11G0KM9 ) . Clinical metadata were obtained from both the Merge Clinical Level one and Clinical Pick Tier 1 Level four data sets . Gene expression values were obtained from the Level 3 RSEM genes ( normalized ) data set and log10-transformed prior to analysis ( a constant of 1 added to preserve zeros ) . Samples identified as primary solid tumors and of non-neuroendocrine origin were used . Specifically , samples with the following values in the ‘patient . histological_type_other’ field were rejected: ‘82463 neuroendocrine carcinoma nos’ , ‘moderately differentiated ductal adenocarcinoma 60% + neuroendocrine 40%‘ , ‘neuroendocrine’ , ‘neuroendocrine carcinoma’ , and ‘neuroendocrine carcinoma nos’ . The background set of genes were defined as those with Bru-seq RPKM greater than 0 . 5 in at least one replicate of both NY8 and NY15 cells and which mapped to either gene symbol or entrez gene ID in the TCGA expression data . Cox proportional hazards survival models were created using the R package survival ( v2 . 40–1 ) . For permutation testing against a particular set of HNF1A-related genes , random sets of genes of the same size were selected from the background set and the percent of genes significantly associated with reduced or increased survival ( using FDR thresholds of 0 . 1 and 0 . 25 ) were calculated . In order for the estimated error of the estimated p value to be less than 10% ( at significant level α = 0 . 05 ) , we set the number of permutations ( N ) to 10 , 000 . The following methods are specific to analysis of the data represented in Figures 1–6 and Figure 1—figure supplement 2 , Figure 2—figure supplement 1 , Figure 3—figure supplement 1 , Figure 4—figure supplement 1 , Figure 6—figure supplement 1 , and Figure 6—figure supplement 2 . Data are expressed as the mean ±SEM . Statistically significant differences between two groups was determined by the two-sided Student t-test for continuous data , while ANOVA was used for comparisons among multiple groups . Significance was defined as p<0 . 05 . GraphPad Prism six was used for these analyses . All animal protocols were approved by University Committee for the Use and Care of Animals ( UCUCA ) at University of Michigan . The animal welfare assurance number for this study is A3114-01 . Patient samples were collected under a protocol approved by the IRB at the The University of Michigan . All patients gave informed consent . The human assurance number for this study is FWA00004969 .	Pancreatic ductal adenocarcinoma is the most common form of pancreatic cancer . It is also one of the deadliest types of cancer: fewer than one in ten patients live for five years after being diagnosed with the disease . Several reasons can explain this poor outcome including that the cancer is often diagnosed late , when tumor cells have already spread , and that there are not many effective treatments for it . Pancreatic tumors contain different types of cancer cells with different properties . Among these are the so-called pancreatic cancer stem cells . These aggressive cells produce copies of themselves , contributing to tumor growth and spread . They can also help tumors to resist chemotherapy and radiotherapy . New treatments that specifically target cancer stem cells could therefore prove important for treating pancreatic cancer . It is still not clear what makes pancreatic cancer stem cells so aggressive , or how they differ from the rest of the cells in a tumor . Abel et al . therefore looked for proteins that were more abundant in human pancreatic cancer stem cells than in other , less aggressive cancer cells with the idea that these proteins are likely to be important for the behavior of the pancreatic cancer stem cells . Abel et al . found that a protein called HNF1A is enriched in pancreatic cancer stem cells . Experimentally reducing the levels of HNF1A in cells taken from human pancreatic cancers caused the cells to grow less well and form smaller tumors when injected into the pancreases of mice . These tumors contained few cancer stem cells , suggesting that HNF1A is important for maintaining the stem cell state . Further experiments showed that HNF1A increases the amount of many other proteins inside cells , including one that controls the activity of normal stem cells . Given the importance of HNF1A to pancreatic cancer stem cells , finding ways to prevent this protein from working could lead to new treatments for pancreatic cancer . At the moment there are no drugs that target HNF1A . Further research is therefore needed to develop new drugs that work against HNF1A or one of the other proteins that it affects .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine", "cancer", "biology" ]	2018	HNF1A is a novel oncogene that regulates human pancreatic cancer stem cell properties
Wnt signaling is essential for tissue homeostasis and its dysregulation causes cancer . Wnt ligands trigger signaling by activating Frizzled receptors ( FZDRs ) , which belong to the G-protein coupled receptor superfamily . However , the mechanisms of G protein activation in Wnt signaling remain controversial . In this study , we demonstrate that FZDRs activate G proteins and trigger non-canonical Wnt signaling via the Dishevelled-binding protein , Daple . Daple contains a Gα-binding and activating ( GBA ) motif , which activates Gαi proteins and an adjacent domain that directly binds FZDRs , thereby linking Wnt stimulation to G protein activation . This triggers non-canonical Wnt responses , that is , suppresses the β-catenin/TCF/LEF pathway and tumorigenesis , but enhances PI3K-Akt and Rac1 signals and tumor cell invasiveness . In colorectal cancers , Daple is suppressed during adenoma-to-carcinoma transformation and expressed later in metastasized tumor cells . Thus , Daple activates Gαi and enhances non-canonical Wnt signaling by FZDRs , and its dysregulation can impact both tumor initiation and progression to metastasis . The Wnt signaling pathway plays a crucial role in embryonic development , in tissue regeneration , and in many other cellular processes including cell fate , adhesion , polarity , migration , and proliferation . Dysregulated expression of components within the Wnt pathway triggers many diseases , and most importantly , heralds cancer ( Klaus and Birchmeier , 2008 ) . Of the multiple known Wnt proteins , some preferentially trigger the well-characterized canonical pathway , which enhances the stability , nuclear localization and activity of β-catenin , and the downstream activation of genes targeted by the TCF/LEF transcription machinery . Other Wnts , for example , Wnt5a deviate from this canonical paradigm , and trigger so-called non-canonical pathways ( Kühl et al . , 2000; Niehrs , 2001; Winklbauer et al . , 2001 ) . Among other events , these non-canonical pathways induce the elevation of intracellular Ca2+ and activation of the small G proteins RhoA and Rac1 , which regulate polarized cell movements and the planar polarity of epithelial cells ( Sheldahl et al . , 1999; Kühl et al . , 2000; Mayor and Theveneau , 2014 ) . Of critical importance , non-canonical Wnt signaling antagonizes the canonical Wnt pathway ( Torres et al . , 1996; Olson and Gibo , 1998; Ishitani et al . , 2003 ) , although it is unclear how this occurs . Despite the lack of molecular mechanisms , dysregulation of the non-canonical Wnt pathway is widely believed to drive cancer via a two-faceted mechanism ( McDonald and Silver , 2009 ) — ( 1 ) Non-canonical Wnt signaling suppresses tumorigenesis by antagonizing the canonical β-catenin/TCF/LEF pathway , and inhibition of non-canonical Wnt signaling heralds neoplastic transformation ( Ishitani et al . , 2003; Medrek et al . , 2009; Grumolato et al . , 2010 ) ; ( 2 ) Hyperactivation of non-canonical Wnt signaling enhances cancer invasion/metastasis by activation of Rac1 and remodeling of the actin cytoskeleton ( Yamamoto et al . , 2009 ) and by upregulating CamKII and PKC ( Weeraratna et al . , 2002; Dissanayake et al . , 2007 ) . Little is known as to how such dysregulation of non-canonical Wnt signaling , that is , early inhibition and late hyperactivation is orchestrated during cancer progression . Non-canonical Wnt signaling is initiated by the binding of Wnt ligands to receptors of the Frizzled ( FZDR ) family . These receptors belong to the G protein-coupled receptor ( GPCR ) superfamily , which classically activate trimeric G proteins . However , the interplay between FZDR and G proteins in Wnt signaling is very controversial—on one hand , there is a wealth of evidence indicating that trimeric G proteins regulate Wnt signaling ( Malbon , 2004; Katanaev et al . , 2005; Liu et al . , 2005; Schulte and Bryja , 2007; Koval et al . , 2011 ) . On the other hand , definitive evidence for the direct activation of trimeric G proteins by FZDR's is elusive . The experimental difficulties and controversies in the field have led to provocative speculations that FZDRs may not bind G proteins directly , but do so indirectly via other intermediates within the Wnt signaling pathway ( Schulte and Bryja , 2007 ) , but such intermediate ‘linker’ molecules have not been identified . Recent advances in the field of trimeric G protein signaling have important implications in this regard . It has become increasingly clear that the activity of trimeric G proteins is regulated by a plethora of accessory proteins ( Siderovski and Willard , 2005; Sato et al . , 2006; Blumer and Lanier , 2014 ) beyond classical activation by GPCRs . Among these accessory proteins , a subset of proteins called non-receptor Guanine nucleotide exchange factors ( GEFs ) are uniquely positioned to fulfill the role of an intermediate to trigger G protein signaling upon Wnt stimulation because they are cytoplasmic factors capable of activating G proteins ( Tall et al . , 2003; Lanier , 2004; Natochin et al . , 2005; Lee and Dohlman , 2008; Garcia-Marcos et al . , 2009 , 2011b; Oner et al . , 2013 ) . Here , we identified Daple , a previously described binding partner of the Wnt signaling protein Dishevelled ( Dvl ) ( Oshita et al . , 2003; Kobayashi et al . , 2005 ) , as a non-receptor GEF for trimeric G proteins . We demonstrate that a novel G protein regulatory motif enables Daple to couple G protein activation to FZDRs , which in turn initiates non-canonical Wnt signaling pathways . We also demonstrate how bimodal dysregulation in Daple expression modulates non-canonical Wnt signaling during cancer progression . We recently discovered the first GEF motif for trimeric G proteins , that is , the Gα-binding and activating ( GBA ) motif , in the C-terminal region of the non-receptor protein GIV ( Garcia-Marcos et al . , 2009 ) . We showed that GIV binds and activates Gα subunits of the Gi subfamily via its GBA motif and regulates signal transduction . GIV is one of the 3 members of the CCDC88 family , which have in common an N-terminal HOOK domain followed by a long coiled-coil region but are highly divergent in their C-terminal region ( Le-Niculescu et al . , 2005; Enomoto et al . , 2006 ) : CCDC88b ( aka GIPIE ) completely lacks this C-terminal region , whereas the C-terminal region of CCDC88c ( aka Daple ) shows significant divergence ( 15% identity , 26% similarity ) compared to CCDC88a's ( i . e . , GIV ) ( Figure 1A ) . The divergence in the C-terminal sequence allows CCDC88 proteins to associate with different proteins and regulate diverse biological processes ( Le-Niculescu et al . , 2005; Enomoto et al . , 2006 ) , for example , a PDZ-binding motif ( PBM ) is found exclusively in Daple , at its extreme C-terminus , which binds the PDZ domain of Dvl and regulates Wnt signaling ( Oshita et al . , 2003; Kobayashi et al . , 2005 ) . Despite these apparent sequence differences among CCDC88 family members , a more detailed analysis of the C-terminal sequences of GIV and Daple from different vertebrate species revealed a cryptic GBA motif in Daple localized within the otherwise highly divergent C-terminal region ( Figure 1A ) . This putative GBA motif ( aa 1668–1683 ) in Daple shares a high degree of similarity to previously reported GBA motifs found in proteins ( Garcia-Marcos et al . , 2009 , 2011b ) and synthetic peptides ( Johnston et al . , 2005; Austin et al . , 2008 ) with GEF activity towards Gαi proteins ( Figure 1B ) . As a first step to investigate the functionality of this GBA motif , we carried out co-immunoprecipitation ( IP ) experiments , which revealed that full-length endogenous Daple in HEK293 cells interacts with the trimeric G protein Gαi3 ( Figure 1C ) . We next investigated if the interaction between Daple and G proteins presents the biochemical properties previously reported for other GBA motif sequences , that is , they bind directly to the G protein with submicromolar to low-micromolar affinity when it is in the inactive but not active conformation ( Tall et al . , 2003; Ghosh et al . , 2008 ) . Recombinant purified GST-Gαi3 bound robustly to purified His-Daple CT ( aa 1650–2028 , containing the GBA motif ) when loaded with GDP ( inactive ) but not when loaded with GDP/AlF4− or GTPγS ( both mimic the GTP-bound active G protein ) ( Figure 1D ) . Equivalent results were obtained when lysates of mammalian cells expressing full-length Daple were used in the pulldown assays ( Figure 1E ) . Binding of His-Gαi3-GDP to GST-Daple CT was saturable , and fitting of the data to a one-site binding curve revealed a submicromolar equilibrium dissociation constant ( Kd = 0 . 11 ± 0 . 03 µM , n = 4 ) , indicating a slightly higher affinity of the G protein for Daple than for GIV ( Kd = 0 . 24 ± 0 . 03 µM , n = 4 ) ( Figure 1F ) . 10 . 7554/eLife . 07091 . 003Figure 1 . Daple contains a GBA motif . ( A ) Phylogenetic sequence analysis reveals a conserved motif in Daple similar to GIV's Gα-binding and activating ( GBA ) motif within an otherwise highly divergent C-terminal domain . Sequences of GIV and Daple from different species were aligned and the degree of identity at each position plotted . A high degree of identity is observed in the N-terminal region ( <∼aa 1400 ) , whereas the C-terminal domain ( >aa 1400 ) is highly divergent . The peak of highest identity ( red box ) within the C-terminal domain corresponds to the GBA motif ( enlarged on the right ) . ( B ) Daple's putative GBA motif is similar to known GBA sequences . Alignment of the putative GBA motif of Daple with the natural GBA sequences of GIV , Calnuc and NUCB2 , and the synthetic GBA sequences of KB-752 and GSP peptides . Consensus is shown below ( ψ = hydrophobic , x = any ) . ( C ) Full-length Daple binds to Gαi3 in cells . Equal aliquots of lysates of HEK293 cells expressing Gαi3-FLAG were incubated with anti-FLAG mAb or control IgG and protein G beads . Immune complexes were analyzed for Daple and Gαi3 ( FLAG ) by immunoblotting ( IB ) . Gβ was monitored as positive Gαi3-binding control . ( D ) Purified Daple binds directly to inactive but not active Gαi3 . Purified , recombinant GST-Gαi3 preloaded with GDP ( inactive ) , GDP + AlF4− ( active ) , or GTPγS ( active ) and immobilized on glutathione-agarose beads was incubated with purified His-Daple-CT ( aa 1650–2028 , containing the putative GBA motif ) as indicated . Resin-bound proteins were eluted , separated by SDS-PAGE and analyzed by Ponceau S-staining and IB with the indicated antibodies . No binding to GST alone was detected . ( E ) Full-length Daple expressed in cells binds preferentially to inactive vs active Gαi3 . Purified , recombinant GST-Gαi3 preloaded with GDP ( inactive ) or GDP + AlF4− ( active ) and immobilized on glutathione-agarose beads was incubated with cell lysates of Cos7 cells expressing full-length myc-Daple as indicated . Bound proteins were analyzed for Daple ( myc ) and Gβ by IB as in D . Binding of Gβ to inactive but not active Gαi3 was used as positive control . No binding of myc-Daple or Gβ to GST alone was detected . ( F ) Daple and GIV bind to Gαi3 with comparable submicromolar affinities . Inset , Purified GST-Daple-CT and GST-GIV-CT ( aa 1671–1755 , containing the GBA motif ) immobilized on glutathione-agarose beads were incubated with increasing amounts ( 0 . 01–3 µM ) of purified His-Gαi3 ( GDP-loaded ) and binding analyzed by IB as described in ( D ) . No binding to GST alone was detected at the highest His-Gαi3 concentration tested . Graph , Gαi3 binding was quantified by measuring band intensities and data fitted to a single-site binding hyperbola ( Daple = BLUE , GIV = RED ) to determine the equilibrium dissociation constants ( Kd ) . Mean ± S . E . M of four independent experiments . ( G ) Daple binds to all three Gαi subunits . Binding of His-Daple-CT to GST-fused Gαi1 , Gαi2 , or Gαi3 in the inactive or active conformations was analyzed exactly as described in ( D ) . ( H ) Daple selectively binds to Gαi , but not Gαo . Binding of His-Daple-CT to GST-fused Gαi3 or Gαo in the inactive or active conformations was analyzed exactly as described in ( D ) . ( I ) Daple binds to Gαi3 mutants that do not bind to other GBA proteins . Table summarizing the binding properties of Gαi3 K248M and W258F mutants to Daple ( from Figure 1—figure supplement 1 ) and GIV or Calnuc ( Garcia-Marcos et al . , 2010 , 2011b ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 00310 . 7554/eLife . 07091 . 004Figure 1—figure supplement 1 . Daple binds mutants of Gαi3 that do not bind GIV ( W258F ) or Calnuc ( K248M ) . Purified , recombinant GST-Gαi3 ( WT and mutants ) preloaded with GDP and immobilized on glutathione-agarose beads was incubated with purified His-Daple-CT ( aa 1650–2028 ) as indicated . Resin-bound proteins were eluted , separated by SDS-PAGE , and analyzed by Ponceau S-staining and IB with anti-His antibodies . No binding to GST alone was detected . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 004 Another common feature among previously reported GBA motifs is their high-G protein specificity , that is , they not only bind preferentially to Gi subfamily members but can discriminate within this subfamily by binding to Gαi subunits but not to the close homologue Gαo ( ∼75% overall similarity to Gαi1/2/3 subunits ) ( Slep et al . , 2008 ) . We found that this is also the case for Daple because it interacts with Gαi1 , Gαi2 , and Gαi3 ( although binding to Gαi2 is partially reduced compared to Gαi1 and Gαi3 ) ( Figure 1G ) but not with Gαo ( Figure 1H ) . Despite these biochemical properties shared with related GBA motifs , we found that binding of Daple to Gαi has unique structural determinants that differentiate it from other proteins with a GBA motif , that is , GIV and Calnuc . We found that mutants of Gαi3 that were previously shown ( Garcia-Marcos et al . , 2010 , 2011b ) to be incapable of binding to GIV or Calnuc ( i . e . , W258F or K248M , respectively ) retain their ability to bind Daple ( Figure 1I , Figure 1—figure supplement 1 ) . This result indicates that the Daple–Gαi3 interface has unique molecular features that provide specificity by making it different from other GBA motif-G protein interactions . Taken together , these results demonstrate that Daple possesses a GBA motif , and that its interaction with G proteins presents all the biochemical features , that is , G protein activation status dependence , affinity and specificity , characteristic of a GBA motif-containing protein . To gain insights into the interface between Daple and Gαi proteins , we took advantage of the previously published atomic structure of KB-752 , a synthetic GEF peptide similar to the GBA motif ( Figure 1A ) , in complex with Gαi1 ( Johnston et al . , 2005 ) . We used this structure as a template to build a homology model of the complex between the GBA motif of Daple and Gαi3 ( Figure 2A ) . Our first prediction based on this model was that Daple would bind to a hydrophobic cleft on the G protein located between the switch II ( SwII ) region and the α3 helix . This seemed to be the case because two molecules known to bind onto the SwII/α3 cleft , that is , the synthetic GEF peptide KB-752 ( Figure 2—figure supplement 1A ) and His-GIV-CT ( aa 1660–1870 , containing its GEF motif ) ( Figure 2—figure supplement 1B ) , competed with His-Daple-CT for binding to GST-Gαi3 . We further substantiated the identity of the binding pocket using site-directed mutagenesis . Analysis of our homology model suggested that a major molecular contact is established by the hydrophobic interaction between the aromatic residues W211 and F215 located in the SwII region of Gαi3 and Daple's F1675 ( Figure 2A ) . Binding of His-Daple-CT to GST-Gαi3 was dramatically impaired upon mutation of W211 or F215 to Alanine ( Ala; A ) ( Figure 2B ) , indicating that these hydrophobic residues of the SwII/α3 cleft serve as a docking site for Daple . Importantly , W211A and F215A mutations have been previously shown not to disturb the native biochemical properties of Gαi proteins ( Thomas et al . , 2004 ) , and therefore , their inability to bind Daple is not a consequence of an overall defect in G protein folding or function . Furthermore , mutation of Daple's F1675 , the residue in its GBA motif predicted to interact with W211 and F215 of the G protein ( Figure 2A ) to Ala abolished GST-Gαi3 binding to either recombinant His-Daple-CT ( Figure 2C ) or full-length myc-Daple expressed in mammalian cells ( Figure 2D ) . Equivalent results were obtained in co-IP experiments in that binding of full-length myc-Daple and Gαi3 co-expressed in mammalian cells was dramatically impaired upon mutation of F1675 to A ( Figure 2E; henceforth referred to as FA ) . Taken together , these results demonstrate that Daple utilizes its GBA motif to bind onto the SwII/α3 hydrophobic cleft of Gαi3 . 10 . 7554/eLife . 07091 . 005Figure 2 . Daple binds and activates Gαi3 in vitro and in vivo via its GBA motif . ( A ) Prediction of molecular contacts critical for the Daple-Gαi interaction . Homology-based model of Daple's GBA motif ( Red ) bound to Gαi3 ( green = Switch II , blue = ras-like domain , yellow , all-helical domain ) with an enlarged section depicting a putative hydrophobic contact between Daple's F1675 and Gαi3's W211/F215 . ( B ) Mutation of residues in the SWII region of Gαi3 disrupts Daple binding . Binding of His-Daple-CT to GST-Gαi3 WT , W211A , or F215A was analyzed exactly as described in Figure 1D . ( C ) Mutation of Daple F1675 to A abrogates Gαi3 binding . Binding of His-Daple-CT WT or F1675A ( FA ) to GST-Gαi3 was analyzed exactly as described in Figure 1D . ( D ) F1675A mutation disrupts binding of full-length Daple expressed in cells to Gαi3 . Myc-Daple WT or F1675A ( FA ) was expressed in Cos7 cells and binding to GST-Gαi3 analyzed exactly as described in Figure 1E . ( E ) Binding of full-length Daple to Gαi3 in cells is abolished upon F1675A mutation . Lysates of Cos7 cells expressing Gαi3-FLAG and myc-Daple-WT or F1675A ( FA ) were incubated with anti-FLAG mAb and subsequently with protein G beads . Immune complexes were analyzed for Daple ( myc ) and Gαi3 ( FLAG ) by IB . Gβ was monitored as positive Gαi3-binding control . ( F ) Daple accelerates the rate of Gαi3 steady-state GTPase activity . The steady-state GTPase activity of His-Gαi3 alone ( black ) or in the presence of 2 µM His-Daple-CT ( blue ) was determined by measuring the production of [32P]Pi at different time points as described in ‘Materials and methods’ . One experiment representative of 3 is shown . ( G ) Daple WT but not F1675A ( FA ) accelerates the rate of Gαi3 steady-state GTPase activity in a dose-dependent manner . The steady-state GTPase activity of His-Gαi3 was determined in the presence of increasing concentrations ( 0–2 µM ) of His-Daple-CT WT ( blue ) or His-Daple-CT FA ( red ) by measuring the production of [32P]Pi at 15 min . Mean ± S . E . M of five independent experiments . ( H ) Daple WT but not F1675A dose-dependently accelerates the rate of GTPγS binding to Gαi3 . GTPγ35S binding to His-Gαi3 at 15 min was determined in the presence of increasing concentrations ( 0–2 µM ) of His-Daple-CT WT ( blue ) or His-Daple-CT FA ( red ) . Mean ± S . E . M of four independent experiments . ( I ) Schematic for the Gαi1-intYFP and Gβ1-CFP constructs used as paired Fӧrster resonance energy transfer ( FRET ) probes in J , K , and L . ( J–L ) Heterotrimers of Gi1 ( Gαi1 and Gβ1γ2 ) are dissociated at the plasma membrane ( PM ) in control ( J , sh Luc ) , but not Daple-depleted ( K , sh Daple 1 ) HeLa cells after Wnt5a stimulation . Control ( Left ) or Daple-depleted ( Right ) HeLa cells ( sh Daple 1 described in Figure 2—figure supplement 1A , B ) cotransfected with Gαi1-intYFP , Gβ1-CFP , and Gγ2 were maintained overnight in 0 . 2% FBS and subsequently stimulated with 0 . 1 mg/ml Wnt5a and analyzed for FRET by confocal microscopy . Representative freeze-frame images from live-cell movies are shown , which display intensities of acceptor emission due to FRET in each pixel . Activation of Gi , as determined by the loss of interaction ( i . e . , FRET ) between Gαi1 and Gβ1γ2 was observed exclusively after ligand stimulation ( compare t0 and t5 ) in control ( J ) , but not in Daple-depleted HeLa cells ( K ) . ( L ) Bar graphs display differences between FRET intensities observed in control vs Daple-depleted cells in ( J , K ) . Error bars representing mean ± S . D . of 5 randomly chosen regions of interest ( ROIs ) at the PM per cell , from 4 to 5 cells per experiment , from three independent experiments . ( M ) HeLa cells expressing Daple-WT , but not Daple-F1675A activate Gαi3 in response to Wnt5a stimulation , as determined by immunoprecipitation ( IP ) with conformationally-sensitive anti-Gαi:GTP antibodies . Daple-depleted HeLa cells transiently transfected with myc-Daple WT or F1675A ( FA ) were serum-starved and treated ( + ) or not ( − ) with 0 . 1 mg/ml Wnt5a for 20 min were subjected to immunoprecipitation with antibodies that selectively recognize active Gαi subunits in their GTP-bound state . Immune complexes ( top ) and lysates ( bottom ) were analyzed for active Gαi3:GTP and total Gαi3 by immunoblotting ( IB ) with anti-Gαi3 antibody . ( N ) HeLa cells expressing Daple-WT , but not Daple-F1675A inhibit cAMP in response to Wnt5a stimulation , as determined by radioimmunoassay . HeLa cells transiently transfected with myc-Daple WT or F1675A ( FA ) incubated with forskolin and PDE inhibitors for 10 min , treated ( + ) or not ( − ) with 0 . 1 mg/ml Wnt5a for 20 min and cAMP levels quantified as detailed in ‘Materials and methods’ . Mean ± S . D . of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 00510 . 7554/eLife . 07091 . 006Figure 2—figure supplement 1 . Binding of Daple to Gαi triggers activation of Gi at the PM after Wnt5a stimulation . ( A , B ) Daple competes for binding to Gαi3 with peptides/proteins that dock onto the switchII/α3 cleft of the G protein . ( A ) Purified , recombinant GST-Gαi3 preloaded with GDP and immobilized on glutathione-agarose beads ( ∼0 . 3 mM ) was incubated with a fixed concentration ( ∼0 . 2 mM ) of purified His-Daple-CT ( aa 1650–2028 ) in the presence of the indicated concentrations of KB-752 or a control peptide . Resin-bound proteins were eluted , separated by SDS-PAGE , and analyzed by Ponceau S-staining and IB with anti-His antibodies . No binding to GST alone was detected . One experiment representative of 3 is shown . ( B ) Analogous experiments to those described in A were carried out using His-GIV-CT WT and His-GIV-CT F1685A ( as negative control ) instead of peptides . ( C–E ) Daple is essential for activation of trimeric Gi at the PM after Wnt5a stimulation . ( C , D ) Two independent shRNA sequences targeting the 3′ UTR of the gene efficiently ( <90% ) deplete Daple mRNA ( A ) and protein ( B ) from HeLa cells . The knock-down efficiency was assessed by comparing Daple mRNA by quantitative PCR ( qPCR ) ( C ) or protein by IB ( D ) on HeLa cells stably expressing two Daple-targeting ( shDaple1 and shDaple2 ) or control ( shRNA targeting luciferase [shLuc] ) shRNA sequences . ( E ) Control ( Luc shRNA ) or Daple-depleted ( Daple shRNA1 ) HeLa cells co-transfected with Gαi1-YFP , Gβ1-CFP , and Gγ2 were starved overnight in media containing 0 . 2% FBS prior to stimulation with Wnt5a and analyzed for FRET using confocal microscope . Representative freeze-frame images from live-cell movies are shown , which display acceptor ( Gαi1-YFP ) , donor ( Gβ1-CFP ) and intensities of acceptor emission due to FRET in each pixel ( from left to right ) . Activation of Gi , as determined by the loss of interaction ( i . e . , FRET efficiency ) between Gαi1-YFP and Gβ1-CFP is observed at the PM exclusively after ligand stimulation ( compare t0 and t5 ) in Luc shRNA treated control cells , but not in Daple-depleted cells . Red circle = ROI at the PM . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 006 GEFs are defined by their ability to accelerate the rate of nucleotide exchange . To determine if binding of Daple to Gαi3 accelerates the rate of nucleotide exchange on the G protein , we carried out two well-established enzymatic assays—the steady-state GTPase assay , which indirectly reflects the rate of nucleotide exchange ( Mukhopadhyay and Ross , 2002 ) , and the GTPγS-binding assay , which directly measures the rate of nucleotide exchange . We found that incubation of His-Gαi3 with His-Daple-CT accelerated the rate of steady-state GTP hydrolysis ∼threefold over the basal activity ( Figure 2F ) . This acceleration of Gαi3 steady-state GTPase activity by Daple was dose-dependent , with an EC50 of 0 . 25 ± 0 . 06 µM ( similar to the estimated Kd for the Daple–Gαi3 interaction , Figure 1F ) , and was greatly diminished ( >90% ) in parallel reactions in which His-Daple-CT WT was replaced by the Gαi3 binding-deficient mutant F1675A ( Figure 2G ) . We further validated that Daple is a bona fide GEF for Gαi using GTPγS-binding assays , which showed that the initial rate of nucleotide binding by His-Gαi3 was increased by His-Daple-CT in a dose-dependent manner , but it was not significantly affected by His-Daple-CT FA ( Figure 2H ) . Thus , Daple activates Gαi proteins in vitro by virtue of a GEF activity associated to its GBA motif . Next , we asked whether Daple activates G proteins in mammalian cells responding to Wnt5a . To this end , we generated HeLa cells stably expressing Daple-targeting shRNA sequences under the control Cre recombinase activity ( see Supplemental Materials for the rationale behind the choice of this cell type and others in subsequent sections ) . Upon Cre treatment , two independent shRNA sequences reduced Daple mRNA levels by >80% ( Figure 2—figure supplement 1C ) and the Daple protein to virtually undetectable levels ( Figure 2—figure supplement 1D ) compared to cells expressing a control shRNA targeting luciferase ( shLuc ) . We used these cells in a previously validated assay in which activation of Gi is monitored by dissociation of fluorescently tagged Gαi and Gβγ subunits with a resultant loss of Förster resonance energy transfer ( FRET ) ( Janetopoulos et al . , 2001; Bunemann et al . , 2003; Gibson and Gilman , 2006 ) ( Figure 2I–L ) . When control HeLa cells co-expressing Gαi1-YFP ( internal tag ) , CFP-Gβ1 ( N-terminal tag ) , and Gγ2 ( untagged ) were stimulated with Wnt5a , we observed a significant loss of FRET , that is , Gi heterotrimer dissociated into Gαi-YFP and CFP-Gβγ subunits at the plasma membrane ( PM ) within 5 min as determined by a significant drop in FRET efficiency from 0 . 36 ± 0 . 08 to 0 . 17 ± 0 . 06 ( Figure 2J , L , Figure 2—figure supplement 1E ) , indicating that Gi is activated in response to Wnt5a . No significant drop in FRET was observed in Daple-depleted cells ( Figure 2K , L; Figure 2—figure supplement 1E ) , indicating that donor-CFP-Gβγ and acceptor-Gαi-YFP subunits continued to interact ( i . e . , Gi heterotrimers remained intact ) at the PM regardless of Wnt5a stimulation , and that Gαi remained inactive . These results demonstrate that Daple is essential for activation of Gi upon Wnt5a stimulation . Next , we asked if the GBA motif in Daple is essential for activation of Gαi in cells responding to Wnt5a . To this end , we analyzed activation of Gαi in HeLa cells expressing Daple-WT or FA using an anti-Gαi:GTP mAb that specifically recognizes Gαi in a GTP-bound active conformation ( Lane et al . , 2008a ) . Previous work by others ( Lane et al . , 2008a ) and by us ( Lopez-Sanchez et al . , 2014 ) has demonstrated that this antibody can specifically recognize active Gαi in cells . When we immunoprecipitated Gαi from HeLa cells , active Gαi3 was immunoprecipitated exclusively after Wnt5a stimulation in cells expressing Daple-WT ( Figure 2M ) , but not in those expressing Daple-FA . These results indicate that an intact GBA motif is essential for Daple to activate Gαi3 after Wnt5a stimulation . To further substantiate this , we determined the intracellular levels of cAMP as a measure of the activity of adenylyl cyclase , which is directly inhibited by active Gαi subunits . We found that Wnt5a stimulation suppressed cAMP levels by ∼50% in HeLa cells expressing Daple-WT , but no such suppression occurred in cells expressing Daple-FA ( Figure 2N ) . Taken together , these results demonstrate that Daple is a bona fide GEF that activates Gαi proteins in vitro and in cells responding to Wnt5a via its GBA motif . In addition to modulation of cellular cAMP , another major consequence of activating Gαi subunits is the release of free Gβγ subunits , which in turn modulates a wide array of signaling pathways ( Smrcka , 2008 , 2013 ) . Comparative analysis of the crystal structure of the Gαi1·βγ trimer and the homology model of Daple's GBA motif bound to Gαi3 revealed that Gβγ and Daple have overlapping binding sites on Gαi subunits ( Figure 3A ) . Based on this , we reasoned that binding of Daple to Gαi will displace Gβγ from trimeric Gαi·βγ complexes . We found that is indeed the case because His-Daple-CT WT , but not the FA mutant ( which cannot bind Gαi ) , displaced Gβγ from a pre-assembled complex with GST-Gαi3 ( Figure 3B ) . The IC50 for this displacement was 0 . 16 ± 0 . 01 µM ( Figure 3C ) , which is consistent with the estimated affinity of Daple for Gαi3 ( Figure 1F ) . 10 . 7554/eLife . 07091 . 007Figure 3 . Daple's GBA motif triggers the release of ‘free’ Gβγ subunits , which in turn enhance Rac1 and PI3K-Akt signaling . ( A ) Daple's GBA motif and Gβγ subunits are predicted to dock onto an overlapping binding site on Gαi . Binding areas ( in red ) for Daple ( left ) or Gβγ ( right ) on Gαi ( solid gray ) were extracted from a homology-based model of Daple-Gαi3 and the crystal structure of the Gαi1·Gβγ complex ( Protein Data Bank [PDB]: 1GG2 ) , respectively . ( B , C ) Daple displaces Gβγ subunits from Gαi3 via its GBA motif . GST-Gαi3·Gβγ preformed complexes immobilized on glutathione beads were incubated with increasing concentrations of His-Daple-CT WT or F1675A ( FA ) . Bound proteins were analyzed by IB ( B ) and Gβγ binding data fitted to a single-site competition curve ( C ) . Mean ± S . E . M . of three independent experiments . ( D , E ) Activation of Rac1 is impaired in Daple-depleted HeLa cells . Control ( shLuc ) or two clones of Daple-depleted HeLa cell lines ( sh Daple 1 and 2 ) ( described in Figure 2—figure supplement 1A , B ) were incubated in 2% serum media ( D ) or starved and treated ( + ) or not ( − ) with Wnt5a ( 0 . 1 mg/ml ) for 5 min ( E ) and analyzed for Rac1 activation by pulldown assays using GST-PBD . ( F ) Activation of Rac1 is impaired in cells expressing Daple-F1675A ( FA ) mutant compared to those expressing Daple-WT . Daple-depleted ( sh Daple 1 ) HeLa cells transiently transfected with myc-Daple-WT or FA were starved and stimulated with Wnt5a and analyzed for Rac1 activation as in E . ( G , H ) Daple's GBA motif is required for activation of PI3K-Akt signaling in HeLa cells , as determined by phosphorylation of Akt at S473 . Daple-depleted ( sh Daple 1 ) HeLa cells transiently transfected with myc-Daple WT or F1675A ( FA ) were incubated in a 2% serum media ( G ) or in a 0 . 2% serum media overnight and treated ( + ) or not ( − ) with 0 . 1 mg/ml Wnt5a for 5 min ( H ) prior to lysis . Equal aliquots of whole-cell lysates were analyzed for Akt phosphorylation ( pAkt S473 ) by IB . ( I , J ) Inhibition of Gβγ signaling impairs Daple-dependent activation of Rac1 and Akt . Daple-depleted ( sh Daple 1 ) HeLa cells transiently transfected with myc-Daple WT were treated with DMSO , 10 µM of the Gβγ inhibitor gallein or its inactive analog fluorescein for 6 hr , as indicated , and analyzed for Rac1 ( I ) or Akt ( J ) activation by IB or pulldown assays , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 007 To determine if the ‘free’ Gβγ released by Daple's GBA motif modulated cellular signaling , we analyzed two signaling pathways , Rac1 and PI3K-Akt because previous studies have demonstrated a direct and critical role of ‘free’ Gβγ subunits in enhancement of these signals ( Leopoldt et al . , 1998; Welch et al . , 2002; Niu et al . , 2003; Ueda et al . , 2008; Xu et al . , 2012 ) , and because they represent major signals downstream of the non-canonical Wnt pathway ( Kawasaki et al . , 2007; Nishita et al . , 2010; Anastas et al . , 2014 ) . Rac1 activity , as determined in pulldown assays using the p21 binding domain ( PBD ) of PAK1 ( Knaus et al . , 2007 ) , was suppressed in Daple-depleted HeLa cells both at steady-state in the presence of low serum ( Figure 3D ) as well as after Wnt5a stimulation ( Figure 3E ) . Furthermore , Wnt5a triggered activation of Rac1 in cells expressing Daple-WT , but not the FA mutant ( Figure 3F ) . These findings indicate that Daple and its GBA motif are required for the efficient activation of Rac1 activity . Similarly , we found that activation of Akt , as determined by phosphorylation of the kinase at Ser473 was enhanced in cells expressing Daple WT , but not the FA mutant , both at steady-state in the presence of low serum ( Figure 3G ) , as well as after Wnt5a stimulation ( Figure 3H ) , indicating that Daple's GBA motif is essential for enhancement of PI3K/Akt signaling . To pinpoint whether the enhanced Rac1 and Akt signals are triggered directly by ‘free’ Gβγ subunits that are released by Daple , we used a Gβγ inhibitor , that is , Gallein , that selectively blocks the interaction between Gβγ with key downstream effectors ( Bonacci et al . , 2006; Lehmann et al . , 2008; Smrcka et al . , 2008; Urano et al . , 2008; Seneviratne et al . , 2011 ) . We found that incubation of HeLa cells expressing Daple WT with Gallein effectively inhibited both Rac1 ( Figure 3I ) and Akt ( Figure 3J ) activities to levels observed in cells expressing Daple FA , whereas the inactive analog , Fluorescein had no such effect . These results indicate that Daple enhances Rac1 and Akt signaling at least in part by facilitating the release of ‘free’ Gβγ subunits , which subsequently trigger signaling via downstream intermediates . In summary , these results indicate that the dissociation of Gαi·βγ heterotrimers triggered upon Wnt5a stimulation by Daple's GBA motif sets off at least two major immediate events within the G protein signaling cascade— ( 1 ) GTP-loading of Gαi subunits , which subsequently inhibits the adenylyl cyclase/cAMP pathway and ( 2 ) release of Gβγ subunits that trigger the activation of non-canonical Wnt signaling pathways including Rac1 and PI3K-Akt . Because Daple enhances non-canonical Wnt signaling that is initiated by FZDRs , we asked how Daple may modulate signals downstream of these receptors and wondered if they interact . We tested several purified GST-tagged FZDR cytoplasmic tail proteins for their ability to bind Daple from Cos7 lysates ( Figure 4—figure supplement 1A–C ) . More specifically , we tested FZDRs 1–7 , which belong to 3 evolutionary distinct subfamilies within the FZDR superfamily ( Figure 4—figure supplement 1A ) containing divergent sequences in their C-terminus that determine which regulatory proteins are assembled ( Schulte , 2010; Dijksterhuis et al . , 2014 ) . Daple bound robustly to FZD7R , and only weakly to others , indicating that Daple may engage preferentially with FZD7R ( Figure 4—figure supplement 1B , C ) . Based on this result , we used FZD7R in all subsequent assays to further analyze the interaction between Daple and FZDR . We found that both endogenous and exogenously expressed Daple and Gαi3 co-immunoprecipitated with FZD7R exclusively after Wnt5a stimulation ( Figure 4A , Figure 4—figure supplement 2A ) , indicating that Daple and Gαi3 form complexes with ligand-activated FZD7R . Immunofluorescence studies revealed that in starved HEK293 cells , Daple is cytosolic in distribution , but in cells responding to Wnt5a Daple is localized at the PM , where it colocalized with FZD7R ( Figure 4B ) . These findings suggest that the ligand-dependent interaction between FZD7R and Daple we see in 4A occurs at the PM . 10 . 7554/eLife . 07091 . 008Figure 4 . The C-terminus of Daple directly binds ligand-activated FZDRs and triggers the assembly of FZDR-Gαi complexes at the PM . ( A ) Daple and Gαi3 co-immunoprecipitate with FZD7R after Wnt5a stimulation . HeLa cells cotransfected with myc-Daple WT and HA-FZD7 were starved and stimulated with Wnt5a as in 3G . Equal aliquots of lysates ( bottom ) were then incubated with anti-HA mAb and subsequently with protein G beads . Immune complexes ( top ) were analyzed for myc ( myc-Daple ) and endogenous Gαi3 by IB . ( B ) Daple is recruited to the PM after Wnt5a stimulation , where it colocalizes with FZD7R . HEK293 cells expressing FZD7-CFP ( pseudocolored green ) were grown on cover slips coated with Poly-D-Lysine , starved for 24 hr ( 0% FBS ) and treated with 0 . 1 mg/ml Wnt5a as in 4A . Cells were fixed and stained for Daple ( red ) and analyzed by confocal microscopy . ( C ) The C-terminal region ( 1650–2028 aa ) is sufficient for Daple to bind FZD7R . Lysates of Cos7 cells expressing full-lenght myc-Daple-WT or myc-Daple-CT ( 1650–2028 aa ) were incubated with recombinant GST-FZD7-CT immobilized on glutathione-agarose beads in pulldown assays . Bound Daple ( myc ) was analyzed by IB . ( D ) Daple directly binds FZD7R and the extreme C-terminus ( 1881–2028 ) is essential for the interaction . His-Daple-CT ( 1650–2028 aa ) or a shorter fragment of Daple-CT ( 1650–1880 aa ) was incubated in pulldown assays with immobilized GST-FZD7-CT exactly as above . Bound Daple-CT ( His ) was analyzed by IB . ( E ) Daple's GBA motif is required for enhanced binding of Gαi3 to cytoplasmic tails of FZD7R in vitro . His-Gαi3 preloaded with GDP was incubated with immobilized GST-FZD7-CT , either alone ( lane 2 ) or in the presence of His-Daple-CT ( 1650–2028 aa ) WT ( lane 3 ) or FA ( lane 4 ) in pulldown assays as described in D . Bound Gαi3 and Daple-CT were detected by IB . ( F ) Daple's GBA motif is essential for the co-IP of Gαi3 with ligand-activated FZD7Rs . HeLa cells cotransfected with HA-FZD7 and myc-Daple-WT or FA were starved and subsequently stimulated with Wnt5a prior to lysis as in A . Equal aliquots of lysates ( bottom ) were incubated with anti-HA antibodies and subsequently with protein G beads . Immune complexes were analyzed for the presence of Gαi3 by IB . ( G–I ) Wnt5a stimulates formation of FZD7R-Gαi3 complexes at the PM in HEK293T cells . ( G ) Schematic of the FRET probes used in H . ( H ) HEK293 cells were cotransfected with FZD7-CFP and Gαi3-YFP , starved , and subsequently stimulated with Wnt5a and analyzed for FRET using confocal microscopy . Image panels display CFP , YFP , and intensities of acceptor emission due to FRET in each pixel . FRET was observed after Wnt5a stimulation ( right ) . ( I ) Bar graphs display FRET efficiency observed at the PM in starved vs Wnt5a stimulated cells in H . Error bars represent mean ± S . D . The analysis represents 5 randomly chosen ROIs at the PM per cell , from 4 to 5 cells per experiment , from three independent experiments . ( J , K ) Daple's GBA motif is essential for the assembly of FZD7R-Gαi3 complexes at the PM . HEK293T cells were cotransfected with FZD7-CFP , Gαi3-YFP and myc-Daple ( WT or FA ) , starved , and subsequently stimulated with Wnt5a prior to fixation . Fixed cells were stained for Daple ( 632 nm , far red; see Figure 4—figure supplement 2 ) and analyzed for FRET using confocal microscope . Image panels display the intensities of acceptor emission due to FRET in each pixel . FRET was observed in cells expressing Daple-WT , but not in cells expressing Daple-FA . ( K ) Bar graphs display the FRET efficacy observed in Daple WT vs Daple FA cells before ( − ) and after ( + ) Wnt5a stimulation . Error bars representing mean ± S . D . The analysis was done exactly as in H , I . ( L ) Schematic summary . Upon stimulation with Wnt5a , Daple's C-terminus enables the formation of FZD7R-Daple-Gαi3 complexes at the PM . Two distinct interaction modules present in-tandem within the C-terminus of Daple , the GBA motif , and the FZD-binding domain are essential for the formation of such complexes . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 00810 . 7554/eLife . 07091 . 009Figure 4—figure supplement 1 . Daple preferentially binds the cytoplasmic tail of the FZD7R . ( A ) A sequence homology-based cluster tree of vertebrate Frizzled receptors ( FZDRs ) is shown . The FZD ( IUPHAR nomenclature ) family roughly clusters into four distinct families based on sequence identity ( modified from Verkaar and Zaman , 2010 ) : ( I ) FZD 1 , 2 , and 7; ( II ) Frizzled-5 and Frizzled-8; ( III ) Frizzled-3 and Frizzled-6; ( IV ) Frizzled-4 , Frizzled-9 , and Frizzled-10 . The Smoothened ( SMO ) receptor is a distant relative of FZD receptors that functions in Hedgehog signal transduction . ( B , C ) Lysates of cells expressing myc-Daple full length was used as source of Daple in pulldown assays with immobilized recombinant GST-tagged C-termini of various FZDRs . Bound proteins were analyzed for Daple by IB . Full-length Daple binds preferentially to the cytoplasmic tail of FZD7R , to an intermediate extent to the cytoplasmic tail FZD6R and only weakly the cytoplasmic tails of other FZDRs . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 00910 . 7554/eLife . 07091 . 010Figure 4—figure supplement 2 . Daple binds to the C-terminus of FZD7R and links Gαi to ligand-activated receptors . ( A ) HEK cells expressing HA-tagged FZD7R were starved for 24 hr ( 0% FBS ) and stimulated with Wnt5a for 5 min as indicated prior to lysis . IP was carried out on lysates with anti-HA or control mouse IgGs and protein G beads . Equal aliquots of lysates ( bottom ) and immune complexes ( top ) were analyzed for Daple , Gαi3 , FZD7R ( HA ) , and tubulin by IB . Endogenous Daple and Gαi3 are recruited to FZD7R exclusively after Wnt5a stimulation . ( B ) Lysates of Cos7 cells expressing myc-tagged Daple-WT or GBA-deficient ( FA ) and PBM-deficient ( ΔPBM ) mutants were used as source of Daple in pulldown assays with GST-FZD7-CT immobilized on glutathione beads . Bound proteins were analyzed for Daple by IB . Mutant Daple proteins bound FZD7 as efficiently as Daple-WT . ( C ) HEK cells co-transfected with FZD7-CFP , Gαi3-YFP , and Daple-WT or FA were starved and subsequently stimulated with Wnt5a and analyzed for FRET using confocal microscope . Representative freeze-frame images from live-cell movies are shown , which display ( from left to right ) donor ( FZD7-CFP ) , acceptor ( Gαi3-YFP ) , Daple ( far-red; 632 nm ) and intensities of acceptor emission due to FRET in each pixel . Interaction ( i . e . , FRET ) is observed exclusively after Wnt5a stimulation in cells expressing Daple-WT , but not Daple-FA . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 010 Next , we asked which region of Daple interacts with FZD7R and whether the binding is direct . We found that the C-terminal ∼380 aa of Daple ( aa 1650–2028 ) was sufficient to interact with GST-FZD7R-CT as efficiently as the full-length Daple ( Figure 4C ) . Pulldown assays with the purified , recombinant His-tagged identical segment ( aa 1650–2028 ) of Daple revealed that the binding is direct ( Figure 4D ) . A shorter C-terminal fragment of Daple ( aa 1650–1880 ) , which lacks the ∼150 aa at the extreme C-terminus does not ( Figure 4D ) . Furthermore , the GEF-deficient ( FA ) and the ΔPBM-deficient mutants bound GST-FZD7R-CT as efficiently as Daple WT ( Figure 4—figure supplement 2B ) . These findings demonstrate that— ( 1 ) the FZD7R-Daple interaction is direct; ( 2 ) that the aa 1650–2028 in the C-terminus of Daple is sufficient to mediate the interaction; ( 3 ) that the extreme C-terminal ∼150 aa within the C-terminus ( 1881–2029 ) is essential for the interaction , whereas both the GBA and PBM motifs are dispensable . Because Gαi3 co-immunoprecipitated with ligand-activated FZD7R-Daple complexes ( Figure 4A ) , we asked if the interaction observed is direct or mediated by Daple . We first carried out GST pulldown assays with recombinant His-Gαi3 and the GST-tagged cytoplasmic tail of FZD7R . We found that Gαi3 bound weakly to GST-FZD7R-CT ( Figure 4E; lane 2 ) ; however , binding was increased ∼fivefold in the presence of recombinant Daple-CT WT , but not the FA mutant . This raised the possibility that the ligand-dependent interaction between Gαi and FZD7 we see in cells ( Figure 4A ) is indirect and mediated by the GBA motif in Daple . Indeed , ligand-dependent recruitment of Gαi3 to FZD7R occurred exclusively in cells expressing full-length Daple-WT ( where GBA motif is intact ) , but not the FA mutant ( Figure 4F ) . Next , the spatiotemporal dynamics of ligand-dependent complex formation between FZD7R and Gαi3 was analyzed in HEK293 cells by FRET imaging ( Figure 4G ) . We found that the probe-pair FZD7R-CFP and Gαi3-YFP interact at the PM within 5 min after ligand stimulation ( FRET efficiency = 0 . 25 ± 0 . 06 ) ( Figure 4H , I ) . No such interaction was observed in starved cells ( FRET efficiency = 0 . 04 ± 0 . 01 ) , demonstrating that Wnt5a triggers the assembly of complexes between ligand-activated FZD7R and Gαi3 at the PM . Furthermore , ligand-dependent assembly of such complexes occurred in cells expressing Daple-WT , but not the FA mutant ( Figure 4J , K; Figure 4—figure supplement 2C ) , further confirming that Daple serves as in intermediate protein that couples FZD7R to Gαi3 . Although the interaction between ligand-activated FZD7R and Daple does not require the GBA motif ( Figure 4—figure supplement 2B ) , the recruitment of Gαi into the complex requires a functionally intact GBA to trigger the formation of FZD7R ( active ) -Daple-Gαi complexes . Thus , two non-overlapping modules in-tandem within Daple's C-terminus cooperate to facilitate the assembly of FZD7R ( active ) -Daple-Gαi ternary complexes ( Figure 4L ) — ( 1 ) a GBA motif that binds Gαi and ( 2 ) a stretch of C-terminus ( aa 1681–2024 ) is essential for binding to FZD7R . Previous studies have demonstrated that Dvl , a key scaffold protein in the Wnt signaling pathway , interacts with both FZDRs ( Schulte and Bryja , 2007 ) and Daple ( Oshita et al . , 2003 ) and shapes both canonical and non-canonical Wnt signals . Furthermore , Dvl interferes with the engagement of Gi proteins with ligand-activated FZDRs ( Kilander et al . , 2014 ) , suggesting a possible interplay between Dvl and the FZDR-Daple-Gi signaling axis we define here . First , we investigated how the ligand-dependent Daple-FZD7R interaction affects Dvl's ability to bind Daple . We found that Daple co-immunoprecipitated with Dvl exclusively in starved cells and that such complexes were undetectable after stimulation with Wnt5a ( Figure 5A ) , indicating that the dissociation of Daple-Dvl complexes coincides with the assembly of Daple-FZD7R complexes we observed in Figure 4A . Next , we investigated how Daple affects the interaction between Dvl and FZDR . We found that expression of Daple in HEK293 cells reduces Dvl association with FZD7R in pulldown ( Figure 5—figure supplement 1A ) and co-IP experiments ( Figure 5B ) , suggesting that Daple and Dvl may compete with each other for binding to FZD7R . Furthermore , immunofluorescence studies confirmed that localization of Dvl at the PM in cells expressing FZD7R was reduced within 5 min after Wnt5a stimulation ( Figure 5—figure supplement 1B ) , which coincides with the ligand-dependent recruitment of Daple ( Figure 4B ) . We found that Daple and Dvl actually compete for binding to FZD7R because increasing amounts of purified His-Daple-CT ( 1650–2028 ) , but not a shorter fragment ( His-Daple 1650–1880 , which lacks the FZD7R-binding region ) increased the formation of Daple-FZDR complexes and reduced DvlFZD7R complexes ( Figure 5C ) . Furthermore , immunofluorescence studies revealed , that in cells without Daple , stimulation with Wnt5a does not trigger the loss of Dvl from the PM observed in control cells ( Figure 5D ) , suggesting that the competition we observe in vitro ( Figure 5C ) may occur also in cells . Taken together , these results indicate that Daple determines the relocalization of Dvl upon Wnt5a stimulation by displacing the latter from FZDRs . 10 . 7554/eLife . 07091 . 011Figure 5 . Daple competes with Dvl for binding to FZD7R and inhibits the canonical β-catenin/TCF/LEF signaling pathway via the GBA motif . ( A ) Dishevelled ( Dvl ) –Daple complexes are disrupted upon Wnt5a stimulation . HeLa cells cotransfected with myc-Daple-WT and Dvl were incubated in a 0 . 2% serum media overnight , and treated ( + ) or not ( − ) with 0 . 1 mg/ml Wnt5a for 5 min prior to lysis . Equal aliquots of lysates ( bottom ) were incubated in the presence of anti-Dvl mAb and subsequently with protein G beads . Immune complexes ( top ) were analyzed for Daple ( myc ) , Dvl , and Gαi3 by IB . ( B ) Dvl and Daple compete for recruitment to FZD7 receptor in cells . Equal aliquots of lysates of HEK293 cells cotransfected with FZD7-HA with Dvl and/or myc-Daple-WT were incubated with anti-HA mAb and subsequently with protein G beads . Immune complexes were analyzed for Daple and Dvl by IB . ( C ) Daple can displace Dvl bound to the cytoplasmic tail of FZD7R in vitro . Dvl expressed in HEK cells was pre-bound to GST or GST-FZD7CT and subsequently incubated with increasing amounts of recombinant His-Daple-CT proteins as indicated . Bound proteins were analyzed for Daple ( His ) and Dvl by IB . ( D ) Daple is required for the ligand-stimulated dissociation of Dvl from the PM . Control ( sh Luc ) and Daple-depleted ( sh Daple 1 ) Hela cells coexpressing Dvl and FZD7R were starved and stimulated with Wnt5a prior to fixation as in 4B . Fixed cells were stained for Dvl ( red ) and analyzed by confocal microsocpy . Bar = 10 µM . ( E ) Gαi competes with Dvl for binding to Daple in vitro . Equal aliquots of GST or GST-Dvl-PDZ ( immobilized on glutathione beads ) and Daple-CT ( WT or FA ) recombinant proteins were incubated with increasing amounts of purified His-Gαi3 as indicated . Bound ( top ) and unbound ( supernatant; lower ) proteins were analyzed for Daple-CT and Gαi3 ( His ) by IB . GST and GST-Dvl-PDZ were visualized by ponceau staining . ( F ) Depletion of Daple increases the levels of β-catenin . Whole-cell lysates of control ( shLuc ) and Daple-depleted ( shDaple1 and 2 ) HeLa cells were analyzed for β-catenin by IB . ( G ) Bar graphs display quantification of β-catenin in F . Error bars represent mean ± S . D of three independent experiments . ( H ) Daple's GBA motif is required for suppression of β-catenin expression/stability . Whole-cell lysates from HeLa cells transfected with myc-Daple-WT or FA were analyzed for β-catenin expression by IB . Two biological replicates are shown . ( I ) Bar graphs display quantification of β-catenin in H . Error bars represent mean ± S . D of three independent experiments . ( J , K , L ) Daple's GBA motif is required for suppression of Wnt target genes . HeLa cells transfected with myc-Daple-WT or FA were analyzed for SFRP-1 , OPN , AXIN-2 mRNA by qPCR . Results were normalized internally to mRNA levels of the housekeeping gene , GAPDH . Bar graphs display the fold change in each RNA ( y axis ) in cells expressing Daple-FA normalized to the expression in cells expressing Daple-WT . Error bars represent mean ± S . D of three independent experiments . ( M ) Schematic of working model . ( From left to right ) In the absence of Wnt5a ligand , Dvl remains at the PM complexed to inactive FZD7Rs , whereas Daple remains in the cytosol in complex with cytosolic Dvl , and Gαi/βγ trimers at the PM are largely inactive . Upon ligand stimulation , Dvl-Daple complexes dissociate and Daple is recruited to the cytoplasmic tails of activated receptors , Dvl is displaced from the receptor tail by Daple , Daple favors the assembly of receptor-Gαi complexes and triggers the activation of Gαi within these complexes . Activated Gαi and Gβγ subunits trigger signaling via their respective downstream intermediates ( Rac1 , PI3K , and cAMP ) . Another major consequence of these signaling events is suppression of the canonical β-catenin/TCF/LEF signaling pathway , which regulates the transcription of Wnt target genes . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 01110 . 7554/eLife . 07091 . 012Figure 5—figure supplement 1 . Daple competes with Dvl for binding to FZD7R and inhibits the canonical β-caenin/TCF/LEF signaling pathway . ( A ) Equal aliquots of lysates from Cos7 cells expressing Dvl1 alone ( lane 2 ) , myc-Daple alone ( lane 3 ) , or coexpressing both ( lanes 1 , 4 ) were used as source of Daple and Dvl in GST pulldown assays with recombinant , immobilized GST or GST-FZD7-CT . Bound proteins were analyzed for Dvl1 and Daple by immunoblotting ( IB ) . Binding of each protein was higher when expressed alone ( lanes 2 , 3 ) than when co-expressed ( lane 4 ) . ( B ) Dvl loses colocalization with FZD7R at the PM after Wnt5a stimulation . HEK293 cells expressing FZD7-CFP were grown on cover slips coated with Poly-D-Lysine , starved overnight , and treated with 0 . 1 mg/ml Wnt5a as in 4B . Cells were fixed and stained for endogenous Dvl ( red ) and analyzed by confocal microscopy . ( C , D ) Generation and characterization of DLD1 7TGP cell lines stably expressing Daple . ( C ) DLD1 7TGP cell lines stably expressing Daple-WT or FA were starved and stimulated analyzed for Daple expression and phosphorylation of Akt by immunoblotting ( IB ) . ( D ) Images display representative fields from monolayers of DLD1 cells grown in 0 . 2% FBS by fluorescence microscopy . The intensity of eGFP signals denotes Wnt transcriptional activity . Inset shows immunoblots ( IB ) of equal aliquots of whole-cell lysates of DLD1-7TGP cells expressing control vector , Daple-WT , or Daple-FA . Compared to DLD1 cells expressing Daple-WT , those expressing Daple-FA also express higher levels of GFP protein , indicative of higher Wnt transcriptional activity . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 01210 . 7554/eLife . 07091 . 013Figure 5—figure supplement 2 . Daple and its GBA motif do not affect canonical Wnt signaling . ( A–D ) Daple does not activate Gi after Wnt3 stimulation . ( A , B ) Control ( Luc shRNA ) or Daple-depleted ( Daple shRNA ) HeLa cells were cotransfected with Gαi1-YFP , Gβ-CFP , and untagged Gγ , serum starved overnight ( 0 . 2% FBS ) and subsequently stimulated with either Wnt3 and analyzed for FRET by confocal microscopy . Representative freeze-frame images from live-cell movies are shown ( A ) , which display intensities of acceptor emission due to FRET in each pixel . Activation of Gi was insignificant , as determined by continued interaction ( i . e . , continued FRET ) between Gαi1 and Gβ1γ2 both before and after Wnt3 stimulation ( compare t0 and t5 ) both in control ( Luc shRNA ) and in Daple-depleted HeLa cells . Bar graphs ( B ) display FRET intensities observed in control ( Luc shRNA ) vs Daple-depleted HeLa cells . Error bars representing mean ± S . D . of 5 randomly chosen ROIs at the PM per cell , from 2 to 3 cells per experiment , from three independent experiments . These results are in striking contrast to the findings after Wnt5a stimulation ( see Figure 2I–L in manuscript ) . ( C , D ) Control ( Luc shRNA ) or Daple-depleted ( Daple shRNA ) HeLa cells were serum-starved ( 0 . 2% FBS ) and treated ( + ) or not ( − ) with Wnt3 ( C ) or Wnt5a ( D ) for 15 min prior to lysis . Equal aliquots of lists were subjected to IP with antibodies that selectively recognize active Gαi subunits in their GTP-bound state . Immune complexes ( top ) and lysates ( bottom ) were analyzed for active Gαi3:GTP and total Gαi3 by IB . Wnt5a robustly activates Gαi3 , and this activation is abolished upon Daple depletion , whereas Wnt3 marginally activates Gαi3 and this activation is not diminished upon Daple depletion . ( E ) Myc-Daple is translocated to the PM after Wnt5a , but not after Wnt3 stimulation . HeLa cells were transfected with myc-tagged Daple-WT , serum starved overnight ( 0 . 2% FBS ) and subsequently stimulated with either Wnt5a or Wnt3 as indicated . Cells were fixed at 5 min after ligand stimulation and analyzed for localization of myc-Daple ( Green ) by immunofluorescence . Myc-Daple was found in cytosolic distribution prior to ligand stimulation in starved cells . Upon stimulation with Wnt5a Daple was found to localize sharply at the PM . Upon stimulation with Wnt3 myc-Daple remained in cytosolic location . ( F ) Endogenous Daple is translocated to the PM after Wnt5a , but not after Wnt3 stimulation . HEK cells transfected with CFP-tagged FZD7R were serum starved for 24 hr ( 0% FBS ) , and subsequently stimulated with either Wnt5a or Wnt3 as indicated . Cells were fixed at 5 min after ligand stimulation and analyzed for localization of endogenous Daple by immunofluorescence . Daple was found in cytosolic distribution prior to ligand stimulation in starved cells ( see Figure 4B ) . Upon stimulation with Wnt3 , Daple remained in cytosolic location , however , upon stimulation with Wnt5a Daple was found to localize at the PM , where it colocalized with FZD7R ( see Figure 4B ) . ( G ) Daple's GBA motif does not affect Wnt3-dependent stabilization of β Catenin . HEK293 cells were transfected with Daple-WT or FA mutant , serum starved ( 0% FBS ) for 24 hr , and subsequently stimulated with Wnt3 for 4 hr ( lanes 1–6 ) , 8 hr ( lanes 7–12 ) , or 20 hr ( lanes 13–18 ) prior to lysis . Equal aliquots of cytoplasmic extracts were analyzed for β Catenin , Daple , and tubulin by IB . β Catenin was stabilized ( increased , compare even lanes with odd lanes ) in each condition tested , without significant differences between Daple-WT vs Daple-FA at any time points observed . A representative experiment from a total of four independent experiments is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 013 Because the interplay between of Daple and Dvl is modulated by Wnt5a and the GBA motif of Daple regulates Wnt5a-signaling responses , next , we examined if/how the Daple-Gαi interaction affects the interaction between Dvl and Daple . In in vitro competition assays with recombinant proteins , we found that binding between Daple and Dvl was reduced with increasing amounts of His-Gαi3 ( Figure 5E ) . No such reduction was noted when the Daple-CT-WT was replaced by the GBA-deficient FA mutant ( that cannot bind G proteins ) in the above assays . These findings indicate that Gαi3 competes with Dvl for binding to Daple-CT , and that an intact GBA motif is essential for such competition . Together , these results suggest that the Gαi-Daple and Daple-FZD7R interactions we describe here have at least two major effects on the interplay between Daple , Dvl , and FZD7R: ( 1 ) Daple and Dvl compete for binding to the C-terminus of FZD7R and ( 2 ) Gαi and Dvl compete for binding to the C-terminus of Daple . Consequently , stimulation with Wnt5a triggers the dissociation of Daple-Dvl and FZD7R-Dvl complexes and favors the assembly of FZD7R-Daple-Gαi signaling complexes at the PM in detriment of FZD7R-Dvl complexes . Next , we asked what might be the consequences of replacing Dvl with Daple and activation of G proteins in the vicinity of ligand-activated FZD7R on β-catenin/TCF/LEF signaling . Prior studies have demonstrated that activation of G proteins downstream of FZDRs is sufficient for antagonistic suppression of β-catenin-dependent signaling ( Slusarski et al . , 1997a , 1997b ) . Others have implicated binding of Dvl to FZDRs is required for the enhancement of the β-catenin/TCF/LEF pathway of signaling ( Gao and Chen , 2010 ) . We asked if activation of G proteins via Daple's GBA motif may antagonize β-catenin stability/signaling . We found that HeLa cells without Daple ( Figure 5F , G ) or those expressing the GEF-deficient Daple FA mutant ( Figure 5H , I ) had increased levels of β-catenin protein compared to respective controls , indicating that Daple and its GBA motif are required for maintenance of low levels of β-catenin , and that in their absence β-catenin is stabilized . Consistently , increased stability of β-catenin was also associated with enhanced transcription of downstream target genes SFRP-1 , Osteopontin , and Axin-2 ( Figure 5J–L ) . Similar results were obtained when we analyzed the β-catenin/TCF/LEF pathway in DLD1 colon cancer cells stably expressing Daple-WT or FA mutant ( Figure 5—figure supplement 1C ) using 7-TGP , an eGFP expressing Wnt activity reporter construct ( Fuerer and Nusse , 2010 ) . Wnt activity was enhanced in cells expressing Daple-FA , but not Daple-WT ( Figure 5—figure supplement 1D ) , consistent with our prior findings in HeLa cells . Finally , we found that Daple specifically functions within the non-canonical Wnt signaling cascade and not within the canonical Wnt pathway , for example , stimulation of the canonical Wnt pathway with Wnt3a did not require Daple to activate Gi ( Figure 5—figure supplement 2A–D ) , did not trigger the recruitment of Daple to the PM ( Figure 5—figure supplement 2E , F ) , and did not affect the stabilization of β-catenin ( Figure 5—figure supplement 2G ) . These results suggest that the repressive effects of Daple we observe on the β-catenin/TCF/LEF pathway ( Figure 5J–L ) are likely due to enhancement of the antagonistic non-canonical Wnt signaling pathway . Taken together , these results support an overall model ( Figure 5M ) in which Daple orchestrates non-canonical Wnt signaling by favoring the recruitment and activation of G proteins and displacement of Dvl from activated FZDRs upon Wnt5a stimulation . This leads to enhancement of Akt and Rac1 signaling ( via ‘free’ Gβγ ) and suppression of cellular cAMP ( via Gαi:GTP ) , which is accompanied by diminished activity of the β-catenin/TCF/LEF pathway . Next , we investigated how non-canonical Wnt signaling via the Wnt5a/FZDR-Daple-Gαi axis impacts cancer cell behavior . We first analyzed the cellular phenotypes that are modulated by Wnt5a and non-canonical Wnt signaling during different stages of cancer progression ( McDonald and Silver , 2009 ) . In the normal mucosa , this pathway serves as a tumor-suppressor , by antagonizing the canonical Wnt-β-catenin signaling pathway ( Torres et al . , 1996; MacLeod et al . , 2007; Ying et al . , 2007 , 2008; Chien et al . , 2009 ) , whereas in advanced tumors it triggers cell migration/invasion by enhancing PI3K-Akt and Rac1 pathways and the formation of actin stress fibers ( Nishita et al . , 2010; Liu et al . , 2013; Zhang et al . , 2014 ) . Consistent with the role of Daple's GBA motif in enhancement of Akt and Rac1 activities ( Figure 3 ) , we found that monolayers of Daple-depleted HeLa cells stably expressing Daple-WT , but not Daple FA efficiently closed wounds and generated actin stress fibers ( Figure 6—figure supplement 1A–C ) and migrated efficiently along a gradient of Wnt5a in chemotaxis assays ( Figure 6A ) . To determine if Daple can trigger cell invasion through basement membrane proteins , we carried out 3-D matrigel invasion assays . Non-invasive NIH3T3 cells ( Albini et al . , 1987 ) stably expressing Daple-WT , Daple-FA , or vector control were grown into tumor spheroids and subsequently analyzed for cell invasion through matrix ( Figure 6B , C ) . Enhanced invasion ( as determined by the area of invasion; Figure 6—figure supplement 1B ) was detected exclusively in the presence of Daple-WT , but not in cells expressing control vector or Daple-FA , indicating that Daple is sufficient to trigger cell invasion , and that a functionally intact GBA motif is essential . Compared to cells expressing Daple-FA , those expressing Daple-WT had significantly higher expression of Lox-L3 and Vimentin , two genes commonly associated with epithelial–mesenchymal transition ( EMT ) ( Figure 6D , E ) , indicating that higher invasiveness was accompanied by an EMT gene signature . 10 . 7554/eLife . 07091 . 014Figure 6 . Daple enhances cell migration and invasion via its GBA motif . ( A ) Daple WT , but not FA triggers chemotactic migration towards Wnt5a . Daple-depleted HeLa cells ( sh Daple 1 ) stably expressing Daple-WT or Daple-FA were analyzed for their ability to migrate towards Wnt5a ( + ) or vehicle control ( − ) in transwell assays . Cells were allowed to migrate for 24 hr , fixed and stained with Giemsa . The number of migrating cells was averaged from 20 field-of view images per experiment . Data are presented as mean ± SEM; n = 3 . HPF = high-power field . Lysates of cells used in this assay were analyzed for Daple expression by IB ( see Figure 6—figure supplement 1C ) . ( B , C ) Daple WT , but not FA triggers cell invasion . Spheroids ( S ) of NIH3T3 cells expressing vector control , myc-Daple-WT , or FA were analyzed for their ability to invade matrigel in response to serum stimulation using a Cultrex-3D Spheroid Invasion Kit ( Trevigen ) . An increase of invading cells ( arrowheads; B ) were noted only from the edge of tumor spheroids formed by cells expressing myc-Daple-WT , but not FA . Area of invasion was quantified using ImageJ ( as shown with interrupted blue line in Figure 6—figure supplement 1D ) . ( C ) Bar graphs display area of invasion observed in Daple WT and Daple FA expressing cells . Error bars representing mean ± S . D of three independent experiments . ( D , E ) Daple-WT , but not Daple-FA enhances the expression of genes that trigger epithelial–mesenchymal transition ( EMT ) . mRNA expression of the EMT markers , LOXL3 , and Vimentin were analyzed by qPCR . Results were normalized internally to mRNA levels of the housekeeping gene , GAPDH . Bar graphs display the fold change in each RNA ( y axis ) normalized to the expression in cells expressing vector control . Error bars represent mean ± S . E . M of three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 01410 . 7554/eLife . 07091 . 015Figure 6—figure supplement 1 . Daple enhances cell migration , promotes formation of actin stress-fibers , and triggers invasion , all via its GBA motif . ( A ) Daple-FA , but not Daple-WT inhibits 2-D cell migration . Confluent monolayers of HeLa cells transiently transfected with myc-Daple WT or FA ( ∼90% efficacy of transfection confirmed by immunofluorescence ) or control vector were scratch-wounded and incubated for 24 hr in a 0 . 2% serum media . Wound closure was monitored and quantified as detailed in ‘Materials and methods’ . % wound closure ( y axis ) in various cell lines are displayed as bar graphs . For each cell line , ∼3–5 scratch-wounds were analyzed in each assay . Expression of Daple-FA significantly delays wound closure . Error bars represent mean ± S . E . M of three independent experiments . ( B ) Daple-WT , but not Daple-FA triggers formation of actin stress fibers . Daple-depleted HeLa cells transiently transfected with myc-Daple WT or FA were grown on cover slips in the presence of 0 . 2% FBS , fixed , and subsequently analyzed for actin cytoskeleton patterns by staining with Phalloidin ( red ) . Abundance of stress fibers running across the cell bodies was seen in cells expressing Daple-WT . Blue = DAPI/nucleus . Bars = 10 µm . ( C ) Whole-cell lysates of HeLa cell lines used in transwell chemotaxis assays in Figure 6A were analyzed for Daple expression by IB . ( D ) Daple WT , but not FA triggers cell invasion . Spheroids of NIH3T3 cells expressing myc-Daple WT and FA were analyzed for their ability to invade matrigel in response to serum stimulation using a Cultrex-3D Spheroid Invasion Kit ( Trevigen; see ‘Materials and methods’ ) . Tracks created by invading cells were noted only in cells expressing myc-Daple WT . Area of invasion was quantified using ImageJ ( as shown with interrupted blue line ) . Bar graphs showing the quantification of the area of invasion are shown in Figure 6C . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 015 Next , we investigated the role of Daple and its GBA motif in the modulation of other key cellular phenotypes regulated by non-canonical Wnt signaling during tumorigenesis , that is , cell proliferation , transformation , and growth ( Niehrs and Acebron , 2012; Jamieson et al . , 2014 ) . For this , we used three cell lines: HeLa cell lines , the constitutively active Ras-transformed NIH3T3 cells , and the DLD1 colorectal cancer cells in which transformation is driven by hyperactive β-catenin signaling in addition to active Ras mutations . We chose to study DLD1 colorectal cancer cells because Daple is virtually undetectable in these cells compared to normal colon ( data not shown ) , thereby allowing us to reconstitute expression exogenously and analyze the effect of WT and mutant Daple constructs without significant interference due to the endogenous protein . Expression of Daple-WT reduced the number of colonies formed by Ras-transformed NIH3T3 in soft-agar by ∼65% ( Figure 7A; Figure 7—figure supplement 1A ) , indicating that Daple's GBA motif is required for suppressing neoplastic transformation . The mitotic index , as determined by the presence of phosphorylated Histone H3 in the nucleus ( Hans and Dimitrov , 2001 ) , was higher in HeLa cells expressing Daple-FA compared to those expressing Daple-WT ( Figure 7—figure supplement 1B ) , indicating that Daple's GBA motif suppresses mitosis . When we assessed the tumor-suppressive effect of Wnt5a on HeLa cells in anchorage-dependent tumor growth assays , we found that tumor growth was suppressed in the control cells , but such suppression was lost in cells depleted of endogenous Daple ( Figure 7B ) . This loss of tumor-suppressive effect of Wnt5a was restored by expressing Daple-WT but not by expressing the Daple-FA mutant ( Figure 7C ) , indicating that a functionally intact GBA motif in Daple is essential for Wnt5a to exert its tumor suppressive effects . Daple-WT also inhibited anchorage-independent tumor growth of DLD1 cells by ∼50% ( Figure 7D–F ) , and inhibited anchorage-dependent tumor growth of DLD1 cells by ∼90% ( Figure 7G , H ) , demonstrating that Daple suppresses cellular transformation and growth across all assays . This tumor suppressive effect was mediated via the GBA motif because , compared to Daple-WT , expression of Daple-FA not only failed to inhibit cell transformation ( Figure 7A ) and growth ( Figure 7C , H ) but also enhanced oncogenicity ( Figure 7E ) . Noteworthy , expression of a Daple mutant that cannot bind Dvl ( Daple-ΔPBM ) but has an intact GBA motif retained the tumor suppressive properties of Daple-WT across all assays , whereas a mutant that lacks both the GBA and the Dvl-binding PBM motifs ( Daple-2M ) mirrored the phenotype of the FA mutant , indicating that the G protein regulatory GBA motif , and not the Dvl-binding PBM motif is essential for the tumor suppressive function of Daple . Taken together , these findings demonstrate that Daple inhibits cell transformation and proliferation during tumor growth , but enhances cell motility and cytoskeletal remodeling during invasion; both require the GBA motif , which regulates G protein activity ( Figure 7I ) . 10 . 7554/eLife . 07091 . 016Figure 7 . Daple suppresses proliferation and tumorigenesis via its GBA motif . ( A ) Daple's GBA motif is required for inhibition of cell transformation induced by oncogenic KRas . NIH3T3 cells stably expressing HA-KRas G12V alone or coexpressing HA-KRas G12V with myc-Daple-WT or various mutants were analyzed for their ability to form colonies in soft agar prior to staining with MTT . The top panel displays representative images of colony-containing plates . Bar graphs in the lower panel shows % inhibition of colony formation ( y axis ) by each Daple construct compared to NIH3T3 cells transformed with KRas G12V alone . Lysates of NIH3T3 cells were analyzed for Daple and Ras constructs by IB ( see Figure 7—figure supplement 1B ) . ( B ) Daple is required for inhibition of anchorage-dependent tumor growth by Wnt5a . Control ( shLuc ) and Daple-depleted ( sh Daple 1 ) HeLa cells were analyzed for their ability to form colonies on plastic plates in the presence ( + ) or absence ( − ) of Wnt5a during a 2-week period prior to fixation and staining with crystal violet . Left panel shows the photograph of the crystal violet-stained wells of a 6-well plate . The number of colonies was counted by ImageJ ( Colony counter ) . Right panel shows bar graphs that display the % inhibition of colony formation ( y axis ) seen in each condition normalized to control ( shLuc ) HeLa cells . ( C ) Daple's GBA motif is required for inhibition of anchorage-dependent tumor growth by Wnt5a . Daple-depleted ( sh Daple 1 ) HeLa cells stably expressing either Daple WT or FA were analyzed for their ability to form colonies on plastic plates in the presence ( + ) or absence ( − ) of Wnt5a prior to fixation and staining with crystal violet , photographed and analyzed as in B . Left panel shows the photograph of the crystal violet-stained wells of a 6-well plate . Right panel shows bar graphs that display the % inhibition of colony formation ( y axis ) seen in each condition normalized to control ( shLuc ) HeLa cells . ( D–F ) Daple's GBA motif is required for inhibition of anchorage-independent tumor growth . DLD1 cells expressing either control vector or various myc-Daple constructs were analyzed for their ability to form colonies in soft agar for 2–3 weeks . In panel D , representative fields photographed at 20× magnification are shown . The number of colonies was counted by light microscopy throughout the depth of the matrix in 15 randomly chosen fields . In panel E , bar graphs display the number of colonies ( y axis ) seen in each cell line in D . In panel F , lysates of DLD1 cells used in D were analyzed for Daple constructs by IB . ( G , H ) Daple's GBA motif is required for inhibition of anchorage-dependent tumor growth . DLD1 cells used in D were analyzed for their ability to form adherent colonies on plastic plates during 2–3 weeks prior to fixation and staining with crystal violet . In panel G , photograph of the crystal violet-stained 6-well plate is displayed . The number of colonies was counted by ImageJ ( Colony counter ) . In panel H , bar graphs display the % inhibition of colony formation ( y axis ) seen in each cell line in G normalized to control DLD1 cells . ( I ) Schematic summary . Modulation of G protein activity by Daple's GBA motif is a key determinant of cellular phenotype ( s ) triggered by Wnt5a . In cells expressing Daple-WT , a functionally intact GBA motif ( + ) can activate Gαi , enhance PM-based motogenic signals ( PI3K-Akt and Rac1 activation ) , trigger EMT and cell migration/invasion . In cells expression Daple-FA , without the functional GBA motif ( − ) G protein remains inactive , non-canonical Wnt signaling is suppressed , which increases stability of β-catenin and upregulation of Wnt target genes , resulting in increased transformation , proliferation , and tumor cell growth . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 01610 . 7554/eLife . 07091 . 017Figure 7—figure supplement 1 . Daple suppresses cell proliferation via its GBA motif . ( A ) Compared to cells expressing Daple-WT , those expressing Daple-FA have higher mitotic index , as determined by nuclear localization of phosphorylated histone H3 . HeLa cells expressing myc-Daple WT or FA were grown on cover slips in the presence of 0 . 2% FBS , fixed , and stained for phospho-histone H3 and DAPI . Bar graphs display % cells with nuclear phospho-histone H3 ( y axis ) . Error bars representing mean ± S . D . of three independent experiments . ( B ) Lysates of NIH3T3 cells NIH3T3 cells used in Ras-induced transformation assays ( see Figure 7A ) were analyzed for Daple and Ras constructs by IB . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 017 Because Wnt5a and the non-canonical Wnt pathway is known to be dysregulated during cancer progression ( i . e . , suppressed early during neoplastic transformation and upregulated later during metastasis ) ( McDonald and Silver , 2009 ) , next , we asked whether the expression of Daple is similarly altered during oncogenesis in the colon . Analysis of several publicly available microarray databases revealed expression of Daple mRNA was reduced by ∼twofold in adenocarcinomas of the colon or rectum compared to matched normals ( Figure 8A; Figure 8—figure supplement 1A , B; Figure 8—source data 1 ) . When we analyzed Daple mRNA in another cohort of patients by quantitative PCR ( qPCR ) , we confirmed that Daple is indeed downregulated in cancers ( Figure 8B ) , but not in the precancerous advanced polyps ( defined as any adenoma with >25% villous features , or ≥1 . 0 cm in size , or high-grade dysplasia ) ; the latter showed a modest upregulation in Daple mRNA ( Figure 8B ) . This suggests that the suppression of Daple is fairly late during oncogenesis coinciding with late adenoma-to-cancer progression . Meta-analysis of various microarray databases at The Cancer Genome Atlas ( TCGA; www . cancergenome . nih . gov ) further revealed that expression of Daple mRNA is significantly suppressed in microsatellite stable ( MSS ) colorectal tumors , which account for ∼85% of all colorectal cancers and are characterized by the presence of chromosomal instability ( CIN ) ( Figure 8C , Figure 8—figure supplement 1C; Figure 8—source data 2 ) , whereas tumors with high degree of microsatellite instability ( MSI-high ) express at levels similar to normal colon ( Figure 8C; Figure 8—figure supplement 1C ) . Among MSS tumors , the degree of suppression of Daple correlated with the degree of CIN ( Figure 8D ) . Furthermore , as shown previously in the case of other tumor suppressors ( Pino and Chung , 2010 ) , we found that suppression of Daple mRNA in the primary tumors at the time of diagnosis was associated with disease progression , as determined by formation of distant metastasis in a cohort of patients with stage II colorectal cancers ( Figure 8E ) . Taken together , these results indicate that expression of Daple is frequently reduced during oncogenesis , that such reduction is more common in the setting of CIN , and that reduced expression of Daple in primary tumors may predict disease progression . 10 . 7554/eLife . 07091 . 018Figure 8 . Expression of Daple mRNA is suppressed during oncogenesis by copy number loss , but expressed later during metastasis . ( A ) Daple mRNA is downregulated in colorectal cancers . A meta-analysis was performed using all the available high-throughput microarray data from Genomic Spatial Event ( GSE ) database ( see Figure 8—source data 1 ) to compare the levels of expression of Daple mRNA in colorectal cancer vs matched normal controls . Bar graphs display the results of such meta-analysis as fold change in Daple mRNA ( y axis ) in colorectal carcinomas normalized to matched normal controls . ( B ) Daple mRNA is downregulated during the adenoma-to-carcinoma step of oncogenesis in the colon . Daple mRNA was analyzed by qPCR in normal colon , advanced adenomas , and colorectal carcinomas . Bar graphs display the relative levels of Daple mRNA normalized to GAPDH , as determined by the calculation 2 − ΔCT with reference to an absolute baseline CT of 40 cycles . Error bars represent mean ± S . D . ( C ) Daple mRNA is downregulated in microsatellite stable ( MSS ) , but not microsatellite unstable ( MSI ) colorectal cancers . A meta-analysis was performed using all the available high-throughput microarray data from GSE database ( see Figure 8—source data 2 ) to compare the levels of expression of Daple mRNA in MSI vs MSS colorectal cancers vs their respective matched normal controls . Bar graphs display the results of such meta-analysis as fold change in Daple mRNA ( y axis ) in colorectal carcinomas normalized to normal controls . ( D ) Downregulation in Daple mRNA in MSS colorectal cancers directly correlates with the degree of chromosomal instability ( CIN ) in the tumor . High-throughput microarray data from GSE database ( PMID: 22547595 , GSE: 30 , 540 ) were analyzed for the levels of expression of Daple mRNA in MSS colorectal cancers ( stages II and III ) with varying degrees of CIN [CIN-low ( LOH ratio <33% ) and CIN-high ( LOH ratio ≥33% ) ] and compared to MSI tumors . Bar graphs display the results of such analysis as fold change in Daple mRNA ( y axis ) in CIN-low or CIN-high colorectal carcinomas compared to MSI tumors . ( E ) Downregulation of Daple mRNA in the primary tumor early during cancer progression prognosticates tumor recurrence/metastasis . High-throughput microarray data from GSE database ( PMID: 22917480 , GSE: 37 , 892 ) were analyzed for the levels of expression of Daple mRNA in 130 stage II MSS tumors without ( No Mets ) or with ( Mets ) tumor recurrence/metastatic progression . ( F ) Loss of copy number for CCDC88C ( DAPLE gene ) occurs at the late stages of adenoma-to-carcinoma progression . Array comparative genomic hybridization data from GSE database were analyzed for ccdc88c copy number variations ( CNVs ) in 41 progressed adenomas ( i . e . , adenomas that present a focus of cancer ) . Progressed adenomas were analyzed for CNVs relative toploidy level in the DNA in laser-microdissected adenoma and carcinoma fractions and compared to adjacent normal epithelial fractions as matched controls . ( G ) Cell-free mRNA transcripts of Daple are detected in patients with colorectal cancer , but not in normal control subjects . Microarray data from GSE database ( PMID: 18843029 , GSE: 10 , 715 ) were analyzed for Daple mRNA expression in peripheral blood samples of healthy subjects ( n = 11 ) and of 121 patients with early ( Dukes A , B ) or late ( Duke's C , D ) stages of colorectal cancer . ( H ) Levels of Daple mRNA are frequently elevated in EpCAM ( epithelial cell adhesion molecule ) immunoisolated circulating tumor cells ( CTCs ) from patients with metastatic colorectal cancer , compared to normal subjects . Immunoisolated CTC fractions from the peripheral blood of 51 patients with metastatic ( stage IV ) colorectal cancer or from healthy subjects were analyzed for Daple mRNA by Taqman qPCR and adjusted for leukocyte contaminants by normalizing to CD45 . Scatter-plots display the level of Daple expression in each patient within each group . A normality test confirmed that data sets in both groups were distributed normally . No significant differences were observed in the CD45 levels between two groups ( not shown ) . ( I , J ) High levels of Daple mRNA expression in CTCs are associated with poorer progression-free ( PFS; I ) and overall ( OS; J ) survival in patients with metastatic colorectal carcinoma . Optimal cut-off values for Daple mRNA expression were statistically derived ( see detailed ‘Materials and methods’ ) to generate subgroups of patients with high- or low-expression levels . Time-dependent survival probabilities were estimated with the Kaplan–Meier method , and the log-rank test was used to compare the subgroups . ( K ) Schematic summarizing profile of Daple expression during oncogenic progression in the colon . Degree of upregulation ( green ) or downregulation ( red ) in Daple mRNA is indicated by increasing shades of each color during the normal-to-adenoma-to-carcinoma progression in the colon is shown . ( L ) Proposed model for how a bimodal dysregulation of tumor suppressor Daple , and resultant deregulation of non-canonical Wnt signaling may propel oncogenic progression in the colon . Daple's ability to modulate G proteins via its GBA motif exerts a potent tumor suppressive effect in the normal mucosa . Early during oncogenesis ( top , from left to right ) , downregulation of Daple ( marked by ‘X’ ) occurs at the step of adenoma to cancer conversion , in part by DNA copy loss ( bottom ) due to focal deletion affecting the long arm of Chr 14 . Consequently , low expression of Daple mRNA and protein triggers transformation and tumor growth/progression . Later during cancer invasion , expression of Daple is triggered via unknown mechanisms , which favors ( green arrow ) tumor recurrence and prognosticates poor survival . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 01810 . 7554/eLife . 07091 . 019Figure 8—source data 1 . Meta-analysis of Daple mRNA expression in colorectal cancer vs matched normal controls . The publicly available GSE database , a system to store , retrieve , and analyze all types of high-throughput microarray data was used to compare the levels of expression of Daple mRNA in colorectal cancer vs matched normal controls . From left to right , the columns indicate the GSE series ID , the PMID number for the respective source manuscripts , total samples analyzed in each study , fold change in Daple mRNA observed , and the significance ( p-value ) of any changes observed . A meta-analysis combining the p-values from these studies was analyzed by Fisher's method and displayed as bar graphs in Figure 8A . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 01910 . 7554/eLife . 07091 . 020Figure 8—source data 2 . Meta-analysis of Daple mRNA expression in microsatellite unstable ( MSI ) vs stable ( MSS ) colorectal cancers . The publicly available GSE database was used to compare the levels of expression of Daple mRNA in MSI vs MSS colorectal cancers . From left to right , the columns indicate the GSE series ID , the PMID number for the respective source manuscripts , total samples analyzed in each study , fold change in Daple mRNA observed , and the significance ( p-value ) of any changes observed . A meta-analysis combining the p-values from these studies was analyzed by Fisher's method and displayed as bar graphs in Figure 8C . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 02010 . 7554/eLife . 07091 . 021Figure 8—source data 3 . Daple expression in CTCs correlates with markers of EMT . Expression of Daple , ZEB2 , and LOXL3 mRNA were analyzed in CTCs immunoisolated from 50 patients with metastatic colorectal cancer . An analysis of the Pearson's correlation coefficient for each pair of genes shows that higher expression of Daple is significantly associated with higher expression of ZEB2 and LOXL3 , two genes implicated in triggering EMT . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 02110 . 7554/eLife . 07091 . 022Figure 8—figure supplement 1 . Expression of Daple mRNA is suppressed in colorectal cancers , in part by copy number loss . ( A , B ) Publicly available Kaiser Colon database was analyzed for Daple mRNA expression in adenocarcinomas of the colon ( A ) and rectum ( B ) and their respective normal controls . Daple mRNA expression levels are displayed using log2 median-centered ratio boxplots for normal vs cancer that were generated using the UCSC Cancer Genome Browser . Numbers in parenthesis represent total number of samples analyzed . ( C ) The TCGA colon cancer database was analyzed for Daple mRNA expression in 246 colorectal adenocarcinomas . Daple mRNA expression levels are displayed as heat maps generated using the UCSC Cancer Genome Browser . Red = High Daple; green = Low Daple . Samples are arranged by sample type ( normal vs cancer ) and microsatellite status ( MSI low or high vs MSS ) as indicated on the right margin of the heat map . ( D ) Schematic of chromosome 14 is shown . Ccdc88c gene , which encodes Daple ( red arow ) is located within a frequently deleted region of Chr 14 ( blue box ) . ( E , F ) Publicly available TCGA database was analyzed for number of copies of Daple gene in adenocarcinomas of the colon ( E ) and rectum ( F ) compared to matched normal mucosa and in blood cells . Copy number units of ccdc88c ( Daple ) in various matched samples are displayed using log2 median-centered ratio boxplots for that were generated using the UCSC Cancer Genome Browser . Numbers in parenthesis represent total number of samples analyzed . Compared to matched normal mucosa or peripheral blood , lower copy numbers of Daple gene were observed in adenocarcinomas of colon and rectum . ( G ) The TCGA colon cancer database was analyzed for the relationship between Daple copy number loss and microsatellite status in 461 tumor samples . Daple copy number in each tumor is displayed as heat map ( blue = loss; red = gain ) generated using the UCSC Cancer Genome Browser . Samples are arranged by microsatellite status ( MSI low or high vs MSS ) as indicated on the right margin of the heat map . A large majority of tumors had copy number loss ( blue ) , but not gain ( red ) . Tumors that had a loss of copy for the Daple gene ( blue ) are invariably MSS tumors , or MSI-low tumors . Copy number loss is virtually absent among MSI-high tumors . DOI: http://dx . doi . org/10 . 7554/eLife . 07091 . 022 While seeking clues into how Daple might be downregulated in some tumors , but not all , we noted that ccdc88c , the gene that encodes Daple is located in a region of Chr 14 ( 14q32 . 11 ) that is most frequently deleted in early onset ( <50 y ) colorectal tumors ( Figure 8—figure supplement 1D ) . In fact , 14q deletions are most often associated with significant copy number variations that occur during adenoma-to-carcinoma conversion ( Tsafrir et al . , 2006 ) . An analysis of microarray-based comparative genomic hybridization obtained from polyps that had progressed to cancer revealed that significant loss of Daple copy number was observed in the carcinoma portion , but not in the adenoma portion of these advanced polyps compared to matched normal tissue ( Figure 8F ) . Loss of Daple copy number was noted in adenocarcinomas of both the colon and the rectum ( Figure 8—figure supplement 1E , F ) , and this phenomenon was invariably associated with CIN in MSS tumors ( Figure 8—figure supplement 1G ) . These findings indicate that focal deletions of Chr 14 with resultant loss of copy number may in part contribute to downregulation of Daple we observe in colorectal cancers . Next , we asked how Daple expression changes in disseminated tumor cells and serum . Compared to normal subjects , Daple mRNA was elevated in both cell-free RNA samples ( Figure 8G ) and in tumor cells ( Figure 8H ) isolated from peripheral circulation of patients with colorectal cancer . We found that expression of Daple in circulating tumor cells ( CTCs ) of patients with metastatic colorectal cancer was associated with progression of disease/recurrence ( Figure 8I ) and poor survival ( Figure 8J ) . Furthermore , higher Daple expression in CTCs correlated positively with increased expression of genes that are known to trigger EMT ( Figure 8—source data 3 ) . These results indicate that Daple is expressed in disseminated tumor cells and that higher expression is associated with EMT and poorer clinical outcomes . Taken together , these results define the profile of dysregulated Daple expression during oncogenic progression in the colon ( Figure 8K ) : Daple is first suppressed during adenoma-to-carcinoma progression and expressed later in disseminated tumor cells . The major finding in this work is the discovery of a G protein regulatory function in Daple , which activates trimeric G proteins downstream of FZDRs . We provide biochemical and in-cellulo evidence for the presence of a GBA motif that activates Gαi and an independent domain within the C-terminal region of Daple , which directly binds the cytoplasmic tail of FZDRs . Such a coexistence allows Daple to link G protein activation to ligand-activated FZDRs within ternary FZDR-Daple-Gαi complexes at the PM . We also demonstrate that FZDRs and Gαi come within close proximity of each other ( ∼10 nm based on FRET imaging studies ) within these complexes , suggesting a direct interaction between them on the Daple platform . In cells without Daple , or in cells expressing a mutant in which the GBA motif is selectively disrupted , FZDRs and G proteins do not approach each other and G protein is not activated , demonstrating an obligatory role for Daple's GBA motif in the assembly of FZDR-Gαi complexes . These findings provide a new perspective on the role of G proteins in Wnt signaling because previous work has widely debated the fundamental question whether the 7-TM FZDRs can directly bind and activate G proteins . Arguments that have favored the classification of FZDRs as GPCRs are supported by experimental evidence that FZDRs indeed signal via G proteins , for example , structure-based bioinformatic prediction , pertussis toxin sensitive signaling pathways , genetic linkage with G proteins , and ability to bind β-arrestin for subsequent internalization ( Slusarski et al . , 1997a; Liu et al . , 2001 , 2005; Ahumada et al . , 2002; Katanaev et al . , 2005; Gao and Wang , 2006; Ma and Wang , 2006 ) . Arguments that refute such classification highlight the lack of direct experimental proof of G-protein interaction with FZDRs , and that most studies use experimental models ( overexpressed receptors or gain-of-function ) , which do not necessarily implicate necessity ( Schulte and Bryja , 2007 ) . Our work breaks the impasse in the field by the discovery of an alternative mechanism of G protein activation by FZDRs: we propose that the C-terminus of Daple is the long sought molecular linker that couples FZDRs to efficient G protein activation by virtue of its ability to simultaneously bind receptors and activate G proteins . However , that some FZDRs may directly couple with other G proteins under certain circumstances cannot be ruled out ( see below ) . Here , we demonstrate that Daple is a new member of a family of non-receptor activators of G protein , thereby adding to the growing evidence that trimeric G proteins can be activated by mechanisms differing from classical GPCR-mediated activation . We demonstrated that Daple activates Gi via a signature sequence , that is , the GBA motif that allows proteins and synthetic peptides to exert GEF activity on G proteins and provides a structural basis for non-receptor mediated activation of G proteins ( Johnston et al . , 2005; Austin et al . , 2008; Garcia-Marcos et al . , 2009 , 2011b ) . Daple shares overall homology with GIV , the prototype GBA motif-containing protein , and both of them are classified as members of the CCDC88 family . Interestingly , the C-terminal domains of these two proteins , in which their conserved GBA motifs are located , share very little overall similarity . These observations suggest that Daple and GIV arose from a common ancestor protein and that the GBA function was selectively preserved , while the rest of the C-terminal domain diverged in evolution . Daple has the biochemical features of a GEF: it binds preferentially to inactive , GDP-bound Gαi subunits and accelerates the rate of nucleotide exchange . The GEF activity of Daple , that is , its ability to accelerate the exchange of nucleotide , is more robust than that previously reported for FZDRs in similar in vitro assays ( ∼2 . 5–3-fold activation compared to ∼1 . 5-fold ) ( Koval and Katanaev , 2011 ) . Inefficient activation of G proteins by FZDRs ( ∼5–20% efficacy compared to that observed for a ‘classical’ GPCR , i . e . , Adenosine 2B receptor ) has also been documented in yeast ( Nichols et al . , 2013 ) , which lack homologues of Daple . Our studies measuring G protein activation in Daple-depleted cells or in cells without a functional GBA motif in Daple help establish an obligatory role of Daple as a bona fide G protein activator , which enables FZDRs to indirectly activate Gi to robust levels . These findings cannot rule out other possibilities , for example , that FZDRs may directly activate Gi to a lesser extent under certain circumstances , or that Daple and FZDRs may activate different subsets of G proteins . The latter possibility is exemplified by Gαo , which has been most widely reported as a target for FZDRs ( Liu et al . , 1999 , 2001; Katanaev et al . , 2005; Bikkavilli et al . , 2008; Katanaev and Buestorf , 2009; Egger-Adam and Katanaev , 2010 ) but not for Daple ( this work ) . The marked preference of Daple for α-subunits of the Gi family is a common feature shared with previously described GBA proteins; Daple , GIV , Calnuc , NUCB2 ( Garcia-Marcos et al . , 2009 , 2011b ) , or the synthetic peptides KB-752 and GSP ( Johnston et al . , 2005; Austin et al . , 2008 ) can exquisitely distinguish between Gαi and Gαo proteins , despite their being closely related and sharing 75% sequence homology . Although the biochemical properties of Daple as a G protein regulator are similar to those of other proteins with a GBA motif , we provide evidence that the coupling between Daple and Gαi-subunits has unique structural determinants . Daple can bind to two Gαi3 mutants , W258F and K248M , that abolish binding to GIV and Calnuc , respectively . Moreover , we have previously shown ( Garcia-Marcos et al . , 2010 , 2011b ) that these mutants are able to discriminate between GIV and Calnuc ( K248M binds GIV , but not Calnuc , and W258F binds Calnuc , but not GIV ) , which further suggest that different GBA-Gαi interactions have unique properties that impart a high degree of specificity . The validated homology models of these GBA-Gαi interactions ( Garcia-Marcos et al . , 2009 , 2011b ) offer some clues into the origin of such specificity: despite docking onto the SwII/α3 hydrophobic cleft of Gαi , all GBA-motif containing proteins make additional and unique contacts with Gαi , which generate specificity for each GBA motif . We conclude that the Daple:Gαi interface has unique features that distinguish it from GIV:Gαi or Calnuc:Gαi interfaces , and exploiting such structural specificity may help devise strategies to selectively target the Daple:Gαi interface , and thereby , modulate Wnt signaling . We demonstrate that recruitment of Daple-Gαi complexes to the cytoplasmic tail of ligand activated FZDRs dictates several closely intertwined spatial and temporal aspects of post-receptor signaling events within the non-canonical Wnt pathway . At the immediate post-receptor level , Daple competes with Dvl , the major signaling scaffold for Wnt signaling ( Gao and Chen , 2010 ) , for binding to FZDR , and recruits and activates Gαi in close proximity to activated receptors at the PM . That Dvl and Daple/Gαi complexes may compete for binding to FZDR is in keeping with others' findings that overexpression of Dvl interferes with the engagement of Gi proteins with ligand-activated FZDRs ( Kilander et al . , 2014 ) , and that Dvl is unlikely to directly link G proteins to FZDRs , as proposed by some ( Schulte and Bryja , 2007 ) . Once recruited , Daple's GEF activity triggers Gi activation , which leads to inhibition of cellular cAMP via Gαi:GTP and activation of non-canonical Wnt signaling pathways involved in cell motility ( e . g . , PI3K and Rac1 ) via ‘free’ Gβγ . The consequences of these signaling mechanism are enhanced formation of actin stress fibers , 2D-cell migration after wounding , 3D-invasion through basement membrane proteins , and upregulation of genes that trigger EMT; all phenotypes that have been previously attributed to enhancement of non-canonical Wnt signaling ( Minami et al . , 2010 ) . We also show that the FZD7R-Daple-Gi axis suppresses responses associated with tumorigenesis , for example , β-catenin/TCF/LEF signaling , oncogenic transformation , anchorage independent growth , and anchorage-dependent colony formation; all attributable to its ability to activate G proteins via its GBA motif . Because the FZDR-Daple-Gi axis specifically modulates non-canonical Wnt signals , but has no effect on canonical Wnt responses , we conclude that Daple suppresses the canonical β-catenin/TCF/LEF pathway primarily by enhancing the antagonistic non-canonical Wnt pathway ( Torres et al . , 1996; MacLeod et al . , 2007; Ying et al . , 2007 , 2008; Chien et al . , 2009 ) . Although the mechanism ( s ) by which the non-canonical Wnt pathway inhibits the canonical β-catenin/TCF/LEF pathway remains unclear , and some have proposed that such decisions are made at the level of the receptors ( Logan and Nusse , 2004 ) , how Daple-dependent G protein signaling in the vicinity of the receptors may affect this process remains unclear . Our finding that Daple binds preferentially to some FZDRs , and not others , could influence the decision of canonical vs non-canonical Wnt signaling , or alternatively , activation of Gαi and inhibition of cellular cAMP by Daple could directly antagonize a previously described role of the adenylate cyclase/cAMP/PKA pathway in phosphorylating and stabilizing β-catenin ( Hino et al . , 2005 ) . Regardless of the mechanism ( s ) involved , activation of Gi and enhancement of non-canonical Wnt signaling are accompanied by the suppression of the canonical β-catenin pathway in cells expressing Daple-WT , which correlates with all the key anti-growth and anti-transformation phenotypes that define a tumor suppressor/anti-oncogene ( Cooper , 2000 ) . Although it is possible that some of the effects of the FZDR-Daple-Gi axis in tumor suppression are mediated by the destabilization of β-catenin , further investigations are required to clarify this point . We conclude that the G protein regulatory function of Daple is essential for enhancing at least two major cellular phenotypes previously attributed to non-canonical Wnt signaling ( McDonald and Silver , 2009 ) , suppression of cell transformation and growth , and enhancement of cell invasion . As for potential implications in other key cellular processes that are deregulated during cancer progression , it is noteworthy that non-canonical Wnt signaling has also been demonstrated to play a crucial role in planar cell polarity and asymmetric cell division in stem cells ( Bentzinger et al . , 2014 , 2013 ) . Several studies have shown that the Wnt7A/FZD7 pathway establishes front-rear cell polarity and directional migration of human myogenic progenitors and facilitate the extension of satellite stem cells , all by activating PI3K/Akt pathway and Rac1 ( Bentzinger et al . , 2014 , 2013 ) . Because one of the major roles of the FZDR-Daple-Gi axis is enhancement of PI3K and Rac1 activities , it is possible that this axis also aids in the establishment of cell polarity and/or the maintenance of stem-ness via enhancement of the non-canonical Wnt pathway . Further studies are required to determine if such is the case . We showed here that Daple is downregulated during oncogenesis at the step of conversion from adenoma to carcinoma , and that lower expression of Daple in the primary tumor is associated with higher frequency of cancer recurrence . We describe that expression of Daple in CTCs correlated with an increased EMT signature , disease progression ( growth of current metastasis or formation of new metastasis ) , and poorer survival . This bimodal dysregulation ( suppressed first , expressed later ) and bi-faceted role ( tumor suppressor in the normal epithelium , but enhancer of tumor invasion in cancer cells ) during cancer progression mirrors what was previously unequivocally documented for the non-canonical Wnt5a signaling ( McDonald and Silver , 2009 ) —Wnt5a signaling is suppressed earlier to allow cellular transformation and tumor growth , and enhanced later during tumor invasion . However , molecular mechanisms for such bimodal deregulation of the bi-faceted non-canonical Wnt pathway remain poorly understood . Such phenomenon is not restricted to Daple or the Wnt pathway , because major signaling programs like the TGFβ-SMAD pathway have also been shown to display similar bimodal deregulation and a bi-faceted role ( Akhurst and Derynck , 2001 ) , and a similar phenomenon is observed in the case of Daple's closely related orthologue , GIV ( Ghosh et al . , 2010 ) : downregulation of GIV by alternative splicing triggered proliferation early during tumor growth , whereas an increase in GIV by transcriptional upregulation enhanced cell invasion later during oncogenesis . Because Daple serves as a bona fide enhancer of the non-canonical Wnt pathway , we conclude that upregulation or downregulation in Daple expression contributes , at least in part , to the bimodal deregulation of the Wnt5a signaling pathway observed in cancers . We also demonstrate that the mechanism for downregulation of Daple in cancers follows typical tumor suppressor genetics during neoplastic transformation ( Payne and Kemp , 2005 ) . Downregulation of Daple mRNA coincided with adenoma-to-carcinoma transition , and the frequency of such downregulation in the primary tumor directly correlated with the degree of CIN . A loss of copy number of Daple DNA , and consequent downregulation of gene expression and function was noted in the primary tumors , predominantly among the tumors with CIN . This pattern is in keeping with the well-documented role of CIN in generating loss of heterozygosity ( LOH ) and haploinsufficiency of other tumor suppressors ( Sotillo et al . , 2009 ) . In the case of Daple , such insufficiency is likely to increase the fitness of cells that have undergone such a LOH because depletion of Daple suppresses non-canonical Wnt signaling and allows unrestricted propagation of canonical Wnt pathways . Consequently , proliferation/growth is triggered , which enables these cells to rapidly outcompete the remaining population . Based on the location of ccdc88c ( Daple gene ) at a site on the long arm of Chr 14 , which is known to be frequently deleted in a variety of cancers ( Suzuki et al . , 1989; Hu et al . , 2002; Rouault et al . , 2012 ) , we conclude that tumors harboring a focal deletion at that site are in part driven by insufficient expression of the tumor suppressor Daple . Additional mechanisms , for example , alternative splicing may further contribute to oncogenesis via dysregulation of Daple expression , as described in a rare and fatal human developmental anomaly ( Ekici et al . , 2010 ) . This anomaly was attributed to deregulation of Wnt signaling due to a loss of Daple's 29th exon , which contains the G protein regulatory GBA motif . Although many other mechanisms may be involved , loss of Daple expression , or a selective loss of its G protein regulatory function has emerged as a final common pathway , which disrupts Daple-Gαi axis of Wnt signaling and derails tissue homeostasis . The precise molecular mechanism ( s ) that enhances Daple expression or function and consequently triggers an EMT signature and cell invasion during cancer progression remains unclear . Transcriptional compensation for loss of an allele ( Guidi et al . , 2004 ) or gain-of-function mutations ( van Oijen and Slootweg , 2000 ) is possible mechanisms , as shown previously in the case of other tumors suppressors . In this regard , it is noteworthy that although Daple bound the cytoplasmic tails of several FZDRs to varying extent , the preference for FZD7R was striking and may provide some clues as to why/how Daple may enhance tumor progression . Although all FZDRs promiscuously interact with more than one of the many Wnt isoforms to activate canonical and/or non-canonical Wnt signaling ( King et al . , 2012 ) , FZD7 stands out as a receptor that functions at the cross-roads of canonical and non-canonical Wnt signaling pathways in a unique way . FZD7R is a downstream target of β-catenin in cancer cells ( Barker and Clevers , 2006 ) , and consequently , enhanced canonical Wnt signaling upregulates FZD7R expression during cancer progression . It has been proposed that such increased FZD7R expression due to aberrant canonical Wnt signals may serve as a positive forward-feedback mechanism to perpetuate Wnt/β-catenin signaling , thus , facilitating colorectal cancer progression and metastasis . Because Daple appears to be upregulated during cancer invasion and in circulating cancer cells ( and such upregulation is associated with worse prognosis ) and enhances non-canonical Wnt signaling downstream of FZD7R , it is possible that Daple's functional interaction with this receptor further enhances prometastatic signaling via amplification of the non-canonical Wnt pathway , which synergizes with the previously proposed forward-feedback canonical Wnt signaling loop during cancer progression . We conclude that such preferential signaling downstream of FZD7R and the temporal profile of expression of Daple are well-poised to suppress or enhance non-canonical Wnt signaling and aid in different steps of tumor progression ( see legend for Figure 8L ) . In conclusion , we have defined Daple as a novel regulator of G protein activity , which directly binds FZDRs and enables these 7-TM receptors to recruit and activate Gi , and trigger non-canonical Wnt signaling to suppress tumorigenesis and enhance tumor invasion . These findings set a new paradigm for the long-debated mechanisms by which FZDRs are coupled to G protein activation . As a potent tumor suppressor with multiple intriguing domains , for example , the newly identified GBA and the Frizzled-binding domain , Daple presents many signaling interfaces that could be developed as targets for modulating Wnt signaling . Because its levels of expression in primary tumors , circulating cell-free transcripts and in CTCs may indicate tumor characteristics , Daple presents many avenues for further development as clinically useful diagnostic and prognostic biomarkers . Unless otherwise indicated , all reagents were of analytical grade and obtained from Sigma–Aldrich ( St . Louis , MO ) . Cell culture media were purchased from Invitrogen . All restriction endonucleases and Escherichia coli strain DH5α were purchased from New England Biolabs ( Ipswich , MA ) . E . coli strain BL21 ( DE3 ) , phalloidin-Texas Red were purchased from Invitrogen ( Grand Island , NY ) . Genejuice transfection reagent was from Novagen ( Madison , WI ) . PfuUltra DNA polymerase was purchased from Stratagene ( La Jolla , CA ) . Recombinant Wnt3a and Wnt5a were purified as previously described ( Willert , 2008 ) . Briefly , conditioned media ( CM ) were collected the day after confluence was reached . WNT proteins were purified from 6 liters of CHO CM . CM was complemented with 1% Triton X-100 ( vol/vol ) , 20 mM Tris-Cl pH 7 . 5 , and 0 . 01% NaN3 . Goat anti-rabbit and goat anti-mouse Alexa Fluor 680 or IRDye 800 F ( ab′ ) 2 used for immunoblotting ( IB ) were from Li-Cor Biosciences ( Lincoln , NE ) . Mouse anti-His , anti-FLAG ( M2 ) , anti-α tubulin , and anti-actin were obtained from Sigma; anti-Myc and anti-HA were obtained from Cell Signaling Technology ( Beverly , MA ) and Covance ( Princeton , NJ ) , respectively . Rabbit anti-pan-Gβ ( M-14 ) , anti-Gαi3 , anti-DVL , and anti-β-catenin were obtained from Santa Cruz Biotechnology ( Dallas , TX ) ; anti-Akt and phospho-Akt ( S473 ) were obtained from Cell Signaling; anti-Rac1 was obtained from BD Transduction Laboratories ( San Jose , CA ) . Anti-Daple antibodies were generated in collaboration with Millipore ( Carlsbad , CA ) using the C-terminus of Daple ( aa 1660–2028 ) as an immunogen . Cloning of N-terminally tagged myc-Daple was carried out in two steps by piece-meal assembly . A fragment of hDaple obtained from Kazusa ( KIAA1509; clone fh14721 , inserted into pBluescript II SK [+] ) was used as a source of 3′ nucleotide bp 2131–6087 . The N-terminus of hDaple was artificially synthesized ( Genscript , San Diego ) and used as a source for the 5′ nucleotide bp 1–2130 . The full-length hDaple gene ( corresponding to the Ref Seq NM_001080414 . 3 [mRNA] and NP_001073883 . 2 [protein] ) was assembled by inserting 5′ and 3′ fragments into pcDNA 3 . 1 between NotI/EcoRI and EcoRI/BamHI , respectively . The EcoRI cloning site in the middle of the Daple sequence was eliminated by mutagenesis . The entire gene length was sequenced prior to cloning it into myc-pcDNA 3 . 1 ( + ) between KpnI/EcoRI to generate myc-Daple . All subsequent site-directed mutagenesis and truncated constructs ( myc-Daple full-length F1675A ( FA ) , myc-Daple deleted from aa 2025–2028 ( ΔPBM ) , myc-Daple FA+ΔPBM ( 2M ) , and myc-Daple CT 1650–2028 aa ) were carried out on this template using Quick Change as per manufacturer's protocol . The GST-Daple-CT WT , His-Daple-CT WT , and FA constructs ( 1650–1880 aa and 1650–2028 aa ) used for in vitro protein–protein interaction assays were cloned from myc-Daple pcDNA 3 . 1 and inserted within the pGEX-4T or pET28b vectors , respectively , between NdeI/EcoRI restriction sites . The HA-tagged FZD7R construct was generated by cloning the human receptor ( ATCC# 10658884; Gen Bank BC015915 . 1; Ref Seq: NM_003507 . 1 ) in pcDNA 3 . 1 between HindIII/EcoRI and by subsequently inserting a HA tag at the C-terminus by mutagenesis . FZD7R-CFP construct was a generous gift from Carl-Philip Heisenberg ( Institute of Science and Technology , Austria ) ( Witzel et al . , 2006 ) . Gαi3-YFP and Gαi1-YFP ( internally tagged Gαi subunits: the coding sequence for YFP was inserted in the αb–αc loop after Ala-121 of Giα1 and Ala-114 of Giα3 , which does not affect their biochemical properties ) , CFP-Gβ1 and untagged Gγ are a generous gift from Moritz Bunemann ( Philipps-Universität Marburg , Germany ) ( Bunemann et al . , 2003; Gibson and Gilman , 2006 ) . Mouse Dvl1 and HA-Ras G12V were generous gifts from Mikhail V Semenov ( Harvard Medical School ) and Robert Hayward ( London , UK ) , respectively . Cloning of rat Gα-proteins into pGEX-4T-1 ( GST-Gαi3 , GST-Gαi1 , GST-Gαi2 , and GST-Gαo ) , GST-Gαi3 K248M and W258F; His-Gαi3; Gαi3-FLAG; Gαi3-HA; and GST-GIV CT 1671–1755 aa has been described previously ( Ghosh et al . , 2008; Garcia-Marcos et al . , 2009 , 2010 , 2011b; Ghosh et al . , 2010 ) . GST-tagged C-termini of FZDRs 3–7 ( Yao et al . , 2004 ) were generous gifts from Ryoji Yao ( JFCR research institute , Japan ) . The C-terminal cytoplasmic tails of human FZD1 ( aa 614–647 ) and mouse FZD2 ( aa 537–570 ) were cloned into the BamHI/ EcoRI sites of pGEX-4T-1 to generate the plasmids for bacterial expression of GST-FZD1-CT and GST-FZD2-CT , respectively . GST-PBD was a generous gift from Gary Bokoch ( The Scripps Research Institute , La Jolla ) . Daple shRNA constructs were created using the following approach . Promising targets at the 3′ UTR region of human Daple ( NM_001080414 ) were identified using the pSicoOligomaker software . The two most promising hits were chosen based on favorable score ( >7 ) . Duplexed oligos were designed against those targets and cloned in pSico Puro vector between HpaI and XhoI . Details of targets for hDaple sequence and oligos used are provided below: Targets for hDAPLE 3′ UTR ( coding DNA sequence is from bp 155–6241 ) 6570 GTAGAACACTCATTTGCAA ( shRNA 1 ) 6929 GCACCTGCCTTCCTAGATT ( shRNA 2 ) hDaple sh1 forward 5′ TGTAGAACACTCATTTGCAATTCAAGAGATTGCAAATGAGTGTTCTACTTTTTTC hDaple sh1 reverse 5′ TCGAGAAAAAAGTAGAACACTCATTTGCAATCTCTTGAATTGCAAATGAGTGTTCTACA hDaple sh2 forward 5′ TGCACCTGCCTTCCTAGATTTTCAAGAGAAATCTAGGAAGGCAGGTGCTTTTTTC hDaple sh2 reverse 5′ TCGAGAAAAAAGCACCTGCCTTCCTAGATTTCTCTTGAAAATCTAGGAAGGCAGGTGCA GST and His-tagged recombinant proteins were expressed in E . coli strain BL21 ( DE3 ) ( Invitrogen ) and purified as described previously ( Ghosh et al . , 2008 , 2010; Garcia-Marcos et al . , 2011a ) . Briefly , bacterial cultures were induced overnight at 25°C with 1 mM isopropylβ-D-1-thio-galactopyranoside ( IPTG ) . Pelleted bacteria from 1 l of culture were resuspended in 20 ml GST-lysis buffer ( 25 mM Tris·HCl , pH 7 . 5 , 20 mM NaCl , 1 mM Ethylenediaminetetraacetic acid ( EDTA ) , 20% [vol/vol] glycerol , 1% [vol/vol] Triton X-100 , 2 × protease inhibitor mixture [Complete EDTA-free; Roche Diagnostics] ) or in 20 ml His-lysis buffer ( 50 mM NaH2PO4 [pH 7 . 4] , 300 mM NaCl , 10 mM imidazole , 1% [vol/vol] Triton X-100 , 2 × protease inhibitor mixture [Complete EDTA-free; Roche Diagnostics] ) for GST or His-fused proteins , respectively . After sonication ( three cycles , with pulses lasting 30 s/cycle , and with 2 min interval between cycles to prevent heating ) , lysates were centrifuged at 12 , 000×g at 4°C for 20 min . Except for GST-FZD and GST-PBD constructs ( see in vitro GST pulldown assay section ) , solubilized proteins were affinity purified on glutathione-Sepharose 4B beads ( GE Healthcare ) or HisPur Cobalt Resin ( Pierce ) , dialyzed overnight against PBS , and stored at −80°C . Tissue culture was carried out essentially as described before ( Ghosh et al . , 2008 , 2010; Garcia-Marcos et al . , 2011a ) . We used a total of five different cell lines in this work , each chosen carefully based on its level of endogenous Daple expression and the type of assay . All these cell lines were cultured according to ATCC guidelines . Cos7 cells were primarily used for transient overexpression of tagged Daple or Dvl proteins and lysates of these cells were used as source of proteins in various protein–protein interaction ( IP and pulldown ) assays . We chose to carry out these assays in Cos7 cells because they are easily and efficiently transfected ( >90% efficiency ) with most constructs . The added advantage is that they have no detectable endogenous Daple ( by IB and qPCR ) and provide a system to selectively analyze the properties of WT vs mutant Daple constructs without interference from endogenous Daple . HeLa cells were primarily used to study the in-cellulo dynamics of interaction between Daple and FZD7R during non-canonical Wnt signaling because those cells have been extensively used to study Wnt5a-stimulated non-canonical signaling by various groups ( Yamamoto et al . , 2007; Sato et al . , 2010 ) . We noted that HeLa cells have low amounts of endogenous Daple , and that it was an adequate system to study the role of Daple in cells because Wnt5a stimulation could trigger the previously described downstream signaling responses in our hands ( Yamamoto et al . , 2007; Sato et al . , 2010 ) . Noteworthy , the efficiency of transient transfection of various Daple constructs in these cells was >90% , as determined by immunofluorescence staining . HEK293T cells were used exclusively for FRET and co-IP studies involving FZD7R/G proteins because these cells are widely used and preferred for such studies involving GPCR/G protein signaling due to several reasons . HEK293 cells are the single most widely used cell line for heterologous expression ( both transient and stable expression ) of GPCRs ( Thomas and Smart , 2005 ) because they allow a robust expression of functional receptors compared to most cells ( Massotte , 2003; Thomas and Smart , 2005 ) . Microarray analyses have confirmed that they have an adequate transcriptome that supports various elements of GPCR/G protein signaling pathways , for example , GPCR ligands , trimeric G proteins , scaffolding components that mediate receptor endocytosis , kinases , and phosphatases that phosphoregulate GPCR functions , and so on ( Atwood et al . , 2011 ) . We have confirmed that they express endogenous Daple as a full-length protein , at physiologic levels , and the localization of Daple ( as determined by immunofluorescence ) is primarily at the PM ( data not shown ) , where FZDRs are activated . Low passage NIH3T3 fibroblasts were used exclusively in 3-D Matrigel invasion assays and in neoplastic transformation assays to study the role of Daple in suppressing growth in soft agar upon Ras-mediated transformation . The rationale for their use in invasion assay lies in the fact that non-transformed NIH3T3 fibroblasts are poorly invasive in vitro and non-tumorigenic and non-metastatic in animal studies ( Bondy et al . , 1985; Hill et al . , 1988; Chambers et al . , 1990; Tuck et al . , 1991 ) . It is because of this reason , NIH3T3 cells are widely used to study proteins that can trigger a gain in invasive properties ( Leitner et al . , 2011 ) . For the neoplastic transformation assays , we used Ras-transformed NIH3T3 cells because this is the gold standard assay used to study the role of a gene/protein in tumor transformation ( Egan et al . , 1987 ) . The rationale for using NIH3T3 in both the above assays is further strengthened by the fact that they are highly transfectable ( ∼80% transfection efficiency with myc-Daple ) and express Daple at very low-endogenous levels ( as determined by IB and qPCR ) compared to normal colonic epithelium . Such expression pattern allows us to study the effect of various mutant Daple constructs without significant interference due to the endogenous protein . DLD1 were primarily used to study the effect of Daple on cancer cell growth properties ( anchorage-dependent and independent ) and to assess the effect of Daple on the classical Wnt signaling pathway ( β-catenin/TCF/LEF ) . There are several reasons why this cell line was chosen: ( 1 ) We focused on colorectal cancer in this study , and DLD1 cells were appropriate to translate our findings because they are human colorectal cancer cells; ( 2 ) We determined that levels of Daple are significantly lower ( ∼10-fold ) in these cells compared to normal colon ( data not shown ) , thereby allowing us to study the effect of various mutant Daple constructs without significant interference due to the endogenous protein; ( 3 ) These cells have been extensively characterized with respect to most oncogenes ( ATCC database ) and are highly tumorigenic in 2-D and 3-D cultures due to a mutation in KRAS ( G13D ) ( Shirasawa et al . , 1993; Ahmed et al . , 2013 ) ; ( 4 ) They are a sensitive model to study how various manipulations of the non-canonical Wnt signaling pathway oppose the canonical Wnt pathway during tumor growth because they constitutively secrete Wnt ligands to maintain high levels of the canonical signaling ( Voloshanenko et al . , 2013 ) within the growth matrix . Production and secretion of endogenous ligands bypasses the need to add exogenous ligands repeatedly during prolonged assays that last ∼2 weeks . Transfection was carried out using Genejuice ( Novagen ) for DNA plasmids following the manufacturers' protocols . HeLa and DLD1 cell lines stably expressing Daple constructs were selected after transfection in the presence of 800 µg/ml G418 for 6 weeks . The resultant multiclonal pool was subsequently maintained in the presence of 500 µg/ml G418 . Daple expression was verified independently using anti-Myc and anti-Daple antibodies by IB and estimated to be ∼5× the endogenous level . Unless otherwise indicated , for assays involving serum starvation , serum concentration was reduced to 0 . 2% FBS overnight for HeLa cells and 0% FBS for Cos7 , HEK293T , and DLD1 cells . Whole-cell lysates were prepared after washing cells with cold PBS prior to resuspending and boiling them in sample buffer . Lysates used as a source of proteins in IP or pull-down assays were prepared by resuspending cells in Tx-100 lysis buffer ( 20 mM HEPES[4- ( 2-hydroxyethyl ) -1-piperazineethanesulfonic acid] , pH 7 . 2 , 5 mM Mg-acetate , 125 mM K-acetate , 0 . 4% Triton X-100 , 1 mM Dithiothreitol ( DTT ) , supplemented with sodium orthovanadate [500 µM] , phosphatase [Sigma] , and protease [Roche] inhibitor cocktails ) , after which they were passed through a 28G needle at 4°C , and cleared ( 10 , 000×g for 10 min ) before use in subsequent experiments . For immunoblotting , protein samples were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis ( SDS-PAGE ) and transferred to polyvinylidene difluoride ( PVDF ) membranes ( Millipore ) . Membranes were blocked with phosphate buffer saline ( PBS ) supplemented with 5% non-fat milk ( or with 5% bovine serum albumin ( BSA ) when probing for phosphorylated proteins ) before incubation with primary antibodies . Infrared imaging with two-color detection and band densitometry quantifications were performed using a Li-Cor Odyssey imaging system exactly as done previously ( Garcia-Marcos et al . , 2010 , 2011a , 2011b , 2012; Ghosh et al . , 2010 ) . All Odyssey images were processed using ImageJ software ( NIH ) and assembled into figure panels using Photoshop and Illustrator software ( Adobe ) . Purified GST-Gαi3 or GST alone ( 5 µg ) was immobilized on glutathione-Sepharose beads and incubated with binding buffer ( 50 mM Tris-HCl [pH 7 . 4] , 100 mM NaCl , 0 . 4% [vol:vol] Nonidet P-40 , 10 mM MgCl2 , 5 mM EDTA , 30 µM GDP , 2 mM DTT , protease inhibitor mixture ) for 90 min at room temperature as described before ( Ghosh et al . , 2008 , 2010; Lin et al . , 2011; Garcia-Marcos et al . , 2011a ) . Lysates ( ∼250 µg ) of Cos7 cells expressing appropriate myc-Daple constructs or purified His-Daple-CT ( aa 1650–2028 ) protein ( 3 µg ) were added to each tube , and binding reactions were carried out for 4 hr at 4°C with constant tumbling in binding buffer ( 50 mM Tris-HCl [pH 7 . 4] , 100 mM NaCl , 0 . 4% [vol:vol] Nonidet P-40 , 10 mM MgCl2 , 5 mM EDTA , 30 µM GDP , 2 mM DTT ) . Beads were washed ( 4× ) with 1 ml of wash buffer ( 4 . 3 mM Na2HPO4 , 1 . 4 mM KH2PO4 [pH 7 . 4] , 137 mM NaCl , 2 . 7 mM KCl , 0 . 1% [vol:vol] Tween 20 , 10 mM MgCl2 , 5 mM EDTA , 30 µM GDP , 2 mM DTT ) and boiled in Laemmli's sample buffer . In some experiments , the ‘active’ conformation of the G protein was stabilized by replacing GDP in the binding and wash buffers with 30 µM GTPγS or a mixture of 30 µM GDP/30 µM AlCl3/10 mM NaF . Immunoblot quantification was performed by infrared imaging following the manufacturer's protocols using an Odyssey imaging system ( Li-Cor Biosciences ) . GST-FZD7-CT and GST-PBD constructs were immobilized on glutathione-Sepharose beads directly from bacterial lysates by overnight incubation at 4°C with constant tumbling . Next morning , GST-FZD7-CT immobilized on glutathione beads were washed and subsequently incubated with His-tagged Daple-CT or Gαi3 proteins at 4°C with constant tumbling . Washes and IB were performed as previously . For IP , cell lysates ( ∼1–2 mg of protein ) were incubated for 4 hr at 4°C with 2 μg of appropriate antibody , anti-HA mAb ( Covance ) for HA-Gαi3 or HA-FZD7 , anti-FLAG ( M2 from Sigma ) mAb for FLAG-Gαi3 , or their respective pre-immune control IgGs . Protein G ( for all mAbs ) Sepharose beads ( GE Healthcare ) were added and incubated at 4°C for an additional 60 min . Beads were washed in PBS-T buffer ( 4 . 3 mM Na2HPO4 , 1 . 4 mM KH2PO4 , pH 7 . 4 , 137 mM NaCl , 2 . 7 mM KCl , 0 . 1% [vol:vol] Tween 20 , 10 mM MgCl2 , 5 mM EDTA , 2 mM DTT , 0 . 5 mM sodium orthovanadate ) , and bound proteins were eluted by boiling in Laemmli's sample buffer . The structure of the synthetic peptide KB-752 bound to Gαi1 ( PDB:1Y3A ) was used as the template to generate the modeling project in Deep View/Swiss-PdbViewer v3 . 7 for Daple ( aa 1668–1679 ) in complex with Gαi3 . The modeling project was submitted to the Swiss-Model Server ( http://swissmodel . expasy . org//SWISS-MODEL . html ) ( Schwede et al . , 2003 ) , and model images were generated by MolsoftICM ( San Diego , CA ) . Under the experimental conditions of steady-state GTPase assays , GTP hydrolysis occurs as a two-step reaction , that is , ( 1 ) GDP is released from the G protein and exchanged for GTP and ( 2 ) the GTP loaded is hydrolyzed . Nucleotide exchange is the rate limiting step in this process because it is ∼50–100 times slower than GTP hydrolysis by Gαi subunits ( Mukhopadhyay and Ross , 2002 ) . Thus , the steady-state GTPase activity reflects the rate of nucleotide exchange and was performed as described previously ( Garcia-Marcos et al . , 2010 , 2011b , 2012 ) . Briefly , His-Gαi3 ( 100 nM ) was preincubated with different concentrations of His-Daple-CT ( aa 1650–2028 ) for 15 min at 30°C in assay buffer ( 20 mM Na-HEPES , pH 8 , 100 mM NaCl , 1 mM EDTA , 2 mM MgCl2 , 1 mM DTT , 0 . 05% [wt:vol] C12E10 ) . GTPase reactions were initiated at 30°C by adding an equal volume of assay buffer containing 1 µM [γ-32P]GTP ( ∼50 c . p . m/fmol ) . For the time course experiments , duplicate aliquots ( 50 μl ) were removed at different time points and reactions stopped with 950 μl ice-cold 5% ( wt/vol ) activated charcoal in 20 mM H3PO4 , pH 3 . For the dose–dependence curve experiments , reactions were stopped at 15 min . Samples were then centrifuged for 10 min at 10 , 000×g , and 500 μl of the resultant supernatant was scintillation counted to quantify released [32P]Pi . For the time course experiments , data were expressed as raw c . p . m . For the dose–dependence curve experiments , the background [32P]Pi detected at 15 min in the absence of G protein was subtracted from each reaction and data expressed as percentage of the Pi produced by His-Gαi3 in the absence of His-Daple-CT . GTPγS binding was measured using a filter binding method as described previously ( Garcia-Marcos et al . , 2010 , 2011b ) . His-Gαi3 ( 100 nM ) was preincubated with different concentrations of His-Daple-CT ( aa 1650–2028 ) for 15 min at 30°C in assay buffer ( 20 mM Na-HEPES , pH 8 , 100 mM NaCl , 1 mM EDTA , 25 mM MgCl2 , 1 mM DTT , 0 . 05% [wt:vol] C12E10 ) . Reactions were initiated at 30°C by adding an equal volume of assay buffer containing 1 µM [35S] GTPγS ( ∼50 c . p . m/fmol ) . Duplicate aliquots ( 25 μl ) were removed at different time points , and binding of radioactive nucleotide was stopped by addition of 3 ml ice-cold wash buffer ( 20 mm Tris-HCl , pH 8 . 0 , 100 mm NaCl , 25 mm MgCl2 ) . The quenched reactions were rapidly passed through BA-85 nitrocellulose filters ( GE Healthcare ) and washed with 4 ml wash buffer . Filters were dried and subjected to liquid scintillation counting . To determine the specific nucleotide binding , the background [35S] GTPγS detected in the absence of G protein was subtracted from each reaction and data expressed as percentage of the [35S] GTPγS bound by His-Gαi3 in the absence of His-Daple-CT . FRET experiments were performed using the classical ECFP- and EYFP-tagged proteins as donor and acceptor FRET-probe pairs , respectively . Previously validated and published FZD7-CFP construct was a generous gift from Carl-Philip Heisenberg ( Witzel et al . , 2006 ) . Previously validated Gαi3-YFP and Gαi1-YFP ( internally tagged Gαi subunits ) and CFP-Gβ1 were generous gifts from Moritz Bunemann ( Bunemann et al . , 2003; Gibson and Gilman , 2006 ) . Interaction of FZD7-CFP and Gαi3-YFP proteins was studied in HEK293T cells using a Leica inverted laser scanning confocal microscope . Axial scans of 0 . 5 µ thickness that resolved most of the PM from a single cell were chosen for imaging and the signal in the donor and acceptor channels was ensured to be in mesoscopic regime to avoid inhomogeneity's between samples ( Midde et al . , 2014 ) . Loss of FRET upon Gi activation and heterotrimer dissociation was measured between Gαi1-YFP and CFP-Gβ1 proteins co-expressed in living HeLa cells using Olympus FV1000 inverted confocal laser scanning microscope equipped with a 60× 1 . 49 N . A oil immersed objective designed to minimize chromatic aberration and enhance resolution for 405–605 nm imaging as described previously ( Midde et al . , 2015 ) . Images were sequentially acquired through Donor , FRET , and acceptor channels using 405 and 488 laser lines to excite CFP and YFP , respectively . FRET efficiency was calculated on a pixel by pixel basis from ratiometric images obtained in individual channels ( donor , acceptor , and FRET ) through a RiFRET plugin in ImageJ software ( Roszik et al . , 2009 ) . All images are corrected for the spectral cross-talk obtained from cells transfected with either donor or acceptor probes alone . Regions of interest were randomly drawn at the PM ( an example is shown in Figure 2—figure supplement 1E; red circle ) to compute FRET efficiency . For IP of active Gαi3 , freshly prepared cell lysates ( 2–4 mg ) were incubated for 30 min at 4°C with the conformational Gαi:GTP mouse antibody ( 1 μg ) ( Lane et al . , 2008b ) or with control mouse IgG . Protein G Sepharose beads ( GE Healthcare ) were added and incubated at 4°C for additional 30 min ( total duration of assay is 1 hr ) . Beads were immediately washed three times using 1 ml of lysis buffer ( composition exactly as above; no nucleotides added ) , and immune complexes were eluted by boiling in SDS as previously described ( Lopez-Sanchez et al . , 2014 ) . HeLa cells were transfected with Daple-WT or Daple-FA , serum starved ( 0 . 2% FBS , 16 hr ) and incubated with isobutylmethylxanthine ( IBMX , 200 µM , 20 min ) followed by Wnt5A stimulation ( 100 ng/ml , 20 min ) and Forskolin ( 10 μM , 10 min ) . To stop the reaction , cell medium was replaced with 150 μl of ice-cold TCA 7 . 5% ( wt/vol ) . cAMP content in TCA extracts was determined by radioimmunoassay and normalized to the amount of protein ( determined using a dyebinding protein assay [Bio-Rad] ) per sample as previously described ( Ostrom et al . , 2001 ) . This assay was performed as described previously ( Garcia-Marcos et al . , 2009 ) . Briefly , GST alone or GST-Gαi3 proteins immobilized on glutathione-agarose beads were incubated overnight at 4°C with HEK293T cell lysates in binding buffer ( 50 mM Tris-HCl , pH 7 . 4 , 100 mM NaCl , 0 . 4% [vol:vol] NP-40 , 10 mM MgCl2 , 5 mM EDTA , 2 mM DTT , protease inhibitor cocktail supplemented with 30 µM GDP ) . Unbound Gβγ-subunits were washed twice with the same buffer and proteins bound to the glutathione-agarose beads divided into equal aliquots containing ∼5 µg ( ∼0 . 4 µM ) GST-fusion proteins . Aliquots were incubated with increasing concentrations ( 0 . 05–1 µM ) of purified His-Daple-CT ( 1650–2028 ) wild-type or 1 µM His-Daple-CT F1675A in binding buffer supplemented with GDP ( ∼200 µl ) for 5 hr at 4°C . Glutathione-agarose beads were washed and bound proteins eluted by boiling in Laemmli sample buffer and separated by SDS-PAGE . Rac1 activity in HeLa cells lines was monitored using GST-tagged PAK1-binding domain ( PBD; pGEX-PBD ) as described previously ( Benard and Bokoch , 2002 ) . Briefly , E . coli strain BL21 bacteria transformed with pGEX-PBD were grown at 37°C , and GST-PBD expression was induced at OD600 with 1 mM IPTG for 3 hr at 37°C with shaking . Bacterial lysates were prepared as described above in protein purification section , cleared of debris by centrifugation and subsequently aliquots of lysates were stored at −80°C until use . Aliquots of bacterial lysates were thawed , cleared of precipitated proteins by centrifugation at 14 , 000×g for 20 min , and the cleared supernatant was subsequently incubated with glutathione beads overnight at 4°C with constant tumbling to prepare purified bead-bound GST-PBD freshly for each assay . To analyze the role of Daple in regulation of Rac1 activity , we used HeLa cells . For assays done on cells at steady-state , cells were maintained overnight in a media containing 2% or 0 . 2% FBS prior to lysis . Lysis was carried out first in RIPA buffer ( 20 mM HEPES pH 7 . 4 , 180 mM NaCl , 1 mM EDTA , 1% Triton X-100 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , supplemented with 1mMDTT , sodium orthovanadate [500 μM] , phosphatase [Sigma] , and protease [Roche] inhibitor mixtures ) for 15 min on ice , and then for an additional 15 min after addition of an equal volume of Triton X-100 lysis buffer ( 20 mM Hepes [pH 7 . 2] , 5 mM Mg-acetate , 125 mM K-acetate , 0 . 4% Triton X-100 , 1 mM DTT , supplemented with sodium orthovanadate [500 μM] , phosphatase [Sigma] , and protease [Roche] inhibitor mixtures ) . During the second 15 min of incubation , cells were broken by passing through a 28-gage needle at 4°C and lysates were subsequently cleared ( 10 , 000×g for 10 min ) before use . For assays done with/without ligand stimulation , HeLa cells serum-starved ( 0 . 2% FBS ) overnight and subsequently treated or not with 100 ng/ml Wnt5a for 5 min at prior lysis as above . Equal aliquots of lysates were incubated with bead-bound GST-PBD for 1 hr at 4°C with constant tumbling . Beads were washed in PBS-T buffer ( 4 . 3 mM Na2HPO4 , 1 . 4 mM KH2PO4 , pH 7 . 4 , 137 mM NaCl , 2 . 7 mM KCl , 0 . 1% [vol:vol] Tween 20 , 10 mM MgCl2 , 5 mM EDTA , 2 mM DTT , 0 . 5 mM sodium orthovanadate ) and bound proteins were eluted by boiling in Laemmli's sample buffer . HeLa cell lines were fixed at room temperature with 3% paraformaldehyde for 20–25 min , permeabilized ( 0 . 2% Triton X-100 ) for 45 min , and incubated for 1 hr each with primary and then secondary antibodies as described previously ( Ghosh et al . , 2008 ) . Dilutions of antibodies and reagents were as follows: Myc ( 1:500 ) ; Phalloidin ( 1:1000 ) ; DAPI ( 1:2000 ) ; goat anti-mouse ( 488 and 594 ) Alexa-conjugated antibodies ( 1:500 ) ; anti-phospho-Histone H3 ( Ser28 ) ( 1:150 ) . Cells were imaged on a Leica SPE confocal microscope using a 63× oil objective and 488 , 561 , and 405 laser lines for excitation ( Lopez-Sanchez et al . , 2014 ) . All individual images were processed using ImageJ software and assembled for presentation using Photoshop and Illustrator software ( Adobe ) . These assays were carried out using the well-established reporter 7xTcf-eGFP ( 7TGP ) ( Fuerer and Nusse , 2010 ) . Stable cells lines expressing this reporter were generated by lentiviral transduction and subsequent selection using standard procedures . Lentiviral infection and selection were performed according to standard procedures . Briefly , 10-cm plates DLD1 cells at 70% confluency were incubated with media containing 8 µg/ml polybrene and 10 µl of lentivirus for 6 hr . After 24 hr post-infection , selection of puromycin-resistant clones was initiated by adding the antibiotic at 2 µg/ml final concentration . The resultant DLD1-7TGP stable cells were subsequently transfected with various myc-Daple constructs and selected for G418 resistance as described earlier in methods . The DLD1-7TGP cells stably expressing myc-Daple were incubated overnight at 0 . 2% FBS , analyzed by fluorescence microscopy , and photographed prior to lysis . Whole-cell lysates samples were then boiled in Laemmli's sample buffer , and GFP protein expression was monitored by IB . Scratch-wound assays were done as described previously ( Ghosh et al . , 2008 ) . Briefly , monolayer cultures ( 100% confluent ) of HeLa cells expressing Daple WT or Daple FA were scratch-wounded using a 20-μl pipette tip and incubated in 2% FBS media . The cells were subsequently monitored by phase-contrast microscopy over the next 24 hr . To quantify cell migration ( expressed as percent of wound closure ) , images were analyzed using ImageJ software to calculate the difference between the wound area at 0 hr and that at 12 hr divided by the area at 0 hr × 100 . Chemotactic cell migration assays were performed using Corning Transwell plates according to the manufacturer's protocol . HeLa cells were trypsinized , counted , and placed in a Transwell with media containing 0 . 2% FBS ( 75000 cells/well ) . Media in the bottom chamber of each well were supplemented with 0 . 2% FBS and 100 ng/ml Wnt5a to trigger chemotactic migration . Cells were allowed to migrate for 24 hr and fixed prior staining . Cells that had successfully migrated to the side of the permeable membrane facing the bottom chamber were visualized by staining the membrane with Giemsa . Cell migration ( expressed as number of cells/high-power field ) was quantified by analyzing 15–20 random fields per membrane insert per condition for the number of Giemsa stained cells . NIH3T3 cell invasion assay in 3D culture was performed according to the manufacturer's protocol ( Trevigen , Cultrex 3D Spheroid BME Cell Invasion Assay , catalog # 3500-096-K ) . Briefly , non-invasive NIH3T3 cells ( ∼3000 cells ) transfected with empty vector ( control ) or myc-Daple constructs were incubated first in the Spheroid Formation extracellular matrix containing 0 . 2% FBS for 3 days . Invasion matrix was then added and layered on top with media containing FBS . Serum-triggered cell invasion was photographed under light microscope everyday for 10 days , and fresh media ( FBS concentration is increased each time in order to maintain a gradient ) were replenished every 48 hr . Photographs were analyzed and pseudocolored by ImageJ to reflect cell density . The mitosis rate of HeLa cells stably expressing Daple-WT and Daple-FA was measured by phospho-Histone H3 ( Ser28 ) ( mitotic index ) exactly as we did previously ( Ghosh et al . , 2010 ) . Mitotic index was determined by dividing the number of positively stained cells/the total number of DAPI-stained nuclei × 100 . Neoplastic transformationin Ras-transformed NIH3T3 fibroblasts were analyzed using standard assays of colony formation in soft agar as described previously ( Clark et al . , 1995 ) . Low-passage NIH3T3 cells ( ∼5000 ) stably co-transfected with appropriate myc-Daple construct ( 2 µg cDNA ) and HA-Ras G12V ( 1 µg cDNA ) were analyzed for their ability to form tumor foci in soft agar plates . Plates were incubated in 5% CO2 at 37°C for ∼2 weeks in growth media supplemented with 2% FBS . They were finally incubated with 0 . 1% ( wt/vol ) 3- ( 4 , 5-dimethylthiazol-2-yl ) 2 2 , 5-diphenyl tetrazolium bromide ( MTT; Sigma ) in PBS for 1 hr to visualize colonies . The remaining NIH3T3 cells not used for this assay were lysed and analyzed for myc-Daple and Ha-Ras G12V expression by IB . Anchorage-independent growth of DLD1 cells was analyzed in agar as described previously ( Provost et al . , 2012 ) . Briefly , petri plates ( 60 mm ) were pre-layered with 3 ml 1% Bacto agar ( Life Technologies ) in Dulbecco's Modified Eagle's medium ( DMEM ) containing 10% Fetal Bovine Serum ( FBS ) . Approximately ∼5000 DLD1 cells stably expressing various Daple constructs were then plated on top in 3 ml of 0 . 3% agar–DMEM with 10% FBS . All assays were carried out using three replicate plates at a seeding density of ∼5000 cells/plate . Following overnight incubation in 5% CO2 incubator , 1 ml DMEM supplemented with 2% FBS was added to maintain hydration . After 2 weeks of growth , colonies were stained with 0 . 005% crystal violet/methanol for 1 hr and subsequently photographed by light microscopy . The number of colonies in ∼15–20 randomly-selected fields was counted under 10× magnification . The remaining DLD1 cells were lysed and analyzed by IB to confirm Daple construct expression . Each experiment was analyzed in triplicate . Anchorage-dependent growth was monitored on solid ( plastic ) surface as described previously ( Franken et al . , 2006 ) . Briefly , anchorage-dependent growth was monitored on solid ( plastic ) surface . Approximately ∼1000 DLD1 cells stably expressing various Daple constructs were plated in 6-well plates and incubated in 5% CO2 at 37°C for ∼2 weeks in 0 . 2% FBS growth media . Colonies were then stained with 0 . 005% crystal violet for 1 hr . The remaining DLD1 cells were lysed and analyzed by IB to confirm Daple construct expression . Each experiment was analyzed in triplicate . 51 patients with metastatic colorectal cancer from the Complexo Hospitalario Universitario de Santiago de Compostela , Spain were enrolled ( Barbazán et al . , 2012 ) . All participants signed an informed consent specifically approved for this study by the Ethical Committee of the Complexo Hospitalario Universitario of Santiago de Compostela ( code of approval: 2009/289 ) . Inclusion criteria were the presence of measurable metastatic colorectal cancer ( stage IV ) and an Eastern Cooperative Oncology Group ( ECOG ) performance status not greater than 2 . Disease progression , evaluated by computerized tomography , was defined following RECIST 1 . 1 guidelines ( 1 ) as an increase in the number of metastatic lesions , growth of existing lesions in more than 20% or both during treatment . Furthermore , 24 healthy individuals with similar age ranges to those of patients were included as negative controls . CTCs were isolated using an EpCAM-based immunoisolation ( dynabeads ) using the CELLection Epithelial Enrich kit ( Life Technologies ) , and CTC RNA was purified with the Qiamp Viral kit ( Qiagen ) as previously described ( Barbazán et al . , 2012 ) . Briefly , Superscript III based cDNA synthesis ( Life Technologies ) was carried out to preamplify a region within the coiled-coil domain of Daple to maximize posterior detection rates ( TaqMan Preamp kit , Applied Biosystems ) . Preamplified samples were subsequently subjected to TaqMan real-time PCR amplification ( Applied Biosystems ) ( probe numbers Hs00380245_m1 and Hs00325884_m1 ) . Non-specific blood cells in the CTC-enriched isolates were accounted for by analyzing the expression of CD45 as a lymphoid cell marker ( not present in cancer cells ) . All the results for Daple are normalized with the expression of CD45 ( in all sample types ) . Briefly , the Ct value ( coming from qPCRs ) for Daple and CD45 are subtracted to 40 ( maximum number of cycles in qPCR ) to get an intuitive value ( more value , more expression ) . Daple 40-ct values are normalized with those from CD45 , afterwards . Total RNA was isolated using an RNeasy kit ( QIAGEN ) as per the manufacturers' protocol . First-strand cDNA was synthesized using Superscript II reverse transcriptase ( Invitrogen ) , followed by ribonuclease H treatment ( Invitrogen ) prior to performing quantitative real-time PCR . Reactions omitting reverse transcriptase were performed in each experiment as negative controls . Reactions were then run on a real-time PCR system ( ABI StepOnePlus; Applied Biosystems ) . Gene expression was detected with SYBR green ( Invitrogen ) , and relative gene expression was determined by normalizing to GAPDH using the ΔCT method . Primer sequences are listed as follows:GeneForwardReverseDaple-CC5′-TGA CAT GGA GAC CCT GAA GGC TGA-3′5′-TTTCATGCGGGCCTCACTGCTGA-3′GAPDH5′-TCA GTT GTA GGC AAG CTG CGA CGT-3′5′-AAGCCAGAGGCTGGTACCTAGAAC-3′LOXL35′-ATGGGTGCTATCCACCTGAG-3′5′-GAGTCGGATCCTGGTCTCTG-3′Vim5′-AAGAGAACTTTGCCGTTGAA-3′5′-GTGATGCTGAGAAGTTTCGT-3′SFPR-15′-GAGTTTGCACTGAGGATGAAAA-3′5′-GCTTCTTCTTCTTGGGGACA-3′AXIN-25′-GAGTGGACTTGTGCCGACTTCA-3′5′-GGTGGCTGGTGCAAAGACATAG-3′OPN5′-TTGCAGCCTTCTCAGCCAA-3′5′-GGAGGCAAAAGCAAATCACTG-3′ Advanced adenomas were collected and analyzed as described previously ( Toiyama et al . , 2013 ) . All patients provided written informed consent and the study was approved by institutional review boards of Baylor University Medical Center , Dallas , USA and the Okayama University Hospital , Okayama , Japan . Colorectal carcinomas used in this work were derived from a previously well-characterized , chemo-naive , stage II colorectal cancer cohort from Munich ( Nitsche et al . , 2012 ) . The ethics committee of the Klinikum rechts der Isar , Munich , Germany , approved collection of the patient samples ( #1926/07 , and #5428/12 ) . All samples were obtained after prior informed written consent . For each sample , 20 to 30 mg of frozen tumor tissue was removed for further analysis using a cryostat microtome ( CM3050 S , Leica Microsystems , Wetzlar , Germany ) . Histology-guided sample selection ( Maak et al . , 2013 ) was performed by a pathologist to ensure a sufficient amount of tumor cells ( good cellularity and >30% tumor cells ) . RNA was obtained using the Qiagen AllPrep DNA/RNA Mini Kit ( Qiagen GmbH , Hilden , Germany ) according to the manufacturer's protocol . Subsequently qPCR was performed as described above . All experiments were repeated at least three times , and results were presented either as one representative experiment or as average ±SD or SEM . Statistical significance was assessed with two-tailed Student's t-test . Statistical evaluation for CTC studies were performed using IBM SPSS Statistics Version 19 ( SPSS Inc . , IBM Corporation , Somers , New York , USA ) . In order to derive optimal cut-off values of Daple expression levels , maximally selected log-rank statistics performed by R Software version 2 . 13 . 0 ( R Foundation for Statistical Computing , Vienna , Austria ) were used . To consider multiple test issue within these analyses , the R-function maxstat . test was employed ( Hothorn and Zeileis , 2008 ) . Time-dependent survival probabilities were estimated with the Kaplan–Meier method , and the log-rank test was used to compare independent subgroups . All statistical tests were performed two-sided , and p-values less than 0 . 05 were considered to be statistically significant .	Our cells need to be able to communicate with each other to coordinate many processes in the body , including the formation and maintenance of our organs . A system called Wnt signaling allows cells in different tissues to communicate . During Wnt signaling , one cell releases a protein called Wnt that then binds to a receptor protein known as Frizzled on the surface of another cell . This triggers a cascade of signaling events in the second cell , which leads to changes in the activity of particular genes . Wnt signaling is vital to many processes in cells , and any defects can cause cancer and other severe diseases . Frizzled is a member of a large family of receptor proteins known as the G protein-coupled receptors ( or GPCRs for short ) . These proteins can bind to other proteins called G proteins . When a GPCR is active , it can activate the G protein , which can then interact with several other signal proteins to amplify the signal from the GPCR . However , there is currently no firm evidence that Frizzled can directly bind to G proteins . Some researchers have suggested that it may interact with G proteins via another ‘linker’ protein , but no such protein has yet been identified . Here , Aznar et al . investigated how Frizzled can activate G proteins in human cells . The experiments revealed that a protein called Daple can bind to both Frizzled and the G proteins when the cells are exposed to Wnt to activate the G proteins . Aznar et al . show that Daple can act as a ‘tumor suppressor’ that reduces the risk of healthy cells becoming cancerous and can inhibit the growth of tumors . However , the amount of Daple increases in some tumor cells in the later stages of cancer , which makes it easier for these cells to spread around the body . Aznar et al . also observed that Daple was present at different levels in the late-stage tumor cells taken from a variety of cancer patients . Patients with higher levels of Daple were less likely to have a positive outcome from their cancer treatment , and their illness progressed more rapidly than patients with lower levels of Daple . The next challenge is to understand what causes Daple to switch from its tumor suppressor role to one that promotes the spread of tumors in the later stages of disease .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "structural", "biology", "and", "molecular", "biophysics" ]	2015	Daple is a novel non-receptor GEF required for trimeric G protein activation in Wnt signaling
The ability to isolate rare live cells within a heterogeneous population based solely on visual criteria remains technically challenging , due largely to limitations imposed by existing sorting technologies . Here , we present a new method that permits labeling cells of interest by attaching streptavidin-coated magnetic beads to their membranes using the lasers of a confocal microscope . A simple magnet allows highly specific isolation of the labeled cells , which then remain viable and proliferate normally . As proof of principle , we tagged , isolated , and expanded individual cells based on three biologically relevant visual characteristics: i ) presence of multiple nuclei , ii ) accumulation of lipid vesicles , and iii ) ability to resolve ionizing radiation-induced DNA damage foci . Our method constitutes a rapid , efficient , and cost-effective approach for isolation and subsequent characterization of rare cells based on observable traits such as movement , shape , or location , which in turn can generate novel mechanistic insights into important biological processes . Characterization of biological samples relies heavily on microscopy where , in response to various stimuli , molecular probes and a myriad of contrast reagents are routinely used to identify and label individual live cells of interest . These methods often require prior knowledge of cellular markers or use of elaborate reporter constructs . On the other hand , based solely on visual inspection or using image processing algorithms , it is possible to distinguish rare cells which exhibit distinct biological properties from among thousands of counterparts within a microscopy field . Such visually discernable traits include movement , shape , intracellular protein distribution , and location within the sample , and in turn can reflect important physiological features of individual cells . For example , cell migration ( movement ) is an essential determinant in normal embryonic development , wound healing , immune responses , tumor progression , and vascular disease ( Kurosaka and Kashina , 2008 ) . Moreover , changes in cellular morphology ( shape ) constitute biomarkers of cellular growth , division , death , and differentiation , as well as of tissue morphogenesis and disease ( Prasad and Alizadeh , 2019 ) . Cell-to-cell contacts ( location ) or distance to sources of chemical cues such as senescent cells , inflammation or necrotic tissue are critical factors in chemokinesis , differentiation , neural function , and immune responses ( Garcia et al . , 2018 ) . Finally , expression and visualization of fluorescent fusion proteins permits the identification of cells presenting molecular behaviors of interest , such as differential relocalization of proteins to subcellular compartments or structures upon various stimuli . Unfortunately , however , isolation and expansion of single cells characterized by such easily-observable features is technically challenging , and indeed has not been accomplished to date . We recently developed a method termed Cell Labeling via Photobleaching ( CLaP ) ( Binan et al . , 2016 ) allowing the arbitrary tagging of individual cells among a heterogeneous population within a microscopy field . This is accomplished by crosslinking biotin molecules to their plasma membranes with the lasers of a confocal microscope , followed by use of fluorescent streptavidin conjugates to reveal the marked cells . In this manner , the same instrument used for imaging can also be adapted to label particular cells based on any visible trait that distinguishes them from the ensemble . Importantly , previous knowledge of surface markers or transfection of reporter genes are not required . Tags can be added with single-cell precision and the incorporated label displays convenient tracking properties to monitor location and movement . The mark is stable , non-toxic , retained in cells for several days , and moreover , does not engender detectable changes in cell morphology , viability , or proliferative capacity . Moreover , gene expression profiling indicated no major changes associated with the procedure ( Binan et al . , 2016 ) . Nevertheless , a technology for the efficient isolation and expansion of CLaP-tagged cells is still lacking . The fact that cell populations are often highly heterogeneous underscores the need for new approaches to capture and clonally expand individual cells of interest for further characterization . However , as mentioned above , current sorting techniques cannot efficiently isolate such rare cells ( Pappas and Wang , 2007 ) ; indeed , classical protocols like Fluorescence and Magnetic Activated Cell Sorting ( FACS and MACS ) are typically optimized for high throughput at the expense of capture efficiency and specificity , and require large numbers of cells ( Pappas and Wang , 2007 ) . Small cell populations representing 10−3 of the total , which have been defined as rare , or ultrarare in the case of 10−5 , can only be effectively captured and purified with repeated cycles of sorting and cell expansion protocols ( Pappas and Wang , 2007 ) . Starting with rare and hence precious cell populations , highly conservative gating strategies are needed , which can at best achieve approximately 45% purity ( Kuka , 2013; Shields et al . , 2015 ) . Time-consuming manipulations , cost , hardware footprint , and handling complexity ( Takahashi and Okada , 1970 ) make approaches based on microfluidics ill-suited for capturing small numbers of cells , which are often masked within tens of thousands . Here , we report a novel technology , termed Single-Cell Magneto-Optical Capture ( scMOCa ) , for isolating cells based purely on visual traits from within large heterogenous populations . After tethering biotin moieties to their membranes , cells of interest are targeted with streptavidin-coated ferromagnetic beads and captured with high efficiency using a simple magnet . The procedure is fast , uses low-cost commercially available reagents and only requires access to a standard confocal microscope . As proof-of-principle for the utility and power of this novel approach , we used scMOCa to i ) : capture and expand individual cells that differ in their capacity to resolve ionizing radiation ( IR ) -induced foci of the DNA repair protein 53BP1 , ii ) purify rare multinucleated cells , and iii ) isolate cells that differentiated into adipocytes and accumulated lipid vesicles . Overall , the ease of use and affordability of our method is expected to facilitate the characterization of phenotypes of interest occurring in a small fraction of cell populations . To demonstrate the utility of scMOCa , we sought to isolate and expand cell populations based on their ability to resolve ionizing radiation ( IR ) -induced 53BP1 DNA damage foci , a well-characterized indicator of DNA double strand break ( DSB ) repair capacity ( Asaithamby and Chen , 2009 ) . For this , we used U2OS osteosarcoma cells harboring a construct permitting doxycycline-inducible expression of 53BP1 fused to Green Fluorescent Protein ( GFP ) . 53BP1 is directly involved in DSB repair and is rapidly recruited to DSB sites where it forms foci that can be readily detected by fluorescence microscopy in live-cells ( Mirzayans et al . , 2018 ) when fused with GFP . Foci of 53BP1 are resolved gradually as cells repair DSB , and within approximately 3 hr post-irradiation with 0 . 5 Gy most are expected to disappear ( Mirzayans et al . , 2018 ) . We exposed cells to 0 . 5 Gy of IR and imaged GFP-53BP1 foci . We first characterized focus formation and resolution by measuring the average number of foci before and after IR in 500 cells . At 45 min post-irradiation an average of 10 . 2 ± 2 . 5 ( mean ± standard deviation ) foci per cell was detected . At 2 hr post-irradiation , a second set of images was acquired , and the average number of foci was reduced to 7 . 6 ± 2 . 3 . Since on average cells resolved approximately 25% of their foci within 2 hr , we defined cells in which more than 85% of foci have disappeared after 2 hr as ‘fast resolving’ . Such fast resolving cells , represented approximately 1% of the population . In all following experiments , we compared both sets of images to search for fast-resolving cells ( two such cells are shown in Figure 5A ) and used scMOCa to tag , capture and expand them . We emphasize that FACS or similar approaches are not suitable for sorting based on focus resolution , even if the fraction of target cells was relatively large , as the overall fluorescence signal originating from cell nuclei does not reflect the local distribution of protein . Indeed , we observed no change in global protein abundance or average intensity of GFP-53BP1 upon focus resolution: the average intensity of nuclei showed no correlation with the number of 53BP1 foci ( Pearson coefficient of −0 . 15 ) . Because we used very stringent selection criteria for focus resolution , we tagged only 5 and 3 ‘fast-resolving’ cells in two independent experiments , which were subsequently isolated using scMOCa , pooled and expanded to generate Populations #1 and #2 . We next compared the kinetics 53BP1 focus resolution in Populations #1 and #2 vs . the parental cell population . The resolution of foci was quantified using ( i ) live-cell imaging of GFP-53BP1 ( Figure 5B ) and also ( ii ) following immunostaining with anti-53BP1 antibody ( when GFP-53BP1 expression was not induced ) to evaluate focus formation involving the endogenous untagged protein ( Figure 5C ) . Images were acquired at 45 , 60 , 75 , 90 and 120 min post-irradiation with 1Gy for the two populations and the distribution of DNA foci per cell compared with that of the parental cell line . We used Matlab to program a fully automated algorithm for focus quantification ( Figure 5D ) and analyzed approximately 1800 cells per time-point . This allowed the unbiased evaluation of large datasets as Figure 5B and C taken together represent the behavior of more than 21 , 000 cells . Figure 5B and C shows normalized histograms ( probability density functions ) of the number of foci per cell at each time-point . Importantly , all three populations exhibited similar numbers of foci per cell 45 min after irradiation , indicating that the initial formation of 53BP1 foci is comparable between all cell populations . However , we found that the progeny of captured cells ( Populations #1 and #2 ) retained the original visually detected phenotype of fast focus resolution . These cells resolved foci at least 1 . 5 times more rapidly than parental counterparts , as the median number of GFP-53BP1 foci per cell 60 min post-IR for Populations#1 and #2 ( 17 and 15 respectively ) is equal to the median number of foci that parental cells exhibit at 90 min post-IR . After 75 min , these numbers of foci are already statistically different ( p-values from student T-tests comparing the parental cells to Populations #1 and #2 are respectively 10−75 and 10−39 ) . Such differences in focus resolution dynamics is particularly striking in cells for which the expression of GFP-53BP1 is induced ( Figure 5B ) but is clearly observable as well using immunofluorescence of the endogenous protein in non-induced fixed cells ( Figure 5C ) . To rule out the possibility that resolution of 53BP1 foci might be due to increased degradation upon IR or to globally decreased levels of the protein , we monitored 53BP1 levels by immunoblotting at different time points post-IR . No changes in the levels of either endogenous 53BP1 or GFP-tagged version was observed ( Figure 5E ) . Finally , FACS analysis shows that all populations exhibit similar ratios of cells in each cell cycle phase ( Figure 5F ) . Therefore , the observed focus resolution differences between populations is unlikely to be attributable to cell cycle-related effects . We next sought to illustrate of the utility of scMOCa to capture cells based on their morphology , which have so far proven challenging to sort using currently available technologies . For example , multinucleated cells constitute a rare subpopulation ( Mirzayans et al . , 2017; Coward and Harding , 2014 ) that does not express specific markers and cannot be differentiated from mononucleated polyploid cells using DNA-specific stains in a FACS experiment . However , multinucleated cells can be easily identified visually even without DNA staining . In the context of cancer , such cells have been ( i ) described as generally being more aggressive and metastatic than mononucleated counterparts , and ( ii ) proposed to be prone to acquisition of drug resistance and cancer relapse ( Mirzayans et al . , 2017; Mittal et al . , 2017; Weihua et al . , 2011; Green and Meuth , 1974 ) . Moreover , even though multinucleated cells do not undergo classical cytokinesis , they can generate mononucleated progeny by budding ( Mirzayans et al . , 2017; Weihua et al . , 2011 ) and influence neighboring cells by secreting factors that promote stemness , as well as by transmitting sub-genomes ( Mirzayans et al . , 2017 ) . Multinucleated cells were isolated using scMOCa and kept in culture for 4 days to evaluate their viability and metabolic activity ( Figure 6 ) . We used WGA-alexa647 to stain plasma membranes , and Hoechst for the nuclei ( Figure 6 ) and Mitotracker green FM to tag polarized mitochondrial membranes , indicating that scMOCa preserves the viability of isolated cells ( see Figure 6—figure supplement 1 ) . As another example of a visual phenotype that can be sorted using scMOCA , we evaluated the differentiation of 3T3 cells into adipocytes . These cells are amongst the most common models to study metabolic disorders , for example , obesity ( Armani et al . , 2010; Majka et al . , 2014 ) . When cultured for 2 days in medium containing dexamethasone , insulin and isobutylmethylxanthin ( IBMX ) , an inhibitor of cyclic nucleotide phosphodiesterases , and 3 days in medium containing insulin , a fraction of 3T3 cells differentiate and lipid vesicles accumulate in their cytoplasm . In order to obtain pure adipocyte cultures , flow cytometry sorting based on granularity requires several steps to select cells of interest and then remove false positives , such as debris and cell aggregates ( Nagrath et al . , 2007 ) , whereas scMOCA may provide a much simpler approach to isolated live adipocytes , especially when these are present in very low abundance . We used scMOCA to capture differentiated adipocytes and then kept them in culture for a week ( Figure 6 ) . Sorted cells remained viable and maintained their ability to store lipids in vesicles that appear as clear spheres on Figure 6 , while the magnetic beads that remained attached to cells membranes appear as dark spheres . To the best of our knowledge , scMOCa is the only technology that permits isolation , and subsequent clonal expansion , of extremely small numbers of cells from relatively large heterogeneous populations based solely on visual criteria . scMOCa is highly efficient , as the fraction of tagged cells collected in the top chamber exhibits minimal capture losses and high specificity . Rare false positive cells , presumably attached by cell junctions to true positive cells , can be eliminated by repeating the sorting procedure to reach 100% purity . The most widely used cell sorting technique , FACS , is not optimized for sorting rare cells . Adaptations needed for capturing cell populations representing <1% of the sample with high specificity make FACS experiments cumbersome and inefficient . Moreover , repetition of flow cytometry sorting to obtain pure samples of a given cell type imposes can only be performed with robust cell types due to reduced survival and proliferation capacity ( Pappas and Wang , 2007 ) . More refined procedures have been developed to sort rare cells via binding to microfluidic channels coated with antibodies against specific surface markers of interest ( Antfolk et al . , 2017 ) . However , this requires high-affinity antibodies that are specific to the target cell types and leads to dilution of cells in laminar flows within microfluidics chips ( Moon et al . , 2011; Kang et al . , 2012 ) , which can become a drawback for downstream applications . Techniques based on magnetism display an improved capacity to isolate rare cells without dilution ( Tham et al . , 2014 ) . Nevertheless , while the majority of protocols that use magnetic fields can capture cells of interest with efficiency near 90% , their specificity remains a major challenge , as published results vary between 10% and 80% purity for captured cells ( Miltenyi et al . , 1990 ) , generally closer to 50% ( Zborowski and Chalmers , 2011; Pamme and Wilhelm , 2006; Radbruch et al . , 1994; Khojah et al . , 2017 ) . Finally , only a handful of approaches allow label-free cell sorting , where intrinsic physical properties , such as size ( Zhao et al . , 2017; Monti et al . , 2017 ) or magnetic susceptibility ( Moon et al . , 2011; Pamme and Wilhelm , 2006; Hosokawa et al . , 2010 ) differentiate the target population . Filtration , for example , relies on porous membranes to capture cells based on size and deformability ( Davis et al . , 2006; Gascoyne et al . , 2009 ) and can achieve 80% efficiency . Dielectrophoresis exploits natural differences in dielectric properties of cell types for discrimination and circulates cells in microfluidics channels , deviating target cells within an electric field ( Hu et al . , 2005; Landry et al . , 2015 ) . The application we introduced here is focused on magnetic separation , but the same concept of adding particles to individual live cells may open the door to novel strategies where other actionable properties can be exploited in a simple and straightforward manner . For example , fluorescence or electron density can be manipulated on single cells ( Binan et al . , 2016 ) , and recent advances in cellular nanotechnologies such as scattering and plasmon resonance using gold nanoparticles , thermal capacity with nanoshells , or electrical properties using carbon nanotubes can now be modulated only on chosen cells using low-cost commercially available reagents . ScMOCa presents critical advantages over more traditional sorting techniques . It allows isolation of live cells without previous knowledge of surface markers and can simply be based on morphological traits such as the presence of nuclear foci or lipid vesicles and the number nuclei . More importantly , it has the potential to sort based on time-dependent characteristics such as migration speed or foci resolution . In addition , because sorting is carried out in small chambers of similar size , there is no sample dilution . This prevents cells from sustaining strong shear stress upon passing through microfluidic tubing ( Miltenyi et al . , 1990 ) , and allows their use in downstream applications such as cell culture , reinjection , or even lysis prior to transcriptomic or proteomic analysis . ScMOCa crosslinks biotin to cell membrane and the strength of the ensuing biotin-streptavidin bond is extremely high ( Kd = 10−15M ) . In comparison , the bonds utilized in immunochemistry are much weaker , from 10−12 to 10-9 39 , 40 , which may cause tags to detach from cells because of shear stress within the microfluidics tubing ( Wooldridge et al . , 2009 ) . Another example is provided by ligands targeting the major histocompatibility complex ( MHC ) on immune cells where binding strength is so weak that ligands usually need to be grouped in tetramers for increased strength ( Tsai et al . , 2004; van der Toom et al . , 2017 ) . Finally , while the precise mechanisms influencing 53BP1 focus resolution was not investigated in our proof-of-principle experiments , our data demonstrates that markers used for identification need not be exposed on the membrane since the spatial distribution of fluorescent signal originating from the nucleus were used here as a reporters . Simplicity is a key advantage of scMOCa , as it does not require highly specialized software , or hardware such as microfluidic chips . Indeed , a standard confocal microscope with no modification , simple handmade chambers and low-cost magnets are all that is needed to sort single cells of choice from among tens of thousands . The main limitation of scMOCa is that high throughput implementations would depend on efficient image processing tools for cell detection . While automated detection and tagging are possible on motorized microscopy systems , the duration of the procedure is roughly proportional to the number of target cells . Thus , even if laser illumination of a single cell typically requires one second , this might become a limitation for applications that deal with large cell numbers . The capacity of scMOCa to isolate and profile individual cells within a large population based purely on visual phenotypes constitutes a powerful tool for understanding cellular heterogeneity . We envision that one potential application of high interest would combine scMOCa with single cell sequencing to characterize the molecular basis of differential metastatic potential among particular cells within a tumour ( Navin et al . , 2011; Valastyan and Weinberg , 2011; Shapiro et al . , 2013; Tirosh et al . , 2016; Heitzer et al . , 2013; Gierahn et al . , 2017 ) . Indeed , scMOCa can easily be combined with currently available techniques that allow sequencing RNA from single cells captured in wells ( Brennecke et al . , 2013 ) and microfluidic chips ( Wu et al . , 2014; Tan et al . , 2017 ) . More generally , it is becoming increasingly obvious that the capacity to analyze rare cells in heterogeneous populations will be useful in designing personalized treatments for cancer ( Hood et al . , 2004; Pugia et al . , 2017 ) as well as for inflammatory , autoimmune , and neurologic disorders ( Miltenyi et al . , 1990; Weissleder , 2009; Hesketh et al . , 2017 ) . U2OS osteosarcoma cells , MDCK ( dog ) cells , and IMCD ( mouse ) cells were grown in DMEM/F12 medium supplemented with 10% FBS and antibiotics , all purchased from Thermofisher Scientific . One day prior to the experiment , cells were detached and seeded on either collagen-coated glass coverslips or circular pieces of Aclar ( polychlorotrifluoroethylene ) coated with collagen , onto which polydimethylsiloxane ( PDMS ) chambers had been placed ( see below ) . A U2OS cell line with inducible expression of GFP-tagged 53BP1 was constructed as previously described ( Al-Hakim et al . , 2012 ) using pcDNA5-FRT/TO-eGFP-53BP1 ( Fradet-Turcotte et al . , 2013 ) ( Addgene plasmid #60813 ) and the U2OS Flip-In TREX host cell line ( Brown et al . , 1997 ) ( both generous gifts from Dr . Daniel Durocher , University of Toronto ) . Cells were selected in medium supplemented with 200 µg/mL hygromycin and 5 µg/mL blasticidin . GFP-53BP1 expression was induced by addition of 5 µg/mL doxycycline for 48 hr . H226 cells were grown in RPMI medium supplemented with 5% FBS and antibiotics ( Thermofisher Scientific ) . Four days prior to the experiment , cells were exposed to 6 µg/mL cytochalasin B for 24 hr . Low-passage primary human lung fibroblasts ( LF-1 ) were a kind gift from Dr John Sedivy ( Talbot et al . , 2015 ) . Cells were grown in Eagle's MEM ( Corning ) containing 15% FBS , essential and nonessential amino acids , vitamins , L-glutamine , and antibiotics ( Life Technologies ) . HUVECS were grown in Endogro TM ( Millipore ) supplemented with VEGF . Primary dorsal root ganglion ( DRG ) neurons were harvested from IsI-Gcamp6 x TRPV1-cre mice and cultured in plastic bottom dishes ( as detailed elsewhere [Bélanger et al . , 2018] ) one day prior to the sorting . Pre-adipocyte 3T3-L1 cells were grown in DMEM medium supplemented with 10% FBS ( Gibco ) , 2 mM glutamine ( Wisent ) and 1% Penicillin/Streptomycin ( Biobasic ) . For adipogenic differentiation of 3T3L1 , the cells were plated at confluency and media was changed to induction media containing 10% FBS , 1% Penicillin/Streptomycin , 1 μM Dexamethasone , 1 μg/ml Insulin and 500 μM IBMX ( Sigma ) . Two days post-induction , the medium was changed to maintenance media containing 10% FBS ( Gibco ) , 1% Penicillin/Streptomycin ( Biobasic ) , 1 μg/ml Insulin . After 3 days post-induction , 10 , 000 cells were plated on homemade chambers for sorting . Mouse Embryonic Stem cell ( mES ) culture mES cells were grown in DMEM medium supplemented with 15% FBS ( embryonic stem cell qualified , Wisent ) , 1 X non-essential amino acids ( Sigma ) , 100 μM 2-Mercaptoethanol ( Gibco ) , 1000 Units/mL Leukemia inhibitory factor ( LIF , Stemcell ) , 2 mM glutamine ( Wisent ) and 1% Penicillin/Streptomycin ( Biobasic ) on 0 . 1% porcine gelatin-coated plastic dishes ( Sigma ) . About 10 , 000 cells were plated for sorting as above . PDMS chambers were prepared by pouring a mix of resin and curing agent ( 10:1 ratio ) in a petri dish to achieve a gel thickness of 2 mm . The dish was degassed overnight in a vacuum chamber and the resin allowed to polymerize at room temperature for 2 days . Square pieces were cut with a blade , circular wells of 5 mm diameter were made using a biopsy punch from Miltex ( 33-38 ) ( see Figure 1B and C ) and placed on either glass or Aclar coverslips ( onto which PDMS naturally adheres ) . Cells were incubated in regular medium with 40 μg/mL biotin-4-fluorescein ( Sigma ) on glass coverslips or Aclar substrates . A spot within each cell of interest was illuminated at 473 nm with the laser of a confocal microscope at 75 μW for 2 s with 10 × 0 . 4 NA objective . The sample was then thoroughly rinsed in PBS , and medium containing 8 μL of streptavidin-coated ferromagnetic beads of 2 . 8 μm in diameter ( Thermofisher , 65305 and 11533D ) was added . When beads were attached to a whole area rather than a single cell ( Figure 1B and Figure 1—figure supplement 1 ) the sample was scanned with a 700 µW laser scanned at 0 . 2 mm/s with a 0 . 4 NA objective in a succession of lines 0 . 005 mm apart to form a pattern generated from a binary image . Beads were pulled down in contact with the cells and re-suspended 3 times , attracted by a magnet placed alternatively below or above the sample . Cells were then rinsed thrice with PBS and a magnet was positioned above the sample to remove unbound beads . After this , very few beads remain in the dish ( Figure 1C ) . Cells are detached using 0 . 25% trypsin ( Thermofisher , 25200072 ) for magnetic capture . The resulting cell suspension is then subjected to a magnetic field that attracts positive cells upwards to a collection chamber , while negative cells settle by gravity in the original chamber , regardless of the total number of cells in the sample . More specifically , once the original PDMS culture chamber contains a suspension of individual cells in trypsin , a second identical PDMS culture chamber is placed on top of the first one as depicted in Figure 2A . The structure that holds the top chamber in place can be built with Lego bricks ( Figure 1—figure supplements 3 and 4 ) : the collection chamber is positioned between two Lego bricks that maintain it at 6 mm above the cells ( Figure 1C ) . While magnetic attraction of tagged cells toward the collection chamber is quick , negative cells require 4 min to settle down to the original chamber before the top chamber is separated , flipped , and the magnets removed . This procedure needs to be performed slowly to minimize turbulence and to avoid capture of negative cells . These manipulations are repeated three times to attain maximum specificity ( Figure 2C ) . The collection chamber is always filled with trypsin solution to avoid rapid cell adhesion , and gentle up and down pipetting can be performed to prevent cell clumping . Only for the last capture is the collection chamber filled with medium in which the cells will be expanded . The entire procedure is summarized in Figure 2C . Experimental conditions need to be fine-tuned for different cell types . The most important parameters that need to be optimised are surface coating of both donor and collection chambers , duration and number of repeats of the sorting steps . The collection chamber should provide optimal plating efficiency to maximize cell survival of very few cells while the donor chamber should allow strong adhesion of the cells to allow thorough rinsing of free magnetic beads . In our experience collagen coating provides strong cell attachment , but also generates extracellular fibers where beads and negative cells can be entangled and captured . Gelatin solves the issue of collagen fibers , but cell adhesion is slightly reduced , which may cause cell loss during rinsing . Uncoated substrates are an easy solution for cells like U2Os but many cell types including primary cells do not proliferate well on such surfaces . Plastic bottom chambers allow better cell adhesion and survival , but their reduced optical quality may hamper the precise observation of selection criteria . In this respect , Aclar possesses excellent optical properties and represents an excellent alternative . For most cell types , longer incubations ( approximately 4 min ) allow negative cells to settle down in the donor chamber , reducing the number of repeats required for optimal purity . On the contrary , experimentation with cells that adhere rapidly ( e . g . MDA-MB-231 ) , require the capture protocol to be performed as quickly as possible and more repeats may be needed . In our hands , the best results were obtained using 10 magnets each generating a 1 . 2 Gauss magnetic field and 2 mm deep PDMS wells . In this condition , it is important that the distance between the bottom of each chamber is kept at 6 mm to allow the magnetic field to attract all tagged cells against gravity to the collection chamber while preventing the turbulence generated by the separation of the chambers to bring negative cells into the collection chamber . Increasing this distance requires a stronger magnetic field , which in turn reduces viability of captured cells . The diameter of the chambers should also be 5–6 mm , to ensure the necessary surface tension that allows merging and splitting the media in both donor and collections chambers . 30 , 000 U2Os cells were plated in our homemade chambers 1 day prior to sorting . On the day of the experiment , 30 cells were arbitrarily chosen and tagged in three independent experiments . We manually counted and verified that the right number of cells ( 30 ) were covered with magnetic beads in each dish . Commercial MACS columns were washed with PBS containing 0 . 5% BSA and 2 mM EDTA as indicated by the manufacturer . Cells were detached using 60 µL trypsin and then diluted in 500 µL of the same buffer and placed in the column in the magnets from Miltenyi Biotec . Columns were rinsed three time with buffer , then removed from magnets and washed with 5 mL buffer . Cells were then centrifuged , resuspended in 70 µL medium and placed in new homemade chambers for observation and counting under the microscope . Any cell that had visible magnetic beads on its membrane was considered as a positively selected cell , while cells free of beads were counted as negative cells . Forty-eight hours after induction of GFP-53BP1 with doxycycline , U2OS cells were irradiated with 0 . 5 Gy of IR . A first set of images was acquired with a 40X , 0 . 95 NA objective 45 min post irradiation , to detect focus formation . Cells that displayed a > 85% reduction in the number of foci at the second time point ( 2 hr ) were considered ‘fast-resolving’ . Biotin-4-fluorescein ( 0 . 04 mg/mL ) was then added to the medium , and such cells were illuminated for 2 s through a 10 × 0 . 4 NA objective with 75 µW of laser intensity at 473 nm . Immunofluorescence was performed to evaluate levels of endogenous 53BP1 foci . Briefly , cells were rinsed with PBS , and fixed 15 min with 4% paraformaldehyde in PBS . Cells were then permeabilized for 10 min with 0 . 5% Triton X-100 in PBS , rinsed twice in PBS and twice in PBS + 0 . 05% Tween-20 and then blocked in PBS + 3% BSA and 0 . 05% Tween20 . Rabbit anti-53BP1 antibody ( Santa-Cruz ) was diluted 1:500 and incubated on the cells for 3 hr . Cells were rinsed in PBS + 0 . 05% Tween-20 and incubated with Alexa-488 anti-rabbit for 1 hr , washed three additional times , and finally imaged for focus quantification . An image processing pipeline was programmed to fully automate DNA focus detection as we have previously done ( Bélanger et al . , 2016; Otsu , 1979 ) . Cell nuclei were detected using the background signal of remaining free GFP-53BP1 protein by Otsu thresholding63 . This initial detection was used to create a mask , where objects were filtered for their size , signal saturation , and shape . A band-pass filter was used to enhance the signal generated by objects the size of a 53BP1 focus . Local maxima were then detected using a threshold automatically calculated for each nucleus . Sorted multinucleated H226 cells were stained 2 and 4 days after their isolation . Mitotracker green FM ( Thermofisher Scientific , M7514 ) was used at 150 mM for 20 min , followed by a 5-min incubation in Hoechst 33342 to stain nuclei , and WGA-alexa 647 to stain plasma membranes . Images were acquired with a 60 × 1 . 35 NA objective . Cell selection and CLaP were performed on an Olympus IX71 microscope ( Olympus Corp . ) with the appropriate epifluorescence filters , in medium at 37°C , 5% CO2 , with a 10 × 0 . 4 NA objective and an Orca Flash 4 . 0 camera ( Hamamatsu Photonics ) . Images of irradiated GFP-53BP1 expressing cells were taken at two time points using a 40X , 0 . 85NA objective and compared to identify outliers . Since laser tagging was performed with a 10 × 0 . 4 NA objective , cells were identified in a new live image at different magnification during tagging . Automatic acquisition of immunostained samples for characterization of large numbers of cells from purified cell populations was performed with an automated Zeiss AxioObserver Z1 Epifluorescence microscope , at room temperature in PBS with Zen Blue software and a 20 × 0 . 85 NA objective . Exponentially growing cell cultures were trypsinized , fixed with 70% ethanol , and stored at −20°C until use . Fixed cells were washed with PBS and treated with 0 . 5% triton X-100 for 10 min at room temperature . After washing with PBS , cells were resuspended in PBS containing 2 μg/mL propidium iodide and 0 . 2 mg/mL RNase A and incubated for 30 min at room temperature . Samples were analysed by flow cytometry on a FACSCalibur instrument ( Becton-Dickinson ) . Data was analyzed with FlowJo v10 software , and cell cycle phases were determined using the Watson algorithm . U2OS cells were plated at a density of 2 million cells per 10 cm dish . 24 hr later , medium was removed and filtered through a 0 . 2 µm filter to ensure sterility and remove any floating cells . Conditioned medium was always prepared fresh . 10 cells were resuspended in 40 µL of water and boiled for 10 min . Samples were subjected to 24 PCR cycles using Agilent Herculase II with primer sets specific for the mitochondrial gene Cytb of either dog or mouse . 2 µL of each reaction were then used for PCR or qPCR with each primer set . Total genomic DNA from either dog or mouse cells were used as controls . The primers used are Cytb1L ( 5′- CATAGCCACAGCATTCATGG −3′ ) , Cytb1R ( 5′- GGATCCGGTTTCGTGTAGAA −3′ ) , and Cytb2L ( 5′- CCTCAAAGCAACGAAGCCTA −3′ ) , Cytb2R ( 5′- TCTTCGATAATTCCTGAGATTGG −3′ ) , which amplify fragments of 247 nt and 196 nt from the Cytb gene of dog and mouse , respectively . Quantitative PCR was performed with the above primer pairs using the 2X SYBR Green Master Mix ( Bimake ) and an ABI7500 instrument ( ThermoFisher ) . The amount of dog and mouse DNA in each sample was calculated using standard curves made from serial dilutions of genomic DNA isolated from each cell type . Immunoblotting was performed with total cellular extract using standard protocols . Antibodies used were rabbit anti-53BP1 ( Santa-Cruz , sc-22760 ) and rat anti-tubulin ( Abcam , ab6161 ) .	When scientists use microscopes to look at cells , they often want to then isolate certain cells based on how these look like . For example , researchers may want to select cells with specific shapes , movements or division rates , because these visual clues give important information about how the cells may be behaving in the body . However , it remains difficult to precisely pick a few live cells within a bigger sample . To address this problem , Binan et al . created a new approach , called single cell magneto-optical capture ( scMOCa ) , to set aside specific cells within a larger population . The technique uses the lasers present on confocal microscopes to attach tiny metallic beads to the surface of chosen cell . Then , a magnetic field is applied to gently pull the cell to a new location . The method is cheap – it relies on commonly available research tools – and it works on a broad variety of cells . In the future , scMOCa could be used to capture and then grow cells that can only be recognized by how they look or behave , which will help to study them in greater details .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "tools", "and", "resources", "cancer", "biology" ]	2019	Opto-magnetic capture of individual cells based on visual phenotypes
α-Xenorhabdolysins ( Xax ) are α-pore-forming toxins ( α-PFT ) that form 1–1 . 3 MDa large pore complexes to perforate the host cell membrane . PFTs are used by a variety of bacterial pathogens to attack host cells . Due to the lack of structural information , the molecular mechanism of action of Xax toxins is poorly understood . Here , we report the cryo-EM structure of the XaxAB pore complex from Xenorhabdus nematophila and the crystal structures of the soluble monomers of XaxA and XaxB . The structures reveal that XaxA and XaxB are built similarly and appear as heterodimers in the 12–15 subunits containing pore , classifying XaxAB as bi-component α-PFT . Major conformational changes in XaxB , including the swinging out of an amphipathic helix are responsible for membrane insertion . XaxA acts as an activator and stabilizer for XaxB that forms the actual transmembrane pore . Based on our results , we propose a novel structural model for the mechanism of Xax intoxication . Pore-forming toxins ( PFTs ) are soluble proteins produced by bacteria and higher eukaryotes , that spontaneously form pores in biomembranes and act as toxins ( Dal Peraro and van der Goot , 2016 ) . Dependent on their transmembrane region , which is formed either by α-helices or β-strands , PFTs are classified as α-PFTs and β-PFTs ( Iacovache et al . , 2008; Bischofberger et al . , 2012 ) . A common trait of all PFTs is the conversion from a soluble monomer to a membrane-embedded oligomer ( Bischofberger et al . , 2012 ) ; however , a different mechanism has been recently found for ABC toxins ( Gatsogiannis et al . , 2016 ) . Specific targeting of the PFTs to the host membrane involves mostly recognition of specific proteins , glycans or lipids on the target membrane . Conformational changes resulting in the oligomerization and membrane perforation are triggered by receptor binding , catalytic cleavage , pH change or other factors ( Iacovache et al . , 2008 ) . The sequential order of oligomerization and membrane penetration including the formation of an oligomeric pre-pore is still a matter of debate ( Cosentino et al . , 2016 ) . The size of the oligomers ranges from tetrameric pores in Cry1Aa ( Gómez et al . , 2014 ) and heptameric pores in the anthrax protective antigen ( Jiang et al . , 2015 ) to 30–50-meric pores in cholesterol-dependent cytolysins ( CDCs ) ( Dal Peraro and van der Goot , 2016; Hotze and Tweten , 2012 ) . PFTs can be further divided into two groups ( Iacovache et al . , 2010 ) . PFTs of the first group perforate membranes by forming stable pores resulting in an uncontrolled influx and efflux of ions and other biomolecules . This destroys ion gradients and electrochemical gradients at the membrane . The toxins of the second group also perforate the membrane , but use the transmembrane channel to specifically translocate toxic enzymes into the host . Binary toxins , also called AB toxins ( Odumosu et al . , 2010 ) and also recently characterized ABC toxins ( Meusch et al . , 2014 ) belong to the latter group . A prominent AB toxin is the anthrax toxin ( Collier and Young , 2003 ) , where component B , the protective antigen , forms a translocation pore through which lethal factor or edema factor , different A components , are translocated . The members of α-PFTs show a high structural diversity . They include proteins mainly consisting of α-helical structures ( bax , colicins ) or β-strand motifs with a single helix responsible for membrane insertion ( actinoporins ) ( Dal Peraro and van der Goot , 2016; Parker and Feil , 2005 ) . Their transmembrane regions are all composed of hydrophobic or amphipathic regions buried within the core structure of the soluble monomer . Therefore , a conformational change that exposes the hydrophobic or amphipathic region is required for successful membrane insertion . The structures of cytolysin A ( ClyA ) ( Wallace et al . , 2000; Mueller et al . , 2009 ) and fragaceatoxin C ( FraC ) ( Wallace et al . , 2000; Mueller et al . , 2009 ) of both the monomer and oligomer gave the first structural insight into the mechanism of action of this class of PFTs . In contrast to α-PFTs , the structures of many β-PFTs , such as members of the cholesterol-dependent cytolysins ( Hotze and Tweten , 2012 ) , hemolysin and aerolysin family ( Dal Peraro and van der Goot , 2016 ) , have been determined in their monomeric and pore conformation . The transmembrane β-strands in the soluble monomers of β-PFTs are typically amphipathic with small hydrophobic patches that upon oligomerization form a hydrophobic membrane-spanning β-barrel . α-Xenorhabdolysin is a PFT that has been first isolated from the bacterium Xenorhabdus nematophila ( Ribeiro et al . , 2003 ) . Xenorhabdolysins are also found in other entomopathogenic bacteria , such as Photorhabdus luminescens , and human pathogenic bacteria , such as Yersinia enterocolitica and Proteus mirabilis ( Vigneux et al . , 2007 ) . They are composed of two subunits , namely XaxA ( 45 kDa ) and XaxB ( 40 kDa ) and are only active when the two components act together ( Vigneux et al . , 2007 ) . Xenorhabdolysins , which were suggested to be binary toxins ( Vigneux et al . , 2007; Zhang et al . , 2014 ) , perforate the membranes of erythrocytes , insect granulocytes and phagocytes and induce apoptosis ( Vigneux et al . , 2007; Zhang et al . , 2014 ) . The mechanism of action of xenorhabdolysins including the interaction between components A and B , oligomerization and pore formation has remained enigmatic so far . Structural prediction using the PHYRE2 server ( Kelley and Sternberg , 2009 ) does not yield any significant similarities for XaxB . XaxA cytotoxins , however , are predicted to be similar to two pore-forming cytolysins , Cry6Aa from Bacillus thuringiensis ( Huang et al . , 2016 ) and binding component B of hemolysin BL ( Hbl-B ) from Bacillus cereus ( Madegowda et al . , 2008 ) . The best characterized cytolysin is probably ClyA from Escherichia coli and Salmonella enterica strains . The structure of ClyA has been determined in its soluble form ( Wallace et al . , 2000 ) and pore conformation ( Mueller et al . , 2009 ) and the mechanism of pore formation mechanism has been extensively studied ( Roderer and Glockshuber , 2017 ) . However , in contrast to XaxAB , ClyA only contains one component . Thus , despite the structural similarity , the mechanism of action must be different . So far structural data on xenorhabdolysins are missing limiting our understanding of these important type of toxins . Here , we used a hybrid structural biology approach combining X-ray crystallography and electron cryomicroscopy ( cryo-EM ) to determine the crystal structures of XaxA and XaxB from Xenorhabdus nematophila as soluble monomers and the cryo-EM structure of the XaxAB pore complex . In two different experiments , we independently expressed and purified XaxA and XaxB ( Materials and methods ) . The protein quantity and quality of both proteins was sufficient ( Figure 1—figure supplements 1a–b and 2a–b ) to perform crystallization experiments . We obtained well diffracting crystals of both XaxA and XaxB in their soluble monomeric form and solved their structures to 2 . 5 and 3 . 4 Å , respectively ( Figure 1a–b , Table 1 ) . Both XaxA and XaxB have a long rod-shaped structure and are mainly composed of α-helices ( Figure 1a–b ) . XaxA and XaxB have a similar domain organization . They both contain a tail domain that is formed by a five-helix bundle ( αA , αB , αC , αG and αH ) and elongated neck and head domains . The five-helix bundle motif has so far only been described for ClyA and ClyA-type toxins ( Roderer and Glockshuber , 2017 ) . Like in the case of ClyA-type toxins the N-terminal helices ( αA ) of XaxA and XaxB are significantly shorter than in ClyA , where it plays a crucial role in pore formation ( Figure 1—figure supplement 4 ) . Interestingly , XaxA contains two large loops connecting the helices , a big hook-shaped loop ( lp1 , aa 136–169 ) between helices αB and αC at the top of the tail domain and an additional loop ( lp2 , aa 202–215 ) dividing helix αC ( Figure 1a ) . The four XaxB molecules in the asymmetric unit differ considerably in their tail domain ( Figure 1—figure supplement 5 ) . Especially , helices αB and αC that protrude slightly from the five-helix-bundle take different positions . Although this might be due to tight crystal packing , it also indicates a certain degree of flexibility of the tail domain of XaxB . A long coiled-coil structure , composed of a continuous helix ( αG ) and another one that is divided into three ( XaxA: αC1 , αC2 and αD ) or two ( XaxB: αC , αD ) segments , form the backbone of XaxA and XaxB . It connects all domains and forms in both XaxA and XaxB the neck and head domain . The neck domain , that is approximately 35 Å in length , does not exist in ClyA-type toxins , which are in general more compact ( Figure 1—figure supplement 4a ) . In XaxA , the tip of the coiled-coil , predicted as hydrophobic transmembrane region is not resolved in our crystal structure , however , secondary structure predictions for this region suggest a continuation of the coiled-coil ( Figure 1—figure supplement 6 ) . In contrast to XaxA the head domain of XaxB contains in addition to the central coiled-coil a helix-loop-helix motif , dividing helix αE and a 21-residue long amphipathic helix ( αF ) . The highly conserved hydrophobic face of helix αF is oriented toward helices αD and αG and thereby shielded from the solvent ( Figures 1b and 2 ) . The conformation of the head domain is stabilized by conserved hydrophobic as well as electrostatic interactions , including putative salt bridges ( Figure 2 ) . In general , the overall fold of the soluble monomers is similar to that of ClyA from Escherichia coli ( Wallace et al . , 2000 ) or ClyA-type toxins , such as Cry6Aa from Bacillus thuringiensis ( Huang et al . , 2016 ) , non-hemolytic enterotoxin A ( NheA ) ( Ganash et al . , 2013 ) , and binding component B of hemolysin BL ( Hbl-B ) from Bacillus cereus ( Madegowda et al . , 2008 ) ( Figure 1—figure supplement 4a ) . An important feature of ClyA and ClyA-type cytotoxins is the typical tongue motif that inserts into the membrane during pore formation ( Mueller et al . , 2009 ) ( Figure 1—figure supplement 4a , b ) . In ClyA , Hbl-B , NheA , and Cry6Aa the tongue is formed by a hydrophobic or amphipathic β-hairpin or a large hydrophobic loop ( Wallace et al . , 2000; Huang et al . , 2016; Madegowda et al . , 2008 ) . Interestingly , in the case of XaxB the tongue is formed by an amphipathic helix , while XaxA does not contain such motif . Comparing the structure of XaxA with that of its pore conformation ( see below ) suggests that XaxA is already in its extended conformation as soluble monomer . It is tempting to speculate that the function of the N-terminal helix and β-tongue in ClyA has been evolutionary compensated in multicomponent toxins , such as XaxAB , NheABC and Hbl-ABC that only contain a short N-terminal helix . In the case of XaxAB , the hydrophobic helices of XaxA that enter the membrane in the pore , are functionally equivalent to the hydrophobic β-tongue of ClyA , The β-tongue likely inserts first into the membrane , where it rearranges into two α-helices ( Mueller et al . , 2009 ) . Similar to XaxA , these helices only span half the membrane . αF and the helix-turn-helix motif αE of XaxB , that span the complete membrane in the XaxAB pore , would substitute the N-terminal membrane-spanning helix of ClyA . To investigate the pore complex formed by XaxA and XaxB , we planned to induce pore formation in vitro and analyze the structure of the complex by single particle electron cryomicroscopy ( cryo-EM ) . We first mixed both soluble monomers , incubated them with a variety of detergents and analyzed the pores by negative stain electron microscopy . We could indeed observe pore formation in most cases; however , the choice of detergent greatly influenced the size and homogeneity of the observed crown-shaped pore complexes . Some detergents induced the formation of star-like aggregates or differently sized pores ( Figure 3—figure supplement 1 ) . We observed the most homogenous distribution of XaxAB pore complexes , that appear as crown-shaped structures , after incubating the monomers with 0 . 1% Cymal-6 ( Figure 3—figure supplement 1c , Figure 3—figure supplement 2b ) . The average diameter of the pores was ~250 Å . However , the pores had the tendency to aggregate and were not suitable for further structural investigations . Interestingly , when we incubated soluble monomers of XaxA and XaxB in the absence of detergents at room temperature , we observed the formation of higher oligomers but not of complete pores ( Figure 1—figure supplement 3a , b ) . This is not the case when XaxA and XaxB are not mixed ( Figure 1—figure supplements 1 and 2 ) . This indicates that heterodimerization and oligomerization of XaxAB can happen independently of the hydrophobic environment provided by detergents or a lipid bilayer and may happen prior to pore formation also in vivo . To improve the homogeneity of our XaxAB pore complexes , we exchanged Cymal-6 with amphipols and separated the amphipol-stabilized XaxAB pores from the aggregates by size exclusion chromatography ( Figure 3—figure supplement 2c , d ) . The thus obtained pore complexes were homogeneous and suitable for single particle cryo-EM . Analyzing the single particles by two-dimensional clustering and sorting in SPHIRE ( Yang et al . , 2012; Moriya et al . , 2017 ) revealed populations of XaxAB pores with different numbers of subunits ( Figure 3—figure supplement 3 ) . Most of the complexes contain either 12 , 13 , 14 or 15 subunits . We separated the different populations by multi-reference alignment and solved the structure of the different complexes in SPHIRE ( Moriya et al . , 2017 ) ( Materials and methods , Figure 3—figure supplement 4–5 ) . The average resolutions of the reconstructions were 5 , 4 , 4 . 2 and 4 . 3 Å for do- , tri- , tetra- , and pentadecameric pores , respectively ( Figure 3—figure supplements 4–5 ) . We used the highest resolved cryo-EM density of the tridecameric pore complex to build an atomic model of XaxAB ( Video 1 , Figure 3 , Table 2 ) . The high quality of the map allowed both models to be almost completely built , except for the first residues of the N-terminal helix αA in XaxA ( aa 1–40 ) and XaxB ( aa 1–12 ) . These regions are also not resolved in the crystal structures indicating a high flexibility of the N-termini . The pore complexes have a total height of 160 Å and depending on the number of subunits a diameter of 210 to 255 Å . Each subunit consists of a XaxAB heterodimer with XaxA bound to the back of XaxB . This results in a localization of XaxA on the periphery of the pore , whereas XaxB resides more at the center of the complex lining the inner pore lumen ( Figure 3 , Video 1 ) . Interestingly , the transmembrane helices of XaxA that fortify the inner ring of helices of XaxB , do not completely span the membrane ( Figure 3—figure supplement 6 ) . The arrangement of the components clearly shows that XaxAB is not a binary toxin as suggested ( Vigneux et al . , 2007; Zhang et al . , 2014 ) , but rather a bi-component toxin , such as BinAB from Lysinibacillus sphaericus ( Colletier et al . , 2016 ) and leukocidin A and B ( LukGH and SF ) from Staphylococcus aureus ( Badarau et al . , 2015 ) where both proteins contribute to building the pore . Depending on the number of subunits , the inner diameter of the pore narrows down from 140 to 170 Å at the membrane-distal part to 40–55 Å at the transmembrane region . The inner surface of the pore is hydrophilic and mostly negatively charged suggesting a preference for positively charged ions and molecules ( Figure 3—figure supplement 7 ) . At the outside , the pore complex has a conserved highly hydrophobic band of 40 Å corresponding to the transmembrane region ( Figure 3—figure supplement 7 ) . The hydrophobic band merges into a positively charged stretch that is formed by the conserved arginine and lysine residues of XaxA ( R290 , K291 , K293 , K295 , K301 ) ( Figure 3—figure supplements 7–8 ) . These residues likely interact with negatively charged lipid head groups of target membranes and thereby stabilize the pore complex in the lipid bilayer . When comparing the shape of XaxAB with that of the pores of FraC and ClyA , we found that the crown-like structure of XaxA is shared by actinoporin FraC ( Tanaka et al . , 2015 ) but not by ClyA ( Mueller et al . , 2009 ) , where the extramembrane regions form a cylinder ( Figure 3—figure supplement 9a ) . In agreement with the smaller number and size of subunits in FraC and ClyA , these pores have a smaller diameter than the XaxAB pore , and , in addition , FraC contains large β-sheets in the extramembrane region ( Figure 3—figure supplement 9b ) . Interestingly , the lumen of all pores is negatively charged ( Figure 3—figure supplement 9c ) , suggesting the same preference for positively charged molecules . The tail domains of XaxA and XaxB do almost not differ between the oligomeric pore conformation and soluble monomers . The neck and head domains of XaxA are also arranged similarly to the crystal structure , however , the coiled-coil is twisted by 15 Å and interacts with helices αB and αC of the adjacent XaxB ( Figure 4a , c , Figure 5 ) . The neck and head domains of XaxB , however , differ considerably in comparison to the soluble monomer . The amphipathic helix αF and the helix-loop-helix motif fold out , thereby extending helices αD and αG forming the transmembrane region ( Figure 4b , d , Figure 5 ) . The tail and head domains of XaxA and XaxB mediate interactions between the proteins in the heterodimer . We identified four major interfaces , two in the tail and two in the head domain region . The interfaces between the tail domains are stabilized by several putative hydrogen bonds and electrostatic interactions between helices αG and the C-terminal helix αH of XaxA and helices αB , αC and the C-terminal helix αH of XaxB ( Figure 5a–c ) . Dimerization of XaxA and XaxB probably helps stabilizing the tail domain of XaxB , which takes different positions in the crystal structure ( Figure 5a , Figure 1—figure supplement 3 ) . The first interface between the head domains is formed by helices αD and αG of XaxA that interact with helices αD and αG of XaxB via a putative hydrogen network and salt bridges ( Figure 5d ) . The second one is mediated by hydrophobic interactions between helices αF and αE of XaxB with αD and αG of XaxA ( Figure 5f ) . A prominent feature is the high accumulation of aromatic residues at this interface ( Figure 5e ) . Interestingly , some of these residues are also involved in stabilizing the soluble XaxB monomer ( Figure 2 ) . Since most of the interfaces between XaxA and XaxB in the heterodimer locate to the tail domain and do not differ between the soluble monomer and pore conformation , we suggest that heterodimer formation precedes membrane insertion . The heterodimers are linked manifold in the oligomeric pore . One XaxA interacts simultaneously with XaxA and XaxB of the adjacent heterodimer . The same is true for XaxB that interacts with both XaxA and XaxB of the adjacent heterodimer ( Figure 6 ) . Two major interfaces are mediated by the tail domains of XaxA and XaxB ( Figure 6a–c ) . The residues K45 , N50 , E398 , E402 and D333 that are conserved in XaxA form an extensive putative hydrogen network and salt bridges with helices αB ( R48 , Y52 ) and αC ( D138 , R147 ) of the adjacent XaxB ( Figure 6b ) . A second putative hydrogen network between helices αC2 and αG in XaxA and helices αD and αG in XaxB likely contributes to the stabilization of the oligomer ( Figure 6c ) . The oligomer is further stabilized by a putative salt bridge between two XaxAs . Glutamate E206 in the loop between αC1 and αC2 of XaxA of one subunit interacts with lysine K405 in the C-terminal helix of neighboring XaxA ( Figure 6d ) . The fourth interface is formed by the head domains of two XaxBs via several putative hydrogen bonds ( D197/S241 N192/K245 ) ( Figure 6e ) . Taken together the heterodimeric subunits of the complex and the heterodimer itself are stabilized by strong interactions that guarantee a stable pore complex inside the membrane . There are at least two concerted or consecutive steps during pore formation of PFTs , namely oligomerization and membrane penetration ( Cosentino et al . , 2016 ) . In bi-component toxins , where both proteins contribute to building the pore , the two components first dimerize into a heterodimer prior to oligomerization ( Colletier et al . , 2016; Badarau et al . , 2015 ) . In several cases , PFTs have been shown to oligomerize and insert spontaneously into membranes in vitro ( Iacovache et al . , 2008 ) . However , membrane insertion in vivo depends on the specific interaction with lipids or proteins on the membrane surface of the host ( Ros and García-Sáez , 2015 ) . To better understand the process of dimerization , membrane insertion and pore formation of XaxA and XaxB , we performed in vitro reconstitution assays with and without liposomes . XaxA alone has the tendency to form small aggregates by interacting with its head domain ( Figure 1—figure supplement 1d–e ) . Since the short hydrophobic region of the head domain resides inside the membrane in the pore complex , we believe that the clustering of XaxA monomers is caused by mild hydrophobic interactions of these regions . This again suggests that already the soluble monomeric form of XaxA has a certain affinity to the hydrophobic environment of biomembranes . To test , whether XaxA can spontaneously insert into or associate with membranes , we incubated it with 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine ( POPC ) or brain polar lipids ( BPL ) liposomes and analyzed its incorporation by size exclusion chromatography and negative stain electron microscopy ( Figure 7a–d ) . Interestingly , the protein was not incorporated into the liposomes and no larger structures could be observed on the vesicles ( Figure 7c , d ) . This indicates that albeit the hydrophobic tip of the head domain , XaxA cannot spontaneously insert blank membranes in vitro . The same is true for XaxB alone . When incubated with liposomes the protein neither perforates membranes nor oligomerizes on the liposomes ( Figure 7a , b , e , f ) . When both XaxA and XaxB were added to liposomes , they spontaneously associated with the vesicles and formed the typical crown-shaped structures ( Figure 7g–i ) as we have observed them in detergents ( Figure 3—figure supplement 1c ) . Notably , this is independent of the sequence of mixing , that is XaxA can be added before XaxB or vice versa , suggesting that dimerization of XaxA and XaxB is necessary for spontaneous association of the proteins with the membrane and subsequent pore formation . Importantly , association with liposomes happens without specific lipids , such as cholesterol , or protein receptors at the membrane surface . At this point , we cannot distinguish between association with and insertion into the membranes . Thus , formation of a pre-pore before membrane insertion cannot be excluded . In general , the transition from the soluble monomer to the protomer does not involve major structural rearrangements of the whole molecule . Only the conformation of the head domains changes considerably . Besides the described twist of XaxA ( Figure 4a ) , the α-helical tongue αF of XaxB folds out forming the transmembrane region . Interestingly , the conformation of the coiled-coil backbone in XaxB remains unaltered ( Figure 4b ) . This is in direct contrast to ClyA ( Wallace et al . , 2000 ) but similar to FraC ( Tanaka et al . , 2015 ) , the only other two α-PFTs , for which a structure of the soluble and pore complex has been determined at high resolution . In ClyA , not only the head domain but also the tail domain undergoes considerable conformational changes ( Mueller et al . , 2009 ) . In order to better understand the conformational changes during dimerization , oligomerization and pore formation , we compared the structures of the soluble and pore forms of XaxA and XaxB . When the crystal structures of XaxA and XaxB are overlaid with the respective XaxAB structure , it becomes obvious that the neck and head domains of the proteins would not interact ( Video 2 ) . In agreement with our reconstitution assays such a dimer would probably not be able to spontaneously insert into membranes . In the XaxAB pore conformation , however , helices αD and αG of XaxA , forming the coiled-coil backbone twist by 15 Å toward XaxB ( Figure 4a , Figure 5 , Video 2 ) . As described above , through this conformational change a stronger interaction with XaxB is created . Interestingly , without the conformational change in XaxA , oligomerization of XaxAB would not be possible because of prominent steric hindrances ( Video 3 ) . This movement is therefore crucial for complex formation . If we assumed that only XaxA and not XaxB changed its conformation during dimerization and oligomerization ( Video 2 , Video 3 ) , the transmembrane region of XaxA would sterically clash with the loop between αF and αG of XaxB from the adjacent subunit ( Videos 2 and 3 ) . This could in principle trigger conformational changes in XaxB that activate its head domain for membrane insertion . To better analyze these hypothetical conformational changes in detail , we created a heterodimer model comprising the cryo-EM structure of XaxA ( XaxAprot ) and the crystal structure of XaxB ( XaxBmon ) and analyzed interfaces and residues that might trigger membrane insertion ( Figure 8 ) . We identified two hinge regions that facilitate the swinging out movement of αF in XaxB ( Figure 8 , Video 2 , Video 3 ) . One hinge region is located in the hydrophobic loop between the short helices of the helix-loop-helix motif . It contains a highly conserved proline residue ( P204 ) that is also involved in stabilization of the soluble monomer ( Figure 2 ) . The second hinge is located in the loop connecting αF and αG , including the conserved residue G243 ( Figure 2 , Figure 8 ) . A cluster of aromatic residues at the bottom of the head domain of our XaxAprot-XaxBmon heterodimer model suggests that this region could be crucial in triggering the conformational changes in XaxB when exposed to a lipid membrane . In the heterodimer of the XaxAB pore complex , most of these residues build a hydrophobic cluster between the transmembrane domain of XaxA and the reorganized helix αE of XaxB ( Figure 8 ) . Aromatic residues have been shown to be important for membrane insertion of many PFTs and responsible for conformational changes induced by their interaction with membranes or detergents ( Mueller et al . , 2009 ) . Interactions with the membrane likely destabilize this region , inducing stronger conformational changes in the rest of the domain . Our atomic model of XaxA and XaxB in solution as well as in the pore conformation provides important insights into the interaction and function of these proteins . Although the structural record is lacking intermediate states , we can use the information provided by our structural data to define critical steps in the action of XaxAB toxins and suggest the following mechanism . Although XaxA and XaxB are not homologous , their structure is very similar . The two components of the xenorhabdolysin form heterodimers , 12 to 15 of which assemble into membrane-perforating pores . In contrast to previous predictions ( Vigneux et al . , 2007; Zhang et al . , 2014 ) , XaxAB is therefore not a typical binary toxin , but rather a bi-component α-PFT . So far , only structures of bi-component β-PFTs have been reported . Our structure of the XaxAB pore represents the first structure of a bi-component α-PFT . Our results show that XaxA and XaxB together form higher oligomers in the absence of detergent or membranes . In addition , XaxA likely activates XaxB during oligomerization by inducing conformational changes . We therefore propose that XaxA and XaxB dimerize ( Figure 9a–c , Figure 9—figure supplements 1a–c and 2a–c ) and oligomerize ( Figure 9d , Figure 9—figure supplements 1d and 2d ) in solution . Dimerization happens probably spontaneously since the conformation of domains located at the heterodimer interface in the tail domains of XaxA and XaxB is not different compared to the monomers . The conformational change in the neck and head domain of XaxA ( Figures 9b and 5d–f ) further stabilizes the interaction and is crucial for oligomerization ( Figure 9d ) . During oligomerization XaxA sterically clashes with the loop connecting helices αF and αG in XaxB . We therefore propose that XaxA induces conformational changes in XaxB that do not immediately result in exposing its hydrophobic domain but rather put XaxB in an activated state for membrane insertion ( Figure 9b–c ) . When interacting with a lipid membrane , aromatic residues at the bottom of the head domain of XaxB likely trigger the conformational change resulting in membrane perforation ( Figure 9d ) . Our reconstitution assays in liposomes showed that neither XaxA nor XaxB strongly interact with liposomes . Thus , neither the interaction of the aromatic residues of XaxB nor the hydrophobic domain of XaxA are able to enter membranes on their own and dimerization and induced conformational changes during oligomerization are crucial for membrane insertion . Since XaxB is the component that finally forms the pore , XaxA that only partially enters the membrane , acts like an activator of XaxB and stabilizes it in the pore complex . Recently , a new mechanism for ClyA membrane permeation has been suggested in which a conformational change in a ClyA monomer initiates the assembly of dimers and higher oligomers on the membrane forming a homo-dodecameric pre-pore complex that ultimately enters the membrane after an additional conformational change ( Benke et al . , 2015 ) . Although , we never observed structures at high resolution that would indicate a pre-pore complex , we cannot exclude that such a complex exists as intermediate state on liposomes before permeation ( Figure 9e , Figure 9—figure supplements 1e and 2e ) . Obviously , more evidence is needed before our proposed mechanism of XaxAB action can be regarded as established . Thus , additional structures of intermediate states are needed to fully understand the process . In summary , our results provide novel insights into the mechanism of action of xenorhabdolyins and serve as a strong foundation for further biochemical experiments to fully understand the molecular mechanism of xenorhadolysin intoxication . During the revision of our work , crystal structures and a low-resolution cryo-EM structure of the human pathogenic homolog YaxAB from Yersinia enterocolitica as well as a crystal structure of PaxB from Photorhabdus luminescens have been published ( Bräuning et al . , 2018 ) . The crystal structures of YaxA and XaxA , as well as YaxB ( PaxB ) and XaxB are very similar although their sequences are only 54 . 5 and 36 . 0% ( 56 . 7% ) identical , respectively ( Figure 4—figure supplement 1a , c , d ) . Importantly , YaxA does not contain the hook-shaped loop ( lp1 ) , which is a prominent feature of XaxA ( Figure 4—figure supplement 1a , b ) . The neck and head domain of XaxA and YaxA as well as the head of XaxB and PaxB differ in their position indicating that these domains are flexible in solution . This is supported by the fact that the head domain of XaxA and YaxB are not resolved in the structures . Similarly , the tip of the tail domain takes different positions in XaxB and is not resolved in YaxB ( Figure 4—figure supplement 1c , d ) . Although the authors used the same detergent for pore assembly and also stabilized the pores in amphipols , the YaxAB pore complex comprises 8 to 12 heterodimers in contrast to the 12 to 15 heterodimers in our XaxAB pore . This suggests a species-dependent size variability . The protomer structures of YaxAB and XaxAB are very similar ( RMSD of 1 . 145 and 1 . 252 , respectively ) . Significant differences , however , can be seen in the head domains which could in principle indicate structural differences between the proteins ( Figure 4—figure supplement 1b , e ) . However , since the relatively low resolution of the YaxAB pore structure ( 5 . 5 Å ) impedes an accurate building of the atomic model , only a high-resolution structure of YaxAB would enable a proper comparison . Interestingly , YaxA associates directly with erythrocyte membranes . This is in direct contrast to our findings showing that XaxA does interact with liposomes in vitro . This results in different models . Whereas Bräuning et al . hypothesize that YaxA enters the membrane first and then recruits YaxB , we propose that XaxA and XaxB already heterodimerize/oligomerize in solution and then associate with the membrane as heterodimers or oligomers . The genes coding for C-terminally His6-tagged XaxA and N-terminally His6-tagged XaxB were introduced into a pET19b vector and expressed in the E . coli BL21 RIPL ( DE3 ) expression strain . Both constructs contained a PreScission cleavage site . For the expression culture , 2 l of LB media containing 125 μg/ml ampicillin were inoculated with the preculture and cells were grown at 37°C until an OD600 of 0 . 5–0 . 8 was reached . Selenomethionine-substituted XaxB was expressed in the E . coli BL21 RIPL ( DE3 ) strain in M9 minimal medium with the addition of 100 mg/l L-lysine , 100 mg/l L-phenylalanine , 100 mg/l L-threonin , 50 mg/l L-isoleucine , 50 mg/l L-leucine 50 mg/l L-valine and finally 60 mg/l l-selenomethionine ( SeMet ) . Afterwards , protein production was induced by adding 0 . 4 mM of isopropyl-β-D-thiogalactopyranoside ( IPTG ) and incubated for 20 hr at 20°C . The cells were harvested and the bacterial pellet homogenized in a buffer containing 50 mM HEPES , pH 7 . 5 , and 200 mM NaCl . After cell disruption , the lysate was centrifuged at 38 , 000 rpm , 4°C and XaxA and XaxB was purified using Ni-NTA affinity and size-exclusion chromatography ( Superdex 200 26/600 , GE Healthcare ) . Crystallization experiments were performed using the sitting-drop vapor diffusion method at 20°C . XaxA crystals formed by mixing 0 . 1 µl of 40 mg/ml purified XaxA with 0 . 1 µl reservoir solution containing 0 . 2 M sodium chloride , 0 . 1 M phosphate citrate pH 4 . 2% and 10% PEG 3000 over a period of 3 weeks . SeMet-labeled XaxB ( 40 mg/ml ) was mixed in a 1:1 ratio with reservoir solution containing 0 . 2 M NaBr , 0 . 1 KCl and 20% PEG 3350 with a final drop size of 2 µl . Prior to flash freezing in liquid nitrogen , the crystals were soaked in reservoir solution containing 20% glycerol as cryo-protectant . X-ray diffraction data for XaxA was collected at the PXIII-X06DA beamline at the Swiss Light Source ( 24 datasets ) and at the DESY PETRA III beamline P11 ( 3 datasets ) from one crystal . The datasets were merged and used for phase determination . Data collection for XaxB was performed at the PXII-X10SA beamline . Datasets were indexed , integrated and merged with the XDS package ( Kabsch , 2010b , 2010a ) . XaxA crystallized in orthorhombic space group P212121 with a unit cell dimension of 67 × 90 × 153 Å and two molecules per asymmetric unit ( AU ) . Phases were determined using the anomalous sulfur signal of the merged datasets and HKL2MAP ( Pape and Schneider , 2004 ) , the graphical interface for SHELX C/D/E ( Sheldrick , 2010 ) . The obtained phases combined with the given sequence and a few placed α-helices in the density with COOT ( Emsley et al . , 2010 ) were sufficient enough for phenix autobuild ( Terwilliger et al . , 2008 ) to almost completely build the structure of XaxA . The structure was refined with the datasets collected at the DESY PETRA III beamline P11 . XaxB also crystallized in orthorhombic space group P212121 with a unit cell dimension of 89 × 99 × 194 Å and four molecules per AU . The diffraction data of XaxB was processed with the XDS package and SeMet atoms were determined using the CRANK2 pipeline ( Ness et al . , 2004; Pannu et al . , 2011 ) in the CCP4 software package . SHELX C/D ( Winn et al . , 2011 ) was used in the substructure detection process , while REFMAC ( Skubák et al . , 2004 ) , SOLOMON and PARROT ( Abrahams and Leslie , 1996 ) were used for phasing and substructure refinement and density modification for hand determination , respectively . BUCANEER ( Cowtan , 2012 ) gave the best results for the initial model-building step . This model was first optimized with phenix autobuild ( Terwilliger et al . , 2008 ) . The rest of the model was built in COOT ( Emsley et al . , 2010 ) using the anomalous peaks of the SeMet residues to determine the amino acid sequence due to the limited resolution . The structures were optimized by iteration of manual and automatic refinement using COOT ( Emsley et al . , 2010 ) and phenix refine implemented in the PHENIX package ( Adams et al . , 2010 ) to a final Rfree of 28 and 30% for XaxA and XaxB , respectively ( Table 1 ) . Stock solutions of 10 mg/ml 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine ( POPC ) and brain extract polar lipids ( BPL ) ( Avanti Polar Lipids ) were prepared in buffer containing 20 mM Tris-HCl pH 8 , 250 mM NaCl and 5% w/v n-octyl-β-D-glucopyranoside ( Antrace ) . 10 µM XaxA and XaxB were mixed with a final lipid concentration of 2 mg/ml and incubated for 30 min at room temperature . For reconstitution , the mixture was dialyzed against a buffer containing 20 mM HEPES pH 7 . 5 and 200 mM NaCl . The sample was then analyzed by size exclusion chromatography with a Superose 6 10/300 GL column ( GE Healthcare Life Sciences ) and by negative stain electron microscopy . XaxAB pore complexes were prepared by incubating equimolar concentrations of XaxA and XaxB with 0 . 1% cymal-6 ( Antrace ) at room temperature overnight . For a more homogenous and stable distribution of XaxAB pore complexes , the detergent was exchanged to amphipols A8-35 ( Antrace ) . Amphipols were added in fivefold molar excess and the solution was incubated at room temperature for 20 min . For detergent removal , the mixture was dialyzed against a buffer containing 20 mM HEPES pH 7 . 5 , 200 mM NaCl overnight at room temperature . Subsequently , aggregates and XaxAB pore complexes with higher molecular weight were separated by size exclusion chromatography on a Superose 6 10/300 GL column ( GE Healthcare Life Sciences ) . The quality of the XaxAB pore complexes was evaluated by negative stain electron microscopy before proceeding to cryo-EM grid preparation . 4 µl of a 0 . 01 mg/ml XaxAB solution in amphipols were applied to a freshly glow-discharged copper grid ( Agar Scientific; G400C ) coated with a thin carbon layer and incubated for 45 s . After sample incubation , the solution was blotted with Whatman no . four filter paper and stained with 0 . 75% uranyl formate . The digital micrographs were acquired with a JEOL JEM-1400 TEM equipped with an acceleration voltage of 120 kV , and a 4000 × 4000 CMOS detector F416 ( TVIPS ) with a pixel size of 1 . 33 Å/pixel . For sample vitrification , XaxAB pore complexes were concentrated to a final concentration of 1 mg/ml and 4 µl sample was applied onto freshly glow-discharged holey carbon grids ( C-flat 2/1 , Protochips ) , incubated for 45 s , blotted for 2 . 5 s and plunged into liquid ethane with a CryoPlunge3 ( Cp3 , Gatan ) at 90% humidity . The grids were then stored in liquid nitrogen . A cryo-EM dataset of XaxAB in amphipols was collected with a Cs-corrected TITAN KRIOS electron microscope ( FEI ) , with a XFEG and operated at an acceleration voltage of 300 kV . Images were acquired automatically using EPU ( FEI ) and a Falcon III ( FEI ) direct detector operated in counting mode at a nominal magnification of 59 , 000 x corresponding to a pixel size of 1 . 11 Å/pixel on the specimen level . In total 4746 images were collected with 180 frames , an exposure time of 60 s resulting in a total dose of ~44 e- Å−2 and a defocus range of 1 . 0–2 . 6 µm . Motion correction was performed using the MotionCor2 program ( Zheng et al . , 2017 ) . All image-processing steps were carried out with the SPHIRE software package ( Moriya et al . , 2017 ) ( Figure 3—figure supplement 4 ) . Initially , micrographs were manually screened for bad ice or high drift and discarded accordingly . The remaining 3617 motion-corrected sums without dose weighting were evaluated in aspect of defocus and astigmatism in CTER ( Moriya et al . , 2017 ) and low-quality images were discarded using the graphical CTF assessment tool in SPHIRE ( Moriya et al . , 2017 ) . 186 , 700 single particles were automatically picked from motion-corrected sums with dose weighting using gautomatch ( http://www . mrc-lmb . cam . ac . uk/kzhang/ ) . 2-D class averages were generated as a template for gautomatch by manually picking 200 micrographs with EMAN2 boxer ( Tang et al . , 2007 ) . Pre-cleaning of the dataset and reference-free 2-D classification were performed with the iterative stable alignment and clustering approach ISAC2 ( Yang et al . , 2012 ) in SPHIRE with a pixel size of 4 . 97 Å/pixel on the particle level . Refined and sharpened 2-D class averages with the original pixel size and exhibiting high-resolution features were generated with the Beautifier tool implemented in SPHIRE ( Figure 3—figure supplements 3 and 5b ) . The quality of the 2-D class averages were examined in regard of high-resolution features and completeness of the XaxAB pore complexes . According to observed oligomerization states of XaxAB pore complexes in the class averages , five initial 3-D models with c12 , c13 , c14 , c15 and c16 symmetry were generated with RVIPER . Particles were then sorted against the five RVIPER models using the 3-D-mulrireference projection matching approach ( sxmref_ali3d ) . The clean dataset was split into four datasets according to the number of XaxAB subunits in the complex: c12: 4409 particles , c13: 53 , 546 particles , c14: 46 , 596 particles and c15 34 , 542 particles . The sixteen-fold symmetry was discarded due to low number of particles ( 193 ) . The subsets containing particles with 13- , 14- and 15-fold symmetry were further cleaned with ISAC and subsequently subjected to 3-D refinements in MERIDIEN with a mask excluding amphipols and applying c12 , c13- , c14- , and c15-symmetry , respectively ( Moriya et al . , 2017 ) . In the following only the results of the map with the highest resolution will be described in detail . SPHIRE’s PostRefiner tool was used to combine the half-maps , to apply a tight adaptive mask and a B factor of −170 Å2 . The estimated average resolution according to the gold standard FSC@0 . 5/0 . 143 criterion between the two masked half-maps was 4 . 5/4 Å for the c13-symmetry ( Figure 3—figure supplement 5f ) . The estimated accuracy of angles and shifts at the final iteration of the 3-D refinement was 0 . 55 degrees and 0 . 6 pixels , respectively . The ‘Local Resolution’ tool in SPHIRE ( Figure 3—figure supplement 5e ) was used to calculate and analyze the local resolution of the c13 density map . The resulting colored density map showed a local resolution of up to 3 . 4 Å at the lower tail domain region , whereas the tip of the spikes at the top of the XaxAB pore and at the end of the transmembrane region showed the lowest resolution ( 5–6 . 7 Å ) ( Figure 3—figure supplement 5e ) . The final density was locally filtered according to the estimated local resolution using the ‘LocalFilter’ tool in SPHIRE . Details related to data processing are summarized in Table 2 . The atomic model of the XaxAB pore complex was built by isolating the EM density of a XaxAB dimer and rigid body fitting the crystal structure of XaxA into the EM density map using UCSF Chimera ( Pettersen et al . , 2004 ) . XaxA was further fitted into the dimer density using IMODFIT ( Lopéz-Blanco and Chacón , 2013 ) . For XaxA only the transmembrane region ( aa 254–283 ) had to be manually built , which was missing in the crystal structure . The final model of the XaxA protomer covers residues 41–405 of the full-length sequence with residues 1–40 missing at the N-terminal helix αA . XaxB was built by placing helix fragments into the remaining density with COOT ( Emsley et al . , 2010 ) , generating first a polyalanine model and subsequently determining the correct sequence by the identification of bulky side chains . The full sequence of the XaxB protomer is also almost covered in the final model ( aa 13–350 ) with the first 12 residues missing at the N-terminal helix αA . The XaxAB dimer was then rigid-body fitted into the XaxAB pore complex using UCSF Chimera ( Pettersen et al . , 2004 ) and the full model refined using PHENIX real-space refinement ( Adams et al . , 2010 ) . Finally , the overall geometry of the refined model was evaluated with MOLPROBITY ( Williams et al . , 2018 ) . The data statistics are summarized in Table 2 .	Some bacteria make toxins that punch large holes into the membranes of host cells , destroying them like a puncture destroys a football . These “pore-forming toxins” allow many bacterial species to infect a variety of organisms , from insects to humans . Some sophisticated pore-forming toxins , such as the anthrax toxin , do not only form a pore but also use it to flood lethal toxins into the cell to kill it . One bacterium called Xenorhabdus nematophila punctures the membranes of insect cells , using the same type of pore-forming toxins that other bacteria use to infect humans . Previous research has shown that two proteins – components A and B – form these pore-forming toxins . Given this two-protein formation , some scientists predicted these pore-forming toxins might act like those of the anthrax bacterium: one component forms the pore; the other component poisons the cell . But without detailed images of this pore-forming toxin’s structure , understanding exactly how these two components work together is almost impossible . To explore how components A and B operate within X . nematophila , Schubert et al . captured images of the molecular structure of the two proteins . Common methods reliant on X-rays and electron microscopes revealed the layouts of both components . By visualizing the proteins at different stages , Schubert et al . observed key structural changes that enable them to form the pore and puncture a host cell . Component A binds to component B’s back , forming a subunit – twelve to fifteen of which then conjoin as the pore-forming toxin . Schubert et al . conclude that component A stabilizes each subunit on the membrane and activates component B , which then punctures the membrane by swinging out its lower end . Unlike the anthrax pore-forming toxin , both components collaborate to form the pore complex and puncture the membrane . These results provide a foundation of knowledge about what these toxins look like and how they operate . More research building upon this structural analysis may help scientists develop antibiotics that prevent bacteria from destroying human cells .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "structural", "biology", "and", "molecular", "biophysics", "microbiology", "and", "infectious", "disease" ]	2018	Membrane insertion of α-xenorhabdolysin in near-atomic detail
Antiviral development is plagued by drug resistance and genetic barriers to resistance are needed . For HIV and hepatitis C virus ( HCV ) , combination therapy has proved life-saving . The targets of direct-acting antivirals for HCV infection are NS3/4A protease , NS5A phosphoprotein and NS5B polymerase . Differential visualization of drug-resistant and -susceptible RNA genomes within cells revealed that resistant variants of NS3/4A protease and NS5A phosphoprotein are cis-dominant , ensuring their direct selection from complex environments . Confocal microscopy revealed that RNA replication complexes are genome-specific , rationalizing the non-interaction of wild-type and variant products . No HCV antivirals yet display the dominance of drug susceptibility shown for capsid proteins of other viruses . However , effective inhibitors of HCV polymerase exact such high fitness costs for drug resistance that stable genome selection is not observed . Barriers to drug resistance vary with target biochemistry and detailed analysis of these barriers should lead to the use of fewer drugs . In a recent triumph of modern science and medicine , patients chronically infected with hepatitis C virus ( HCV ) now receive multidrug regimens that are often curative and have low toxicity ( Lawitz et al . , 2013; Afdhal et al . , 2014; Dhaliwal and Nampoothiri , 2014 ) . Over the past two decades , researchers have developed and tested thousands of antiviral compounds with varying efficacies and toxicity profiles that have ultimately lead to the FDA approval of powerful combination therapies ( Lawitz et al . , 2013; Scheel and Rice , 2013 ) . Several different direct-acting antivirals ( DAAs ) that target the NS3/4A protease , NS5A phosphoprotein , or NS5B RNA-dependent RNA polymerase of HCV have been approved for use in the clinic ( Afdhal et al . , 2014; Dhaliwal and Nampoothiri , 2014; Manns and von Hahn , 2013; Younossi et al . , 2015 ) . Ideally , the knowledge gained in developing HCV antivirals that are effective and not prone to the outgrowth of drug resistance will be applied to other viruses as well . The emergence of drug-resistant variants follows basic evolutionary principles , requiring spontaneous mutations as well as selective pressure , so that beneficial mutations increase the progeny size of genomes that bear them . The genetic diversity in RNA viral genomes results from the high error frequencies incurred by RNA-dependent RNA polymerases , which occur at approximately 4 × 10−5 errors for each nucleotide synthesized ( Sanjuán et al . , 2010 ) . Given the iterative copying of positive and negative strands , much higher cumulative error frequencies are observed , even during a single cycle of infection ( Sanjuán et al . , 2010; Acevedo et al . , 2014 ) . When more than one mutation is required to confer drug resistance , the outgrowth of drug resistance can be delayed ( Bloom et al . , 2010 ) . As a result , treatment with combinations of drugs can be extremely effective at suppressing drug resistance , because the number of mutations required for resistance to multiple drugs is ideally the sum of the number of mutations needed for each drug alone . Combination therapies have proven invaluable in reducing the frequency of drug resistance in both microbiology and oncology ( Fillat et al . , 2014; Falade-Nwulia et al . , 2017; Kerantzas and Jacobs , 2017 ) . Other strategies to suppress viral drug resistance accept the inevitability of drug-resistant mutations , but seek to decrease selection for their outgrowth . Examples of antivirals for which resistance comes with a high fitness cost include integrase inhibitors of HIV ( Mesplède et al . , 2015 ) , protease inhibitors of coronaviruses ( Deng et al . , 2014 ) and certain nucleoside inhibitors of HCV NS5B polymerase ( Lawitz et al . , 2013 ) . As was first shown for 2’-C-methyl CTP , selected drug-resistant HCV variants grow poorly and retain their low fitness upon passage ( Dutartre et al . , 2006 ) . Sofosbuvir , the FDA-approved NS5B polymerase inhibitor , has dramatically increased the efficacy of HCV treatment , and also generates little outgrowth of resistant variants . The few HCV variants observed in patients are nearly inviable ( Svarovskaia et al . , 2016 ) . Understanding the mechanisms by which this kind of fitness cost is enforced would greatly facilitate future antiviral design . Another approach to decrease the selection of drug-resistant variants is termed dominant drug targeting . This applies to antiviral targets for which the drug-bound products of pre-existing drug-susceptible genomes are dominant-negative inhibitors of new drug-resistant progeny ( Crowder and Kirkegaard , 2005; Tanner et al . , 2014; Mateo et al . , 2015 ) . Recently , this has been demonstrated for the capsid proteins of poliovirus and dengue virus ( Tanner et al . , 2014; Mateo et al . , 2015 ) , but other potential dominant drug targets have also been identified ( Crowder and Kirkegaard , 2004 ) . When a drug-resistant genome is in its cell of origin , it coexists with its drug-susceptible parents and siblings . If the drug target is , for example , a subunit of an oligomeric complex and subunits from different genomes have the opportunity to mix , chimeric oligomers often form . At the time of its creation , the drug-resistant genome will be a minority species , and such chimeras would be predominantly composed of the drug-bound , susceptible components thus incapacitating the entire oligomeric structure . Such ‘phenotypic masking’ was originally invoked to explain the very low frequency of foot-and-mouth-disease escape variants following selection with neutralizing antibodies when passaged at high multiplicities of infection ( MOIs ) ( Holland et al . , 1989 ) . Our goal was to screen the HCV-encoded viral proteins that are current targets of antiviral compounds to determine the intracellular dominance relationships between drug-resistant and drug-susceptible genomes . The high cost to viral fitness of Sofosbuvir-resistant variants is sufficient to explain its high barrier to resistance . There are currently no antivirals directed against HCV core protein; however , it is likely to be a dominant drug target . We used differential hybridization of RNA probes to detect two different genomic RNAs in a single cell by confocal microscopy and by flow cytometry . This analysis showed the cis-dominance of HCV viruses that are resistant to inhibitors of either NS3/4A protease or NS5A phosphoprotein , consistent with the rapid outgrowth of drug-resistance in patients of these two inhibitor classes . Newly mutated drug-resistant genomes first arise within cells that are pre-populated by drug-susceptible genomes . To mimic such mixed infections , we have previously employed co-infection of cultured cells with drug-susceptible and drug-resistant viruses at high MOIs to ensure mixed infection ( Tanner et al . , 2014; Mateo et al . , 2015 ) . For HCV , it is not practical to use high MOIs to achieve co-infection due to the difficulty of obtaining sufficiently high-titer viral stocks . Thus , we needed to develop an approach to distinguish between uninfected , singly infected and co-infected cells in relatively sparsely infected cell populations ( Figure 1A ) . To detect individual genomes in infected cells , a single-molecule fluorescence in situ hybridization ( FISH ) approach was used . A recently developed branched DNA probe technology allows the generation of sufficiently sensitive RNA probes to identify single molecules within cells , but requires approximately 1000 nucleotides of differential probe hybridization to achieve specificity ( Affymetrix Inc , 2016 ) . To create a viral strain with this extreme dissimilarity from wild-type virus , we tested the viability of three different codon-altered versions of the JFH1 variant of HCV ( Figure 1B ) . Each mutated version contained 200–300 nucleotide changes that did not alter the protein sequence ( Figure 1—figure supplements 1–3 ) . Of these codon-altered ( CA ) variants , CA-1 was inviable , CA-2 showed reduced viral protein accumulation , and CA-3 showed accumulation of both viral protein and RNA to abundances equivalent to those of the wild-type virus ( Figure 1C ) . Recently , detailed analysis of covarying nucleotides within the HCV coding region has identified the location of several previously unknown functional RNA secondary structures ( Pirakitikulr et al . , 2016 ) . Interestingly , CA-1 contains two such regions and CA-2 contains one , which correlates with decreasing viability , while CA-3 contains no such regions ( Figure 1—figure supplements 1–3 ) ( Pirakitikulr et al . , 2016 ) . Thus , subsequent experiments were performed only with CA-3 . This variant , now termed ‘CA’ virus , contains 247 synonymous mutations over a 918-nucleotide region that spans the coding sequences for most of NS2 and the N-terminus of NS3 ( Figure 1—figure supplement 3 ) . To test the sensitivity of RNA FISH probes generated against the positive- and negative-strands of wild-type ( WT ) and codon-altered ( CA ) viruses , both confocal microscopy and flow cytometry analyses were employed . Branched DNA technology allowed the labeling of each target RNA with as many as 8000 fluorophores ( Figure 2A ) ( Affymetrix Inc , 2016 ) . Huh-7 . 5 . 1 cells were infected with either WT or CA viruses , subjected to FISH and visualized by confocal microscopy . WT and CA probe sets specifically targeted either the positive-sense ( Figure 2B ) or the negative-sense vRNA ( Figure 2C ) of their corresponding virus . Additionally , we tested whether flow cytometry efficiently resolved cells transfected with different vRNAs; transfection was used to maximize the yield of each population . We resolved cells transfected with WT vRNA ( Figure 2Di ) , transfected with CA vRNA ( ii ) , a mixture of these two cell types ( iii ) and cells co-transfected with both WT and CA vRNAs ( iv ) . Thus , specific RNA probes could be used to monitor the fate of drug-susceptible and drug-resistant viruses in co-infected cells . To test the genetic properties of viruses that are resistant to NS3/4A inhibitors , we employed the original NS3/4A inhibitor , BILN-2061 ( Figure 3A ) ( Lamarre et al . , 2003 ) . Like other NS3/4A inhibitors , BILN-2061 treatment rapidly allows the selection of drug resistant variants both in tissue culture and in patients ( Lamarre et al . , 2003; Lin et al . , 2004 ) . Given the ease of outgrowth of drug-resistant variants , we hypothesized that NS3/4A was not a dominant drug target and that drug resistance would be genetically dominant . NS3-D168A is the prototypic mutation associated with resistance to NS3/4A inhibitors . Asp168 is in close proximity to the protease active site ( Figure 3B ) . The ability of the NS3-D168A mutation to confer resistance to BILN-2061 was confirmed in both the WT and CA backgrounds ( Figure 3—figure supplement 1 ) . As shown schematically in Figure 3D , the ability to track cells that are uninfected ( U ) , singly infected with drug-susceptible virus ( S ) , infected with both susceptible and resistant virus ( S + R ) and singly infected with drug-resistant virus ( R ) , can reveal dominance relationships during co-infection . In the absence of a drug , all viral populations should be present . However , in the presence of a drug , three outcomes are possible depending on the genetic outcome within the R + S population . If drug resistance were trans-dominant ( Figure 3E ) , the drug-resistant virus would rescue the drug-susceptible genomes and all viruses in R + S cells would survive in the presence of the drug . S cells would drop into the U population , and R cells would survive . If drug resistance were cis-dominant ( Figure 3F ) , only the R viruses in the R + S cells would survive , because the drug-resistant proteins would be unable to rescue the S viruses in the same cell . Consequently , the R + S cells would drop into the R population . If drug susceptibility were dominant ( Figure 3G ) , all viruses in the R + S cells would be cleared , and the R + S cells , like the S cells , would drop into the U population , and only the R cells would continue to replicate . To determine the dominance relationship between BILN-2061-susceptible and the BILN-2061-resistant virus , Huh-7 . 5 . 1 cells were infected with CA and WT-D168A viruses ( Figure 3C ) . Cells were infected for 72 hr at MOIs such that all four populations were represented , followed by 36 hr of continued incubation in the absence or presence of 2 μM BILN-2061 . Cells were then harvested , fixed , co-stained with wild-type and codon-altered RNA probe sets and analyzed by flow cytometry . All four cell types appeared in the absence of BILN-2061 ( Figure 3H , I ) . In the BILN-2061-treated samples ( Figure 3J , K ) , the susceptible S population shifted to the U cells as expected . The S + R cells , on the other hand , shifted to the R population upon drug treatment . Thus , the drug-resistant viral genomes in the co-infected cells could replicate , but could not rescue the drug-susceptible ones . Data from this and replicate experiments ( Figure 3I , K ) confirmed the quantitative shift of S + R cells into the R population upon drug treatment . We conclude that , for the NS3/4A target , drug-resistant genomes are cis-dominant for the 1:1 ratio of S and R viruses tested here . We also tested whether over-expressed drug-resistant NS3/4A precursors could rescue BILN-2061-susceptible virus ( Figure 3—figure supplement 2B ) . Salvage of S virus was not observed . Importantly , when cis-acting proteins are drug targets , drug-resistant products will enhance the propagation of only those genomes that encode them , allowing powerful selection for drug resistance . For NS5B polymerase inhibitor Sofosbuvir , the few resistant viral variants that arise in patients are highly attenuated . To investigate whether a related compound , R1479 ( Klumpp et al . , 2006 ) , exacted a similar cost to viral fitness to drug-resistant variants , we attempted to recover R1479-resistant viruses for dominance testing . JFH1 was passaged for multiple rounds of infection in the presence of 25 μM R1479 . Several variants in NS5B ( A336P , D438G , S282T , F427L , T481A ) arose during passage ( Figure 3—figure supplement 3 ) . Each mutation was introduced independently into the JFH1 genome and RNA transfections were performed . The T481A genome was the only variant to show any viral RNA production by 7 or 21 days post-transfection . We noticed that F427L and T481A were always isolated together . To test whether these mutations could together increase viral fitness , JFH1 viruses were generated that contained both mutations . Viruses with the mutations separate or together were passaged extensively in the presence of R1479 . Occasional resistant outgrowths were observed , but none conferred sustained growth ( Figure 3—figure supplement 3C ) . Thus , like Sofosbuvir , the poor viability of mutant viruses resistant to R1479 precludes the ability to perform further genetic analysis but provides an excellent paradigm for antiviral development . NS5A is highly oligomeric ( Sun et al . , 2015; Tellinghuisen et al . , 2005 ) and we were curious as to whether drug resistance or drug susceptibility would be dominant during viral infections . This idea seemed promising because exogenously expressed NS5A has a dominant-negative effect on the growth of HCV replicons ( Graziani and Paonessa , 2004 ) . Additionally , the NS5A inhibitors , as a class , display EC50’s in the low picomolar range ( Gao , 2013 ) , making them among the most potent antiviral compounds ever identified . Assuming uniform inhibitor concentrations in cells and in medium , it has been estimated that only a small fraction of NS5A molecules should be bound to drugs under inhibitory conditions ( Sun et al . , 2015; Gao et al . , 2016 ) . Thus , it seemed mechanistically likely that drug-bound NS5A proteins from drug-susceptible viruses could be dominant inhibitors of NS5A encoded by newly arising drug-resistant ones . However , NS5A inhibitors have generally demonstrated low barriers to resistance in patients . Our goal was gain mechanistic insight into this dichotomy . The structures of two such potent NS5A inhibitors , SR2486 ( also known as BMS-346 ) ( Lemm et al . , 2011 ) and Daclatasvir ( Gao et al . , 2010 ) are shown in Figure 4A . Mutations of Tyr93 to Asp or His confer resistance to a broad array of NS5A inhibitors ( Gao et al . , 2016 ) . Tyr93 is located near an NS5A dimer interface shown in the crystal structure ( Figure 4B ) ( Tellinghuisen et al . , 2005 ) . Thus , this interface is postulated to be part of the binding site for the NS5A inhibitor class . The Y93N and Y93H mutations were introduced into both the wild-type and codon-altered viruses . As shown in Figure 4C , the Y93H mutation conferred resistance to both SR2486 and Daclatasvir while the Y93N mutation conferred resistance only to SR2486 . To test whether susceptibility to NS5A inhibitors was dominant in the context of viral infections , we analyzed U , S , S + R and R cell populations by flow cytometry as previously performed for the NS3/4A inhibitor in Figure 3 . Huh-7 . 5 . 1 cells were coinfected with CA and WT-Y93N viruses for 72 hr ( Figure 4D ) . Cells were then treated with DMSO or 500nM SR2486 for 24 hr , harvested , fixed , co-stained for WT and CA vRNAs and analyzed by flow cytometry . In the absence of the NS5A inhibitor , all four populations , U , S , R + S and R were observed ( Figure 4E , F ) . In the presence of SR2486 , the S population of cells dropped into the U population as expected . As was the case in Figure 3 , the co-infected R + S population of cells dropped into the R population . Thus , resistance to NS5A inhibitor SR2486 in the context of viral infection was genetically dominant and the lack of rescue of the S virus with which it was mixedly infected shows that drug resistance is also cis-dominant . HCV infected cells become resistant to superinfection upon expression of non-structural proteins ( Schaller et al . , 2007; Tscherne et al . , 2007 ) . Due to this superinfection exclusion , it is likely that all coinfected cells arise through nearly synchronous infection throughout the course of the experiment . To control for any effects on selection that may occur due to the differential timing of coinfections that occurs over the initial 72 hr incubation period , we performed the same experiment with higher titer virus and a single cycle of infection in the absence of drug . Huh7 . 5 . 1 cells were infected at an MOI of 1 focus-forming unit ( FFU ) /cell with CA and WT-Y93N viruses and incubated for only 24 hr before drug treatment . Cells were then incubated in the absence and presence of 500nM SR2486 for an additional 24 hr . In this case , we also observe cis-dominance of drug resistant WT-Y93N genomes , indicating that asynchronous coinfection has no effect on selection ( Figure 4G ) . Finally , the cis-dominance of Daclatasvir-resistant WT-Y93H was observed when coinfected with drug susceptible virus ( S ) in the absence and presence of Daclatasvir ( Figure 4H ) . We conclude that NS5A , despite being an oligomeric species is not a dominant drug target . Instead , genomes resistant to NS5A remain drug resistant in co-infected cells but do not rescue drug-susceptible viruses present in the same cell . This is consistent with the observed outgrowth of viruses that are resistant to NS5A both in cultured cells and in patients , and with an earlier report that at least some functions of NS5A act exclusively in cis ( Fridell et al . , 2011 ) . One hypothesis that could mechanistically account for cis-dominant drug resistance is that NS5A molecules expressed from different alleles may not freely associate in mixed oligomers . As previously demonstrated , two different NS5As expressed from the same RNA can associate , while NS5A molecules expressed from different constructs could not ( Berger et al . , 2014 ) . We were curious whether the dominance phenotypes were altered if we forced NS5A alleles to mix . To test whether exogenously expressed drug-susceptible NS5A proteins could co-assemble with drug-resistant NS5A , we utilized the previously described HCV plasmid that expresses HA-tagged and GFP-tagged NS5A within the same polyprotein but does not support genome replication ( Figure 5A ) . Constructs that contained all combinations of drug-susceptible NS5A ( S ) and the drug-resistant Y93N variant ( R ) were created . Upon transfection , all tagged proteins were expressed and can be observed in Figure 5 ( Input , Panels B , C ) . Immunoprecipitation with anti-HA antibodies revealed that the GFP-tagged and HA-tagged NS5A proteins were present in the same complexes in the presence or absence of SR2486 . Therefore , as has been shown previously , mixed oligomers can form upon co-expression within the same polyprotein ( Berger et al . , 2014 ) . Furthermore , these interactions are not disrupted by drug treatment or by drug-resistant mutations ( Figure 5B , C ) . To determine whether there were any functional consequences of mixed oligomer formation , we visualized cells that expressed mixed oligomers using confocal microscopy . All S and R combinations of NS5A co-localized at discrete membrane-associated complexes characteristic of HCV infection in the absence of drug ( Figure 5D , top panel ) . However , in the presence of SR2486 , membrane-associated complex formation was inhibited in R:S and S:S expressing cells and observed only in R:R expressing cells ( Figure 5D , bottom panel ) . The dispersal of NS5A signal upon drug treatment in the presence of S protein makes NS5A protein appear less abundant ( Figure 5D ) . However , the immunoblots demonstrate that no such decrease in expression occurs as we observe equal levels of NS5A protein independent of allele or the presence of drug ( Figure 5B , C ) . One hallmark of HCV infection is the accumulation of cytoplasmic lipid droplets ( Miyanari et al . , 2007; Romero-Brey et al . , 2012 ) . Electron microscopy performed by high-pressure freezing and freeze-substitution , to preserve membrane structure , revealed many lipid droplets in the cytoplasm of cells expressing S:S , R:R and S:R combinations of NS5As in the absence of inhibitor ( Figure 5E , F ) . However , in the presence of SR2486 , only the R:R cells displayed the accumulation of lipid droplets ( Figure 5E , G ) . Therefore , using both assays , the presence of drug-susceptible NS5A prevented drug-resistant phenotypes from being displayed , and thus drug-susceptibility was genetically dominant . This confirmed our original hypothesis that NS5A had the potential to be a dominant drug target . One of the most likely mechanisms for cis-dominance , when the benefit of a particular gene product accrues only to the genome that encodes it , is physical isolation . We hypothesized that HCV genomes co-infecting the same cell might be physically isolated from each other . To test this possibility , confocal microscopy was used to identify and localize negative vRNA strands in genome-specific RNA replication complexes ( Figure 6A ) . The majority of negative strands of the two different viruses were found to be discrete . Identification and quantification the vRNA puncta in coinfected cells was determined computationally using Volocity software . This program determined the number of negative strand puncta per cell per strain and quantified how many puncta overlapped ( Figure 6B ) . This value was low even for the positive stranded vRNAs , which are present in the cytoplasm at much higher frequencies ( Figure 6A , B ) . As a positive control for colocalization , we performed a similar experiment but additionally stained for NS5A or Core in addition to minus strand vRNAs . We would expect minus strand vRNA and NS5A to colocalize strongly , as NS5A is present inside replication complexes . Alternatively , Core is not localized directly within replication complexes , but is present within packaging complexes and on lipid droplets , which are nearby . Volocity was used to count negative strand vRNAs , and then asked , how many of those puncta were touching NS5A or Core . Representative images demonstrating each of the pairwise comparisons demonstrate that nearly 80% of all minus strand vRNAs were touching NS5A while fewer than half of the minus strand vRNAs were touching Core ( Figure 6C , D ) . These data support the hypothesis that , upon co-infection , drug-resistant and drug-susceptible RNA genomes create independent membranous web structures , limiting the mixing of NS3/4A and NS5A proteins and their precursors . This scenario is modeled schematically in Figure 6E . Failure of NS5A proteins to mix during infection is a likely explanation for the cis-dominance of drug resistance observed in cultured cells ( Figure 4 ) . These circumstances account for the ready outgrowth of drug resistance in patients ( Gao et al . , 2016 ) , even though NS5A is highly oligomeric . Even when drug-resistant NS5A was overexpressed in a precursor form , no rescue of drug-susceptible virus was detected ( Figure 3—figure supplement 2C ) . It has previously been shown that when HA- and GFP-tagged NS5A molecules were expressed on different constructs such as those depicted in Figure 5A , no mixed complexes were formed ( Berger et al . , 2014 ) . Thus , even though high-order NS5A oligomers are formed in infected cells , it is unlikely that these are mixed-allele oligomers , preventing dominant inhibition of drug-resistant HCV . Due to the highly mutagenic nature of RNA viruses and the large number of genomes and anti-genomes generated during infection , a high barrier to drug resistance is extremely difficult to achieve . This has led to abandoned usage and development of many otherwise promising antivirals . To decrease the frequency with which drug-resistant variants arise , combinations of antivirals that , individually , exhibit low barriers to resistance are often used . When drug-resistant variants are first formed intracellularly , through error-prone RNA replication , they arise in a population that includes parental and sibling drug-susceptible viruses . Several genetic relationships between drug-resistant and drug-susceptible genomes are possible . First , the drug resistance of the new variants has the potential to be genetically dominant , and rescue both resistant and susceptible viral genomes . Alternatively , drug resistance can be cis-dominant , with the drug-resistant products rescuing only the genomes that encode them . Finally , the drug-resistant genome can fail to benefit any genomes in the cell because the drug-susceptible products present in the same cell are dominant inhibitors . The DAAs targeting NS3/4A protease of HCV were the first to be discovered ( Lamarre et al . , 2003 ) and the first to reach the clinic ( Bacon et al . , 2011; Jacobson et al . , 2011 ) . It was soon realized that , both during the growth of HCV replicons in cultured cells and in phase II clinical trials , drug-resistant viruses were generated rapidly ( Lin et al . , 2004 ) . Nonetheless , in 2011 , further advances led to FDA approval of Telaprevir and Boceprevir ( Jacobson et al . , 2011; Poordad et al . , 2011 ) . The anticipation , which proved to be correct , was that inhibitors of NS3/4A would prove useful in combination therapies ( Feld et al . , 2014; Lawitz et al . , 2014 ) . We have used flow cytometry to identify cell populations that are co-infected with HCV that is susceptible or resistant at a 1:1 ratio , to protease inhibitors . Within these cells , the drug-resistant genomes replicated , but the drug-susceptible genomes did not . We therefore conclude that NS3/4A inhibitor resistance is cis-dominant ( Figure 3 ) , which should allow the rapid and specific selection for outgrowth from its cell of origin . Cis-dominance of drug resistance was also not originally anticipated while targeting NS3/4A . The original characterizations of the NS3/4A protease suggested that cleavage of the NS3/4A junction occurred in cis , but that cleavages at the 4A/4B , 4B/5A and 5A/5B junctions could all occur in trans ( Bartenschlager et al . , 1994 ) . We felt that it was therefore , more likely , that drug-resistant NS3/4A could rescue drug-susceptible virus within the same cell . NS3/4A is not known to assemble into high-order oligomers in the same manner as NS5A , and we therefore did not anticipate drug-susceptible NS3/4A would be trans-dominant . Furthermore , a trans cleavage assay demonstrated that a NS4B-5B polyprotein could be cleaved by NS3/4A supplied in trans ( Romero-Brey et al . , 2015 ) . However , the trans-cleavage system does not result in membranous web formation that would accompany genome sequestration . Other groups have reported a different result , that defective NS3 mutants cannot be rescued in trans by replicons with functional NS3/4A ( Kazakov et al . , 2015; Appel et al . , 2005 ) . Our interpretation of these studies is that NS3/4A is likely physically able to cleave in trans in cells , but requires access to the alternate precursor proteins in order for this to occur . Therefore , cis-dominance of drug-resistance is likely the result of a lack of free-mixing of NS3/4A encoded by different vRNAs within the same cell . NS5A emerged as an HCV drug target through a chemical genetics screen for compounds that inhibited HCV growth but did not target the NS3/4A protease or the NS5B polymerase ( Gao et al . , 2010 ) . The ease with which resistant viruses were selected suggested that drug resistance was either dominant or cis-dominant . This was somewhat surprising , given that NS5A is oligomeric and the NS5A inhibitors are extremely potent , and have been postulated to function at sub-stoichiometric ratios to NS5A protein ( Gao , 2013 ) . Indeed , when drug-susceptible and drug-resistant NS5A protein were co-expressed in the present study , hetero-oligomers formed and the biological phenotypes of the drug-susceptible protein were dominant ( Figure 5 ) . Nonetheless , single-cell analysis of cells co-infected at a 1:1 ratio with NS5A inhibitor-susceptible and -resistant viruses showed , as with the NS3/4A inhibitor , that resistance to both SR2486 and Daclatasvir was cis-dominant ( Figure 4 ) . What does cis-dominant resistance mean mechanistically ? One potential mechanism is physical sequestration of the RNA replication complexes of the two co-infecting genomes . Genome-specific RNA probing of co-infected cells revealed that both the negative strands and positive strands from the two viruses were present at physically distinct locations ( Figure 6 ) . It is therefore highly probable that membrane-associated proteins such as HCV NS3/4A and NS5A do not mix within individual RNA replication complexes . However , not all mutations in a particular viral product should lead to the same defect with the same genetic properties . For example , we show here that viruses that are defective in a function of NS5A in RNA replication complexes are not rescued and have no effect on the outgrowth of drug-resistant variants . However , NS5A also plays an important role in packaging and assembly of mature HCV particles on lipid droplets ( Miyanari et al . , 2007; Boson et al . , 2017 ) . Lipid droplets are large and form adjacent to RNA replication complexes . In Figure 6E , we have depicted the possibility that NS5A molecules encoded by distinct RNA replication complexes might mix on the surface of these lipid droplets . However , if replication of the drug-susceptible genomes is inhibited , contribution of their encoded proteins to any oligomers on the surface of lipid droplets should be minimal . In this vain , a hypothetical NS5A inhibitor that allowed RNA replication but inhibited the function of NS5A in particle assembly might have different genetic properties than the NS5A inhibitors currently in use . Viral capsids have especially interesting genetic properties , often intermixing within co-infected cells . Defective capsid proteins of poliovirus , HBV and HIV have been shown to be dominant inhibitors of wild-type viruses ( Crowder and Kirkegaard , 2005; Tanner et al . , 2016; Tan et al . , 2013; Tan et al . , 2015; Trono et al . , 1989; Pettit et al . , 2005; Lee et al . , 2009; Müller et al . , 2009; Checkley et al . , 2010 ) . Thus , when antiviral targets are capsid proteins , drug susceptibility can be genetically dominant by suppressing the outgrowth of drug-resistant virus within the cell in which it is first generated ( Crowder and Kirkegaard , 2005; Tanner et al . , 2014; Kirkegaard et al . , 2016 ) . For HCV , very few inhibitors of capsid function have been identified , and their inhibition of viral growth is not sufficiently robust to make genetic experiments possible ( Kota et al . , 2010 ) . It is therefore not yet possible to test if , as we hypothesize , drug-susceptible virus will prove to be a dominant inhibitor of drug resistance . Consistent with this hypothesis , however , epitope-tagged HCV core protein can form mixed disulfide-bonded core oligomers ( Kushima et al . , 2010 ) . The success of combination therapy for HCV and the efficacy of the individual constituents illustrate some of the weapons in the arsenal of antiviral strategies . Future directions are likely to include , as well , the rational design of antivirals with high barriers to resistance such as those that hyper-stabilize oligomers and the prediction of DAA targets that impart a high fitness cost to drug resistance . Huh7 . 5 . 1 cells were a gift from Dr . Michael Gale Jr ( University of Washington ) and were cultured in DMEM ( Sigma , St . Louis , MO ) supplemented with 10% fetal bovine serum ( Omega , Tarzana , CA ) , penicillin/streptomycin ( Invitrogen , Grand Island NY ) , non-essential amino acids ( Invitrogen ) , and Glutamax ( Invitrogen ) . Huh7-Lunet-T7 cells were a gift from Dr . Ralf Bartenschlager ( University of Heidelberg ) and were cultured in DMEM supplemented with 10% fetal bovine serum , penicillin/streptomycin , non-essential amino acids , Glutamax and 5 μg/mL Zeocin ( Invitrogen ) . Cell line identification was performed using STR profiling services available through the Stanford Functional Genomics Facility . Alignments were generated using Huh7 as a reference . Cell lines were screened for mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit ( Lonza ) . The plasmid pJFH1 was a gift from Dr . Michael Gale Jr ( Kato et al . , 2006 ) . This plasmid contains a synthesized genome length copy of the JFH1 strain of HCV ( genotype 2a ) . To produce cell-culture-derived HCV particles ( HCVcc ) , pJFH1 was digested with XbaI ( New England Biolabs ) . The linearized plasmid was then used as a template for in vitro transcription with the MEGAscript high yield transcription kit ( Ambion ) . vRNA was purified using Trizol ( Invitrogen ) and electroporated into Huh7 . 5 . 1 cells as previously described to generate HCVcc cultures ( Wakita et al . , 2005 ) . Following a period of amplification , HCVcc cultures were converted to human serum media as described previously ( Steenbergen et al . , 2013 ) . Human serum media comprised DMEM supplemented with 2% heat inactivated human serum ( Omega ) , penicillin/streptomycin , non-essential amino acids and Glutamax . Antibodies recognizing HCV core ( Abcam ) , GAPDH ( Santa Cruz Biotechnologies ) , GFP ( Life Technologies ) and HA ( Genscript ) were purchased from the individual suppliers . Antibodies recognizing NS5A were described previously ( Lindenbach et al . , 2005 ) . To construct codon-altered strains of HCV , we subjected three approximately 1000-nucleotide fragments of the JFH1 genome through the GeneArt codon optimization algorithm offered by Life Technologies . The genome fragments were composed of nucleotides 2613–3530 ( CA-3 ) , 7441–8456 ( CA-2 ) , and 7867–8896 ( CA-1 ) . All three codon-altered genome fragments were synthesized by Life Technologies and cloned into the pJFH1 plasmid by restriction digestion and ligation with T4 DNA ligase ( Invitrogen ) . The resulting plasmids: pJFH1-CA-1 , pJFH1-CA-2 and pJFH1-CA-3 , were used to produce HCVcc cultures as described above . To create drug-resistant HCVcc cultures , two subcloning plasmids were created by PCR by amplifying nucleotides 6395–8670 or 4584–6498 of the pJFH1 plasmid with Taq polymerase ( New England Biolabs ) and ligating the PCR products into pCR2 . 1 ( Invitrogen ) . The resulting plasmids , pCR2 . 1-6395-8670 and pCR2 . 1-4584-6498 were used as templates for site-directed mutagenesis using the QuikChange Site-Directed Mutagenesis kit ( Agilent Technologies ) . pCR2 . 1-6395-8670-Y93N was generated using the forward primer 5’-CCTATCAATTGCAATACGGAGGGCCAGTGCGCGCC-3’ and the reverse primer 5’-GGCGCGCACTGGCCCTCCGTATTGCAATTGATAGG-3’ . pCR2 . 1-6395-8670-Y93H was generated using the forward primer 5’-CCTATCAATTGCCATACGGAGGGCCAGTGCGCGCC-3’ and the reverse primer 5’-GGCGCGCACTGGCCCTCCGTATGGCAATTGATAGG-3’ . pCR2 . 1-4584-6498-D168A was generated using the forward primer 5’-AAATCCATCGCCTTCATCCCC-3’ and the reverse primer 5’-GGGGATGAAGGCGATGGATTTGGC-3’ . These mutated HCV genome fragments were cloned into pJFH1 or pJFH1-CA using restriction digestion and ligation with T4 DNA ligase ( Invitrogen ) . HCVcc cultures were generated as described above . The plasmids pTM_NS3-5B_NS5A-HA_2a_NS5A-gfp_JFH1 ( referred to as pTM-Dual-NS5A ) and pTM_NS3-5B_NS5A-GFP ( referred to as pTM-NS3-5B ) were the generous gifts of Dr . Ralf Bartenschlager ( University of Heidelberg ) . The D168A and Y93N mutations were cloned into the pTM-NS3-5B plasmid using the Quikchange Lightening Mutagenesis kit using the primers described above . The NS5A alleles of the pTM-Dual-5A plasmid were first separated by removing an RsrII fragment containing most of the NS5A-GFP allele to create pTM-Dual-5A-ΔRsrII and pcDNA5-NS5A-GFP-RsrII . Site-directed mutagenesis was performed using the Quikchange Lightening kit on pTM-Dual-5A-ΔRsrII or on pcDNA5-NS5A-GFP-RsrII independently . The RsrII fragments containing wild-type NS5A or NS5A-Y93N were then cloned back into the pTM-Dual-5A-ΔRsrII vectors to create all combinations of wild-type NS5A and NS5A-Y93N pTM-Dual-5A . vRNA was harvested from cells using Trizol ( Invitrogen ) or collected from HCVcc culture supernatants using the QIAamp vRNA mini kit ( Qiagen ) . A standard curve was generated using in vitro transcribed HCV vRNA . qRT-PCR was performed using the QuantiTect Sybr-Green RT-PCR kit ( Qiagen ) and the qRT-PCR forward 5’-CTGGCGACTGGATGCGTTTC-3’ and reverse 5’-CGCATTCCTCCATCTCATCA-3’ primers . Alternatively , the following CA-specific primers were used: forward 5’-GTGGTGTCCATGACCGGCA-3’ and reverse 5’-GGTCACGGGGCCTCTCAGT-3’ , or the following WT-specific primers were used: forward 5’-GTGGTGAGTATGACGGGGC-3’ and reverse 5’-CGTGACCGGACCCCGTAAG-3’ . Samples were analyzed on a 7300 Real-Time PCR Machine ( Applied Biosystems ) . WT vRNA target probes recognizing the NS2 region of either the positive or negative strand were designed and synthesized by Affymetrix . These probes were specifically designed to avoid detection of codon-altered JFH1 viral RNA . Additionally , probes were designed to recognize the corresponding region of the negative or positive strand JFH1-CA vRNA . These CA target probes were specifically designed not to recognize the WT vRNA . Huh7 . 5 . 1 cells were infected with WT or CA HCVcc particles for 72 hr . Infected cells were fixed with 4% formaldehyde solution ( Sigma ) and subjected to RNA in situ hybridization ( ISH ) using the ViewRNA Cell Assay kit ( Affymetrix ) according to the manufacturer’s protocol . Cells were co-stained with both CA and WT vRNA target probe sets in all experiments . Cells were visualized on a Leica SP8 confocal microscope . Protein and vRNA colocalization was performed on cells coinfected with JFH1-CA and JFH1-Y93N for 24 hr . Following infection , cells were fixed and stained using the ViewRNA Cell Plus assay reagents . Core and NS5A were visualized using the antibodies described above at a 1 to 100 dilution followed by the anti-mouse-AlexaFlour-647 secondary antibody at 1 to 200 dilution . Quantification of colocalization was performed using Volocity software ( Perkin Elmer ) . Briefly , we defined vRNA puncta as objects larger than 0 . 1 μm2 . Objects larger than 0 . 25 μm2 were broken into subunits based on total volume . Objects sharing 0 . 05 μm2 of mutual space were quantified as mutual . Due to the localization patterns of core and NS5A , spot counting algorithms were not appropriate . Total vRNA objects and as well as the total number of vRNA objects touching NS5A or Core were quantified . Huh7-Lunet-T7 cells were transfected with pTM-Dual-NS5A constructs using branched polyethylenimine ( Sigma-Aldrich ) at a ratio of 1:3 . At 4 hr post-transfection , cells were treated with 500nM SR2486 or a DMSO control . At 24 hr post-transfection , cells were fixed with 4% paraformaldehyde , stained with anti-HA antibodies and DAPI and visualized on a Leica SP8 confocal microscope . A more detailed description of the RNA FISH microscopy methods are provided in Bio-Protocol ( van Buuren and Kirkegaard , 2018 ) . Huh7-Lunet-T7 cells were transfected with pTM-Dual-NS5A constructs using the polyethylenimine transfection reagent . At 4 hr post-transfection , cells were treated with DMSO or 500nM SR2486 . At 24 hr post-transfection , cells were harvested using an enzyme-free cell dissociation buffer ( Life Technologies ) and FACS sorted for GFP-positivity on a FACS Aria cell sorter . GFP-positive cells were re-suspended in 20% BSA in PBS then placed into a 200 μM deep hat and high-pressure frozen using a Leica EMpact2 . Frozen samples were then freeze substituted in 1% Osmium tetroxide and 0 . 1% uranyl acetate in acetone using a Leica EMAFS at −90°C for 72 hr , warmed to −25°C in 16 . 3 hr at 4°C/hr and held for 12 hr then warmed to 0°C in 5 hr at 5°C/hr and held for 12 hr . The samples were then washed two times in acetone , then in propylene oxide for 15 min each . Samples are infiltrated with EMbed-812 resin ( EMS Cat#14120 ) mixed 1:2 , 1:1 , and 2:1 with propylene oxide for 2 hr each , leaving samples in 2:1 resin to propylene oxide overnight rotating at room temperature . The samples are then placed into EMbed-812 for three hours then placed into TAAB capsules with fresh resin and placed into a 65°C oven overnight . Sections were taken between 75 and 90 nm , picked up on formvar/carbon coated 100 mesh copper grids , then contrast stained for 30 s in 3 . 5% uranyl acetate in 50% acetone followed by staining in 0 . 2% lead citrate for 3 min . Cells were visualized using the JEOL JEM-1400 120kV microscope and photos were taken using a Gatan Orius 4k × 4k digital camera . Huh7 . 5 . 1 cells were either transfected with WT and/or CA vRNA as previously described or infected with WT and/or CA HCVcc particles . Coinfections were performed by infecting with each virus for 24 or 72 hr followed by treatment with either 2 μM BILN-2061 for 36 hr or 500nM SR2486 for 24 hr . Cells were harvested with trypsin and fixed with the FlowRNA Fixation and Permeablization kit . Cells were then costained with CA and WT vRNA target probe sets using the FlowRNA kit ( Affymetrix ) and analyzed on the Scanford FACScan Flow Cytometer . Data was analyzed and processed using Flowjo software .	Viruses are simple organisms that consist of genetic information and a few types of proteins . They cannot replicate on their own , and instead hijack the molecular machinery of a host cell to produce more of themselves . Inside an infected cell , the genetic information of the virus is replicated and ‘read’ to create viral proteins . These components are then assembled to form a new generation of viruses . During this process , genetic errors may occur that lead to modifications in the viral proteins , and help the virus become resistant to treatment . For instance , a viral protein that used to be targeted by a drug can change slightly and not be recognized anymore . Currently , the most efficient way to fight drug resistance is to use combination therapy , where several drugs are given at the same time . This strategy is successful , for example to treat infections with the hepatitis C virus , but it is also expensive , especially for developing countries . An alternative approach is dominant-drug targeting , which exploits the fact that both drug-resistant and drug-susceptible viruses are ‘born’ in the same cell . There , the susceptible viruses can overwhelm and ‘mask’ the benefits of the resistant ones . For example , proteins from resistant strains , which are no longer detected by a treatment , can bind to proteins from susceptible viruses; drugs will still be able to recognize these resulting viral structures . The proteins that operate in such ways are potential dominant-drug targets . However , resistant and susceptible strains can also cohabit without any contacts if their proteins do not interact with each other . Now , van Buuren et al . screen several viral proteins , including one called NS5A , to test whether a dominant drug target exists for the hepatitis C virus . Only a few molecules of a drug that targets NS5A can stop the virus from growing . In theory , drug-bound NS5A proteins could block their non-drug-bound neighbors , but when these drugs have been used on their own , resistance quickly emerged . Experiments showed that NS5A is not a dominant drug target because the drug-resistant and drug-susceptible proteins do not mix . Unless ‘forced’ in the laboratory , NS5A proteins only bind to the ones produced by the same strain of virus . This explains why resistant viruses quickly take over when NS5A drugs are the sole treatment . However , other hepatitis C proteins , such as the HCV core protein , are known to mix during the assembly of the virus , and thus are likely be dominant drug targets .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "microbiology", "and", "infectious", "disease", "genetics", "and", "genomics" ]	2018	Transmission genetics of drug-resistant hepatitis C virus
Utilizing molecular data to derive functional physiological models tailored for specific cancer cells can facilitate the use of individually tailored therapies . To this end we present an approach termed PRIME for generating cell-specific genome-scale metabolic models ( GSMMs ) based on molecular and phenotypic data . We build >280 models of normal and cancer cell-lines that successfully predict metabolic phenotypes in an individual manner . We utilize this set of cell-specific models to predict drug targets that selectively inhibit cancerous but not normal cell proliferation . The top predicted target , MLYCD , is experimentally validated and the metabolic effects of MLYCD depletion investigated . Furthermore , we tested cell-specific predicted responses to the inhibition of metabolic enzymes , and successfully inferred the prognosis of cancer patients based on their PRIME-derived individual GSMMs . These results lay a computational basis and a counterpart experimental proof of concept for future personalized metabolic modeling applications , enhancing the search for novel selective anticancer therapies . Personalized medicine is moving us closer to a more precise , predictable and powerful method of treatment , customized for the individual patient . One field of research in which personalized medicine holds great promise is cancer therapy . The use of molecular data to personalize cancer treatment and differentiate one type of cancer from another can facilitate the use of highly tailored therapies and offers tremendous potential for improved prognoses ( Simon and Roychowdhury , 2013 ) . A fundamental stepping-stone towards this goal is the ability to derive large-scale functional physiological models of specific cells that capture their unique cellular behavior . These models can then be utilized to identify drug targets that differentiate one cancer type from the other , and most importantly , distinguish them from their normal counterparts thus achieving treatment response selectivity . This study addresses these challenges within the growing paradigm of Genome-Scale Metabolic Modeling , a computational framework for studying metabolism on a genome-scale that has been successfully used for a variety of applications ( Burgard et al . , 2003; Oberhardt et al . , 2009; Chandrasekaran and Price , 2010; Jensen and Papin , 2010; Lewis et al . , 2010; Szappanos et al . , 2011; Wessely et al . , 2011; Agren et al . , 2012; Lee et al . , 2012; Lerman et al . , 2012; Pey et al . , 2012; Schuetz et al . , 2012; Oberhardt et al . , 2013 ) . In recent years , two Genome-Scale Metabolic Models ( GSMMs ) of human metabolism were published ( Duarte et al . , 2007; Ma et al . , 2007 ) , and their utility in predicting human metabolic phenotypes has been demonstrated in a wide range of studies ( Shlomi et al . , 2008; Lewis et al . , 2010; Folger et al . , 2011; Frezza et al . , 2011; Agren et al . , 2012; Yizhak et al . , 2013 ) . Recently , more comprehensive versions of the generic human model were published ( Thiele et al . , 2013; Mardinoglu et al . , 2014 ) . While these generic models are not specific to any cell- or tissue-type , they have successfully served both as a basis for generating context-specific models of tissues ( Shlomi et al . , 2008; Jerby et al . , 2010; Agren et al . , 2012 ) and for studying cancer metabolism ( Folger et al . , 2011; Frezza et al . , 2011; Shlomi et al . , 2011; Agren et al . , 2012; Facchetti et al . , 2012; Wang et al . , 2012; Dolfi et al . , 2013; Agren et al . , 2014; Yizhak et al . , 2014 ) . Importantly , methods for building context-specific models do not take into account subtle differences in levels of expression of a particular enzyme , but rather its presence or absence . This coarse discretization makes these methods less applicable for the task of building cell-specific models , in cases where a high similarity in transcriptomics levels of different samples is observed . Namely , when the inter-individual variations in the molecular signatures of different cells are too small , this type of methods would lead to nearly identical models with little specific predictive value . Alternatively , absolute expression levels can be used to constraint the model's solution space , as previously done by E-Flux for studying bacterial metabolism ( Colijn et al . , 2009 ) . Nonetheless , the applicability of E-Flux for studying human metabolism has not been established . In this study we aim to derive cell-specific metabolic models for human cell lines that are capable of predicting metabolic phenotypes in an individual manner . We aimed to construct such models for the human NCI-60 and HapMap cell line collections , where the similarity in expression levels of different cell lines is quite high . We began our investigation by testing the suitability of two existing model-building approaches towards this end . The moderate performance achieved by existing methods ( see next section ) have led us to develop a new cell-specific model building method termed PRIME ( Personalized ReconstructIon of Metabolic models ) , which utilizes both molecular and phenotypic data for tailoring cell-specific GSMMs . We applied PRIME to reconstruct >280 GSMMs of cancer and normal proliferating cells , which are tested by their ability to predict metabolic phenotypes such as proliferation rate , drug response and biomarkers on an individual level . We then utilized the models of normal and cancer cell lines to predict cancer selective drug targets . We validate experimentally that the top predicted gene target , Malonyl-CoA decayboxylase ( MLYCD ) , induces a clear selective effect on cell growth when tested in both leukemia and renal cancer cell lines , vs normal lymphoblast and renal cell lines . Furthermore , we used PRIME to reconstruct personalized metabolic models of breast and lung cancer patients successfully inferring their prognosis . We therefore suggest that PRIME can be applied in the future to a variety of personalized medicine applications where molecular and phenotypic data can be coupled together to find metabolic drug targets . In this study we aim to derive individualized metabolic models for both normally proliferating lymphoblast cell lines ( HapMap dataset ) , and a panel of cancer cell lines ( the NCI-60 collection ) ( Lee et al . , 2007; Choy et al . , 2008 ) . As these datasets contain both gene expression information and growth rate for each cell line , our goal has been to use the gene expression to build cell-specific models that can predict an array of metabolic phenotypes using the measured proliferation rates for initial testing and validation . The difference in the gene expression of HapMap and NCI-60 datasets is very subtle ( mean Spearman R > 0 . 92 , Figure 1A , upper panel ) , which may in turn imply that discretization-based methods would result here with nearly identical models that will fail to differentiate between their phenotypes . We therefore hypothesized that the integration of absolute expression levels would possibly be more suitable for our goal . To this end , we examined the performance of the two representative previously published methods on these datasets , one accepting discretized expression as inputs ( iMAT [Shlomi et al . , 2008] ) and one analyzing the raw , non discretized expression data ( E-Flux [Colijn et al . , 2009] ) . 10 . 7554/eLife . 03641 . 003Figure 1 . The PRIME pipeline and growth rate predictions obtained by different methods . ( A ) Upper panel: Spearman rank correlation between the metabolic gene expression of two representative cell lines in the HapMap ( left ) and NCI-60 ( right ) datatset ( the two cell lines represent the average correlation across the entire datasets ) ; Middle panel: Spearman rank correlation between predicted and measured growth rates in the HapMap ( left ) and NCI-60 ( right ) datatset as predicted by iMAT , a method that utilizes discrete gene expression signature as input; Lower Panel: Spearman rank correlation between predicted and measured growth rates in the HapMap ( left ) and NCI-60 ( right ) datatset as predicted by E-Flux , a method that utilizes absolute gene expression levels as input . ( B ) A schematic overview of PRIME . As input , PRIME gets a GSMM and gene expression measurements for p cells together with their associated phenotypic measurement ( e . g . , proliferation rate ) . ( Step 1 ) : A set of genes whose expression is significantly associated with the phenotype is identified . ( Step 2 ) : A linear transformation from the expression of the phenotype-associated genes , to reactions' upper bound ( maximal flux capacity ) is applied ( ‘Materials and methods’ ) . PRIME outputs a GSMM for each of the p input cells , such that each cell model generates a different feasible flux solution space . See also Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 00310 . 7554/eLife . 03641 . 004Figure 1—figure supplement 1 . Biomass production as a function of flux upper bound . A piecewise linear curve representing the change in biomass production as a function of the upper bound . The latter are modified gradually over the subset of growth-associated reactions , starting from the minimal flux necessary for growth ( as estimated over the subset of essential reactions , ‘Materials and methods’ ) , to the maximal bound in the model . The normalization range is set to the first segment of the piecewise linear curve . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 004 As shown in Figures 1A and 2A , The performance of these methods leaves much to be desired: iMAT , an omics-integration method that defines a subset of active and inactive reactions based on expression data ( Shlomi et al . , 2008 ) , resulted in insignificant or even negative correlations between the actual and predicted proliferation rates for both datasets ( HapMap: Spearman R = 0 . 03 , p-value = 0 . 66; NCI-60: Spearman R = −0 . 07 , p-value = 0 . 59 , Figure 1A middle panel , Figure 2A ) , probably due to the high correlation in metabolic gene expression between samples ( mean pair-wise Spearman R = 0 . 97 and R = 0 . 92 for the HapMap and NCI-60 datasets , respectively; Figure 1A ) . E-flux ( Colijn et al . , 2009 ) similarly failed to obtain significant results in predicting the HapMap cell lines' proliferation rates ( Spearman R in the range of 0 . 1–0 . 11 , p-value > 0 . 07 , Figure 1A lower panel , Figure 2A , Supplementary file 1A ) , but obtained significant results in predicting the NCI-60 cell lines' proliferation rate ( Spearman R in the range of 0 . 43–0 . 44 , p-value > 3 . 6e-4 , Figure 1A lower panel , Figure 2A , Supplementary file 1A ) . 10 . 7554/eLife . 03641 . 005Figure 2 . Growth rate predictions obtained by PRIME . ( A ) The Spearman correlation achieved by the different methods in predicting the individualized growth rates measurements across the HapMap and NCI-60 cell lines . ( CV; Cross-Validation ) . ( B ) Individual predicted vs measured growth rates in the HapMap ( left ) and NCI-60 ( right ) datasets . ( C ) A comparison between mean predicted and measured growth rates across the four HapMap populations . Measured growth rates are represented as bars and the predicted growth rate is represented as a line . PRIME correctly predicts the population-based order of proliferation rates: CEU < YRI < JPT < CHB . ( D ) A comparison between mean predicted and measured growth rates across the nine tumor types composing the NCI-60 collection . Measured growth rates are represented as bars and the predicted growth rate is represented as a line ( Spearman R = 0 . 71 , p-value = 0 . 03 ) ; Leukemia ( LE ) ; Breast ( BR ) ; Central Nervous System ( CNS ) ; Colon ( CO ) ; Renal ( RE ) ; Lung ( LU ) ; Ovarian ( OV ) ; Prostate ( PR ) ; Melanoma ( ME ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 005 We hence turned to develop a new approach termed PRIME that is designed for our specific task ( Figure 1B and Figure 1—figure supplement 1 ) . PRIME aims to reconstruct distinct , phenotype-based cell-specific metabolic models ( PBCS ) based on sample-specific molecular data . This is achieved by setting maximal flux capacity constraints on a selected subset of reactions in the generic species model , according to their associated gene expression levels and phenotypic data . PRIME's starting point is similar to E-Flux . While both methods utilize the rather straightforward notion of adjusting reactions' bounds according to expression levels , few key differences between them help PRIME generate more accurate models: ( 1 ) since modifying the reactions' bounds is considered to be a hard constraint , one should aim to avoid over-constraining the network based on irrelevant or noisy information . Clearly , only a subset of the metabolic genes affects a specific central cellular phenotype . Accordingly , PRIME identifies this set in the wild type unperturbed case and modifies the bounds of only the relevant set of reactions; ( 2 ) while a common assumption is that expression levels and flux rates are proportional , this is known to hold only partially ( Bordel et al . , 2010 ) . PRIME therefore utilizes the additional phenotypic data to determine the direction ( sign ) of this relation and modifies the bounds accordingly ( ‘Materials and methods’ ) ; ( 3 ) PRIME modifies reactions' bounds within a pre-defined range where the modification is known to have the greatest effect on a given phenotype ( ‘Materials and methods’ ) . Importantly , E-Flux has only been utilized to build models of two different bacterial conditions , by aggregating the expression levels of all samples associated with each condition . In this study we employ the principles described above to build individual cell models from the human metabolic model based on a single sample gene expression signature of each cell . PRIME takes three key inputs: ( a ) gene expression levels of a set of samples; ( b ) a key phenotypic measurement ( proliferation rate , in our case ) that can be evaluated by a metabolic model; and ( c ) a generic GSMM ( the human model , in our case ) . It then proceeds as follows: ( 1 ) A set of genes that are significantly correlated with the key phenotype of interest is determined ( Supplementary file 2A ) ; ( 2 ) The maximal flux capacity of reactions associated with the genes identified in ( 1 ) is modified according to the directionality and level of their corresponding gene expression level . Importantly , to assure that bound modifications would have an effect on the models' solution space , reactions' flux bounds are modified within an effective flux range . Accordingly , PRIME outputs a GSMM tailored uniquely for each input cell ( see Figure 1B , Figure 1—figure supplement 1 and the ‘Materials and methods’ for a formal description ) . We first applied PRIME to a dataset composed of 224 lymphoblast cell lines from the HapMap project ( International HapMap Consortium , 2005 ) . This dataset is composed of cell lines taken from healthy human individuals , from four different populations , including Caucasian ( CEU ) , African ( YRI ) , Chinese ( CHB ) and Japanese ( JPT ) ethnicities ( Supplementary file 1B ) . Applying PRIME to the generic human model ( Duarte et al . , 2007 ) , we constructed the corresponding 224 metabolic models , one for each cell line . The correlation between the proliferation rates predicted by these models and those measured experimentally is highly significant ( Spearman R = 0 . 44 , p-value = 5 . 87e-12 , Figure 2A–B , Supplementary file 1C and Supplementary file 2B ) . In addition to capturing the differences between each of the cell lines the models also correctly predict the experimentally observed significant differences between populations' proliferation rates ( CEU < YRI < JPT < CHB ) in the correct order ( Figure 2C and [Stark et al . , 2010] ) . The correlation observed remains significant also after employing a five-fold cross validation process 1000 times , controlling for the ( indirect ) use of proliferation rate in determining the modified reactions' set ( mean Spearman R = 0 . 26 , empiric p-value = 0 . 007 , Figure 2A , ‘Materials and methods’ ) . Specifically , this analysis is performed by utilizing the set of growth-associated genes derived from the train-set to build the models of the test-set , where the correlation between measured and predicted proliferation rates is then evaluated . We further applied PRIME to build individual models and predict the proliferation rates of 60 cancer cell lines , obtaining a highly significant correlation between the measured and predicted proliferation rates ( Spearman R = 0 . 69 , p-value = 1 . 22e-9 , Figure 2A–B , Supplementary file 1C and Supplementary file 2B ) . A four-fold cross-validation analysis resulted with a mean Spearman correlation of 0 . 56 ( empiric p-value = 0 . 006 , Figure 2A , ‘Materials and methods’ ) . Grouping the samples into the nine tumor types found in this dataset and evaluating the mean proliferation rate of each group , a significant correlation is obtained between the measured and actual growth rates of the different tumors ( Spearman R = 0 . 71 , p-value = 0 . 03 , Figure 2D ) . The higher correlation achieved for the cancer cell-lines in respect to that achieved for the normal cell-lines , is a result of the higher correlation found between metabolic gene expression and growth rate in the former datatset ( see Supplementary file 2A ) . To further examine the process employed by PRIME we tested three additional alternatives: ( 1 ) modifying the bounds of all enzyme-associated reactions and not only of those that are growth-related . This process decreased the correlation to Spearman R = 0 . 24 , p-value = 2 . 4e-9 and Spearman R = 0 . 56 , p-value = 2 . 8e-6 for the NCI-60 and HapMap datasets , respectively; ( 2 ) selecting random sets of reactions at the size of the original set and modifying their bounds according to their gene expression . Repeating this process 1000 times resulted with significantly inferior predictive performance in both datasets compared to PRIME ( empiric p-value < 9 . 9e-4 , ‘Materials and methods’ ) ; ( 3 ) permuting the measured proliferation rates in each of the cell lines datasets for a 1000 times and correlating them with those computed by the PRIME models . In this case as well the original growth prediction results were found to be highly superior ( empiric p-value < 9 . 9e-4 , ‘Materials and methods’ ) . PRIME's major goal is to generate cell-specific metabolic models . Therefore , PRIME has the potential to guide pharmacological interventions based on the individual's phenotype , which underlies the basis of personalized medicine . We therefore tested the ability of PRIME to predict the response of each individual cell line to various metabolic drugs , and compared it with the response measured in vitro ( Scherf et al . , 2000; Choy et al . , 2008; Holbeck et al . , 2010; Garnett et al . , 2012; Lock et al . , 2012 ) . In silico drug response is computed according to the biological phenotype measured experimentally , which in this case includes ATP levels , or AC50/IC50 values ( the concentration at which a given compound exhibits half-maximal efficacy or half-maximal inhibition of cell growth , respectively ) . ATP flux production levels can be estimated directly in a metabolic model . The latter measurements ( AC50/IC50 ) were computed by evaluating the flux through the drug's target reaction under 50% of drug maximal efficacy or 50% inhibition of cell maximal growth ( ‘Materials and methods’ and Supplementary file 1D–F ) . As shown in Figure 3A , this analysis yields a significant Spearman correlation ( p-value < 0 . 05 ) between measured and predicted drug response for 12 out of 16 drugs tested in the HapMap and the NCI-60 datasets . Moreover , performing a permutation test in each of the datasets separately by permuting the measured drug response data , a highly significant result is obtained ( empiric p-value < 9 . 9e-4 , ‘Materials and methods’ ) . Applying a partial correlation analysis between in silico predicted and measured drug response while controlling for the experimentally measured proliferation rate ( as growth rate itself has been implicated as a predictor of drug response , e . g . , for cytotoxic drugs ) , we still find a significant association between predicted and measured drug response for the HapMap and CEU datasets , and in some cases even higher than before ( Supplementary file 1D–E ) . These results demonstrate that utilizing a specifically-tailored metabolic model for predicting metabolic drugs response has a clear advantage over utilizing the raw data alone . 10 . 7554/eLife . 03641 . 006Figure 3 . Drug response , biomarkers and selectivity analysis . ( A ) A comparison between measured and predicted drug response for the HapMap , CEU ( Western European ancestry ) and NCI-60 datasets . Overall , significant correlations ( Spearman p-value < 0 . 05 ) were obtained for 12 out of the 16 drugs examined ( those marked with an asterisk ) . The HapMap drugs are 5-fluorouracil ( 5FU ) and 6-mercaptopurin ( 6MP ) ; the CEU drugs are Ethacrynic acid , Hexachlorophene , Digoxin , Azathioprine , Reserpine and Pyrimethamine; The NCI-60 drugs for dataset 1 include Gemcitabine , Methotrexate and Pyrimethamine; For dataset 2 , Trimetrexate and Gemcitabine; For dataset 3 , Methotrexate , Quinacrine HCl and Allopurinol . ( B ) 14 metabolites for which a significant correlation between measured and predicted uptake and secretion rates is achieved . Both the Spearman correlation coefficient ( gray ) and the–log ( p-value ) ( blue ) are shown . The dashed line represents the FDR corrected significance level for α = 0 . 05 . ( C ) Metabolic reaction targets that are predicted to be non-selective ( green ) or selective ( blue ) . The x-axis represents the selectivity score ( ‘Materials and methods’ ) and the y-axis represents the growth inhibition predicted for the normal cell lines . Non-selective targets are predicted to reduce both normal and cancer cell growth by more than 50% . The selective targets are predicted to reduce normal cell growth by less than 20% and cancer cell growth by more than 30% . MLYCD is the third ranked target with a predicted reduction of >90% in cancer cell growth and <10% in normal cell growth . See also Figure 3—figure supplement 1 . ( D ) Growth survival ( in % ) for the HapMap ( normal ) and NCI-60 ( cancer ) cell lines upon MLYCD knock down , as predicted by E-Flux and PRIME . While E-Flux predicts less than 10% reduction in cellular growth for both normal and cancer cell lines in a largely indiscriminate manner , PRIME predicts a cancer selective effect . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 00610 . 7554/eLife . 03641 . 007Figure 3—figure supplement 1 . Core metabolic pathways and their association with selective and non-selective predicted targets . Metabolic enzymes colored green are a subset of known cytostatic drug targets . Metabolic enzymes colored red are those found in current clinical trials ( and thus likely to be more selective than traditional cytotoxic drugs ) , out of which those marked by an asterisk were identified as selective targets in our simulations as well . Metabolic enzymes colored blue denote novel selective predictions according to our simulations . αKG , α-ketoglutarate; Ac-CoA , acetyl CoA; ASN , aspargine; ASNS , asparagine synthetase; ASP , aspartate; 1 , 3BPG , 1 , 3 biphosphoglyxerate; DHF , dehydrofolate; DHFR , dehydrofolate reductase; CDP , cytosine diphosphate; dCDP , deoxycytosine diphosphate; DHAP , dehydroxyacetone phosphate; dTMP , deoxythymidine monophosphate; dUMP , deoxyuridine monophosphate; F6P , fructose-6-phosphate; FBP , fructose-1 , 6-bisphosphate; G3P , glyceraldehydes 3-phospate; G6P , glucose-6-phosphate; Gln , glutamine; Glu , glutamate; HK2 , hexokinase 2; LDHA , lactate dehydrogenase A; Mal-CoA , malonyl coa; MCT1 , monocarboxylate transporter 1 , 4; mTHF , 5 , 10-Methylenetetrahydrofolate; MYLCD , malonyl-CoA decarboxylase; 2PG , glycerate 2-phosphate; 3PG , glycerate 3-phosphate; PEP , phosphoenolpyruvate; PGAM , phosphoglycerate mutase; PKM2 , pyruvate kinase M2 isoform; R5P , ribose-5-phosphate; RRM1 , ribonucleotide reductase M1; SHMT1 , serine hydroxymethyltransferase 1; TCA , tricarboxylic acid; THF , tetrahydrofolate; TYMS , thymidylate synthase; UDP , uridine diphophate; UMP , uridine monophosphate; UPP1 , uridine phosphorylase . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 007 To further validate the NCI-60 PRIME models we have used measured uptake and secretion rates ( Jain et al . , 2012; Dolfi et al . , 2013 ) and compared them to those predicted by our models ( ‘Materials and methods’ ) . We obtained significant Spearman correlations ( Benjamini-Hochberg adjusted p-value with False Discovery Rate ( FDR ) and α = 0 . 05 ) for 14 out of 33 metabolites with a corresponding transporter reaction in the human model ( Figure 3B ) . By performing a permutation test on the measured data a highly significant result is obtained ( empiric p-value < 9 . 9e-4 , ‘Materials and methods’ ) . Importantly , utilizing the models reconstructed by E-Flux for the same task , insignificant results are obtained for all metabolites . The array of models built for both normal and cancer cells provides us with a unique opportunity not only to predict cell-specific drug target effects , but more importantly , to find drug targets that inhibit proliferation across all cancer cells but have no effect on the non-transformed counterpart . To this aim we simulated all knock downs of individual reactions in the 224 normal lymphoblasts and 60 cancer cell models , and quantified their selective effect on cell proliferation ( ‘Materials and methods’ ) . The set of predicted non-selective targets was highly enriched with current cytostatic drugs ( Wishart et al . , 2008; Folger et al . , 2011 ) ( mean hypergeometric p-value = 7 . 28e-4 , Figure 3—figure supplement 1 and Supplementary file 1G ) . Second , the predicted selective targets were enriched with targets of newly developed drugs ( Figure 3—figure supplement 1 ) : Out of the five metabolic enzyme drug targets reported in ( Cheong et al . , 2012 ) , our analysis identified three as being selective ( Hypergeometric p-value = 3 . 98e-4; Supplementary file 2C ) . To further validate these findings , we examined the clinical relevance of our predicted selective targets on a cohort of 1586 breast cancer patients ( Curtis et al . , 2012 ) . A Cox multivariate regression analysis shows that this set is enriched ( Hypergeometric p-value = 2 . 1e-5 ) with genes whose lower expression is significantly associated with improved survival ( Benjamini-Hochberg adjusted p-values with FDR and α = 0 . 1 , ‘Materials and methods’ ) , when examined together with known prognostic variables such as patients' clinical stage , histological grade , tumor size , lymph node status and estrogen receptor status . A similar analysis for the set of predicted non-selective targets yielded either borderline or insignificant results ( Supplementary file 1G ) . A top predicted selective target is Malonyl-CoA Decarboxylase ( MLYCD ) ( Figure 3C ) . While the highest ranked predicted reaction is catalyzed by isoenzymes and therefore more difficult to target experimentally , and the second ranked reaction occurs spontaneously , MLYCD is the first prediction that could be tested from a practical , experimental point of view ( Supplementary file 2C ) . Of note , the knock down of MLYCD is predicted by E-Flux to reduce both normal and cancer cell proliferation by less than 10% , suggesting that without including phenotype-based constraints , this candidate gene would have not been revealed ( Figure 3D ) . Interestingly , this enzyme has been recently proposed as potential anticancer target for breast cancer ( Zhou et al . , 2009 ) , however its selective effects on other tumor types have not been assessed . Therefore , we decided to further investigate the role of MLYCD as selective target for cancer therapy . The prediction of selective targets made by PRIME capitalizes on the non-transformed lymphoblast cell lines HapMap as normal counterpart . Therefore , to experimentally validate the cancer versus normal selectivity , we initially used leukemia cells , the only hematological tumor type in the NCI-60 database . In line with PRIME's predictions , the small interfering RNA ( siRNA ) -mediated silencing of MLYCD significantly inhibited the proliferation of the leukemia cell lines RPMI-8226 and K562 cells , but had no effect on HapMap cells ( Figure 4A–B ) . To further corroborate the cancer versus normal selectivity , we tested the effects of MLYCD depletion on two renal cancer cell lines , TK-10 and CAKI-1 , using the non-transformed renal cell line HK-2 as normal control ( Figure 4C ) . Of note , the silencing of MLYCD suppressed proliferation of renal cancer cell lines without affecting the non-transformed counterpart ( Figure 4D ) . Importantly , the anti-proliferative effects of MLYCD suppression could not be explained by the different expression of the enzyme among the different cell lines ( Figure 4—figure supplement 1 ) . These results substantiated PRIME's prediction that MLYCD is a cancer selective drug target . 10 . 7554/eLife . 03641 . 008Figure 4 . MLYCD depletion on normal and cancer cell lines . ( A ) MLYCD mRNA expression upon nucleofection with Non Targeting Control ( NTC ) and three independent siRNA constructs in HapMap , RPMI-8226 and K562 cells . ( B ) Cell counts after 72 hr of culture of the indicated cell lines . ( C ) MLYCD mRNA expression upon nucleofection with Non Targeting Control ( NTC ) and three independent siRNA constructs in HK2 , TK10 and CAKI1 cells . ( D ) Cell counts after 72 hr of culture of the indicated cell lines . Data are shown as mean ± s . e . m of three independent cultures . p-value<0 . 05 . p-value<0 . 01 . *p-value < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 00810 . 7554/eLife . 03641 . 009Figure 4—figure supplement 1 . Expression levels of MLYCD across multiple cancer and normal cell lines . mRNA levels of MLYCD in the indicated cell lines was measured by qPCR . Data are indicated as ΔCT of MLYCD vs Actin in the indicated cell lines . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 009 We wanted to functionally validate the effect of silencing of MLYCD in cancer cells . To this aim , we generated a leukemia cell line that stably expresses a doxycycline-inducible short hairpin RNA ( shRNA ) targeting MLYCD . The incubation with doxycycline resulted in efficient silencing of MLYCD and led to a significant growth inhibition ( Figure 5—figure supplement 1–2 ) , in line with the siRNA experiments . Previous reports have shown that MLYCD depletion leads to the accumulation of malonyl-CoA , which blocks fatty acid oxidation by allosteric inhibition of the mitochondrial enzyme Carnitine-Palmitoyl-Transferase ( CPT1 ) ( Zhou et al . , 2009 ) . These observations prompted us to investigate the effects of the loss of MLYCD on fatty acid oxidation . To this aim , cells were incubated with 13C16-palmitate and the abundance of 13C-labelled palmitoyl-carnitine and of TCA cycle metabolites was measured by liquid chromatography coupled to mass spectrometry ( LCMS ) ( see Figure 5A for a schematic of the experiment ) . We observed a significant decrease in the 13C-labelling of palmitoyl-carnitine ( Figure 5B ) and of the m+2 isotopologues of TCA cycle intermediates ( Figure 5C , and Figure 5—figure supplement 3 for the full isotopologue analyses of these metabolites ) , indicating that fatty acid oxidation is reduced in MLYCD-depleted cells . Of note , this marked decrease in fatty acid oxidation only partially affected the overall abundance of TCA cycle intermediates ( Figure 5—figure supplement 4 ) . We also noticed a striking accumulation of succinate and a decrease in fumarate and malate in MLYCD-depleted cells ( Figure 5—figure supplement 4 ) . These results are consistent with the inhibition of the TCA cycle enzyme succinate dehydrogenase ( SDH ) , which may be caused by malonyl-CoA-derived malonate . Taken together , these results show that the silencing of MLYCD is sufficient to inhibit fatty acid oxidation and alter TCA cycle . 10 . 7554/eLife . 03641 . 010Figure 5 . Metabolic characterization of MLYCD depletion . ( A ) Schematic representation of isotope tracing experiment with 13C16-Palmitate . Black-filled circles indicate 13C-carbon , whereas the white filled circles represent the unlabeled carbon . The schematic shows the expected composition of labeled carbons of the indicated metabolites . ( B ) Labeling incorporation from 13C-Palmitate into Palmitoyl-carnitine in non-targeting control ( NTC ) and MLYCD-depleted ( shMLYCD ) cells . Data are shown as percentage of 13C16-palmitoylcarnitine to the total pool of Palmitoyl-carnitine . ( C ) Labeling incorporation from 13C16-palmitate into TCA cycle intermediates of the indicated cell lines . Data are shown as percentage of the m+2 isotopologue to the total pool size of each metabolite . ( D ) Schematic representation of isotope tracing experiment with 13C6-Glucose . The distribution of light and heavy carbons is depicted as in A . ( E ) Labeling of Citrate and of ( F ) Palmitate after incubation with 13C6-glucose . Data are shown as percentage of the indicated isotopologue to the total pool size of each metabolite . Isotopologue distribution of citrate is indicated in Figure 5—figure supplement 6 . Palmitate isotopologues above m+10 were not detected ( G ) Schematic representation of isotope tracing experiment with 1 , 2-13C2-Glucose . Ru5p: ribulose-5-phosphate . The distribution of light and heavy carbons is depicted as in A . ( H ) Ratio between m+1 and m+2 isotopologues of Lactate in the indicated cell lines . ( I ) Ratio between reduced ( GSH ) and oxidized ( GSSG ) glutathione in RPMI-8226 cells infected with the indicated constructs . ( J ) Cell counts after 72 hr of culture of the indicated cell lines in the presence or absence of 2 mM N-Acetyl Cysteine . Data are shown as mean ± s . e . m of three independent cultures . p-value<0 . 05 . p-value<0 . 01 . *p-value < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 01010 . 7554/eLife . 03641 . 011Figure 5—figure supplement 1 . Silencing of MLYCD in RPMI-8226 cells using shRNA . MLYCD mRNA expression upon infection with Non Targeting Control ( NTC ) and two independent shRNA constructs targeting MLYCD ( shMLYCD1 and 2 ) in RPMI-8226 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 01110 . 7554/eLife . 03641 . 012Figure 5—figure supplement 2 . Effects of Silencing of MLYCD in RPMI-8226 cells . Cells were treated as indicated in Figure 5—figure supplement 1 and counted after 72 hr of culture were . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 01210 . 7554/eLife . 03641 . 013Figure 5—figure supplement 3 . Isotopologue distribution of TCA cycle intermediates after incubation with 13C16-palmitate . Labeling incorporation from 13C16-palmitate in NTC and shMLYCD2 cells . Isotopologues above m+2 were not detected . Data are presented as percentage of the indicated isotopologue to the total pool of each metabolite . Data are presented as mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 01310 . 7554/eLife . 03641 . 014Figure 5—figure supplement 4 . LCMS analyses of TCA cycle intermediates in MLYCD-depleted cells . Total levels of TCA cycle intermediates in RPMI-8226 cells infected with NTC or shMLYCD2 . Data are presented as relative abundance of total metabolites in shMLYCD2 compared to NTC . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 01410 . 7554/eLife . 03641 . 015Figure 5—figure supplement 5 . A schematic description of the metabolic changes following MLYCD knock down . MLYCD suppression is predicted to divert the accumulated malonyl-CoA to fatty acid biosynthesis , increasing the demand of cells for reducing power . The latter is generated by the oxidative branch of the pentose phosphate pathway , overall leading to increased oxidative stress . Green/red arrows represent increased/decreased flux , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 01510 . 7554/eLife . 03641 . 016Figure 5—figure supplement 6 . TCA cycle activity in MLYCD-depleted cells . Labeling incorporation from 13C6-Glucose in NTC and shMLYCD2 cells . Data are presented as percentage of the indicated isotopologue to the total pool of each metabolite . Data are presented as mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 016 We then used the PRIME-derived models to systematically assess the metabolic changes that occur upon MLYCD inactivation . Of note , the model predicted that upon MLYCD suppression , part of the accumulated malonyl-CoA is diverted to fatty acid biosynthesis . Since this process requires NADPH as source of reducing power , the aberrant activation of fatty acid synthesis caused by the loss of MLYCD would impair redox homeostasis of the cell ( Berg , 2002 ) ( Figure 5—figure supplement 5 and Supplementary file 1H ) . We validated this hypothesis by first assessing fatty acid synthesis . To this aim , cells were incubated with 13C6-glucose and the abundance of 13C-labelled TCA cycle intermediates and palmitate were analyzed by LCMS ( Figure 5D ) . While the labeling of citrate , the main lipogenic precursor , was , if any , slightly decreased ( Figure 5E ) , the m+4 and m+6 isotopologues of palmitate were significantly increased in MLYCD-depleted cells ( Figure 5F ) , suggesting that fatty acid synthesis is accelerated in these cells . Of note , the reduction of the m+2 and m+4 isotopologues of TCA cycle intermediates suggested that the oxidative capacity of the TCA cycle is intact , albeit reduced , in MLYCD-depleted cells ( Figure 5—figure supplement 6 ) . To validate the prediction that MLYCD-depleted cells increase the demand of NAPDH to fuel fatty acid synthesis , we measured the activity of the pentose phosphate pathway ( PPP ) , the major source of cytosolic NADPH ( Fan et al . , 2014 ) . To this end , cells were incubated with 1 , 2-13C2-glucose and the amount of singly ( m+1 ) or doubly ( m+2 ) labeled lactate was used as measure of PPP or glycolysis activity , respectively ( see Figure 5G for a representation of the experiment ) . As predicted by PRIME , PPP flux was increased in MLYCD-depleted cells ( Figure 5H–I ) . Together , these results corroborate the prediction made by PRIME that the loss of MLYCD increases fatty acid synthesis and impinges on the PPP for generation of reducing power . Finally , we tested whether the observed activation of fatty acid synthesis , by draining NADPH , impairs the capacity of cells to maintain redox homeostasis . In line with this hypothesis , MLYCD-depleted cells exhibited a lower GSH/GSSG ratio compared to control cells ( Figure 5I ) . Furthermore , the incubation of cells with the antioxidant N-acetyl-cysteine ( NAC ) fully restored the proliferation defects observed in MLYCD-depleted cells ( Figure 5J ) . Taken together , these results suggest that the suppression of cancer cell proliferation caused by the loss of MLYCD depends , at least in part , on the aberrant activation of fatty acid synthesis , which leads to a reduced ability of cells to maintain redox homeostasis . Overall , this investigation showed the benefits of PRIME to predict and investigate metabolic liabilities of cancer cells , based on cell-specific metabolic models . We next aimed to go beyond predicting targets that are selective with respect to cancer versus normal cell populations as a whole , to study if we can use PRIME to predict the differential response amongst cancer cell lines to specific treatments . To this end we used PRIME models of individual breast cancer cell lines of the NCI-60 panel , and simulated via Minimization of Metabolic Adjustment ( MOMA ) ( Segre et al . , 2002 ) the knock down of all metabolic reactions catalyzed by a single gene , examining their effect on cell growth ( ‘Materials and methods’ ) . We focused on reactions whose knock down yielded highly variable predicted growth rates across the different cell lines studied . 13 genes associated with these top ranked reactions and spanning different metabolic pathways were selected for further experimental investigation ( ‘Materials and methods’ and Supplementary file 2D ) . The effect of each of these genes on cell growth was examined via small interference RNA ( siRNA ) knock down in the two cell lines predicted to have the most differential effect on cell growth . 11 out of the 13 genes studied were found to have an effect on cell growth as predicted by the models ( Figure 6A and Supplementary file 2D , empiric p-value < 0 . 01 , ‘Materials and methods’ ) . A significant correlation is obtained between predicted and measured % inhibition values across all 11 targets ( Spearman R = 0 . 64 , p-value = 1e-3 ) . These data underscore the ability of PRIME to successfully predict individual cell-specific responses of cancer cells to the knock down of metabolic enzymes , at least at a qualitative level . 10 . 7554/eLife . 03641 . 017Figure 6 . Differential growth affects in breast cancer cell-lines and clinical data analysis . ( A ) Four gene/reaction targets showing a differential effect on cancer cell growth ( represented as % of growth inhibition ) according to both PRIME's predictions and experimental validations via siRNA knock downs ( when compared to a negative control , a siRNA that targets luciferase ) . Each gene was tested experimentally in two cell lines in triplicate , where the gene knock down is predicted to have the lowest and highest effect on cell growth . 11 out of the 13 top predictions tested were confirmed experimentally . Data are shown as mean ± s . e . m . For the full list see Supplementary file 2D . The genes GSR and PROSC are predicted to completely suppress the Hs578 t cell line growth ( Supplementary file 2 ) but for presentation appear with a 0 . 05% height bar; ( B ) Kaplan-Meier plots for the two breast cancer datasets and for a lung cancer dataset . In all cases low growth rate ( GR ) is associated with improved survival . DOI: http://dx . doi . org/10 . 7554/eLife . 03641 . 017 Finally , we examined PRIME's ability to build personalized models of cancer patients and predict their prognosis based on gene expression levels collected from biopsy samples . Importantly , growth rate measurements are not available for these datasets . Nonetheless , a possible way to overcome this hurdle and to build personalized metabolic models for cancer patients is to use phenotypic data measured for one set of cells to reconstruct models of a different set of cells or clinical samples . To examine this approach we utilized the set of growth-associated genes derived from the NCI-60 collection to build personalized GSMMs of more than 700 breast and lung cancer clinical samples ( Miller et al . , 2005; Chang et al . , 2010; Okayama et al . , 2012 ) . A Kaplan–Meier survival analysis ( Kaplan and Meier , 1958 ) showed that patients with predicted low growth rate had significantly improved survival compared to those with a predicted high growth rate ( logrank p-values are: 0 . 01 , 1e-3 and 0 . 02 for Miller et al . , Chang et al . and Okayama et al . respectively , Figure 6B , Supplementary file 1I , ‘Materials and methods’ ) . This result was further supported by a Cox univariate survival analysis ( Grambsch , 2000 ) ( p-values are: 1e-3 , 1e-4 and 2e-3 for Miller et al . , Chang et al . and Okayama et al . respectively , Supplementary file 1I ) and by performing a permutation test ( p-values are: 0 . 015 , 2e-3 and 0 . 018 for Miller et al . , Chang et al . and Okayama et al . respectively , ‘Materials and methods’ ) . Of note , estimating the samples growth rates directly from the gene expression data by using multiple linear regression , resulted in inferior performance ( Supplementary file 1J ) , testifying to the added value of personalized GSMMs . Importantly , while iMAT and E-Flux require only ‘omics’ data and can hence be applied directly , they fail to obtain meaningful and significant results in this setting as well ( Supplementary file 1K ) . In this study we present a novel method termed PRIME for building cell-specific GSMMs based on the integration of gene expression and phenotypic data . We apply this method for the reconstruction of metabolic models of both cancer and normal cells . To the best of our knowledge , PRIME is the first method able to generate human cell-specific GSMMs that can predict metabolic phenotypes in an individual manner , including growth rates and drug response . The set of normal and cancer PRIME-derived models is utilized to identify a set of drug targets that can inhibit the proliferation of specific cell lines , as well as metabolic targets that can selectively block cancer but not normal cells growth . The experimental validation that we provide testifies that coupling molecular and phenotypic data for building cell-specific models can enhance the predictive power of GSMMs . As many other computational approaches , PRIME is not devoid of limitations . First , PRIME assumes that cells try to maximize their proliferation , while different objective function ( s ) should be considered for non-proliferating cells . Second , we assume that all models share the same set of enzymes and differ only in their cellular abundance , but different cells may express different coding variants that should be incorporated in future studies . Third , PRIME relies on the measurement of a specific phenotype that is not always available for a given set of cells or samples . Here we introduced a possible way to overcome this hurdle , as demonstrated by PRIME's ability to utilize clinical data and build cell-specific GSMMs tailored for each individual patient . However , while this analysis provided significant results , the obtained signal is mild and the question whether and how best one can identify a universal set of growth-associated genes still requires further study . Given the results obtained , one can confidently expect that follow-up work analyzing richer datasets , and most importantly , incorporating additional kinds of omics data ( such as enzyme sequence data ) will significantly improve the predictive power of PRIME further . In this work we have also experimentally validated the prediction made by PRIME that MLYCD inhibition selectively affects cancer proliferation . MLYCD is an important enzyme of fatty acid metabolism , which role in cancer therapy has been recently suggested ( Zhou et al . , 2009 ) . However , the selectivity across cancer types , and the mechanism of action of its inhibition have not been fully investigated . Our results show that the silencing of MLYCD has an anti-proliferative effect across multiple cancer cell lines but spares the non-transformed counterparts , confirming PRIME's predictions . We have also shed some light on the functional effects of inactivation of MLYCD in cancer cells . The toxic effects of MLYCD inhibition have been previously attributed to the accumulation of malonyl-coA and to the inhibition of fatty acid oxidation ( Zhou et al . , 2009 ) . Our results suggest that , besides turning off fatty acid oxidation and partially deregulating TCA cycle , the loss of MLYCD stimulates fatty acid synthesis , which drains reducing equivalents and sensitize cells to oxidative stress . Therefore , our results not only confirmed the cancer versus normal selectivity of MLYCD inhibition but also elucidated a novel liability of cancer cells based on the pharmacological inhibition of fatty acid metabolism . Of note , both these features were accurately predicted by PRIME . Importantly , in humans , the loss of MLYCD leads to methylmalonic aciduria , an extremely rare autosomal recessive disorder . Nevertheless , in vivo experiments in rodents and pigs ( Dyck et al . , 2004; Wu et al . , 2014 ) , ex vivo experiments in human skeletal muscle ( Bouzakri et al . , 2008 ) , and in MRC-5 non-transformed fibroblasts ( Zhou et al . , 2009 ) suggest that the inhibition of MLYCD is well tolerated , as our results indicate . It is therefore possible that the inhibition of the enzyme has no detrimental effects on normal cells and tissues , and that other factors contribute to the severity of MLYCD deficiency in humans , including a toxic effect of the mutated protein ( Polinati et al . , 2014 ) . In summary , we here show that incorporating gene expression measurements and phenotypic data within a genome-scale model of human metabolism via PRIME results in functional cell-specific models with considerable predictive power . We believe that the demonstrated ability of PRIME to predict the effects of known metabolically-targeted drugs on individual cell proliferation rates will help to pave the way for tailoring specific therapies based on metabolic modeling of cancer biopsies from individual patients . A metabolic network consisting of m metabolites and n reactions can be represented by a stoichiometric matrix S , where the entry Sij represents the stoichiometric coefficient of metabolite i in reaction j ( Price et al . , 2004 ) . CBM imposes mass balance , directionality and flux capacity constraints on the space of possible fluxes in the metabolic network's reactions through the set of linear equations: ( 1 ) S·v=0 ( 2 ) vmin≤v≤vmaxwhere v is the flux vector for all of the reactions in the model ( i . e . , the flux distribution ) . The exchange of metabolites with the environment is represented as a set of exchange ( transport ) reactions , enabling a pre-defined set of metabolites to be either taken up or secreted from the growth media . The steady-state assumption represented in Equation ( 1 ) constrains the production rate of each metabolite to be equal to its consumption rate . Enzymatic directionality and flux capacity constraints define lower and upper bounds on the fluxes and are embedded in Equation ( 2 ) . The biomass function utilized here is taken from ( Folger et al . , 2011 ) . The media simulated in all the analyses throughout the paper is the RPMI-1640 media that was used to grow the cell lines experimentally ( Lee et al . , 2007; Choy et al . , 2008 ) . Gene knock outs are simulated by constraining the flux through the corresponding metabolic reaction to zero . Following , two different approaches can be taken to estimate the effect of a perturbation on the network: ( 1 ) via Flux Balance Analysis ( FBA ) ( Varma and Palsson , 1994 ) where maximization of growth rate is defined as the cellular objective function ( max Vbio ) ; ( 2 ) Minimization of Metabolic Adjustment ( MOMA ) ( Segre et al . , 2002 ) where the minimization of the Euclidean distance between a wild-type flux distribution ( Vwt ) and the post-perturbation flux distribution ( VKO ) is set as the cellular objective function min∑in ( Vwt , i−VKO , i ) 2 . Different wild-type flux distributions are obtained via sampling where each sample is determined based on a FBA analysis maximizing for cellular growth . PRIME is given the following three inputs: ( 1 ) a set of p samples with gene expression levels; ( 2 ) the p samples' corresponding growth rate measurements; and ( 3 ) a generic model ( the human model , in our case ) . Next , the model reconstruction process is as follows:Each reversible reaction is decomposed into its forward and backward direction and the maximal biomass production is evaluated . Next , the upper bound of all the reactions in the network is decreased simultaneously in steps of 0 . 1 . In each step , the maximal biomass production is re-evaluated and the process proceeds as long as the reduction in bound doesn't decrease the maximal production found above by more than an ε ( here we used ε = 1e-4 ) . Finally , the upper bound of all reactions is set to the minimal upper bound allowed by this process . The goal of this step is only to narrow down the solution space and reduce the effect of futile cycles in the simulation of gene perturbation . Next , the correlation between the expression of each reaction in the network and the measured growth rates is evaluated . The expression of a given reaction is defined as the mean expression of its catalyzing enzymes . The significance threshold is corrected by FDR with α = 0 . 05 . The upper bound of each reaction demonstrating a significant correlation to the growth rate ( e . g . , t reactions ) is modified in a manner that is linearly related to its expression value . Specifically , we generate the Exp-matrix ( E ) , a ( t × p ) matrix that embeds the information on the direction and magnitude of change of the upper bound based on the expression data . For each reaction a in sample b we define the Exp-matrix such that: ( 3 ) Ea , b=ρa\|ρa\|·GEa , b In Equation ( 3 ) , GEa , b represents the expression value of reaction a in sample b . Likewise , ρ ( a ) represents the correlation coefficient of reaction a as found in step ( 2 ) . Overall , for reactions whose expression is positively correlated with growth rate , the corresponding values in the matrix increase ( become more positive ) as the expression increases . Alternatively , for negatively correlated reactions , the corresponding values in the matrix decrease ( become more negative ) as the expression increases ( due to the multiplication by ρa\|ρa\| which equals to −1 in this scenario ) . We then apply Equation ( 4 ) to normalize the values of the Exp-matrix and adapt them to the actual upper bounds . In this normalization procedure each reaction a is normalized across its p samples such that the bound associated with the sample having the lowest ( highest ) expression value is assigned the minimal ( maximal ) value of the normalization range , respectively . ( 4 ) UBa , b= ( Ea , b−min ( Ea ) max ( Ea ) −min ( Ea ) · ( maxNormVal−minNormVal ) ) +minNormVal min ( Ea ) and max ( Ea ) refer to the minimal and maximal value of reaction a across all p samples in the Exp-matrix , respectively . The minimal and maximal values of the normalization range ( minNormVal and maxNormVal , respectively ) are determined according to the procedure described in the next section . minNormVal is set to be the minimal flux necessary for biomass production . This value is computed in the following manner: First , the set of essential reactions in the model is identified via Flux Balance Analysis . This set is composed of those reactions that their knock out reduces growth by more than 90% of its maximal rate . Next , the minimal flux through each essential reaction is found via Flux Variability Analysis ( Varma and Palsson , 1994 ) . As each of these reactions is necessary for biomass production , reducing the upper bound below their minimal flux value would result with a lethal phenotype . We therefore set minNormVal to be the maximal value among these values ( Figure 1—figure supplement 1 ) . To define the maximal value of the normalization range ( maxNormVal ) we examine the change in biomass production as a function of the model's upper bounds according to the following steps:A . First , we define the set of reactions in the model that are significantly correlated to the proliferation rate ( as described in step ( 2 ) of PRIME above ) . B . Next , we examine how the biomass production is changed as a function of the model's upper bound . This is done by changing the upper bounds of the growth-associated reactions in steps of 0 . 1 , and in each step re-evaluating the biomass production . C . Lastly , maxNormVal is defined as the maximal value beyond which the change in biomass production decreases ( Figure 1—figure supplement 1 ) . Importantly , applying alternative ranges resulted with less optimal results in all datasets analyzed here . The PRIME code and the generated models are provided as Supplementary file 3 and 4 , respectively . K-fold cross validation analysis is done by splitting the samples of the examined dataset to train- and test-sets . The set of growth-associated reactions found in the train-set is then used to build the models of the test-set . The correlation reported is the mean Spearman correlation achieved by comparing the measured and predicted growth rates of the test-set alone , while repeating this process 1000 times . The empiric p-value is computed by permuting the gene expression 1000 times , in each case building the resulting models and performing the cross-validation analysis as described here . Finally we compared the resulting mean Spearman correlation of each of these models to that obtained with the original data . Generally , all permutation tests are repeated 1000 times . Empiric p-value is then computed as ( n+1 ) /1001 where n equals the number of times a random set of values yields a result which is more significant than the original result obtained when the data is not permuted . Each drug is mapped to its corresponding metabolic reaction through its known enzymatic targets according to DrugBank database ( Wishart et al . , 2008 ) . In this study we focused on drugs that: ( 1 ) have an inhibitory effect; ( 2 ) the majority of their targets are metabolic; ( 3 ) are not associated with dead-end reactions . The drug response data used in this analysis was measured in various ways: ( a ) ATP concentrations ( HapMap dataset ) : In this case the in silico drug response is computed via MOMA in two steps; ( 1 ) obtaining a wild-type flux distribution via Flux Balance Analysis in which the corresponding drug target reaction is initially forced to be active ( the pre-drug condition ) . Enforcing the target reaction to be active is necessary in order to get an effect on the resulting flux distribution following the inhibition simulated in the next step . Here we enforced a positive flux through the target reactions that is 50% of the maximal flux rate it is able to carry ( our results are robust to various activation thresholds; Supplementary file 1D ) . ( 2 ) Next , the knock out flux distribution is computed via MOMA ( Segre et al . , 2002 ) while constraining the flux through the corresponding reactions to zero . This process is repeated for each personalized model separately and the predicted ATP production is used to estimate the cell response to the simulated drug . A robustness analysis is carried out by using 1000 different wild-type flux distributions ( Supplementary file 1D ) ; ( b ) AC50 values ( CEU dataset ) : AC50 values represent the concentration in which the drug exhibits 50% of its maximum efficacy . In this case , in silico AC50 values are calculated by estimating the maximal flux rate carried by the target reaction when the growth rate is set to 50% of the drug's maximal response ( a value that is available in the dataset used [Lock et al . , 2012] ) ; ( c ) IC50 values ( NCI-60 dataset ) : IC50 values represent the concentration of drug needed in order to reduce the growth rate to 50% of its maximal value . In this case , in silico IC50 values are calculated by estimating the maximal flux rate carried by the target reaction when growth rate is set to 50% of its maximal value . In all cases of drug response simulations the permutation test is carried out by permuting the measured data 1000 times and re-estimating the resulting correlation for each permuted vector . We have utilized the CORE data published by Jain et al . ( 2012 ) and normalized to cell size by Dolfi et al . ( 2013 ) , and compared it to uptake and secretion rates as predicted by the NCI-60 models . We have focused on 33 metabolites for which a corresponding exchange reaction exist in the human model and for which a non-zero flux was measured in at least three of the cell-lines . For each of these metabolites we estimated the maximal flux rate through its exchange reaction under at least 90% maximal growth rate , and compared it to that measured experimentally across the 59 cell-lines for which data exist . A similar approach was taken for both the PRIME and the E-Flux models . The permutation test is performed by permuting normalized CORE data 1000 times and repeating the process described above . The effect of a reaction's deletion on cell growth in four breast cancer cell line models ( MDA-MB-231 , Hs578 t , BT549 and MDA-MB-435 ) was simulated via MOMA while enforcing the tested reaction to carry 50% of its maximal flux in the WT state ( as described in the section ‘Drug response simulations’ above ) . The knock down of each tested reaction was simulated by inhibiting the target reaction by at least 75% of its maximal flux , then maximizing cellular growth under this perturbation . To increase specificity , we focused on reactions that are: ( 1 ) catalyzed by a single gene , and ( 2 ) , their catalyzing gene does not catalyze more than three different reactions . Reactions were then ranked based on the variance in their knock down predicted growth rate across the four cell line models . 13 top predicted genes were selected for further experimental validation based on their high ranking in the list ( i . e . , high variance ) and their association with diverse metabolic pathways ( excluding transport reactions which their catalyzing enzymes are less specific ) . Each gene was examined experimentally in the two cell lines predicted to have the lowest and highest affect on cell growth . The permutation test is performed by permuting the models' predicted growth rates ( after reaction knock down ) 1000 times . The effect of reaction's deletion on cell proliferation for the identification of selective treatment was simulated via MOMA with its robustness analysis as described in the section ‘Predicting differential effects on cancer cell growth’ above . The overlap between the set of cytostatic drug targets and the predicted non-selective targets was found to be robust to different thresholds that determine the value ( in percentage ) under which the deletion is considered to effect the cell's proliferation rate ( Supplementary file 1G ) . The set of selective reaction targets is composed of those that reduce the growth of all normal cells by less than 20% and the growth of all cancer cells by more than 30% . Additionally , this set includes only those reactions that exhibit more than 20% difference in growth reduction between the normal and cancer proliferating cells ( Supplementary file 2 ) . Denoting growth inhibition as Gi and growth survival as Gs , where Gs is defined as ( 1−Gi ) , the selectivity score computed for representation in Figure 3B is defined as ( GiNCI60−GiHapMap ) ∗GsHapMap . The association between selective and non-selective targets and clinical survival data is performed by a Cox multivariate regression analysis . Specifically , a p-value for a Cox regression analysis of the expression of each gene and additional prognostic variables including patients' clinical stage , histological grade , tumor size , lymph node status and estrogen receptor status is computed . Each metabolic reaction is then assigned the minimal p-value achieved by its catalyzing enzymes . p-values are adjusted by Benjamini-Hochberg with FDR and α = 0 . 1 . Utilizing the RPMI-8226 model we first sampled the solution space and obtained 1000 wild-type flux distributions under maximal growth rate , in which the MLYCD reaction is forced to be active in a rate that is at least 50% of the maximal flux rate it can carry . Next , the knock down flux distribution is computed via MOMA while constraining the flux through the MLYCD reaction as described in ‘Drug selectivity analysis’ above . Utilizing the 1000 pre- and post-knockout flux distributions we applied a one-sided Wilcoxon ranksum test to determine reactions whose flux has been significantly increased/decreased . Supplementary file 1H summarizes these results . The set of growth-associated reactions identified in the NCI-60 dataset was utilized as input to PRIME in the reconstruction process of the breast and lung cancer patients' models . PRIME then proceeds by adjusting the bounds of this set of reactions according to the specific cell expression levels .	Cancer is not just one disease , but a collection of disorders; as such there is no single general treatment that is effective against all cancers . Different tissues and organs—including the lungs , skin , and kidneys—can get cancer , and each need different treatments . Even two patients with the same type of cancer might respond differently to the same treatment . Being able to distinguish between different cancer types would help doctors personalize a patient's cancer therapy—which would hopefully improve the outcome of the treatment . An important step in developing such personalized treatments is to find out how each type of cancer cell behaves and to see how this behavior differs both from normal , healthy cells and other types of cancer . Countless chemical reactions take place inside living cells , and these reactions essentially dictate how a cell will grow and behave . The chemical reactions occurring inside a cancerous cell can be described as its ‘metabolic phenotype’ and will likely be different to the chemical reactions occurring in a healthy cell . Now Yizhak , Gaude et al . have used a range of data , including gene expression data , to create computer models of the metabolic phenotypes of 60 different types of human cancer cell . The same approach was also used to create metabolic models of over 200 healthy human cells that were dividing normally . Yizhak , Gaude et al . used these metabolic models to predict how quickly the different types of cancer cell would divide and how the cells would respond to drug treatments . It may be possible to reduce the spread of all types of cancer—without also affecting healthy cells—by targeting proteins that help cancerous cells to proliferate . Yizhak , Gaude et al . used all of the models to search for genes that encode such proteins . One gene that was predicted to provide such a drug target encodes an enzyme that is needed to make and break down fatty acid molecules . Experiments confirmed that inhibiting this gene slowed the proliferation of both leukemia and kidney cancer cells , but had less of an effect on the growth of healthy bone marrow or kidney cells . Finally , Yizhak , Gaude et al . generated detailed metabolic profiles of cancer cells taken from over 700 breast and lung cancer patients and were able to use the models to successfully predict the outcome of the diseases in these patients . Yizhak , Gaude et al . 's findings might help future efforts aimed at developing and delivering personalized cancer therapies . The next challenge is to use additional data—such as gene sequencing data—to generate more detailed and more accurate metabolic models for many cancer patients , to both predict their individual responses to available drugs and identify new patient-specific treatments .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "cell", "biology", "cancer", "biology" ]	2014	Phenotype-based cell-specific metabolic modeling reveals metabolic liabilities of cancer
To identify new approaches to enhance innate immunity to bacterial pneumonia , we investigated the natural experiment of gender differences in resistance to infections . Female and estrogen-treated male mice show greater resistance to pneumococcal pneumonia , seen as greater bacterial clearance , diminished lung inflammation , and better survival . In vitro , lung macrophages from female mice and humans show better killing of ingested bacteria . Inhibitors and genetically altered mice identify a critical role for estrogen-mediated activation of lung macrophage nitric oxide synthase-3 ( NOS3 ) . Epidemiologic data show decreased hospitalization for pneumonia in women receiving estrogen or statins ( known to activate NOS3 ) . Pharmacologic targeting of NOS3 with statins or another small-molecule compound ( AVE3085 ) enhanced macrophage bacterial killing , improved bacterial clearance , and increased host survival in both primary and secondary ( post-influenza ) pneumonia . The data identify a novel mechanism for host defense via NOS3 and suggest a potential therapeutic strategy to reduce secondary bacterial pneumonia after influenza . Bacterial pneumonia remains a major cause of morbidity and mortality ( Mizgerd , 2006; Shrestha et al . , 2013 ) . One approach to the problem might be to enhance innate immunity to infection . Normal host defenses are already quite robust , albeit imperfect , as they keep the incidence of pneumonia much lower than possible given the normal nocturnal aspiration of nasopharyngeal bacteria , example Streptococcus pneumoniae ( Gleeson et al . , 1997; Dockrell et al . , 2012; Donkor , 2013 ) . The resident alveolar macrophage ( AM ) functions as a ‘first responder’ phagocyte , ingesting and killing inhaled bacteria ( Green and Kass , 1964; Fels and Cohn , 1985; Hussell and Bell , 2014 ) . The importance of this function of AMs is indicated by greater susceptibility to infection and diminished bacterial clearance after their experimental depletion ( Dockrell et al . , 2003; Ghoneim and McCullers , 2013 ) or their impairment by clinical risk factors such as recent influenza infection ( Sun and Metzger , 2008 ) . To guide investigation of possible targets to improve or restore lung macrophage antibacterial function , we sought to exploit the natural experiment of gender differences in resistance to infections . Experimental models find that female mice show greater systemic resistance to pneumococci ( Weiss et al . , 1973; Kadioglu et al . , 2011 ) and to many other ( but not all ) pathogens ( McClelland and Smith , 2011 ) . Epidemiologic studies of human pneumonia observe a greater incidence of community-acquired pneumonia in males ( Gutiérrez et al . , 2006 ) and show that males are at greater risk than females for pneumonia after admission to hospital after adjusting for other risk factors such as smoking and alcohol use ( Offner et al . , 1999; Andermahr et al . , 2002; Gannon et al . , 2004 ) . We chose to address this problem by using a model of pneumococcal infection that approximates the frequent challenge to lung defenses by small numbers of bacteria and comparing responses in male and female mice . We identified greater female resistance to infection , mediated in large part by estrogen-dependent activation of constitutive AM nitric oxide synthase-3 ( NOS3 ) . Pharmacologic agents that enhance NOS3 function improved resistance in mouse models of both primary lung infection and post-influenza secondary pneumonia , suggesting a strategy to enhance resistance to common and serious lung infections . We tested effects of a relatively small bacteria inoculum size to simulate the common , low-level challenge to the lungs from aspiration of upper airway bacteria . Female mice and estrogen-treated male mice showed greater clearance of bacteria from the lungs and less acute inflammation ( neutrophil influx ) compared to normal or sham-treated males 24 hr after inoculation of S . pneumoniae ( Figure 1A , B ) . Pilot experiments compared efficiency of delivery in male vs female mice by measuring bacterial CFUs 5 min after instillation . The results showed slightly greater initial bacterial loads in female mice ( p < 0 . 02 , n = 12/group ) , indicating that the female advantage does not reflect a lower inoculum due to anatomic or size differences . Greater female resistance was also observed in longer duration survival studies ( Figure 1C ) . The gender differences we observed with 105 colony-forming units ( CFU ) were not seen if a lethal inoculum ( ∼11-fold higher ) of bacteria was used , as both genders showed markedly increased inflammation and lung cytokine levels ( Figure 1D–G ) . 10 . 7554/eLife . 03711 . 003Figure 1 . Females show greater resistance to pneumococcal pneumonia . ( A ) Twenty-four hours after intranasal ( i . n . ) inoculation of S . pneumoniae ( ∼105 CFU ) , lung samples from female mice ( and estrogen-treated male mice via subcutaneous slow-release 17-beta-estradiol pellets , ∼70 µg/day ) contain fewer live bacteria than seen in male mice ( n > 12 , * = p < 0 . 01 vs control or sham-treated males ) and ( B ) show less acute inflammation ( BAL neutrophils , n > 12 , * = p < 0 . 01 ) . ( C ) After i . n . pneumococcus , female mice show significantly greater survival than male mice ( 2 . 5 × 105 CFU , n > 24 , * = p < 0 . 01 ) . Gender differences in pneumonic inflammation are seen with low ( 4 × 105 CFU ) , but not high ( 11 × 105 ) , bacterial inocula , measured as BAL neutrophilia ( D ) or BAL cytokines TNF ( E ) , MIP-2 ( F ) , or IL-6 ( G ) , ( n > 3 , * = p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 00310 . 7554/eLife . 03711 . 004Figure 1—figure supplement 1 . Respiratory burst by male and female alveolar macrophages . Stimulation of normal AMs by antibodies to 2 different surface receptors ( FcR , CD18 ) or with PMA leads to approximately equal increases in H2O2 release in both male and female AMs , indicating absence of gender differences in production of reactive oxygen species . DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 004 To compare the innate antibacterial function of alveolar macrophages from both genders , we measured phagocytosis and killing of pneumococci in vitro by AMs from normal mice and humans . Analysis of bacterial binding and internalization showed no differences between male and female murine AMs ( Figure 2A , B; similar data with human AMs not shown ) . In contrast , killing of internalized bacteria was greater in female AMs than male AMs in mouse samples challenged with S . pneumoniae , as well as with other lung pathogens Staphylococcus aureus , and Escherichia Coli ( Figure 2C–E ) . Similarly , normal human female AMs showed greater killing of ingested pneumococci than their male counterparts ( Figure 2F ) . 10 . 7554/eLife . 03711 . 005Figure 2 . Female alveolar macrophages show better killing of ingested bacteria . Binding ( A ) and internalization ( B ) of S . pneumoniae in normal male and female AMs is similar . Female AMs kill more internalized bacteria than male AMs in assays using pneumococci ( C ) ( n > 11 , * = p < 0 . 01 ) , S . aureus ( D ) or E . coli ( E ) , ( n > 3 , * = p < 0 . 01 ) . ( F ) . Normal human female AMs also show greater killing of internalized pneumococci , ( n > 5 , * = p < 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 005 Mechanisms for killing phagocytosed bacteria include production of reactive oxygen species ( ROS ) , example , superoxide produced by NADPH oxidase , and reactive nitrogen intermediates ( RNI ) , example nitric oxide produced by nitric oxide synthases ( NOS ) . We first used mice genetically deficient in NADPH oxidase ( Morgenstern et al . , 1997 ) to more directly evaluate the importance of this pathway in vivo . Bacterial clearance in NADPH oxidase-null mice was significantly impaired , but did not eliminate the gender difference , since both male and female mice were affected approximately equally ( Figure 3A ) . Additional experiments showed no difference in generation of ROS by male and female AMs stimulated to undergo a respiratory burst in vitro ( Figure 1—figure supplement 1 ) . Similarly , quantitative real-time PCR using primers for NADPH oxidase components ( phox22 , 47 , 91 ) or myeloperoxidase showed no difference in expression ( data not shown ) . 10 . 7554/eLife . 03711 . 006Figure 3 . NOS3 and female resistance to pneumococcal pneumonia . ( A ) NADPH oxidase deficient ( phox91−/− ) mice show comparable reduction in bacterial clearance in both male and female mice ( n = 6 , * = p < 0 . 01 vs wild-type ) . ( B ) In vitro killing of pneumococci by normal mouse female AMs is inhibited by the non-selective NOS inhibitor nitro-L-arginine ( NLA ) , but not by its inactive stereo-isomer , nitro-D-arginine ( NDA ) , nor by the type 2 NOS specific inhibitor 1400W ( n = 3–4 , * = p < 0 . 01 ) . ( C ) Female AMs from Nos3−/− mice lose the in vitro killing advantage of wild-type female AMs and show the same killing rate as wild-type or NOS3 deficient male AMs ( n = 3 , * = p < 0 . 01 vs wild-type ) . ( D ) In vivo , absence of NOS3 reduces , but does not completely eliminate , the female advantage in bacterial clearance ( n = 15 , * = p<0 . 015 vs all 3 other groups ) and results in increased mortality from pneumococcal pneumonia ( E ) ( n = 12 female mice per group , * = p < 0 . 01 ) . Conversely , transgenic male mice with increased expression of human NOS3 show enhanced killing of S . pneumoniae in vivo ( F ) ( lower bacterial survival , n > 5 , * = p < 0 . 01 ) . In this low-dose inoculum model , NOS2 deletion ( G ) or inhibition ( H ) causes reduced bacterial clearance in male , but not female mice ( n = 8 , * = p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 006 To explore the potential role of RNI , we tested the effect of NOS inhibitors on bacterial killing by female murine AMs in vitro . As shown in Figure 3B , the non-selective NOS blocker nitro-l-arginine ( NLA ) caused substantial inhibition of female AM killing , while its stereoisomer controls NDA and the inducible NOS ( iNOS or NOS2 ) -specific inhibitor 1400W had no effect . This suggested a role for another NOS isoform , namely the constitutively expressed endothelial type NOS3 found in alveolar and other macrophages ( Miles et al . , 1998; van Straaten et al . , 1998; Connelly et al . , 2003 ) . This postulate was directly tested using mice genetically deficient in NOS3 . In vitro , deficiency in NOS3 did not alter the killing by male macrophages , but completely abolished the female advantage in bacterial killing ( Figure 3C ) . In vivo , absence of NOS3 had minimal effects on male clearance of bacteria , but reduced the female advantage by ∼50% ( Figure 3D ) . Genetic deletion of NOS3 greatly reduced survival in female mice with pneumonia ( Figure 3E ) , and transgenic overexpression of NOS3 increased bacterial clearance in male mice ( Figure 3F ) , supporting the functional importance of this pathway . The 24 hr in vivo bacterial clearance assay provides enough time for induction of NOS2 , which is known to mediate macrophage antimicrobial function in rodents ( Nathan and Shiloh , 2000 ) . To test the contribution of this isoform , we measured bacterial clearance in NOS2-deficient mice and in mice treated with the NOS2 inhibitor 1400W . The data show that male mice do rely on NOS2 , since clearance of pneumococci was markedly worse in NOS2-deficient or 1400W-treated mice compared to controls ( Figure 3G , H ) . In contrast , female mice showed a trend for improved clearance when Nos2 was genetically deleted ( Figure 3G ) ; similar results were seen in wild-type female mice after pharmacologic inhibition of NOS2 ( Figure 3H ) . Since NO itself can inhibit NOS3 function ( Buga et al . , 1993; Erwin et al . , 2005 ) , the data suggest that low-level basal NOS2 or NOS2 induction actually hampers the beneficial activity of NOS3 in females . To facilitate biochemical analysis of estrogen effects on macrophage function , we tested the effect of estrogen on bacterial uptake and killing by macrophage cell lines . The cell lines were cultured using hormone-free serum . After estrogen treatment , the macrophage cell lines J744A . 1 ( murine ) and U937 ( human ) showed increased killing of internalized bacteria ( Figure 4A , B ) , although no differences in binding or internalization of bacteria were observed ( data not shown ) . This increased killing capacity was inhibited by the NOS inhibitors NLA or L-NMMA , but not their stereoisomer controls . After estrogen treatment ( 0 . 2 ng/ml , 1 hr ) , we measured slight , albeit consistent , increases in nitrite production by bacteria-challenged J774A . 1 macrophages ( p < 0 . 01 , paired t-test , n = 13 ) . The increase ( ∼25 pmoles NO2− per 106 macrophages ) is consistent with the small levels attributed to NOS3 activity in normal rat AMs ( ∼70 pmoles/106 cells ( Miles et al . , 1998 ) ) and with the rate of NO release observed upon estrogen stimulation of human monocytes ( Stefano et al . , 1999 ) . 10 . 7554/eLife . 03711 . 007Figure 4 . Estrogen-mediated activation of macrophage NOS3 . Estrogen treatment of J774A . 1 mouse or human U937 macrophages ( A and B ) increases killing of ingested pneumococci; this increased killing is prevented by the NOS inhibitors NLA or l-NMMA , but not control stereoisomers ( n = 3–4 , * = p < 0 . 01 ) . ( C ) Western blot analysis shows >100-fold NOS3 in macrophages compared to the endothelial cell line bEnd . 1; after 30 min , estrogen-treated ( E2 , estradiol , 0 . 2 ng/ml ) J774A . 1 mouse macrophages show increased phosphorylation of Akt and NOS 3 , while normal female AMs show basally increased pAkt and pNOS3 compared to male AMs; ( D ) basal- and estrogen-enhanced phosphorylation of Akt and NOS3 are inhibited by wortmannin ( Wm , 50 nM ) . ( E ) Inhibition of Akt with 1L-6-hydroxymethyl-chiro-inositol 2- ( R ) -2-O-methyl-3-O-octadecylcarbonate ( 10 µg/ml REF ) prevents estrogen-mediated increased bacterial killing in J774A . 1 cells ( n = 3 , * = p < 0 . 01 ) . ( F ) Aerosol pre-treatment of male mice with albumin-conjugated estrogen 30 min before pneumococcal infection improves bacterial killing ( n = 6 , * = p < 0 . 01 ) . ( G ) In ovariectomy-model of menopause , female mice lose their greater resistance to pneumococcal pneumonia after 10 weeks , an effect reversed by treatment with estrogen prior to infection , n > 8 for control , 10 week groups; n = 3 for 2 and 5 week groups; * = p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 007 Studies of NOS3 in endothelial cell biology have identified rapid activation via phosphorylation of NOS3 at serine 1177 ( Gonzalez et al . , 2002 ) and have shown that estrogen mediates this activation via specific signaling cascades , such as phospho-inositol-3 kinase and protein kinase B , also known as Akt , ( Chambliss and Shaul , 2002; Duckles and Miller , 2010 ) . To investigate the pathway in macrophages , we cultured J774A . 1 macrophages in hormone-free serum . Western blot analysis of responses after addition of estrogen to J774A . 1 cells showed rapid phosphorylation of Akt and NOS3 in response to estrogen ( Figure 4C; fold-increase 1 . 7 + 0 . 3 , 1 . 7 + 0 . 1 , respectively , n = 4–5 , p < 0 . 01 ) and basally increased levels of pNOS3 and pAkt in normal female mouse AMs compared to male AMs ( Figure 4C , fold-increase 1 . 7 + 0 . 4 , 1 . 5 + 0 . 2 , pAkt and pNOS3 respectively , n = 2–3 , p < 0 . 01 ) . The Western blot analysis also shows that the amount of NOS3 in the macrophage cell line and primary AMs is quite low , at least two orders of magnitude lower than present in the positive control bEnd . 1 endothelial cell line samples . A role for PI3K in the signaling cascade was supported by the decrease in pAkt levels , either basally or after E2-treatment , in J774A . 1 macrophages treated with the inhibitor wortmannin ( Figure 4D; fold increase 2 . 3 + 0 . 2 , 1 . 04 + 0 . 5 , pAkt after E2 or E2 + wortmannin , respectively , n = 2–3 , p < 0 . 05 ) . To test this pathway more directly , we also used the Akt inhibitor , 1L-6-hydroxymethyl-chiro-inositol 2- ( R ) -2-O-methyl-3-O-octadecylcarbonate ( Takeuchi and Ito , 2004 ) , to test whether it would block macrophage killing of pneumococci . This agent did not alter bacterial binding or internalization ( not shown ) . As illustrated in Figure 4E , Akt inhibition completely abrogated estrogen-mediated enhanced killing , supporting a functional role for Akt in estrogen-mediated activation of macrophages . To begin testing of the translational potential of targeting NOS3 , we first tested whether local delivery of estrogen to the lungs would improve resistance to bacterial pneumonia . We found that aerosol treatment of male mice before challenge with pneumococci improved the clearance of bacteria ( Figure 4F ) . This was especially true if we used an estradiol conjugated to albumin , which acts to slow absorption of this lipophilic hormone ( systemic uptake being the basis for a form of aerosolized estrogen-replacement therapy formerly used in menopausal women [Studd et al . , 1999] ) . Populations at risk for localized outbreaks of primary pneumococcal pneumonia include elderly ( post-menopausal ) nursing home residents ( Muder , 1998 ) . We evaluated bacterial clearance in a mouse model of menopause that gradually develops several weeks after surgical ovariectomy ( Chakraborty and Gore , 2004 ) . We found a gradual decline in bacterial clearance in ovariectomized female mice to normal male levels at 10 weeks after ovariectomy; aerosolized estrogen treatment of ovariectomized mice reversed this decline and restored the normal , superior female clearance capacity ( Figure 4G ) . Statins can increase levels of NOS3 ( Forstermann and Li , 2011 ) and have been associated with beneficial effects on incidence of hospitalization for pneumonia and subsequent mortality ( Thomsen et al . , 2008; Nielsen et al . , 2012 ) . To study possible effects on innate immunity and initial resistance to pneumonia , we measured the effects of statin therapy and estrogen-replacement therapy on the incidence of hospitalization with pneumonia in a large well-defined female human population in Denmark between 1997 and 2012 . Table 1 shows the characteristics of 28 , 576 female subjects with first-time hospitalized pneumonia and 142 , 880 age-matched female population control subjects . The unadjusted OR results in Table 2 indicate that in crude analyses , estrogen users had a similar pneumonia hospitalization risk than estrogen non-users , and statin users had a slightly higher pneumonia risk than statin non-users . After controlling for comorbidity and other confounding factors associated with drug use , we found that receiving statin therapy reduced incidence of pneumonia requiring hospitalization in females . In addition , estrogen-replacement therapy was associated with a similar reduction , and women receiving both statin and estrogen therapy showed an even greater reduction in pneumonia risk ( adjusted OR 0 . 67 , 95% CI: 0 . 60–0 . 75 ) . 10 . 7554/eLife . 03711 . 008Table 1 . Characteristics of 28 , 576 female subjects with first-time hospitalized pneumonia and 142 , 880 age-matched female population control subjects from Northern Denmark , 1997–2012DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 008Characteristic at time of pneumonia admissionPneumonia casesPopulation controlsTotaln28 , 576142 , 880171 , 456Age 15–39 years2006 ( 7 . 0 ) 10 , 062 ( 7 . 0 ) 12 , 068 ( 7 . 0 ) 40–64 years6383 ( 22 . 3 ) 32 , 023 ( 22 . 4 ) 38 , 406 ( 22 . 4 ) 65–79 years9871 ( 34 . 5 ) 49 , 267 ( 34 . 5 ) 59 , 138 ( 34 . 5 ) ≥80 years10 , 316 ( 36 . 1 ) 51 , 528 ( 36 . 1 ) 61 , 844 ( 36 . 1 ) Charlson comorbidity index Index low ( 0 ) 12 , 031 ( 42 . 1 ) 100 , 297 ( 70 . 2 ) 112 , 328 ( 65 . 5 ) Index medium ( Mizgerd , 2006; Shrestha et al . , 2013 ) 11 , 324 ( 39 . 6 ) 34 , 896 ( 24 . 4 ) 46 , 220 ( 27 . 0 ) Index high ( ≥3 ) 5221 ( 18 . 3 ) 7687 ( 5 . 4 ) 12 , 908 ( 7 . 5 ) Alcoholism-related conditions1398 ( 4 . 9 ) 2225 ( 1 . 6 ) 3623 ( 2 . 1 ) Preadmission medication use Antibiotic use ( ≤3 months ) 12 , 778 ( 44 . 7 ) 20 , 002 ( 14 . 0 ) 32 , 780 ( 19 . 1 ) Statins , current use ( ≤ 6 months ) 3506 ( 12 . 3 ) 16 , 111 ( 11 . 3 ) 19 , 617 ( 11 . 4 ) Estrogen , current use ( ≤ 6 months ) 2592 ( 9 . 1 ) 13 , 225 ( 9 . 3 ) 15 , 817 ( 9 . 2 ) Statins and estrogen , current use ( ≤ 6 months ) 509 ( 1 . 8 ) 2509 ( 1 . 8 ) 3018 ( 1 . 8 ) Statins and estrogen , no current use21 , 969 ( 76 . 9 ) 111 , 035 ( 77 . 7 ) 133 , 004 ( 77 . 6 ) Marital status Married10 , 109 ( 35 . 4 ) 57 , 673 ( 40 . 4 ) 67 , 782 ( 39 . 5 ) Never married2825 ( 9 . 9 ) 13 , 523 ( 9 . 5 ) 16 , 348 ( 9 . 5 ) Divorced or widowed15 , 642 ( 54 . 7 ) 71 , 684 ( 50 . 2 ) 87 , 326 ( 50 . 9 ) 10 . 7554/eLife . 03711 . 009Table 2 . Odd ratios for first-time hospitalized pneumonia associated with current use of statins alone , estrogen alone , and estrogen + statins in combinationDOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 009Statin usePneumonia cases ( n = 28 , 576 ) Matched population controls ( n = 142 , 880 ) Crude OR ( 95% CI ) Adjusted OR ( 95% CI ) †No current use of statins or estrogen21 , 969 ( 76 . 9 ) 111 , 035 ( 77 . 7 ) 1 . 0 ( reference ) 1 . 0 ( reference ) Current use of statins ( ≤ 180 days before admission ) 3506 ( 12 . 3 ) 16 , 111 ( 11 . 3 ) 1 . 12 ( 1 . 07–1 . 16 ) 0 . 82 ( 0 . 78–0 . 85 ) Current use of estrogen ( ≤ 180 days before admission ) 2592 ( 9 . 1 ) 13 , 225 ( 9 . 3 ) 0 . 99 ( 0 . 95–1 . 04 ) 0 . 82 ( 0 . 78–0 . 86 ) Current use of both statins and estrogen ( ≤ 180 days before admission ) 509 ( 1 . 8 ) 2509 ( 1 . 8 ) 1 . 04 ( 0 . 94–1 . 15 ) 0 . 67 ( 0 . 60–0 . 75 ) Matched for age and hospitalization date . †Adjusted for level of Charlson's comorbidity index ( 19 different comorbidities ) , alcoholism-related conditions , antibiotics before admission , and marital status ( see Table 1 ) . To investigate statin effects on macrophage interaction with bacteria , we first found that in vitro treatment of macrophages for 1–3 hr with mevastatin led to increased levels of NOS3 and its phosphorylated isoform ( Figure 5A; fold-increase 2 . 0 + 0 . 2 , 1 . 7 + 0 . 6 , pNOS3 , NOS3 , 3 hr after mevastatin , respectively , n = 2–3 , p < 0 . 05 ) . This was associated with killing of internalized bacteria ( Figure 5B ) . We then measured in vivo clearance of pneumococci in mice treated with statins and found improved clearance in both genders , but this improvement was not seen when mice deficient in NOS3 were used ( Figure 5C ) . Male mice treated with statins showed improved survival with pneumococcal pneumonia ( Figure 5D ) . 10 . 7554/eLife . 03711 . 010Figure 5 . Statins enhance innate immune resistance to S . pneumoniae via NOS3 . ( A ) In vitro treatment of J774A . 1 mouse macrophages with mevastatin ( 5 µM ) increases levels of pNOS3 and NOS3 and ( B ) concomitantly increases killing of internalized bacteria ( n = 4 , * = p < 0 . 01 ) . ( C ) In vivo , pre-treatment of mice with pravastatin ( 50 mg/kg ) significantly improves bacterial clearance in wild-type mice ( n = 8 , * = p < 0 . 01 vs male controls; ** = p < 0 . 01 vs males , males + statin ) , but has no significant effect on either male or female NOS3−/− mice . ( D ) Statin-treated male mice with pneumococcal pneumonia show improved survival ( n = 8 , * = p < 0 . 01 ) . ( E ) AVE3085 , a small molecule activator of NOS3 , increases bacterial killing by mouse macrophages in vitro ( n = 3 , * = p < 0 . 01 ) ( F ) Pre-treatment of male mice with AVE3085 by either subcutaneous or oral route improves in vivo bacterial clearance , an effect not seen in NOS3−/− male mice ( n = 3–8 , * = p < 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 010 AVE3085 is a small-molecule compound that increases NOS3 mRNA and protein levels by mechanisms distinct from statins ( e . g . , by enhancing transcription ) and may provide a more NOS3-specific therapeutic option ( Wohlfart et al . , 2008 ) . We found that AVE3085 treatment improved bacterial killing by macrophages in vitro ( Figure 5E ) and that mice treated with AVE3085 either by oral or subcutaneous administration showed substantially increased clearance of pneumococci in vivo , an effect not observed in Nos3−/− mice ( Figure 5F ) . Secondary pneumococcal pneumonia remains a major problem after primary influenza ( Shrestha et al . , 2013 ) and can be modeled by infecting mice with a non-lethal dose of influenza and then challenging the lungs with pneumococci 7 days later . This is a time period of maximal susceptibility to secondary infection ( Sun and Metzger , 2008; Ghoneim and McCullers , 2013 ) , as illustrated by the remarkably low dose of pneumococci needed to cause pneumonia ( 500 CFU ) compared to the 1 , 00 , 000 CFU used in the primary pneumonia model . We evaluated the pharmacologic agents that proved effective in our primary pneumonia model for potential benefit to reduce susceptibility or severity of secondary pneumonia after influenza . We found that treatment of mice with statins substantially improved survival in a dose-dependent manner ( Figure 6A ) . We observed similar improved survival in mice treated with AVE3085 ( Figure 6B ) and also found this treatment improved bacterial clearance measured at 24 hr after challenge ( Figure 6C ) . 10 . 7554/eLife . 03711 . 011Figure 6 . Statins and AVE3085 improve survival from post-influenza secondary pneumococcal pneumonia . Male mice were allowed to recover 7 days from mild influenza ( PR8 1 PFU i . n . ) and then challenged with S . pneumoniae ( 500 CFU i . n . ) . Pre-treatment with ( A ) pravastatin ( 50 or 100 mg/kg ) or ( B ) AVE3085 ( 0 . 75 mg , s . c . ) caused a significant improvement in survival ( n = 10 , * = p < 0 . 01 ) . ( C ) AVE3085 treatment also lead to improved bacterial clearance 24 hr after pneumococcal challenge in this post-influenza model ( n = 6 , * = p < 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03711 . 011 We sought to identify new approaches to enhancing innate immunity to bacterial pneumonia by investigating the basis for gender differences in resistance to pneumococcal pneumonia . We conclude that estrogen mediates greater host resistance to pneumonia in female mice via effects on the constitutively expressed NOS3 in lung macrophages . The basis for this conclusion includes data in mouse cells or models using pharmacologic agents , inhibitors , and genetically deficient animals . A similar advantage is seen in female human primary AMs and estrogen-treated human macrophages . These observations identify a novel host defense mechanism and a function for the main NOS isoform found in normal human macrophages . Pharmacologic targeting of NOS3 with one type of drug already in clinical use ( statins ) and another small-molecule lead compound ( AVE3085 ) led to improved bacterial clearance and improved survival from secondary pneumococcal pneumonia . The mechanism ( s ) by which macrophage NOS3 contributes to improved killing of ingested bacteria include direct bactericidal mechanisms , either alone or in combination with superoxide ( Fang , 2004 ) , but NOS3 might also act to mediate salutary intracellular signaling events ( Hernansanz-Agustín et al . , 2013 ) . The very low amounts of NO generated by macrophage NOS3 prompt the speculation that additional mechanisms act to amplify its contribution to bacterial killing . These may include increased trafficking of NOS3 after estrogen-mediated phosphorylation to phagosomes , selective triggering of calcium-dependent NOS3 production of NO around bacteria by phagocytosis-associated calcium fluxes , and combination of NO with superoxide to generate peroxynitrite , a potent microbicidal agent ( Darrah et al . , 2000 ) . These possibilities are not mutually exclusive . The data also show an interesting but unresolved finding of incomplete concordance between in vitro and in vivo results using NOS3-deficient mice . Specifically , NOS3 deficiency eliminated the female advantage entirely in studies of bacterial killing by AMs in vitro . In contrast , the female advantage in vivo was diminished by approximately half in NOS3-deficient mice . The mechanism ( s ) and significance of this persistent component of gender difference in mice are unknown and need further study . Another aspect of our study that merits discussion is our use of a low-dose inoculum for the primary pneumonia model . This is intended to model the initial phase of pneumonia that follows the most likely route: aspiration of nasopharyngeal pneumococci or other bacteria . However , the quantity and composition of bacteria that enter the lungs during nocturnal aspiration is unknown . Moreover , while the premise that this is how pneumonia starts in humans is logical , there exists no formal proof . Ultimately , whether this attempt at a pre-clinical model proves relevant will depend on how the results translate to the clinical setting . The potential relevance of these findings to human biology is indeed supported by the epidemiologic findings that both statins and estrogen-replacement therapy reduced risk of lung infection in a large cohort of women . There are also interesting reports of association of NOS3 polymorphisms with risk of community-acquired pneumonia ( Salnikova et al . , 2013 ) . However , we consider that the translational potential of these findings may be most relevant for the problem of secondary pneumonia , specifically the increased susceptibility to bacterial lung infections that follows influenza . Using pre-treatment before the bacterial challenge is arguably a limitation of our studies in the primary pneumonia model; this approach is useful for proof-of-principle but does not mirror likely clinical usage . In contrast , pre-treatment in the secondary pneumonia model does represent a realistic scenario for short-term , prophylactic immunomodulatory therapy . This approach could boost resistance to prevent bacterial pneumonias during intervals of increased susceptibility . This could well benefit patients who are identified to be at high risk of secondary pneumonia , example hospitalized individuals with severe seasonal or pandemic influenza . This study began by seeking to characterize the basis for greater female resistance to pneumococcal pneumonia . A murine model of pneumococcal pneumonia was used to compare male and female bacterial clearance , lung inflammation , and survival in vivo . In vitro assays of macrophage antibacterial function were used to identify killing of internalized bacteria as a critical difference . Pharmacologic inhibitors , genetic mouse models , and in vitro macrophage studies were used to investigate the role of oxidant-generating enzyme systems and estrogen . After identification of NOS3 as an important mediator , we tested pharmacologic agents with translational potential for ability to improve outcomes in primary and post-influenza secondary pneumococcal pneumonia . For in vivo studies , we used male and female mice of the same age , and the studies were not blinded . Sample sizes for all of the experiments were sufficient to detect statistically significant differences between treatment groups , with all measurements included in analyses . To explore whether the effects of estrogen and statins that were observed in the murine models also occur in people , we studied incidence of hospitalized pneumonia associated with the use of estrogen and statins in a population-based case-control study based on medical databases in Denmark . Cases of pneumonia were identified as all women who received a first-time principal hospital diagnosis of pneumonia in the former North Jutland and Aarhus Counties , Northern Denmark ( 1 . 2 million inhabitants ) between 1997 and 2012 . Using the Danish Civil Registration System , each case subject was matched with five population control subjects with same age , female gender , and residence in Northern Denmark on the pneumonia index date . We ascertained use of estrogen and statins in all individuals from population-based prescription databases . To control for confounding by other conditions potentially associated with both estrogen and statin use and pneumonia risk , we retrieved individual-level data on 19 major comorbid disease categories as evidenced in the Charlson comorbidity index , as well as on alcoholism-related conditions , recent antibiotic use , and marital status as a marker of socioeconomic status . We then computed odds ratios ( ORs ) for a first-time pneumonia admission among women with and without estrogen and statin use , using conditional logistic regression analysis to control for confounders . Normal eight- to 12-week old male and female mice C57BL/6 mice from Charles River Laboratories ( Wilmington , MA ) or Jackson Laboratories ( Bar Harbor , ME ) were used . C57BL/6 mice genetically deficient in nitric oxide synthase 2 ( Laubach et al . , 1995 ) , nitric oxide synthase 3 ( Shesely et al . , 1996 ) , and NADPH oxidase ( Morgenstern et al . , 1997 ) were obtained from Jackson Laboratories ( Bar Harbor , ME ) . Mice transgenic for human NOS3 were previously characterized ( Jones et al . , 2003 ) . All animals were housed in sterile microisolator cages in a barrier facility and had no evidence of spontaneous infection . Prior approval for all experimentation was obtained from the institutional animal use review committee . Primary pneumococcal pneumonia was modeled as previously reported ( Arredouani et al . , 2004 ) . Pneumonia was induced by intranasal instillation ( i . n . ) of 25 µl of a bacterial suspension containing approximately 105 colony-forming units ( CFU ) of S . pneumoniae type 3 of mice under short-term anesthesia with inhaled isoflurane . The bacterial suspension was prepared to contain 105 CFU based on optical density calculations . However , results of direct validation of actual CFU in these samples were always performed and showed actual CFU values that ranged from 90–110 % of the calculated 105 CFU value . Hence , the dose delivered is stated as ∼105 CFU . For analysis at 4 or 24 hr post-infection , mice were sacrificed with/by lethal overdose of intraperitoneal sodium pentobarbital ( FatalPlus , Vortech Pharmaceuticals , Dearborn , MI ) or by excess inhaled isoflurane . To measure total lung bacteria counts ( CFU ) , whole lungs were harvested and homogenized in 1 ml sterile water with a tissue homogenizer ( Omni International , Warrenton , VA ) . Serial 10-fold dilutions in sterile water were made from these homogenates , and 100 µl volumes were plated onto sheep-blood agar plates and incubated at 37°C . CFUs were counted after 18–20 hr . In experiments to assess survival , mice were instilled i . n . with a single 25 µl suspension containing a higher dose of S . pneumoniae ( ∼3 × 105 CFU ) and survival followed as reported in ‘Results’ . For analysis of lung inflammation , bronchoalveolar lavage was performed in situ with a 22-gauge catheter inserted into the proximal trachea , flushing the lower airways six times with 0 . 7 ml of phosphate-buffered saline ( PBS ) . The fluid retrieved from the first lavage was kept for ELISA assays . The BAL cells were separated from the BAL fluid by centrifugation , resuspended in PBS and counted . A fraction was cytospun on microscopic slides for staining with Diff-Quick ( Baxter Scientific Products , McGaw Park , IL ) for subsequent differential counts . In initial experiments , male mice were analyzed 20 days after subcutaneous implantation according to manufacturer's instructions of slow-release 17-beta-estradiol pellets ( ∼70 µg/day , Innovative Research , Sarasota , FL ) . In later experiments , mice received a single dose of estradiol by inhalation exposure for 1 hr to an aerosol generated from a solution of E2-BSA β-Estradiol 6- ( O-carboxymethyl ) oxime: BSA ( 75 ng/ml aerosol solution , ∼30 mol steroid per mol BSA , Sigma ) using the mouse exposure system described in Hamada et al . , 2003 . Drug treatments to assess effects on 24 hr bacterial clearance were administered 4 hr before and 8 and 20 hr after bacterial challenge included i . p . pravastatin 50–100 mg/kg , and 1400W 10 mg/kg . The NOS3 activator AVE3085 ( 2 , 2-difluoro-benzo[1 , 3]dioxole-5-carboxylic acid indan-2-ylamide , CAS no . 450348-85-3; empirical formula C17H13F2NO3; ( Wohlfart et al . , 2008 ) ) was administered by gavage at 100 mg/kg/day for 3 days prior to the challenge or by subcutaneous injection of 0 . 75 mg 24 and 3 hr before the challenge . To model post-influenza secondary pneumonia , mice were instilled i . n . on day 0 with a single 25 µl or 50 µl suspension containing influenza virus ( A/PR 8/34; H1N1 ) , under general anesthesia by i . p . ketamine ( 120 mg/kg ) plus xylazine ( 16 mg/kg ) . Influenza-treated animals routinely lost weight and then recovered by day 7 , when the secondary infection was administered . On day 7 after initial influenza infection , mice were subjected to the same anesthesia and then instilled i . n . with a single 25 µl of bacterial suspension containing approximately 500 CFU of S . pneumoniae type 3 . Pravastatin ( 50–100 mg/kg i . p . ) or AVE3085 ( 0 . 75 mg s . c . ) was administered daily starting one day before the pneumococcal challenge . Subsequent analyses were performed as for the primary pneumonia model described above . After euthanasia , mouse AMs obtained by repeated lung lavage with sterile PBS were centrifuged at 150×g and resuspended in HBSS+ ( Hanks' Buffered Salt Solution with calcium , 0 . 3 mM , and magnesium , 1 mM ) . Primary mouse AMs were used immediately without exposure to culture media components that contain estrogen ( e . g . serum ) or ( weakly ) estrogenic phenol red . The mouse and human macrophage cell lines J774A . 1 and U937 were obtained from ATCC and maintained in phenol-free RPMI with 10% FBS . In experiments testing effects of estrogen addition in vitro , a charcoal-stripped FBS ( Hyclone , Logan , UT ) was used for macrophage culture to eliminate exposure to serum hormones . Human alveolar macrophages were obtained from non-smoking volunteers by bronchoalveolar lavage and resuspended in HBSS+ when tested immediately or cultured in phenol-free RPMI with 10% fetal bovine serum ( FBS ) overnight . The potential confounding effects of estrogen in FBS were not observed in human AM samples cultured overnight , that is gender differences observed in cells used immediately after isolation were preserved . All human subject experimentation was conducted under approved protocols reviewed by the institutional review boards . Estrogen was added to macrophage cultures as β-Estradiol-Water Soluble ( Sigma ) at 0 . 05–0 . 3 ng/ml . All other reagents not otherwise specified were obtained from Sigma Chemical , St . Louis , MO . S . pneumoniae serotype 3 , E . Coli , S . aureus ( #6303 , 19138 , 25923 , respectively , ATCC , Rockville , MD ) were cultured overnight on 5% sheep blood-supplemented agar Petri dishes ( VWR # 90001-282 , West Chester , PA ) . A stock suspension was prepared and aliquots kept at −80°C . For each experiment , an aliquot was grown overnight on a blood agar plate and resuspended in sterile PBS . Bacterial concentration of the obtained suspension was estimated by OD600 measurements , comparing to a prior standard curve of OD600 vs CFU . The appropriate dilution was prepared in sterile PBS to be administered to mice , and this estimate was checked in parallel by CFU assay to determine the precise concentration . Macrophage binding , internalization , and killing of internalized bacteria were measured using CFU assays of cell samples lysed by 10-fold excess H2O ( pH 10 , 3 min ) . Macrophages ( 1 . 5 × 106 cells/1 . 5 ml HBSS+ ) were mixed with bacteria ( 15 × 106 CFU: 10 bacteria per 1 macrophage ) for 1 hr at 37°C . After centrifugation the cell pellet was re-suspended and an aliquot taken to measure the total cell-associated CFU ( bound and internalized ) . After brief incubation of the cell suspension with gentamicin ( 200 µg/ml , 15 min at 37°C ) to kill external , bound bacteria , the macrophages were washed and an aliquot taken for lysis and CFU assay to quantitate the number of internalized , live bacteria . The macrophages were then incubated an additional hour to allow killing of internalized bacteria and aliquots taken again for CFU quantitation . The CFU data obtained at various time points allow calculation of the number of bacteria bound , internalized , and killed . Effects of pharmacologic inhibitors or hormones were tested by including agents or their vehicles in the assay buffers . Respiratory burst function in normal mouse AMs was measured by quantitation of the H2O2-catalyzed oxidation of Amplex Red ( Molecular Probes , Eugene , OR ) to a fluorescent product after stimulation with adsorbed antibodies ( anti-FcR , CD18 ) or PMA ( 100 nM ) as previously described ( Józefowski and Kobzik , 2004 ) . Western blot analysis was performed on macrophage lysates lysis buffer 1% NP-40 with protease and phosphatase inhibitors using the protocols described in Gonzalez et al . , 2002 . The endothelial cell line bEnd . 3 ( ATCC CRL-2299 ) was used as a positive control for ( ‘endothelial’ ) NOS3 . Antibodies for NOS3 , Akt , and phosphorylation-state specific isoforms used include: NOS3 rabbit pAb ( C-20 , Santa Cruz Biotechnology ) and phospho-eNOS ( Ser1177 , C9C3 ) rabbit mAb , phospho-Akt ( Ser473 , D9E ) rabbit mAb , Akt rabbit pAb , and Beta-Actin ( 13E5 ) rabbit mAb , all from Cell Signaling Technologies . After incubation with peroxidase-conjugated goat anti-rabbit IgG or goat anti-mouse IgG ( Pierce ) , labeling was detected using chemiluminescence . Quantitation of digitized signal intensity data was performed using ImageJ software ( http://imagej . nih . gov ) . Cytokines were quantitated in BAL fluid using commercially available ELISA kits ( R&D Systems Inc . , Minneapolis , MN ) following the manufacturer's instructions .	Pneumonia is a disease that is commonly caused by a bacterial infection and results in the lungs becoming inflamed . Pneumonia is a serious condition and can lead to hospitalization and sometimes death . However , women—and other female animals—are less likely than males to get pneumonia and are more likely to survive if they do . Understanding this sex-based difference may help to develop treatments or preventive actions that either reduce the number of people who get pneumonia or help infected patients to recover . Bacteria from the nose—including those that cause pneumonia—frequently enter the lungs during sleep . Luckily , the body has very robust defense mechanisms against such invasions; the immune system immediately deploys cells called macrophages as a ‘first response’ to devour and kill invading bacteria in the lungs . However , this system is not perfect , particularly if an individual has a weakened immune system or if they are already suffering with a respiratory infection . Indeed , many individuals with severe influenza infections are hospitalized as a result of pneumonia . Yang et al . studied why females are more able to fend off pneumonia and found that estrogen , the main female sex hormone , boosts the ability of the macrophages to kill bacteria . Treating male mice with estrogen also boosted their immune system's ability to kill off bacteria in the lungs . Investigating further , Yang et al . found that the estrogen worked by increasing the number of proteins produced from one gene called NOS3 . Female mice lacking NOS3 proteins lost their pneumonia-fighting advantage . A widely used class of drugs called statins , which are used to treat cardiovascular disease , boosts the activity of the NOS3 gene . Yang et al . therefore wondered whether treatment with either estrogen or statins might prevent pneumonia , or help patients with pneumonia fight off the infection . Using a large database of information about healthcare in Denmark , Yang et al . assessed the relationship between taking these drugs and the risk of pneumonia . When several confounding factors ( such as unrelated diseases that the patient was suffering from ) are taken into account , the data show that the women were less likely to be hospitalized for pneumonia if they were taking statins or estrogens . Those taking both treatments had an even lower risk . Yang et al . also found that treating mice with statins or an experimental drug that boosts NOS3 activity increased the ability of the animals to fight off pneumonia-causing bacteria—even if they also had influenza—and increased the likelihood that mice already infected with pneumonia would survive . Further studies will be needed to determine if statins or the experimental drug might also help to prevent pneumonia in human patients with influenza .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "immunology", "and", "inflammation" ]	2014	Female resistance to pneumonia identifies lung macrophage nitric oxide synthase-3 as a therapeutic target
Heterogeneity broadly exists in various cell types both during development and at homeostasis . Investigating heterogeneity is crucial for comprehensively understanding the complexity of ontogeny , dynamics , and function of specific cell types . Traditional bulk-labeling techniques are incompetent to dissect heterogeneity within cell population , while the new single-cell lineage tracing methodologies invented in the last decade can hardly achieve high-fidelity single-cell labeling and long-term in-vivo observation simultaneously . In this work , we developed a high-precision infrared laser-evoked gene operator heat-shock system , which uses laser-induced CreERT2 combined with loxP-DsRedx-loxP-GFP reporter to achieve precise single-cell labeling and tracing . In vivo study indicated that this system can precisely label single cell in brain , muscle and hematopoietic system in zebrafish embryo . Using this system , we traced the hematopoietic potential of hemogenic endothelium ( HE ) in the posterior blood island ( PBI ) of zebrafish embryo and found that HEs in the PBI are heterogeneous , which contains at least myeloid unipotent and myeloid-lymphoid bipotent subtypes . Since new cells are generated from pre-existing cells , the frequently asked questions are what progenies are generated from certain pre-existing cells and how they contribute to the organism ( Child , 1906; Hoppe et al . , 2014; Spanjaard and Junker , 2017 ) . To address these questions , fate mapping has been widely used as a crucial methodology to identify progenies of the targeted cells and to trace their location , differentiation and functional dynamics ( Kretzschmar and Watt , 2012 ) . Hematopoiesis , the process of forming blood cells ( Dharampuriya et al . , 2017 ) , is an outstanding paradigm for studying these issues in different animal models ( Höfer et al . , 2016; Jagannathan-Bogdan and Zon , 2013; Orkin and Zon , 2008 ) . Zebrafish has natural advantages on hematopoietic fate mapping owing to the external development , transparent embryo body and its highly conserved hematopoiesis ( Jagannathan-Bogdan and Zon , 2013; Stachura and Traver , 2016 ) . Taking these advantages , permanent genetic marking , photo-convertible labeling and in vivo time-lapse imaging have been employed to monitor the generation , mobilization and lineage specification of hematopoietic stem/progenitor cells ( HSPCs ) in zebrafish ( Murayama et al . , 2006; Jin et al . , 2007; Bertrand et al . , 2010; Kissa and Herbomel , 2010 ) . Recently , the infrared laser-evoked gene operator ( IR-LEGO ) microscope heating system ( Deguchi , 2009; Kamei et al . , 2009 ) has been demonstrated as a powerful tool for bulk cell tracing with high temporal-spatial resolution ( Shimada et al . , 2013; Okuyama et al . , 2013; Xu et al . , 2015; Tian et al . , 2017; Singhal and Shaham , 2017; He et al . , 2018; Henninger et al . , 2017 ) . In this system , an infrared ( IR ) laser is used to generate local heat shock to induce CreER expression in a restricted region of transgenic fish carrying a tissue-specific loxP-DsRedx-loxP-GFP reporter and a hsp70l:mCherry-T2a-CreERT2 . The removal of DsRedx cassette is permanently inherited so that the progenies derived from the targeted tissue will display GFP instead of DsRedx ( Xu et al . , 2015; Tian et al . , 2017; He et al . , 2018 ) . However , it has been realized that heterogeneity broadly exists in multiple cell populations during hematopoiesis ( Tian et al . , 2017; Wilson et al . , 2015; Chen et al . , 2011; Crisan and Dzierzak , 2016; Ye et al . , 2017 ) . The dissection of heterogeneity requires a lineage tracing strategy with single cell resolution . Yet , the previous IR-LEGO techniques ( Deguchi , 2009; Kamei et al . , 2009; Shimada et al . , 2013; Okuyama et al . , 2013; Xu et al . , 2015; Tian et al . , 2017; Singhal and Shaham , 2017; He et al . , 2018; Kawasumi‐Kita , 2015; Suzuki et al . , 2014; Eiji , 2013; Hayashi et al . , 2014; Miao and Hayashi , 2015; Hasugata et al . , 2018 ) face the following fundamental challenges: ( 1 ) high-precision and efficient labeling of the targeted single cell; ( 2 ) fine balance between labeling efficiency and cell viability after heat shock treatment; ( 3 ) permanent marking and long-term tracing of all progeny of the labeled single cell; ( 4 ) rigorous statistical analysis to quantitatively determine the lineages of the labeled single cell under random basal interference . These challenges hamper the wider application of IR-LEGO for cell fate mapping , and it remains unclear whether this technique can indeed be used for long-term tracing of multiple lineages of a single multipotent cell , such as HSPC . Besides IR-LEGO technique , other single-cell lineage tracing methodologies invented in past decade suffer from similar problems . Cell barcode techniques , either by retroviral library infection to insert inheritable DNA barcodes ( Naik et al . , 2014 ) , or by CRISPR/Cas9 system to accumulate random mutations ( McKenna et al . , 2016; Kalhor et al . , 2017 ) , have been used to perform single-cell lineage tracing in hematopoiesis studies ( Gerrits et al . , 2010; Lu et al . , 2011 ) . Single-cell RNA-sequencing is also a prevalent way to depict lineage hierarchy ( Hoppe et al . , 2014; Zhou et al . , 2016; Athanasiadis et al . , 2017 ) . However , these non-imaging based techniques are not suitable for tracing the dynamic behaviors of the targeted cells and their progenies . A multicolor strategy , which stochastically expresses multiple fluorescent reporters in target cells via Cre-mediated recombination , forming dozens of different color modes to distinguish individual cells and their progenies , has been utilized for cell fate mapping ( Livet et al . , 2007; Cai et al . , 2013 ) . Despite its success in some zebrafish hematopoiesis study ( Henninger et al . , 2017; Pan et al . , 2013 ) , this multicolor labeling of cells makes it difficult to directly visualize the development of individual cell lineage . In addition , the lineage hierarchy could be misinterpreted when unrelated cells share the same color or Cre-mediated recombination occurs in the daughter cells . Likewise , photo-convertible protein or caged fluorescent dyes approaches also have limitation for long-term tracing due to the self-degradation and rapid dilution of fluorophores during cell division ( Tian et al . , 2017; Warga et al . , 2009 ) . An optical uncaging method was used to label a targeted cell and its progeny in zebrafish through Cre-based gene recombination ( Sinha et al . , 2010a; Tekeli et al . , 2016 ) . However , the basal uncaging level of the caged compound in zebrafish embryo was as high as 28% ( Sinha et al . , 2010b ) , limiting the application of this technique to trace the long-term development of highly dynamic cells , such as stem cells . Thus , developing a highly precise single-cell labeling method for the long-term in vivo tracing of individual cells will be important for understanding the heterogeneity of HSPCs . To overcome the drawbacks of existing techniques , we develop a high-precision single-cell IR-LEGO technology , in which a two-photon fluorescent thermometer is utilized to measure the temperature rise in vivo to achieve precise single-cell labeling . Using this tool , we document that the hemogenic endothelium ( HE ) cells in the posterior blood island ( PBI ) of zebrafish are heterogeneous in terms of hematopoietic potential . Our study demonstrates that the high-precision single-cell IR-LEGO technology has outstanding capacity to perform single-cell labeling and long-term in-vivo lineage tracing . A 1 , 342 nm diode-pumped solid-state ( DPSS ) IR laser is used as the heat-shock light source in our single-cell IR-LEGO heat-shock microscope system ( Figure 1A ) . The laser at this wavelength provides an appropriate balance between the absorption efficiency of water and penetration depth in tissue ( Appendix 1 ) . The IR laser is integrated with a two-photon microscope , and it is guided by the two-photon fluorescence imaging to heat the targeted cell . A water-immersion objective with a large numerical aperture ( NA ) is used to generate highly localized and stable laser heating at different depths in tissues . The numerical simulation ( Figure 1—figure supplement 1 and Appendix 2 ) shows that heat shock of high spatial resolution can be achieved by generating a point heat source inside tissue , a medium of relatively poor thermal conductivity . A large temperature gradient can be created in the region of about 10 µm size around the heat source ( Figure 1—figure supplement 1D ) , suggesting that the thermal energy produced by a highly focused IR laser heating in tissue could be confined within the single-cell dimensions for efficient single-cell gene induction . Since the heat shock efficiency varies as the type and location of targeted cells , it is of great significance to develop a reliable method that can objectively determine the optimal IR laser heating conditions for single-cell gene induction . Although previous studies demonstrated that temperature-sensitive fluorescent proteins , such as GFP and mCherry , can be used as thermometers to estimate the temperature rise in cells induced by IR laser irradiation ( Kamei et al . , 2009; Singhal and Shaham , 2017 ) , this single-molecular/one-color thermometry has been shown to produce significant errors , likely because of the fluctuation of excitation laser power , or the interference of complex microenvironment on signal intensities of fluorescence emission ( Estrada-Pérez et al . , 2011 ) . In order to precisely characterize the heat diffusion from the highly focused IR laser , we developed a two-photon fluorescent thermometry ( TPFT ) technique to measure the three-dimensional ( 3D ) distribution of temperature rise in the region close to the laser focal point in water , 3% agarose ( a tissue phantom of thermal conductivity similar to typical tissues ) ( Huang et al . , 2004 ) and live zebrafish , respectively . The thermometry measures the temperature rise in tissues noninvasively based on the fluorescence signals of two fluorescent dyes ( Appendix 3 ) ( Estrada-Pérez et al . , 2011; Natrajan and Christensen , 2008 ) . In details , a temperature-sensitive dye ( Figure 1 ) , tetramethylrhodamine ( TAMRA ) which is conjugated with dextran , is adopted as the probe dye in TPFT , while fluorescein ( FITC ) , which is insensitive to temperature and also conjugated with dextran ( Figure 1C ) , is used as a reference dye to eliminate the fluctuation of probe dye fluorescence caused by a variety of interferences ( Appendix 3 ) . The temperature dependencies of fluorescence measured in pure TAMRA and FITC solutions are -0 . 882 ± 0 . 100%/°C and -0 . 165 ± 0 . 098%/°C respectively ( Figure 1—figure supplement 2A and B , Figure 1—figure supplement 2—source data 1 ) . The two-photon excited fluorescence ( TPEF ) intensity ratio of TAMRA and FITC is linearly correlated with the solution temperature due to large difference in temperature sensitivity between two dyes ( Figure 1D and Figure 1—figure supplement 2 ) . The temperature sensitivities of fluorescence intensity ratio are similar in water solution , 3% agarose and zebrafish in vivo ( Figure 1D and Supplementary file 1a ) , indicating that the temperature coefficient of the fluorescent dextran remains stable in different environments . The high consistency between the actual temperature and the measured temperature ( Figure 1E ) demonstrates that TPFT can be used as an effective tool for in vivo measurement of the local temperature rise induced by IR laser heating in tissues . To study the dynamic change of temperature during IR laser heating , firstly we used TPFT to measure the temperature in water solution and 3% agarose with point heating . Low fluorescent dye concentrations were used to avoid self-absorption and fluorescence resonance energy transfer ( FRET ) ( Figure 2—figure supplement 1 and Appendix 4 ) . Using a high-sensitivity EMCCD as the spectra detector , the dynamic temperature change at the heating site can be recorded in real time . We found that the temperature at IR laser focal point increased sharply within 1 ~ 2 s after point heating and remained stable over the exposure time of the IR laser , before decreasing quickly to the ambient temperature as soon as the IR laser was turned off ( Figure 2—figure supplement 2A ) . The depth of IR laser focal point in water and tissue phantom should be over 100 µm to minimize the thermal conduction at the intermedium surface ( Figure 2—figure supplement 2B and C ) . Next , we measured the 3D temperature distributions ( Figure 2A and B , Figure 2—source data 1 ) , and found that high spatial resolution of heat shock could be achieved in 3% agarose . However , a large temperature gradient could not be built in water solution because of its high thermal conductivity and faster convection . Further , we conducted 3D temperature measurement in zebrafish in vivo and the results were compared with the measurement in agarose tissue phantom ( Figure 2C and D , Figure 2—source data 1 ) . Dextran-conjugated TAMRA and FITC were co-injected into fish embryos at one-cell stage . Then the embryos were raised to 1 day post fertilization ( dpf ) . Muscle was chosen as the first tissue to perform temperature measurement because of its relatively uniform structure and simple microenvironment . To avoid laser-induced injury ( Figure 2—figure supplement 3 ) , we applied scan heating on zebrafish muscle and tissue phantom , in which the focused IR laser beam was scanned over an 8 µm × 8 µm region for 32 s instead of staying at a fixed heating point . Results showed that the thermal confinement in zebrafish muscle is higher than in tissue phantom , both laterally and axially ( Figure 2C and D , Figure 2—source data 1 ) . This indicates that the thermal conductivity of zebrafish muscle could be lower than that of 3% agarose . To visualize the thermal confinement clearly , we generated 3D view of the lateral temperature distributions based on the experimentally measured data ( Figure 2E ) . As shown in Figure 2 E1 and E2 , low thermal conductivity and inefficient convection of tissue phantom plays a critical role to confine the thermal energy and achieve single-cell resolution of heat shock . As shown in Figure 2 E2 and E3 , there is no significant difference in the thermal confinement between point and scan heating methods . The results in Figure 2 E3 and E4 demonstrate that TPFT can finely evaluate the thermal distribution in zebrafish muscle in vivo and paves the way for evaluation of single-cell heat shock in different zebrafish tissues . Next , we examined the single-cell labeling efficiency in vivo through the single-cell IR-LEGO system in various kinds of cells , including myocytes in the skeletal muscle , neurons in the brain , and coro1a+ leukocytes in the hematopoietic tissue at the aorta-gonad-mesonephros ( AGM ) and PBI of transgenic zebrafish ( Figure 2F–H ) . Using TPFT , the temperature distribution in the heat shocked tissues was measured . With high IR laser power , the average temperature at the focal point ( P00 ) of IR laser can reach as high as 50°C ( Figure 2—figure supplement 4A , Figure 2—figure supplement 4—source data 1 ) . Although considerable success rates of overall cell labeling can be achieved with this high temperature heat shock , the percentages of single-cell labeling within the overall labeling are relatively low ( Supplementary file 1b ) . This is due to that effective heat shock gene induction can be induced with environmental temperature higher than 38°C ( Shoji and Sato-Maeda , 2008 ) and the heating region of temperature over 38°C was greater than the single cell size . Therefore , in order to increase the efficiency of single-cell labeling , the IR laser power was then optimized to restrain the heat diffusion and limit the effective area of gene induction in a single-cell dimension . With the optimized heat shock condition , the temperature at 10 µm away from the focal point ( P10 ) dropped below 38°C ( Figure 2—figure supplement 4B and C , Figure 2—figure supplement 4—source data 1 ) , preventing unwanted gene induction in neighboring cells . It was demonstrated that , after heat shock with optimized laser condition , successful overall cell labeling in myocytes , neurons and leukocytes can be observed in 36 . 7% , 18 . 6% and 50% of zebrafish , respectively , among which the efficiencies of single-cell labeling for these three types of tissues are 54 . 5% , 100% and 77 . 8% , respectively ( Figure 2—figure supplement 4D and Supplementary file 1b ) . Compared with the high-temperature heat shock , the success rates of overall cell labeling with optimized heat shock condition were decreased , but the efficiency of single-cell labeling was significantly improved . This demonstrates that the single-cell IR-LEGO technology can efficiently and precisely induce heat shock-mediated gene editing within single cell in vivo in multiple tissues under optimized condition . Similar to mammals , the definitive hematopoiesis of zebrafish initiates from the HE in the ventral wall of the dorsal aorta and was thought to give rise to hematopoietic stem cells ( HSCs ) ( Bertrand et al . , 2010; Kissa and Herbomel , 2010; Tian et al . , 2017 ) . Yet , our recent study has shown that in addition to generating HSCs , the HEs in the aorta also produces non-HSC progenitors capable of differentiating into T cells , myeloid and erythroid lineages but not B cells in a transient manner ( Tian et al . , 2017; Bertrand et al . , 2007 ) , highlighting the complexity of endothelial-hematopoietic transition ( EHT ) , a process leading to the formation of blood stem and progenitor cells from the endothelium ( Bertrand et al . , 2010; Kissa and Herbomel , 2010 ) . An important unsolved issue is whether these non-HSC-derived hematopoietic lineages , such as T lymphocytes , myeloid and erythroid cells , arise directly from distinct HE subpopulations that differentiate into different hematopoietic lineages independently or from a uniform HE population , which generates a common progenitor that further differentiates into multiple hematopoietic lineages . To address the issue , we applied the high-precision single-cell IR-LEGO technology to single HE lineage tracing in the PBI region where all three non-HSC-derived hematopoietic lineages but not HSCs are generated ( Tian et al . , 2017; Bertrand et al . , 2007 ) . Specifically , we estimated the physical sizes of HEs by measuring the distance between the nuclei of neighboring endothelial cells , and it shows that the average length of endothelial cells along the aortic floor in the PBI is 24 . 9 µm , with the minimum of 11 . 2 µm ( Figure 3—figure supplement 1 , Figure 3—figure supplement 1—source data 1 ) . Given that we have successfully constrained the effective heat-shock region ( >38°C ) within 10 µm along the ventral wall of aorta ( Figure 2—figure supplement 4B and Supplementary file 1b ) , it is highly feasible to label HE at single-cell resolution using our heat-shock microscope system . In order to label HEs and follow their fates , we generated a double transgenic Tg ( kdrl:loxP-DsRedx-loxP-EGFP;coro1a:loxP-DsRedx-loxP-EGFP ) fish , in which editable genetic reporter loxP-DsRedx-loxP-EGFP is under the control of endothelial-specific kdrl promoter ( Jin et al . , 2005 ) and leukocyte-specific coro1a promoter ( Li et al . , 2012 ) , thus HEs and leukocytes ( including myeloid and lymphoid cells ) were marked by DsRedx ( Figure 3A ) . The double reporter transgenic line was then outcrossed with Tg ( hsp70l:mCherry-T2a-CreERT2 ) fish to obtain a triple transgenic Tg ( kdrl:loxP-DsRedx-loxP-EGFP;coro1a:loxP-DsRedx-loxP-EGFP;hsp70l:mCherry-T2a-CreERT2 ) line ( referred to as ‘triple Tg’ hereinafter ) ( Figure 3A ) . In this triple Tg embryo , single-cell IR-LEGO system-induced heat-shock and 4-OH tamoxifen ( 4-OHT ) treatment would induce and subsequently activate CreER within one HE , resulting in the excision of the DsRedx cassettes . As a consequence , the targeted HE and its leukocyte progenies would be distinguished from the unlabeled cells by EGFP expression ( Figure 3A ) . To ensure only one single HE was labeled in each embryo , we irradiated the 4-OHT treated embryos at 26–28 hpf ( Figure 3B ) prior to the initiation of EHT ( Kissa and Herbomel , 2010; Tian et al . , 2017 ) , then immediately imaged the heat-shocked embryos and control embryos continuously to 48 hpf ( Figure 3B and C , Videos 1 and 2 ) . As shown in Figure 3C and Video 1 , GFP signal began to emerge at ~6 hrs post heat-shock and 4-OHT treatment . The embryos with single GFP+ HE during 20 hrs of time-lapse imaging were then selected for hematopoietic lineage analysis and the contribution of the labeled HE to T lymphocytes and myeloid cells in each of these embryos were determined by counting the coro1a:GFP+ cells at 7 dpf ( Supplementary file 1c and 1d ) . Because T lymphocytes are strictly located in the thymus at 7 dpf , the GFP signals in thymus were analyzed to determine the numbers of T lymphocyte derived from the labeled HE . Whole-mount fluorescent in situ hybridization and antibody staining verified that the GFP+ cells in the thymus were indeed rag1+ T lymphocytes ( Figure 3—figure supplement 2A , B ) . Additionally , co-labeling of thymic epithelium and different hematopoietic cell types using specific transgenic zebrafish lines exhibited that T lymphocytes can also be effectively differentiated from other thymus-resident cells based on their small and round shapes ( Figure 3—figure supplement 2C–F ) . Therefore , the numbers of GFP+ T lymphocyte in thymus can be accurately counted as the contribution of the labeled HE cell to lymphoid lineage ( Figure 3D ) . On the other hand , the GFP+ cells distributed on the embryonic trunk are lyz+ or mpeg1+ myeloid cells ( Figure 3D , Figure 3—figure supplement 3 , Figure 3—figure supplement 3—source data 1 ) . To avoid the disturbance by immature hematopoietic progenitors , GFP+ cells in the hematopoietic tissues including AGM , caudal hematopoietic tissue ( CHT ) and kidney were excluded from the myeloid cell statistics . Results showed that , albeit a small portion of zebrafish show GFP+ background signals in the control group ( Figure 3E ) , the numbers of GFP+ T lymphocytes and myeloid cells in the single HE-labeled group were significantly higher compared with that in the control group ( Figure 3E ) , demonstrating that the high-precision single-cell IR-LEGO system can efficiently trace the progenies derived from a single HE with high fidelity . Considering that the background signals may contribute to the GFP+ cell numbers in the heat shocked zebrafish and interference their lineage interpretation , we applied the maximum likelihood estimation ( MLE ) method ( Shao , 2003 ) to minimize the interference of background signal and to analyze the T lymphoid and myeloid potential of each single HE . The MLE method is a widely acknowledged statistical tool to extract desired information in the presence of noise background . It has been applied in the study of evolution , genetics and lineage tracing ( Tamura et al . , 2011; Sousa and Hey , 2013; Excoffier and Heckel , 2006; Perié et al . , 2014; Chan et al . , 2019; Larsen et al . , 2017 ) . For example , the MLE was used for robustly inferring evolutional trees in molecular evolutionary analysis ( Tamura et al . , 2011 ) , and for population genomic inference with complex demographic models ( Sousa and Hey , 2013; Excoffier and Heckel , 2006 ) , and also to determine the cell lineage pathway by converting barcode relationships into a tree of cell division ( Perié et al . , 2014; Chan et al . , 2019 ) . In this study , the MLE method is used to calculate the lineage distributions that maximize the joint probability density of observed data in both single cell-labeled and control groups . Therefore , it can be used as an unbiased estimator to depict the lineages of a single HE . The details of the MLE model for HE lineage tracing were illustrated in Appendix 5 . Results showed that 43 . 79% of the single HE labeled fish had both lymphoid and myeloid progenies , while 28 . 41% of the fish contained exclusively myeloid progenies ( Figure 4A , B and Supplementary file 1e , 1f ) . Notably , 27 . 8% of the labeled fish showed neither GFP+ T lymphocytes nor GFP+ myeloid cells ( Figure 4A ) , which could due to the fact that the labeled single HE in these fish was bona fide endothelial cells without hematopoietic potential , or contributed to other hematopoietic progenies such as erythroid lineage . Nevertheless , these data indicate that at least two distinct HE subpopulations exist in the aorta of the PBI: one population can give rise to both T lymphocytes and myeloid cells , while the other produces exclusively myeloid progenies ( Figure 4B , C ) . This result demonstrates that combined with comprehensive statistical analysis , the high-precision single-cell IR-LEGO system is a powerful tool to perform in vivo single-cell fate mapping under unperturbed conditions . Despite the contribution of the IR-LEGO heat shock technique in previous bulk fate mapping studies , its spatial resolution does not meet the requirement of single-cell lineage tracing ( Xu et al . , 2015; Tian et al . , 2017; He et al . , 2018; Henninger et al . , 2017 ) . In this work , we developed an advanced single-cell IR-LEGO microscope system equipped with a cutting-edge fluorescent thermometer . By associating temperature with the fluorescence intensity ratio , this dual-dye fluorescent thermometry is immune to the fluctuation of excitation laser power and may resistant to the micro-environmental variations . This advantage leads to a significantly higher signal-to-noise ratio of temperature measurement than the single-probe thermometry ( Deguchi , 2009; Kamei et al . , 2009 ) . Unlike other temperature probes such as nanoparticles ( Alkahtani et al . , 2017; Blakley et al . , 2016 ) , the dextran-conjugated fluorescent dyes used in this TPFT are highly bio-compatible and are ideal options for local temperature probing in live animal models . Benefiting from the intrinsic 3D sectioning capability of two-photon optical microscope , high spatial-resolution ( <1 µm laterally ) temperature measurement can be achieved in tissues . With the equipment of a high-speed EMCCD to detect fluorescence spectra , the temporal resolution can be as high as 0 . 02 s , enabling real-time temperature monitoring during the IR laser heating . Another improvement of this system is the flexible control of heat shock modes . In the old IR-LEGO system used in our previous works ( Xu et al . , 2015; Tian et al . , 2017; He et al . , 2018 ) , a loosely focused IR laser beam was fixed at a single spot of large region in tissues during heat shock . That system can achieve bulk cell labeling without induction of cell death ( Figure 4—figure supplement 1C , Figure 4—figure supplement 1—source data 1 ) , due to the large focal spot size ( 26 µm2 ) and low power density of laser beam . In contrast , in current single-cell IR-LEGO microscope system , we adopted a high-NA objective to tightly focus the IR laser into an extremely small region ( 0 . 4 µm2 ) inside a target cell , which is indispensable for high-resolution single-cell labeling . However , its high power density raises the possibility of cell damage for single-point heating mode ( Figure 2—figure supplement 3 , Figure 4—figure supplement 1 Aand B , Figure 4—figure supplement 1—source data 1 , Video 3 ) . Therefore , to avoid overheating or photochemical damage , our current study utilizes two pairs of scanning galvo mirrors , which enable independent control of the IR laser beam and the fluorescence excitation laser beam , to perform two-dimensional scanning over the targeted cells . In this work , we applied 32 s heating by constantly scanning the IR laser beam over an area of 8 µm × 8 µm in a cell . This scan-heating mode avoids the quick heat accumulation at single point and effectively reduces the cell damage ( Figure 4—figure supplement 1A and B , Figure 4—figure supplement 1—source data 1 , Videos 4 and 5 ) . Additionally , this dual-scanner setup also expedites the characterization of temperature distribution during IR laser heating . Our results show that without strict thermal confinement , the efficiency of single-cell labeling would be very low and the heating effects in different tissues differ largely from each other ( Figure 2—figure supplement 4 and Supplementary file 1b ) . Therefore , optimization of heat shock condition to constrain thermal diffusion in specific tissues is of high necessity for high-throughput single-cell labeling and practical single-cell lineage tracing study . Under optimized condition , our single-cell IR-LEGO technology can simultaneously achieve labeling , visualization and long-term tracing of single cell . Benefiting from this method of high spatial resolution and reliable fidelity , we uncovered the heterogeneity of HEs in the PBI of zebrafish and unveiled the complexity of lineage hierarchy in the definitive hematopoiesis . As a strategy for cell labeling and tracing , our single-cell IR-LEGO technique can be applied on any cell type as long as the IR laser can penetrate through the tissues above the targeted cells . In this work , we have demonstrated that this technique can precisely label single cell in different tissues with various depths and microenvironments , such as muscle , brain and hematopoietic tissue , indicating its great value for many other fields besides developmental biology . It is noticed that the labeling efficiency of single-cell IR-LEGO depends on heat shock conditions and cell types . Our results show that the efficiency of labeling single myocyte and single leukocyte is 54 . 5% and 77 . 8% , respectively , while the success rate of single-neuron labeling can reach 100% , due to their lower density in the brain ( Figure 2—figure supplement 4D and Supplementary file 1b ) . However , the efficiency of single HE labeling in the present lineage tracing assay is relatively low ( 29 . 3%; 27/92 ) , mainly due to the strict criteria ( live imaging for scoring single cell labeling ) we set for scoring . Unlike other cell types , the HEs are highly mobile . Upon EHT , they quickly undergo cell division and differentiate into highly mobile hematopoietic precursors . As a consequence , a portion of heat-shocked single HE would not be counted due to their proliferation ( more than one cells ) and migration ( lost in the circulation ) before the appearance of GFP expression ( induced by heat-shock ) . Thus , the actual efficiency of single HE labeling should be significantly higher than 29 . 3% . Even so , the 29 . 3% labeling efficiency is higher than that of previously reported single cell labeling in Drosophila ( below 20% [Miao and Hayashi , 2015] ) . Although single cell labeling was reported in zebrafish by one study ( Eiji , 2013 ) , its efficiency was not discussed ( the efficiency documented in that study refers to the overall efficiency of both single and multiple cells labeling ) . Moreover , none of these previous works has applied single cell labeling and long-term lineage tracing on highly mobile cells such as HEs . We believe that the single cell labeling strategy based on our newly developed IR-LEGO technology is the first comprehensive and integrated approach demonstrated for reliable and accurate single cell lineage tracing . The background noise , namely the appearance of GFP signals in zebrafish without heat shock , creates challenge to single-cell lineage tracing . In fact , background noise is a common issue among various single-cell lineage tracing technologies , such as high-throughput techniques ( Yuan et al . , 2017 ) and optical tracing based on Cre-LoxP system ( Henninger et al . , 2017 ) . For IR-LEGO technique , the background signals may arise from undesirable activation of the heat shock promoter which occurs occasionally and randomly during zebrafish development . This background noise is unavoidable , especially for long-term lineage tracing study . For bulk-labeling fate mapping , the background signals have little impact on the lineage interpretation because the number of labeled cells and their progenies are much larger than that of background signals . For single-cell lineage tracing , however , the number of progenies derived from a single target cell is limited , thus interference of background signals cannot be ignored . To minimize the effect of background noise , we applied the MLE method , a classic and well-established statistical tool , to analyze all the measurement data and estimate the lineage distribution . In fact , the MLE method has been widely applied in different aspects of biological studies including single cell transcriptomics , genome-wide association study and lineage analysis with DNA barcoding ( Chan et al . , 2017; Austerlitz et al . , 2009; Aulchenko et al . , 2010 ) , showing that the successful integration of mathematic and biological methods can explicitly improve statistical power . For the MLE model in this study , each single HE-labeled zebrafish was classified into different lineage types based on the presence or absence of GFP+ T/myeloid cells , regardless of the specific numbers of GFP+ cells . In this way , the probabilities of each lineage calculated by the MLE method are not affected by the variations of GFP+ cell numbers in both heat shocked and control zebrafish . Our results demonstrate that the MLE statistical method improves fidelity and broadens applicability of single-cell IR-LEGO system in the study of single-cell lineage tracing . Our single-cell tracing study suggested that HEs in the PBI directly give rise to at least two distinct hematopoietic precursors: one capable of generating both T lymphocytes and myeloid cells and the other producing myeloid cells only . Although HEs in the PBI are also known to generate erythrocytes ( Tian et al . , 2017 ) , our current study could not investigate T lymphoid , myeloid and erythroid lineage simultaneously because the coro1a promoter is leukocyte-specific . In principle , a possible solution is to generate globin:loxP-DsRedx-loxP-GFP;coro1a:loxP-DsRedx-loxP-EGFP double reporter line for triple lineage fate mapping analysis . However , a major drawback of this design is that incomplete gene editing , in which the heat shock-induced CreER edits one but not the other reporter cassette , may cause the misinterpretation of the linage potential . Given the fact that the erythroid-myeloid and lymphoid-myeloid progenitors but not lymphoid-erythroid progenitors have been identified in both mouse and zebrafish ( Bertrand et al . , 2007; Perdiguero and Geissmann , 2016; McGrath et al . , 2015; Böiers et al . , 2013 ) , we attend to speculate that the HEs in the PBI may likely give rise to two different progenitor populations with lymphoid-myeloid potential and erythroid-myeloid potential respectively ( Figure 4C ) . Studies in mice suggested that erythro-myeloid progenitors ( EMPs ) and HSCs are derived from distinct subpopulations of endothelial cells ( Chen et al . , 2011 ) , and the mammalian HSCs showed heterogeneity during their emergence from the E11 AGM in mid-gestation embryos ( Ye et al . , 2017 ) . In addition , our previous study has demonstrated that the generation of HSC-independent hematopoietic cells via EHT occurs in both PBI and AGM in zebrafish ( Tian et al . , 2017 ) . Taken together , these studies raise the possibility that the heterogeneity of HEs is not a phenomenon restricted in the PBI region but broadly exists along the aortic floor ( Figure 4C ) . However , we could not exclude the alternative possibility that the lineage biases among hematopoietic progenitors may be acquired by interacting with distinct niches . Indeed , studies in both mammals and zebrafish showed that new-born HSPCs dynamically interact with different cell types in various microenvironments , which is important for the migration , maintenance , proliferation and function of HSPCs ( Tamplin et al . , 2015; Gao et al . , 2018; Li et al . , 2018 ) , and perhaps , is also crucial for lineage commitment . In the future study , it is of great interest to investigate whether the interactions between hematopoietic progenitors and niches contribute to the lineage heterogeneity . In the single-cell IR-LEGO heat shock microscope system ( Figure 1A ) , a femtosecond Ti:sapphire laser ( Chameleon Ultra II , Coherent , Santa Clara , CA ) was used for the excitation of nonlinear optical ( NLO ) signals including TPEF and SHG . A DPSS low-noise CW IR laser ( MLL-H-1342 , Changchun New Industries Tech . Co , . Ltd . ) at 1 , 342 nm wavelength was used for localized heat shock . The femtosecond laser beam was combined with the CW laser beam with a dichroic mirror ( DMSP 1000 , Thorlabs ) and directed into a water-immersion objective ( UAPON 40XW340 , 1 . 15 NA , Olympus ) . Two pairs of galvanometer mirrors were used for x-y scanning of the femtosecond and CW laser beams , respectively . The objective was driven by an actuator for IR laser heating or NLO imaging at different depth . The backscattered NLO signals were collected by the objective and separated from the excitation light by a dichroic mirror ( FF665-Di02 , Semrock ) . In the fluorescent thermometry mode , the TPEF signals were focused into a spectrograph equipped with an EMCCD ( DU-971N , Andor Technology ) , which enabled spectral analysis of the FITC and TAMRA fluorescence at a high resolution ( 0 . 4 nm ) . For the three-dimensional temperature profile measurement , a lens ( L4 , Figure 1A ) on a linear translation stage ( 25 mm of travel ) was moved along the light axis to change the focal plane of the femtosecond laser without changing the focus of the CW laser , allowing the measurement of temperature profile along the axial direction . In the imaging mode , the NLO signals were directed to a spectrograph via a round-to-line fiber bundle . The signals were analyzed by the spectrograph equipped with a linear array of 16 photomultiplier tubes ( PMTs ) and a time-correlated single photon counting ( TCSPC ) module ( PML-16-C-0 and SPC-150 , Becker and Hickl ) . Time-resolved NLO signals were recorded in 16 consecutive spectral bands with a 13 nm resolution , covering the spectral range from 450 nm to 645 nm simultaneously . Spectrally resolved images can be formed with a variety of NLO signals . Accurate co-localization of the focused femtosecond laser ( 830 nm ) and IR laser ( 1 , 342 nm ) beams is critical for the fluorescent thermometry and heat shock gene induction at single-cell resolution . Since IR laser wavelength is beyond the detection range of silicon-based detectors , an CCD camera can not be directly used for precise optical alignment of the probe beam ( 830 nm ) and heat shock beam ( 1 , 342 nm ) . In this study , we painted a thin layer of black ink onto a coverglass and used it to identify the focal point of IR laser . In detail , when the cover glass was placed under objective and the IR laser power was appropriately controlled at low level , the black ink layer could only be vaporated by the laser at its focal point . The focal point without ink became a transparent spot of about 2 µm size that could be visualized in the bright field image captured by a CCD . This allows the focused beam positions of the 830 nm and 1 , 342 nm lasers to be visualized on the CCD camera simultaneously , to achieve a fine optical alignment of the two laser beams . The NLO imaging was then used to guide the IR laser to precisely aim at the targeted single cell for heat shock gene induction . To calibrate the temperature sensitivity of FITC-TAMRA in water and tissue phantom ( 3% agarose ) , the individual or mixed 0 . 006% FITC ( fluorescein and biotin-labeled dextran , 10 , 000 MW , Anionic , Lysine Fixable ( Mini-Emerald ) , D-7178 , Thermofisher Scientific ) and 0 . 004% TAMRA ( tetramethylrhodamine and biotin-labeled dextran , 10 , 000 MW , Lysine Fixable ( mini-Ruby ) , D3312 , Thermofisher Scientific ) solution was injected into a small homemade cuvette with two windows made of coverglasses . The sealed cuvette was mounted in a petri dish filled with warm water through circulation via a water bath cabinet . The water temperature in petri dish was measured with a thermocouple , which was attached to the cuvette . The TPEF spectra of FITC-TAMRA mixture solutions were recorded at different temperatures controlled through the water bath cabinet . FITC and TAMRA fluorescence were decomposed using their individual spectra measured from pure dye solutions . After decomposition of the mixed spectra , the intensity ratios of FITC and TAMRA fluorescence were calculated to measure the temperature sensitivity . To calibrate the temperature sensitivity of FITC-TAMRA mixture in zebrafish in vivo , the dextran-conjugated FITC and TAMRA were injected into zebrafish embryos at the single-cell stage ( ~1–2 nl/embryo ) . The embryos were raised to 1 dpf , 2 dpf and 3 dpf for the calibration of temperature sensitivity in muscle , AGM/PBI and hindbrain tissue , respectively . The zebrafish was mounted in 1% low-melting agarose and placed in an incubation system ( Chamlide TC , Live Cell Instrument ) . A thermocouple was inserted into the agarose and close to the zebrafish to obtain the actual temperature . The FITC-TAMRA TPEF spectra were recorded at different environmental temperatures ( 25–38°C ) controlled through the incubation system and used to measure the temperature sensitivities in corresponding tissues . We used the calibrated fluorescent thermometry to measure the local temperature rise induced by IR laser heating in water solution , tissue phantom ( 3% agarose ) and zebrafish in vivo . Based on the calibrated temperature sensitivity , the local temperature rise was calculated by recording the changes in TPEF intensity ratios before and after IR laser heating . To measure the lateral temperature distribution in water and tissue phantom , the 1 , 342-nm IR laser was fixed at the central position without scanning , while the 830-nm laser beam was scanned from the central position to the furthest distance of 70 µm away from the center to excite the TPEF of the FITC-TAMRA at different lateral positions . For the measurement of axial temperature distribution , the 830-nm probe beam was first co-localized with the 1 , 342-nm heat shock beam on the same focal plane and then separated axially from the heat shock beam by moving the lens ( L4 , Figure 1A ) on a translation stage . For in vivo measurement of temperature distribution in the zebrafish , scan heating was performed to avoid laser-induced tissue injury . The 1 , 342-nm heating beam was scanned in an 8 µm × 8 µm region of the zebrafish tissues , while the 830-nm probe beam was scanned laterally and axially in the same way as in the water solution and tissue phantom to measure the temperature distributions . The 3D views of temperature distributions ( Figure 2E ) were plotted through the two-term Gaussian fitting of the discrete lateral temperature curves . The temperature is the sum of the environmental temperature ( 23°C ) and the temperature rise measured through fluorescent thermometry . For in vivo temperature measurement using fluorescent thermometry , dextran-conjugated FITC and TAMRA were injected into zebrafish embryos at the single-cell stage ( ~1–2 nl/embryo ) . The embryos were raised to the desired stages for fluorescent thermometry measurement . For heat shock gene induction in muscle cells , CreERT2 transgenic fish Tg ( hsp70l:mCherry-T2a-CreERT2 ) ( Hans et al . , 2011 ) were crossed with a Tg ( bactin2:loxP-STOP-loxP-DsRedx ) ( Bertrand et al . , 2010 ) reporter line . The embryos were injected with 1–2 nl PhOTO vector ( Dempsey et al . , 2012 ) at the single-cell stage , which labelled the cell membrane with cerulean and cell nuclei with Dendra2 . The embryos were raised to 1 dpf and then mounted in 1% low-melting agarose for the heat shock experiment . With 1 μM 4-OHT treatment , scan heating was performed on myocyte nuclei for 32 s with the guidance of Dendra2 signals in TPEF imaging . The heat shock gene induction results were examined 24 hrs later by detecting the heat shock-induced DsRedx expression in the myocytes . The SHG signal from the sarcomere of each muscle fiber was used to assist the validation of gene-induced cell numbers ( Figure 2 F2 ) . For heat shock gene induction in tyrosine hydroxylase-positive ( th-positive ) neurons , Tg ( hsp70l:mCherry-T2a-CreERT2 ) fish were crossed with reporter line Tg ( th:loxP-GFP-loxP-DsRedx ) . The embryos were raised to 3 dpf for the experiment . With 1 μM 4-OHT treatment , scan heat shock was performed on a single GFP-labeled th-positive neuron in an 8 µm × 8 µm area for 32 s in the hindbrain of zebrafish with the guidance of TPEF imaging . The heat shock gene induction results were examined 36 hr later by detecting the heat shock-induced DsRedx expression in the neurons . For heat shock gene induction in leukocytes , Tg ( hsp70l:mCherry-T2a-CreERT2 ) fish were crossed with reporter line Tg ( coro1a:loxP-DsRedx-loxP-EGFP ) ( Xu et al . , 2015 ) . The embryos were raised to 2 dpf for the experiment . With 1 μM 4-OHT treatment , scan heat shock was performed on a single DsRedx-labeled leukocyte at the aorta-gonad-mesonephros ( AGM ) or the posterior blood island ( PBI ) region with the guidance of TPEF imaging . The scanning time was set as 32 s and scanning area at 8 µm × 8 µm to cover a single cell . After heat shock , live imaging was conducted to trace the migration of the heat-shocked cell and verify the heat-induced GFP expression in the target cells over the following 24 hrs . For heat shock gene induction in hemogenic endothelium ( HE ) and the subsequent lineage tracing , we constructed transgenic fish Tg ( kdrl:loxP-DsRedx-loxP-EGFP ) , in which the blood vessel endothelium , including HEs , are labeled by DsRedx . To construct this transgene , the coro1a promoter in construct coro1a:loxP-DsRedx-loxP-EGFP was replaced by 6 . 5 kb vessel endothelial-specific kdrl promoter ( Cross et al . , 2003 ) . The adult Tg ( kdrl:loxP-DsRedx-loxP-EGFP ) fish were crossed with Tg ( coro1a:loxP-DsRedx-loxP-EGFP ) fish to acquire double transgenic Tg ( kdrl:loxP-DsRedx-loxP-EGFP;coro1a:loxP-DsRedx-loxP-EGFP ) fish , in which both vessel endothelium and leukocytes are labeled . Then the adult double transgenic fish were crossed with Tg ( hsp70l:mCherry-T2a-CreERT2 ) fish to acquire triple transgenic Tg ( kdrl:loxP-DsRedx-loxP-EGFP;coro1a:loxP-DsRedx-loxP-EGFP;hsp70l:mCherry-T2a-CreERT2 ) fish ( referred to as ‘triple Tg’ hereinafter ) . The triple Tg embryos were raised to 26–28 hpf for the experiment . With 1 μM 4-OHT treatment , scan heat shock was performed on a single kdrl+ HE on the ventral wall of caudal aorta at the PBI region with the guidance of TPEF imaging . The scanning time was set as 32 s and scanning area at 8 µm × 8 µm to cover a single HE . After heat shock , live imaging was conducted to trace the behavior of the heat-shocked HE and verify the heat-induced GFP expression in the target HEs till 48 hpf . For the testing of cell damage/death caused by heat shock , Tg ( hsp70l:mCherry-T2a-CreERT2 ) fish were crossed with reporter line Tg ( kdrl:nls-EOS ) ( Fukuhara et al . , 2014 ) to acquire double transgenic Tg ( hsp70l:mCherry-T2a-CreERT2;kdrl:nls-EOS ) fish . The double Tg embryos were raised to 26–28 hpf , and then the individual nls-EOS+ HEs in the PBI region were exposed to UV laser to convert the EOS protein from green to red . Each converted fish was imaged immediately after photo-conversion to record the number and position of the converted cells . After that , a part of the converted embryos were treated with 4-OHT and the single red-EOS+ HE was heat-shocked either by scanning over 8 µm × 8 µm area for 32 s or by single point irradiation for 32 s with the same laser power ( 80 mW ) . Rest of the converted embryos were only treated by 4-OHT as the control group . After heat shock , live imaging was conducted to record the cell death of the HEs in scanning heat shock group , single point heat shock group and control group , respectively . Cell death was validated based on two criteria . One is the disappearance of cells shortly after heat shock . The other is the observation that cell nuclei burst into fragments during live imaging . The same method was used to test the cell death caused by the old version of IR-LEGO system described previously ( Xu et al . , 2015 ) . In that system , a doublet lens with 60 mm focal length was used as the objective lens to loosely focus the IR laser beam into samples . The converted embryos were treated with 4-OHT and single spot irradiation ( 80 mW ) for 2 min was conducted on the PBI region . Then the cell death was assessed in the heat shocked group and the control group through live imaging . Live imaging was performed according to the previous protocol with minor modifications ( Xu et al . , 2016 ) . Embryos were mounted in 1% low-melting agarose and imaged on a Leica SP8 confocal microscope with a 25 °C thermal chamber . A 20x objective was used to take time-lapse images . The Z step size was set at 1 . 5 µm , with 20–30 planes in each z stack . For each embryo , images were taken every 20 minutes . The single HE-labeled fish were fixed on 7 dpf and then processed to whole mount in situ hybridization following a previously published protocol ( Thisse and Thisse , 2008 ) . Here , we made a few modifications of the protocol . First , we conducted an additional permeabilization step before proteinase K treatment by treating the samples with 100% acetone at −20°C and then washing the samples in PBST for 5 min for 3 times , and we changed the anti-dig-AP antibody with anti-dig-POD and optimized the later fluorescent color reaction ( TSA-cy3 system ) steps according to another published protocol ( Welten et al . , 2006 ) . For antibody staining , briefly , the samples were firstly blocked in 5% FBS in PBST , then incubated with goat-anti-GFP primary antibody ( ab6658; Abcam ) at 4°C overnight . On the second day , the samples were washed in PBST for 30 min for 4 times , and then incubated in donkey-anti-goat-488 secondary antibody ( A11055; Invitrogen ) at 4°C overnight . Finally , the samples were washed in PBST for 30 min . Of note , we inserted the antibody staining steps into the whole mount in situ hybridization steps by adding anti-dig-POD and goat-anti-GFP antibody at the same time , and then firstly completed antibody staining and finally went back to the rest of the whole mount in situ hybridization steps to finish color reaction for POD . In order to confirm whether T lymphocytes can be distinguished from other cell types in thymus based on their morphology , we labeled and compared the size and morphology of potential thymus-residing cell types , including blood vessel endothelial cell , neutrophil , macrophage and T lymphocyte . Specifically , we crossed thymus epithelium-marking line Tg ( foxn1:mCherry ) with blood vessel endothelium-marking line Tg ( kdrl:EGFP ) or neutrophil-marking line Tg ( lyz:EGFP ) or macrophage-marking line Tg ( mpeg1:EGFP ) and imaged thymus regions of 7 dpf fish . In this way , we can observe the distribution of blood vessel endothelium , neutrophils and macrophages in thymus of 7 dpf fish ( Figure 3—figure supplement 2C–E ) . To observe T lymphocytes in thymus , we crossed T lymphocyte-marking line Tg ( lck:loxP-DsRedx-loxP-EGFP ) with Tg ( hsp70I:mCherry-T2a-CreERT2 ) line and performed single spot IR laser illumination at 26 hpf PBI region as described previously to convert a small portion of T cells from DsRedx+ into EGFP+ ( Tian et al . , 2017 ) . After IR laser illumination , the fish were treated with 4-OHT overnight , and were raised to 7 dpf for thymus imaging ( Figure 3—figure supplement 2F ) . As shown in Figure 3—figure supplement 2 , T cells can be effectively distinguished from other cell types in the thymus by their small and round shapes . All the samples were directly fixed in 4% PFA at 4°C overnight , then processed to whole-mount antibody staining as described elsewhere ( Barresi et al . , 2000 ) . The primary antibody used in this study is anti-GFP antibody ( ab6658 , Abcam ) , and the secondary antibody is Alexa 488-anti-goat antibody ( A11055 , Invitrogen ) . After antibody staining , the zebrafish were mounted in 3% methylcellulose and the GFP-positive T lymphocytes as well as myeloid cells of each zebrafish were quantified manually under Nikon Eclipse Ti inverted fluorescent microscope . To capture the representative images of antibody stained samples , the zebrafish were mounted in 1% agarose and imaged with Leica SP8 confocal microscope . In the non-labeling zebrafish of the control group , the GFP+ cell numbers are not in normal distributions . Therefore , a nonparametric test , the Mann–Whitney–Wilcoxon rank-sum test ( also called the Mann-Whitney U test ) , was used for the significance test of the GFP+ cells in the single HE-labeled and control zebrafish groups . The MLE method is used to calculate the lineage distributions of a single HE that maximize the joint probability density of observed data in both single cell-labeled and control groups . The details of the MLE model for HE lineage analysis were illustrated in Appendix 5 .	Animals begin life as a single cell that then divides to become a complex organism with many different types of cells . Every time a cell divides , each of its two daughter cells can either stay the same type as their parent or adopt a different identity . Once a cell acquires an identity , it usually cannot ‘go back’ and choose another . Eventually , this process will produce daughter cells with the identity of a specific tissue or organ and that cannot divide further . Multipotent cells are cells that can produce daughter cells with different identities , including other multipotent cells . These cells can usually give rise to different cell types in a specific organ , and generate more cells to replace any cells that die in that organ . Tracking the cells descended from a multipotent cell in a specific tissue can provide information about how the tissue develops . Hemogenic endothelium cells produce the multipotent cells that give rise to two types of white blood cells: myeloid cells and lymphoid cells . Myeloid cells include innate immune cells that protect the body from infection non-specifically; while lymphoid cells include T cells and B cells with receptors that detect specific bacteria or viruses . It remains unclear whether each of these two cell types originate from a single population of hemogenic endothelium cells or from two distinct subpopulations . He et al . have now developed a new optical technique to label a single hemogenic endothelium cell in a zebrafish and track the cell and its descendants . This method revealed that there are at least two distinct populations of hemogenic endothelium cells . One of them can give rise to both lymphoid and myeloid cells , while the other can only give rise to myeloid cells . These findings shed light on the mechanisms of blood formation , and potentially could provide useful tools to study the development of diseases such as leukemia . Additionally , the single-cell labeling technology He et al . have developed could be applied to study the development of other tissues and organs .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "stem", "cells", "and", "regenerative", "medicine", "developmental", "biology", "tools", "and", "resources" ]	2020	In vivo single-cell lineage tracing in zebrafish using high-resolution infrared laser-mediated gene induction microscopy
Crude oil spills are a worldwide ocean conservation threat . Fish are particularly vulnerable to the oiling of spawning habitats , and crude oil causes severe abnormalities in embryos and larvae . However , the underlying mechanisms for these developmental defects are not well understood . Here , we explore the transcriptional basis for four discrete crude oil injury phenotypes in the early life stages of the commercially important Atlantic haddock ( Melanogrammus aeglefinus ) . These include defects in ( 1 ) cardiac form and function , ( 2 ) craniofacial development , ( 3 ) ionoregulation and fluid balance , and ( 4 ) cholesterol synthesis and homeostasis . Our findings suggest a key role for intracellular calcium cycling and excitation-transcription coupling in the dysregulation of heart and jaw morphogenesis . Moreover , the disruption of ionoregulatory pathways sheds new light on buoyancy control in marine fish embryos . Overall , our chemical-genetic approach identifies initiating events for distinct adverse outcome pathways and novel roles for individual genes in fundamental developmental processes . Catastrophic oil spills , rising water temperatures , ocean acidification , and other large-scale anthropogenic forcing pressures impact the health and survival of myriad marine species in ways that are often unknown . Mechanistic relationships between environmental stress and adverse health outcomes are most readily studied in model laboratory organisms . However , domesticated experimental models may be relatively insensitive to real-world environmental change . Emerging genomic technologies , including high-throughput RNA sequencing ( RNA-seq ) , are providing new opportunities to profile physiological stress in wild , non-model species at a transcriptional level . This approach is premised on the full anchoring of gene expression to physiological or morphological injury phenotypes . The early life stages of marine fish are particularly vulnerable to pollution and other stressors . However , developmental analyses in wild species can be challenging due to limited access to embryos and larvae . Also limited are molecular and cellular tools for imaging-specific structures ( e . g . fluorescent protein-expressing transgenes ) . This includes , for example , a lack of species-specific probes for visualizing gene expression via in situ hybridization . On the other hand , one of the major vertebrate models for studying developmental genetics is a fish . Zebrafish have been a focus for high-throughput experimental techniques for more than three decades . This has yielded a wealth of information on embryonic gene expression patterns , with comparable data often available in chick and mouse embryos . The zebrafish platform therefore provides a powerful mechanistic context for anticipating environmental health impacts in marine fish spawning habitats . Here , we use an RNA-seq approach to assess the effects of crude oil on the early life stages of a cold-water marine species , Atlantic haddock ( Melanogrammus aeglefinus ) . Crude oil spills such as the 1989 Exxon Valdez ( Prince William Sound ) , 2002 Prestige ( Spain ) , and 2010 Deepwater Horizon ( Gulf of Mexico ) continue to threaten fisheries worldwide . Haddock are commercially valuable in Norway and other North Atlantic countries , and they spawn in areas proposed for future crude oil production ( e . g . the Lofoten archipelago in Nordland ) . Similar to many other fish species , haddock embryos are vulnerable to developmental defects from crude oil exposure ( Norwegian Sea crude; [Sørhus et al . , 2016b] ) . Moreover , we recently showed that RNA-seq applied to normally developing haddock clearly anchored organogenesis phenotypes to the expression of genes involved in determination and differentiation ( Sørhus et al . , 2016a ) . Crude oils are complex chemical mixtures , and fish early life stages are especially vulnerable to component polycyclic aromatic hydrocarbons ( PAHs ) and their alkylated homologues ( Carls and Meador , 2009; Adams et al . , 2014 ) . Crude oil-derived PAHs containing three rings disrupt the normal form and function of the embryonic heart , and circulatory failure causes a range of secondary defects ( Incardona et al . , 2004 , 2005 ) . For individual heart muscle cells , the cardiotoxic mechanism involves a blockade of the repolarizing potassium efflux and a reduction in intracellular calcium cycling ( Brette et al . , 2014 ) . The consequent disruption of excitation-contraction ( E-C ) coupling leads to rhythm and contractility defects at the scale of the developing heart ( Incardona et al . , 2009 , 2014; Sørhus et al . , 2016b ) . Mechanisms underpinning morphological defects in other embryonic tissues are still poorly understood . Based on conventional measures of cardiac function and embryolarval anatomy , zebrafish and haddock respond to Norwegian Sea crude oil in ways that are similar and dissimilar . Both species show characteristic abnormalities including bradycardia , reduced chamber contractility , and fluid accumulation in the vicinity of the heart ( edema ) . This suggests an understanding of zebrafish developmental genetics will inform the interpretation of changing messenger RNA ( mRNA ) levels in crude oil-exposed haddock as determined by RNA-seq . This is particularly true for tissue-specific patterns of gene expression that are highly conserved across vertebrates—for example , genes involved in cardiac organogenesis . Relative to zebrafish , however , haddock are sensitive to much lower concentrations of crude oil and also display a distinct suite of craniofacial defects that cannot be attributed to circulatory failure ( Sørhus et al . , 2016b ) . There are several reasons to expect divergent effects of crude oil on marine fish embryos and larvae . These are attributable to differences in physiology and life history . For example , accumulation of cardiac edema is a canonical form of crude oil toxicity in both freshwater and marine species . Yet marine embryos are hyposmotic to the surrounding water and hence expected to lose water with circulatory failure . This suggests that PAHs may have distinct impacts on ionoregulation and related processes . Also , unlike zebrafish , many pelagic marine embryos are buoyant and have a characteristic morphology not found in species with demersal ( sinking ) eggs and larvae . Shelbourne first described a relationship between this unique morphology of pelagic fish larvae , osmoregulation and buoyancy control in the mid-twentieth century ( Shelbourne , 1955 , 1956 , 1957 ) , but there has been little progress in the decades since , particularly at a molecular scale . Understanding cause-effect relationships between exposure to environmental contaminants like crude oil and adverse impacts on organismal health are critical for the construction of adverse outcome pathways ( AOPs ) . The development and application of AOPs is an ongoing movement to improve risk assessments . AOPs are derived from detailed toxicological cause-and-effect relationships that span multiple levels of biological organization , ideally from molecular initiating events to species , community or ecosystem scale responses of regulatory concern ( e . g . reduction in a fisheries abundance target ) . AOPs are widely used in risk assessments for both human and environmental ( ecological ) health ( Ankley et al . , 2010; Kramer et al . , 2011; Villeneuve et al . , 2014; Garcia-Reyero , 2015 ) Our long-term aim is to develop AOPs specific to oil spills and fish populations , premised on well-studied early life stage toxicity . AOPs based on crude oil cardiotoxicity in developing fish are already fairly well constructed ( Incardona and Scholz , 2016 ) but currently lack details at the molecular level at several steps , particularly in relation to cardiac dysmorphogenesis . We anticipate that identification of changes in gene expression associated with oil-induced developmental defects will further complete these AOPs and expand the molecular toolkit for quantifying oil spill impacts . In the present study , we used visible developmental abnormalities as phenotypic anchors for evaluating changes in haddock gene expression . We sequenced the full haddock transcriptome at several time points during and after embryonic and larval crude oil exposures . This approach allowed us to explore gene regulation in association with three distinct phenotypes: ( 1 ) heart form and function defects , ( 2 ) craniofacial deformities , and ( 3 ) fluid balance abnormalities and the characteristic pelagic larval form . We also identify perturbations in cholesterol homeostasis linked to poor absorption of yolk as a novel form of crude oil toxicity in marine fish embryos . Our findings are interpreted in the context of highly conserved gene regulation in zebrafish and other vertebrates . At a rearing temperature of 7°C , haddock embryos began hatching at 12 days post-fertilization ( dpf ) . Unlike zebrafish that complete gastrulation ( epiboly ) before segmentation ( somitogenesis ) begins , haddock embryos begin forming anterior somites at about 50% epiboly ( 3 dpf ) . They subsequently reach the tailbud stage by 6 dpf ( ~30 somites ) , have a regular heartbeat by 8 dpf and completion of organogenesis at 10 dpf ( Fridgeirsson , 1978; Hall et al . , 2004; Sørhus et al . , 2016a ) . Haddock yolk sac larvae have the characteristic morphology associated with ichthyoplankton from pelagic marine habitats ( Figure 1A ) , marked by a large marginal finfold that surrounds the larva nearly completely on both the dorsal and ventral sides . The outer epidermis is thus separated from the brain , main body axis muscles , and internal organs by a voluminous subdermal space . This space is filled with extracellular matrix ( Morrison , 1993 ) and is continuous with an avascular yolk sac sinus , with connections between the dorsal space and the ventral yolk sac in the vicinity of the pectoral fin ( Shelbourne , 1955 ) . The subdermal space acts as a reservoir for water balance in order to maintain larval buoyancy ( Shelbourne , 1955 , 1956 , 1957 ) , with specialized cells regulating ion and water balance ( ionocytes or mitochondria rich cells , MRCs ) distributed throughout the adjacent epidermis ( Shelbourne , 1957; Hirose et al . , 2003; Hiroi et al . , 2005 ) . 10 . 7554/eLife . 20707 . 003Figure 1 . Terminal phenotypes after high dose exposure . Control ( A ) and exposed ( B ) three days post hatch ( dph ) larvae ( 6 days post embryonic exposure ) . Open arrowheads in ( A ) indicate the marginal finfold surrounding the larvae and the white asterisk indicate the location of the connection between the dorsal space and the ventral yolk sac in the vicinity of the pectoral fin . In ( B ) the black arrowhead indicates severely reduced craniofacial outgrowth , while the black arrow indicates yolk sac edema . The ventricle and atrium in control ( C ) and embryonically exposed ( D ) animals are indicated by black and white arrows , respectively . ( E ) Normal craniofacial structure in control , and ( F ) moderate and ( G ) severe craniofacial defects in exposed animals . ( H ) Normal marginal finfold in control , ( I ) exposed animals with severe reduction of anterior marginal finfold ( open arrowheads ) . Yolk mass ( * ) in control ( J ) and embryonically exposed larvae ( K ) . ( L ) Control and ( M ) exposed 18 dph larvae . Open arrowheads indicate increased anterior marginal finfold , black arrowhead indicates reduced upper jaw outgrowth , and black arrow indicates edema formation in the peritoneal cavity in oil-exposed larvae ( M ) . Scale bar: 0 . 2 mm ( C , D; E–G; H–K ) and 1 mm ( A , B and L , M ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 00310 . 7554/eLife . 20707 . 004Figure 1—figure supplement 1 . Normal development of liver and lateral line in the severe phenotypes . Normally developed livers and neuromast cells are indicated by black arrows and arrowheads in control ( A , C ) and severely affected hunchback phenotypes ( B , D ) , respectively . Scale bar 0 . 2 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 004 Haddock embryos were continuously exposed to crude oil at environmentally relevant total PAH concentrations of 6 . 7 ± 0 . 2 µg/L ( high dose ) and 0 . 58 ± 0 . 05 µg/L ( low dose ) , and intermittently at 6 . 1 µg/L ( pulse dose ) . This yielded internal total PAH doses of 3 . 0 ± 1 . 3 µg/g wet weight , 0 . 19 ± 0 . 02 µg/g , and 0 . 22 ± 0 . 06 µg/g , respectively ( see [Sørhus et al . , 2016b] for experimental details ) . Although the low and pulsed exposures led to similar total PAH accumulation in embryos , phenotypes were slightly more severe in the latter due to relatively higher exposure concentrations during critical windows of early development ( Sørhus et al . , 2016b ) . The embryonic exposure began at 2 dpf and ended shortly after the end of organogenesis at 10 dpf , just prior to hatch . Larvae were exposed to the same regimen from day of hatch to 18 days post hatch ( dph ) . As expected , tissue PAH accumulation was lower than for embryos , with 0 . 81 ± 0 . 18 µg/g , 0 . 086 ± 0 . 015 µg/g , and 0 . 081 ± 0 . 024 µg/g , for high , low , and pulse doses , respectively . Except where indicated below , phenotypes were quantified as described in detail previously ( Sørhus et al . , 2016b ) , with 96% of high-dose animals showing abnormal phenotypes , ranging to ~60% for pulse dose and ~35% for the low dose . Representative terminal phenotypes in high dose are shown in Figure 1 for the embryonic exposure at 3 dph and larval exposure at 18 dph . Grossly , as in other species , crude oil exposure led to defects in cardiac function and morphology and accumulation of edema around the heart and in the yolk sac ( Figure 1B ) . Defects in cardiac morphology included a failure to properly loop the atrial and ventricular chambers from a linear to an adjacent orientation , and reduced size of the ventricle ( Figure 1C , D ) . In addition , oil-exposed haddock embryos displayed craniofacial defects in their larval stages that ranged in severity ( Figure 1E–G ) , from marked reductions in upper jaw/skull base structures ( Figure 1F ) to near complete lack of upper and lower jaws ( Figure 1G ) . Moreover , the anterior portion of the dorsal marginal finfold was collapsed or missing and the hindbrain ventricle typically failed to fill with cerebrospinal fluid in embryonically exposed larvae with severe edema ( Figure 1H , I ) . Finally , in more severely affected embryos , a failure of yolk absorption was obvious at 3 dph ( Figure 1J , K ) . Even in mildly affected embryos , yolk absorption was reduced after hatch as assessed by measuring the two-dimensional area of the yolk mass in lateral images ( control yolk area control 0 . 63 ± 0 . 06 mm2 , low-dose group 0 . 90 ± 0 . 11 , high-dose group 1 . 2 ± 0 . 3; mean ± s . d . , ANOVA p<0 . 0001 ) . In contrast , there was no measurable difference in yolk area at day of hatch . After larval exposure , the primary morphological defects were reduced jaw growth and edema accumulation ( Figure 1L , M ) , the latter in the peritoneal cavity . In contrast to embryos , the dorsal anterior subdermal space accumulated fluid in larvae and did not collapse . Abnormal phenotypes relating to the formation of edema , heart function and morphogenesis , craniofacial structure , and yolk absorption manifested at different developmental time points during embryonic exposure and afterwards when embryos were transferred to clean water for hatching ( Figure 2 ) . Samples were collected for transcriptome profiling at four embryonic stages ( E1-4 , Figure 3A ) and at two stages post-hatch ( E5-6 , Figure 3A ) . At 6 dpf ( E2 sampling point; ~30 somites ) , embryos exposed to the high dose were indistinguishable from controls ( Figure 2A ) . By 8 dpf , small accumulations of edema could be observed in the yolk sac of oil-exposed embryos , but their hindbrain ventricles were ‘inflated’ with fluid ( Figure 2B ) . By 10 dpf ( E3 sampling point ) , edema was evident in most embryos , typically filling the space above the yolk between the anterior of the head and the tail , and hindbrain ventricles lacked fluid ( Figure 2C ) . Similarly , at 6 dpf/E2 , heart development appeared unaffected in oil-exposed embryos and was at the un-rotated midline cone stage ( Figure 2D ) . By 9 dpf ( one day before sample E3 ) , hearts in both control and high-dose-exposed embryos had rotated and were beginning to loop , but ventricular walls already appeared slightly thinner ( Figure 2E ) and heart rate was significantly slower ( 20 + 6 beats/min compared to 26 ± 3 beats/min for controls; [Sørhus et al . , 2016b] ) . By day of hatch ( E5 sampling ) , a high percentage ( 54% ) of exposed embryos showed un-looped hearts with small , silent ventricles ( Figure 2F ) . 10 . 7554/eLife . 20707 . 005Figure 2 . Appearance of phenotypes over time . In each panel control and high-dose-exposed embryos are shown on the left and right , respectively . ( A–C ) Lateral overview of whole embryos showing accumulation of edema ( anterior to the left ) . ( A ) 6 dpf/E2 sampling point . ( B ) 8 dpf ( between E2 and E3 sampling points ) . Heart ( h ) and liver bud ( l ) are indicated . White arrowheads indicate outer margins of the yolk sac membranes; asterisk indicates small pocket of edema . Black arrowheads indicate the hindbrain ventricle . ( C ) 10 dpf/E3 sampling point . Arrowheads same as ( B ) ; asterisks indicate expanded yolk sac edema . ( D–E ) High-magnification ventral views of the heart ( anterior at top ) . ( D ) 6 dpf/E2 . Dashed turquoise lines indicate outer border and lumen of midline cardiac cone . ( E ) 9 dpf ( between E2 and E3 ) . Arrows indicate the atrium ( a ) , ventricle ( v ) and bulbus arteriosus ( ba ) . ( F ) 0 dph ( E5 sampling point ) . Chambers indicated as in ( E ) . ( G–I ) Lateral views of the developing head ( anterior to the left ) . ( G ) 8 dpf ( between E2 and E3 ) . ( H ) 0 dph ( E5 ) . Arrow indicates abnormal lower jaw cartilages in oil-exposed larva . ( I ) 3 dph ( E6 sampling point ) . Red bars indicate difference in eye diameter between control and exposed larvae . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 00510 . 7554/eLife . 20707 . 006Figure 3 . Exposure regimes and differentially expressed genes ( DEGs ) during embryonic development . ( A ) Embryos were exposed to a continuous high dose ( black line; 6 . 7 ± 0 . 2 μg/L TPAH ) , a pulsed dose ( red line; 0 . 09 ± 0 . 02–6 . 8 ± 1 . 0 μg/L TPAH ) and a continuous low dose ( blue line; 0 . 58 ± 0 . 05 μg/L TPAH ) of crude oil . Photos indicate normal developmental progress at each of six sampling time points ( E1–E6 ) . Venn diagrams show shared and exclusive DEGs for each of the three oil exposures at E1–E6 . ( B ) Venn diagrams illustrating the number of shared and exclusive DEGs at each stage in development up to hatching for the three exposure regimes . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 00610 . 7554/eLife . 20707 . 007Figure 3—figure supplement 1 . Most regulated KEGG pathways . ( A ) Pathways ( Total ) with highest number of DEGs ≥2 FC during and after embryonic exposure . ( B ) Pathways with the largest fraction of DEGs ≥2 FC/ Total number of genes in pathway ( Normalized ) during and after embryonic exposure . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 00710 . 7554/eLife . 20707 . 008Figure 3—figure supplement 2 . Comparison of mRNA read count data with real-time qPCR for selected genes during and after embryonic exposure . Genes include cp1a ( cytochrome p450 1 a ) , wnt11 ( wingless-type MMTV integration site family member 11 ) , kcnh2 ( potassium voltage-gated channel subfamily H member 2 ) , cac1c ( voltage-dependent L-type calcium channel ) , nac1 ( sodium/calcium exchanger 1 ) , at2a2 ( sarcoplamsic-endoplasmic reticulum calcium ATPase ) . ( A ) Real-time qPCR . ( B ) Read count data from RNA sequencing . Data were normalized as described in Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 008 Onset of craniofacial abnormalities took a longer course . At 8 dpf ( 2 days before sample E3 ) , head structures appeared identical in control and exposed embryos ( Figure 2G ) . At hatch ( E5 sample ) , jaw structures appeared somewhat abnormal ( Figure 2H ) but became much more strikingly severe by 3 dph ( E6 sample; Figure 2I ) . Notably , the eyes appeared smaller in exposed embryos by 3 dph ( Figure 2I ) . We did not quantify this effect , because it was demonstrated earlier in zebrafish that small eyes result from loss of cardiac function by either genetic or chemical means ( Incardona et al . , 2004 ) , and hence , this phenotype is not specific to crude oil toxicity . Other organs and structures were apparently unaffected by oil exposure . For example , the development of the liver and lateral line neuromasts progressed normally even in the most severely impacted larvae that were exposed as embryos ( Figure 1—figure supplement 1A–D ) . Relative to unexposed controls , differently expressed genes ( DEGs ) in oil-exposed fish were defined as having significantly ( p<0 . 05 ) higher or lower levels of expression . The number of exclusive and shared DEGs varied across exposure regime and haddock developmental age ( Figure 3B ) . After 24 hr of oil exposure ( sampling stage E1; 3 dpf ) , relatively few genes were differentially expressed , and most were significantly downregulated ( Supporting dataset 1 , Sørhus et al . , 2017 ) . From sampling point E2 ( 6 dpf ) through E6 ( three days post hatch , dph ) , however , the number of DEGs was substantial , particularly in response to the high dose exposure . With the exception of the initial sampling point ( E1 ) , a total of 28 DEGs were shared across all embryonic stages ( Table 1 ) . As expected , the largest category ( nine DEGs ) included genes associated with stress response and xenobiotic metabolism . The remaining genes play a role in tyrosine catabolism , myofibrillar establishment and cardiac tissue repair , central nervous system ( CNS ) function and degeneration , ATP metabolism , and cholesterol synthesis . 10 . 7554/eLife . 20707 . 009Table 1 . Genes expressed at all stages during and after embryonic exposure ( E2–E6 ) in high dose group . SP , swissprot; GB , genebank; IE; increased expression; DE; decreased expression . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 009Cod ID Swissprot annotation SP ID GB ID Category Regulation ENSGMOG00000018302Fumarylacetoacetasefaaa fah Tyrosine metabolismIEENSGMOG00000000318Cytochrome P450 1A1cp1a1 cyp1a1 xenobiotic metabolism and stressIEENSGMOG00000012518Glutathione S-transferase Pgstp1 gstp1 xenobiotic metabolism and stressIEENSGMOG00000016016Glutathione S-transferase omega-1gsto1 gsto1 xenobiotic metabolism and stressIEENSGMOG000000187523-hydroxyanthranilate 3 , 4-dioxygenase3hao haao xenobiotic metabolism and stressIEENSGMOG000000067963-beta-hydroxysteroid-Delta ( 8 ) , Delta ( 7 ) -isomeraseebp ebp xenobiotic metabolism and stressIEENSGMOG00000007636Glutamine synthetaseglna glul xenobiotic metabolism and stressIEENSGMOG00000015234Heat shock protein HSP 90-alphah90a1 hsp90a . 1 xenobiotic metabolism and stressIEENSGMOG00000012029Peptidyl-prolyl cis-trans isomeraseppia - xenobiotic metabolism and stressIEENSGMOG00000000218Ammonium transporter Rh type A OS=Musrhag rhag xenobiotic metabolism and stressMainly IEENSGMOG00000003353Ferritin , middle subunitfrim - xenobiotic metabolism and stressIEENSGMOG00000018206Filamin-Cflnc Flnc myofibrillar establishment and repairIEENSGMOG00000001317Iron-sulfur cluster assembly enzyme ISCU , mitochondrialiscu Iscu cardiac defectsIEENSGMOG00000010446Fatty acid-binding protein , heartfabph fabp3 cardiac defects and repairIEENSGMOG00000007115Lanosterol 14-alpha demethylasecp51a cyp51a1 Cholesterol syntheisIEENSGMOG00000005565Squalene monooxygenaseerg1 Sqle Cholesterol syntheisIEENSGMOG00000018991Farnesyl pyrophosphate synthasefpps fdps Cholesterol syntheisIEENSGMOG000000057743-hydroxy-3-methylglutaryl-coenzyme A reductasehmdh hmgcr Cholesterol syntheisIEENSGMOG00000015657Epididymal secretory proteinnpc2 npc2 Cholesterol syntheisIEENSGMOG00000001249Putative adenosylhomocysteinasesahh3 ahcyl2 cardiac defectsDEENSGMOG00000013374Peptide Y OS=Dicentrarchuspy - CNS function and developmentIEENSGMOG00000014820Complement C1q-like proteinc1ql2 c1ql2 CNS function and developmentIEENSGMOG00000017148Augurin-A OS=Danio rerioaugna zgc:112443 CNS function and developmentIEENSGMOG00000001072C-4 methylsterol oxidaseerg25 sc4mol CNS function and developmentIEENSGMOG00000013980Fatty acid-binding protein , brainfabp7 fabp7 CNS function and developmentIEENSGMOG00000014938Maltase-glucoamylase , intestinalmga mgam ATP metabolismIEENSGMOG00000003530ADP/ATP translocaseadt3 slc25a6 ATP metabolismDEENSGMOG00000006172IEF0762 protein C6orf58 homologcf058 - not knownIE Haddock larvae under the same three exposure regimes were transcriptionally profiled at five distinct developmental stages ( L1-5; Figure 4A ) . Relative to embryos , transcriptional responses to crude oil-exposed haddock larvae were more subtle at 1 , 2 , and 9 dph ( L1-3 ) . This was followed by marked changes in gene expression at 14 dph ( L4 ) for the high dose and for all treatments at 18 dph ( L5 ) ( Supporting dataset 2 , Sørhus et al . , 2017 ) . In the high-dose group , nearly 1000 of the >3000 DEGs at L4 and L5 were shared ( Figure 4B ) . However , for the high-dose treatment , only eight genes were shared across all larval stages . As expected , five of these genes are involved in xenobiotic metabolism ( Table 2 ) . 10 . 7554/eLife . 20707 . 010Figure 4 . Exposure regimes and differentially expressed genes ( DEGs ) during larval development . ( A ) Larvae were exposed to a continuous high dose ( black line; 7 . 6 ± 0 . 7 μg/L TPAH ) , a pulsed dose ( red line; 0 . 3 ± 0 . 3–6 . 1 ± 0 . 5 μg/L TPAH ) , and a continuous low dose ( blue line; 0 . 65 ± 0 . 08 μg/L TPAH ) of crude oil . Photos indicate normal developmental progress at each of five sampling time points ( L1–L5 ) . Venn diagrams show shared and exclusive DEGs for each of the three oil exposures at L1–5 . ( B ) Venn diagrams illustrating the number of shared and exclusive DEGs at each larval stage for the three exposure regimes . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 01010 . 7554/eLife . 20707 . 011Figure 4—figure supplement 1 . Most regulated KEGG pathways . ( A ) Pathways ( Total ) with highest number of DEGs ≥2 FC during larval exposure . ( B ) Pathways with the largest fraction of DEGs ≥2 FC/ Total number of genes in pathway ( Normalized ) during and after embryonic exposure . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 01110 . 7554/eLife . 20707 . 012Figure 4—figure supplement 2 . Comparison of mRNA read count data with real time qPCR for selected genes during larval exposure . Genes include cp1a ( cytochrome p450 1 a ) , wnt11 , kcnh2 ( potassium voltage-gated channel subfamily H member 2 ) , cac1c ( voltage-dependent L-type calcium channel ) , nac1 ( sodium/calcium exchanger 1 ) , and at2a2 ( sarcoplamsic-endoplasmic reticulum calcium ATPase ) . ( A ) Real-time qPCR . ( B ) Read count data from RNA sequencing . Data were normalized as described in Materials and methods . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 01210 . 7554/eLife . 20707 . 013Table 2 . Genes expressed at all stages during larval exposure ( L1–L5 ) in high-dose group . SP , swissprot; GB , genebank; IE; increased expression; DE; decreased expression . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 013Cod ID Swissprot annotation SP ID GB ID Category Regulation ENSGMOG00000009114Aryl hydrocarbon receptor repressorahrr ahrr Xenobiotic metabolismIEENSGMOG00000020141Cytochrome P450 1B1cp1b1 cyp1b1 Xenobiotic metabolismIEENSGMOG00000006842Cytochrome P450 1B1cp1b1 cyp1b1 Xenobiotic metabolismIEENSGMOG00000019790Cytochrome P450 1B1cp1b1 cyp1b1 Xenobiotic metabolismIEENSGMOG00000000318Cytochrome P450 1A1cp1a1 cyp1a1 Xenobiotic metabolismIEENSGMOG00000014967Keratinocyte growth factorfgf7 fgf7 Myocardial development and tissue repairIEENSGMOG00000020500Forkhead box protein Q1foxq1 foxq1 Transcription factorIEENSGMOG00000000218Ammonium transporter Rh type Arhag rhag Gas transportIE Read count data from the RNA-Seq closely matched expected abundances based on tissue-specific expression patterns for orthologous genes in zebrafish , available in the expression database at www . zfin . org ( Supplementary file 1A ) , and generally correlated with the overall mass of the contributing tissues . Genes known to have tightly restricted cardiac expression generally had read count values below 100 , with bmp10 and kcnh2 just above detection limits ( 10 reads ) . At the onset of expression in zebrafish , bmp10 transcripts are detected by in situ hybridization in perhaps fewer than 100 cells ( Laux et al . , 2013 ) . In contrast , bmp4 is more widely expressed in zebrafish at the segmentation stage , including the eye , tailbud , and epidermis in addition to the heart . As expected , this gene had a correspondingly higher read count ( 265 ) in haddock . Genes expressed more strongly throughout the entire heart tube had read counts above 60 ( e . g . nkx2 . 5 at 85 and fhl2 at 65 ) . The cardiac-specific Serca2 isoform ( atp2a2 ) had a read count of 312 , while the isoform expressed in skeletal muscle ( atp2a1 ) had a read count of 9114 . Similarly , read counts for the atrial-specific isoform of myosin heavy chain ( myh6 ) and the major skeletal muscle isoform myh1 were 176 and 2543 , respectively . Genes expressed in the neural tube , a tissue mass much larger than the heart but less than the myotomes , had intermediate read counts that also fit with their relative expression patterns . For example , pax6 and nkx2 . 2 had read counts of 1773 and 333 , respectively , with pax6 expressed in a fairly wide dorsal domain of the neural tube , and nkx2 . 2 expressed in a narrower ventral region . We characterized pathways affected by oil exposure using three methods: extensive manual curation , KEGG Pathway Mapping , and Ingenuity Pathway Analysis ( IPA; see Materials and methods ) . As detailed in the following sections , our manual curation identified specific patterns of gene expression in the context of cardiotoxicity , craniofacial deformities , disrupted ion and water balance , and disrupted cholesterol homeostasis . These same pathways were identified with statistical rigor using IPA and KEGG . Moreover , IPA demonstrated enrichment for these pathways at developmental time points that preceded the onset of visible phenotypes , and a lack of enrichment for pathways associated with structures that were phenotypically normal ( Table 3 ) . 10 . 7554/eLife . 20707 . 014Table 3 . Time course of pathway enrichment relating to affected and unaffected developmental and functional phenotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 014Phenotype†Development stage3 dpf/E16 dpf/E210 dpf/E311 dpf/E40 Dph/E5three Dph/E6Cardiovascular0 ( 0/8 ) 22 . 4 ( 11/49 ) 5 . 7 ( 4/70 ) 7 . 0 ( 4/57 ) 4 . 7 ( 2/43 ) 2 . 1 ( 1/48 ) Craniofacial0 ( 0/8 ) 12 . 2 ( 6/49 ) 10 ( 7/70 ) 5 . 3 ( 3/57 ) 7 . 0 ( 3/43 ) 2 . 1 ( 1/48 ) Liver12 . 5 ( 1/8 ) 0 ( 0/49 ) 5 . 7 ( 4/70 ) 8 . 8 ( 5/57 ) 0 ( 0/43 ) 0 ( 0/48 ) Eye0 ( 0/8 ) 4 . 1 ( 2/49 ) 20 ( 14/70 ) 48 . 6 ( 17/35 ) 51 . 2 ( 22/43 ) 50 . 0 ( 24/48 ) Osmoregulation--43 . 3 ( 13/30 ) 29 . 3 ( 12/41 ) 15 . 0 ( 3/20 ) 16 ( 4/25 ) --Cholesterol0/300 ( 0/27 ) 27 . 1 ( 13/48 ) 31 . 3 ( 10/32 ) 25 . 5 ( 12/47 ) --Lipid0/3040 . 7 ( 11/27 ) 35 . 4 ( 17/48 ) 50 . 0 ( 16/32 ) 48 . 9 ( 23/47 ) --Percentage of total enriched pathways ( absolute values ) . †Numbers of affected pathways representing Cardiovascular , Craniofacial , Liver and Eye were extracted from the combined Development category in IPA results; numbers of pathways representing osmoregulation/ion transport were extracted from the Molecular Transport category; numbers of pathways affecting Cholesterol/sterol metabolism and other non-cholesterol lipids ( Lipid ) were extracted from the Lipid Metabolism category . At all stages , the IPA subcategory of Organismal Development or Embryonic Development ( henceforth combined as Development ) was in the top 5 Diseases and Bio Functions category under Physiological System Development and Function with p values ranging from 10−3 to 10−19 ( Tables 2 and 3 ) . Counts of the number of pathways specifically involving the cardiovascular system showed that no pathways were affected at E1 ( 3 dpf , 50% epiboly/cardiac progenitor stage ) , while the heart represented 22% of the affected Development pathways at 6 dpf/E2 , the cardiac cone stage at which there was no visible phenotype ( Table 3 ) . The number of cardiovascular pathways fell to 5 . 7% at 10 dpf/E3 , by which point hearts were visibly abnormal , falling to only one or two affected pathways at hatching stages ( E5 and E6 ) . Pathways specifically related to head , face , or skull development were similarly enriched at all stages except 3dpf/E3 , representing 12% and 10% at 6 dpf/E2 and 10 dpf/E3 , prior to visible differences in head structures . Importantly , at 6 dpf/E2 , prior to the onset of both visible cardiac and craniofacial defects , the top 10 enriched pathways under Development included three involving head development and two involving heart development ( p values 10−6 to 10−12; Supplementary file 1B ) . In contrast , pathways relating to liver development were enriched at 5 . 7% and 8 . 8% at only two time points , 6dpf/E3 and 10 dpf/E4 , and these did not appear in the top 10 . Moreover , inspection of individual DEGs associated with those pathways showed genes involved in lipid transport rather than bona fide regulators of liver development ( see below ) . The single pathway relating to the liver at 3 dpf/E1 was represented by a single gene , cyp1a . At these stages , these lipid transport genes are most strongly expressed in the yolk syncytial layer . Notably , IPA also detected larger scale enrichment of eye genes , almost all down-regulated , accounting for roughly 50% of developmental pathways during pigmentation of the retina , but prior to obvious differences in eye sizes after hatch . Genes associated with osmoregulation were identified by IPA under the Molecular Transport category ( Diseases and Bio Functions , Molecular and Cellular Functions ) . We quantified pathways relating to specific ions ( e . g . , Na+ , K+ ) , inorganic ions , and metals ( Table 3 ) . At E1/3 dpf Molecular Transport was not in the top five affected pathways , but became enriched at 43% at E2/6 dpf , with primarily down-regulation of genes prior to the onset of visible edema . These Molecular Transport pathways remained significantly enriched ( 29% , 15% , 16% ) until onset of hatch ( E5 ) . By 3 dph/E6 , Molecular Transport pathways dropped below the top 5 . Pathways related to cholesterol and other lipids ( e . g . phospholipids , fatty acids ) followed a similar pattern as osmoregulation . As for Molecular Transport , Lipid Metabolism was consistently in the top five Molecular and Cellular Functions category . Pathway enrichment was overall at the highest levels for Lipid Metabolism . We separately quantified individual pathways relating to sterols ( e . g . cholesterol synthesis , cholesterol transport , steroid biogenesis ) and other fundamental ( non-signaling ) lipids ( e . g . fatty acid synthesis and transport , glycerolipids , phospholipids ) ( Table 3 ) . General lipid metabolism pathways were highly enriched at 6 dpf/E2 ( 41% ) and remained high ( 35–50% ) until 3 dph/E6 when Lipid Metabolism pathways fell below the top 5 . Cholesterol metabolism pathways were first enriched at 10 dpf/E3 at 27% , remaining at 31% and 26% until 3 dph/E6 , when they also fell below the top 5 . Notably , all Lipid Metabolism pathways were enriched prior to measureable detection of reduced yolk absorption at 3 dph . Pathway enrichment was dose-dependent and clearly associated with the frequencies of abnormal phenotypes ( Supplementary file 1C ) . For example , the combined general Development categories at 6 dpf included 203 , 121 , and 0 pathways ( among the top five general categories ) for the high , pulse , and low doses , respectively . At this stage , Molecular Transport pathways were enriched at levels of 117 , 80 , and 0 for the high , pulse , and low doses , respectively . At 10 dpf/E3 , the Lipid Metabolism category included 115 , 28 , and 5 pathways for the high , pulse , and low doses , respectively . Finally , Cardiotoxicity pathways were prominent for nearly all times points for each dose ( Supplementary file 1C ) . For high , pulse , and low doses , numbers of enriched pathways at 6 dpf/E2 were , respectively , 63 , 33 , and 17; at 10 dpf/E3 , 79 , 10 , and 5; at 11 dpf/E4 , 92 , 40 , and 9; at 0 dph/E5 , 69 , 46 , and 26; and at 3 dph/E6 , 34 , 19 , and 12 . We identified 10 individual genes ( Supplementary file 1D ) and KEGG pathways ( Figure 3—figure supplement 1 and Figure 4—figure supplement 1 ) that were the most highly responsive ( highest positive or negative fold change ) to the high oil treatment regime relative to unexposed control fish . Briefly , in both embryos and larvae , DEGs involved in xenobiotic metabolism and stress response were highly represented . Genes involved in the development and function of neural networks and cholesterol/steroid biosynthesis were upregulated , while genes involved in intracellular calcium signaling were primarily downregulated . In order to investigate both pathways with numerous and few genes , we chose two different approaches for KEGG pathways analysis ( 1 ) Total: Pathways with the highest number of DEGs ≥ 2 FC ( Figure 3—figure supplement 1A and Figure 4—figure supplement 1A ) and ( 2 ) Normalised: Pathways with the largest fraction of DEGs ≥2 FC/ Total number of genes in pathway ) ( Figure 3—figure supplement 1B and Figure 4—figure supplement 1B ) . During embryonic exposure pathways associated with PAH metabolism were represented among the most affected . Indicative of disrupted osmoregulation and ion channel blockade , secretion pathways and calcium signaling showed decreased expression at the earliest stages . Further , we observed increased expression in steroid metabolism and biosynthesis pathways suggesting an effect on cholesterol metabolism ( Figure 3—figure supplement 1B ) . Post exposure , we observed increased expression of several pathways suggestive of an inflammatory response ( protein digestion and degradation , influenza A , antigen processing and presentation ) , while expression of genes in phototransduction pathway was decreased ( Figure 3—figure supplement 1A ) . During larval exposure at the first three sampling stages a small number of genes , and thus , pathways were regulated and most were participating in PAH metabolism . Consistent with total number of DEGs ( Figure 4B ) , stage L4 and L5 included pathways with higher number of DEGs ≥ 2 FC . Most noticeable was the decreased expression in calcium signaling pathway and hypertrophic cardiomyopathy ( HCM ) pathway at 14 dph and decreased expression in pancreatic secretion and protein digestion and absorption ( Figure 4—figure supplement 1A ) and steroid biosynthesis pathways ( Figure 4—figure supplement 1B ) at 18 dph . Finally , four genes stood out as unique for ( 1 ) being highly upregulated in both embryos and larvae , ( 2 ) their non-affiliation with a larger network or pathway , and ( 3 ) their potential connections to visible phenotypes . These included collagen and calcium-binding EGF-like domain 1 ( ccbe1 ) , the ammonia transporter rhag , forkhead box transcription factor foxq1 , and fibroblast growth factor fgf7 ( Table 1 , Supplementary file 1D ) . The following sections identify specific patterns of gene expression in the context of cardiotoxicity , craniofacial deformities , disrupted ion and water balance , and disrupted cholesterol homeostasis . As noted above , early formation of the heart was not affected by crude oil exposure , but morphological defects followed after functional defects were first observed at 9 dpf ( bradycardia ) . In addition , morphology became more severely impacted over time , with later defects including failure of looping and poor ventricular growth becoming prominent by 0 dph/E5 . There are several possible etiologies for ventricular size reduction . For example , although the precise mechanism by which intracellular calcium regulates embryonic cardiomyocyte proliferation is still unknown , a disruption of calcium cycling could reduce proliferation , thereby yielding fish with smaller hearts ( Rottbauer et al . , 2001; Ebert et al . , 2005 ) . We therefore focused on genes involved in cardiac morphogenesis . The earliest alteration was a fourfold increase in the signaling molecule , bmp10 at 6 dpf/E2 while the heart was at the midline cone stage , and appeared unaffected in oil-exposed embryos ( Figure 5 , Supplementary file 1E ) . IPA also identified Bmp signaling as a significantly enriched pathway at this time point , under the Organismal Development category . Elevation of bmp10 was followed by the upregulation of the cardiac transcription factor nkx25 , to a threefold increase at 10 dpf/E4 and a nearly sixfold increase at 11dpf/E5 , when the heart was beating regularly . At hatch ( 0 dph ) , the expression of the transcription factor tbx3 was elevated eightfold . Lastly , atrial natriuretic factor ( nppa ) , a key homeostatic regulator of contractility , was downregulated by 2 . 3-fold in larvae at 3 dph ( Figure 5 , Supplementary file 1E ) . Notably , overexpression of bmp10 , nkx25 , or tbx3 is associated with serious heart defects in other vertebrates ( Chen et al . , 2006; Ribeiro et al . , 2007; Tu et al . , 2009 ) . 10 . 7554/eLife . 20707 . 015Figure 5 . DEGs involved in cardiogenesis . Regulation of genes involved in cardiogenesis during and after embryonic exposure . Purple: increased expression , red: decreased expression in exposed group . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 015 Crude oil exposures caused functional defects in the developing haddock heart , in the form of bradycardia , ventricular asystole and decreased contractility in embryos and partial atrio-ventricular conduction blockade in larvae ( Sørhus et al . , 2016b ) . This is consistent with disruption of the rhythmic fluxes of Ca2+ and K+ ions that regulate E-C coupling in heart muscle cells ( Brette et al . , 2014 ) . We therefore focused on genes associated with cardiomyocyte membrane potential and intracellular Ca2+ cycling—for example , sarcoplasmic reticulum calcium ATPases ( SERCAs ) and the ryanodine receptor ( RyR ) ( [Sørhus et al . , 2016a] , Supplementary file 1F and 1G ) . We found three paralogs for at2a2 ( Serca2 ) that were present at very different read count values ( ~300 , 900 , and 4000 at 6 dpf/E2 ) . The two more abundant paralogs were transiently down-regulated in oil-exposed embryos at 6 dpf/E2 , prior to the onset of functional and morphological defects , while the third paralog was down-regulated at 0 dph/E5 ( Figure 6 , Supplementary file 1H ) . Similarly , there were four nac1 paralogs that were all low abundance , and one was transiently down-regulated with the at2a2 genes at 6 dpf . Finally , the kcnh2 gene contributing to the repolarizing K+ current was down-regulated ninefold at hatch/E5 . There were effects on a few other E-C coupling genes , but these had very high read counts , and are therefore likely to be associated with skeletal muscle . These included two at2a1 ( serca1 ) paralogs that had opposite responses , and atpa . A different picture emerged from the larval exposure . Changes in expression of E-C coupling genes occurred after the onset of functional defects ( AV block arrhythmia ) . No changes in cardiac E-C coupling genes were observed at the L3/9 dph time point when larvae showed AV block . Six days later ( L4 ) , there was fourfold down-regulation of a nac1 paralog with the lowest read count value and an at2a2 paralog with the highest value . At this stage there was up-regulation of two kcnj2 paralogs ( encoding potassium channels ) , two high abundance at2a1 paralogs , and a low abundance scn2a paralog . At 18 dph/L5 , one paralog each of at2a1 and kcnj2 remained elevated , while atpa was elevated , and the kcnj12 potassium channel gene were down-regulated , the latter strongly ( ~6 fold ) . Craniofacial structures that shape the head include cartilage derived from neural crest cells and muscles that develop from paraxial mesoderm . Neural crest cells migrate from the anterior neural tube to form the pharyngeal arches with both dorsal ( upper jaw ) and ventral ( lower jaw ) patterning . They then differentiate into chondrocytes ( Knight and Schilling , 2006; Simões-Costa and Bronner , 2015 ) and grow by processes such as convergence-extension ( Shwartz et al . , 2012; Kamel et al . , 2013 ) . Concurrently , mesodermal cells differentiate into patterned muscle in appropriate association with partner cartilage . Several lines of evidence suggest multidirectional signaling between all associated tissues , including endoderm ( i . e . pharyngeal pouches ) , mesoderm , and overlying ectoderm ( Minoux and Rijli , 2010; Medeiros and Crump , 2012; Kamel et al . , 2013; Kong et al . , 2014 ) . Studies on zebrafish craniofacial mutants have primarily focused on the neural crest cell lineage , with less attention to muscle development or interactions between developing muscle and cartilage ( Lin et al . , 2013 ) . Defects in oil-exposed haddock were marked by a dose-dependent reduction in more anterior cartilages ( Figure 1E–G ) . This affected the basicranium most severely , with progressive loss of more posterior arch derivatives . Where present , craniofacial cartilage appeared small and distorted , without transformation to dorsal or ventral fates . This morphometry superficially aligns to several zebrafish mutants affecting either neural crest cell ( Kimmel et al . , 2001; Nissen et al . , 2006; Lu and Carson , 2009; Kamel et al . , 2013 ) or muscle development ( Hinits et al . , 2011; Shwartz et al . , 2012 ) . We therefore interpreted developmental changes in haddock gene expression in the context of these well-characterised zebrafish mutants . The expression patterns of 12 genes with known roles in neural crest cell-dependent craniofacial development were significantly altered in the highest exposure regime ( Figure 6 ) . Read counts for these genes were all relatively low , consistent with highly restricted tissue-specific expression patterns ( Supporting dataset 1 , Sørhus et al . , 2017 ) . At 6 dpf , prior to visible craniofacial malformation , we observed lower expression levels of foxi1 ( pharyngeal pouches ) wnt9b ( ectoderm ) , fgfr2 ( chondrocytes ) , and fgfr3 ( chondrocytes ) compared to control ( Figure 7A ) . The downregulation of wnt9b and fgfr2 persisted to 10 dpf , together with a downregulation of tgfb3 and two sox9b paralogues , the latter also expressed in neural crest cell-derived chondrocytes . Conversely , edn1 , dlx3b , and dlx5a were upregulated at 10 dpf . In zebrafish embryos the two dlx genes are normally expressed in endoderm and arch neural crest cell-derived mesenchyme ( Talbot et al . , 2010 ) . At 11 dpf , expression levels were down for fgfr2 , fzd7a ( a receptor for Wnt9b; chondrocytes ) , and tgfb3 and up for edn1 and dlx3b . None of these genes were differentially expressed after hatching ( Figure 7A , Supplementary file 1I ) . 10 . 7554/eLife . 20707 . 016Figure 6 . DEGs involved in E–C coupling . Embryonic developmental samples ( E1–6 ) were collected during ( black lettering ) and after ( blue lettering ) crude oil exposure . Oil exposure was continuous across the larval sampling points ( L1–5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 01610 . 7554/eLife . 20707 . 017Figure 7 . DEGs involved in craniofacial development . ( a ) Regulation of genes involved in craniofacial development during and after embryonic exposure . ( b ) Regulation of myosin heavy chain genes . Purple: increased expression , red: decreased expression in exposed group . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 017 Genes controlling craniofacial muscle patterning are poorly characterised; however , muscle determination factors ( e . g . , myod , myf6 [Li et al . , 2014] ) appeared unaffected in oil-exposed haddock . Nevertheless , expression levels were significantly reduced for genes involved in the terminal differentiation of skeletal muscle cells , including several myosin heavy chain genes ( myh ) ( Figure 7B , Supplementary file 1J ) . These included myh4 , myh9 , and myh10 paralogues at 6 dpf , myh9 and another paralogue of myh4 and myh10 at 10 dpf , and myh3 , myh4 and myh9 at 11 dpf . Only myh3 remained significantly downregulated after hatching ( 3 dph ) relative to controls . Notably , expression of myh1 , encoding the major fast myosin heavy chain gene expressed in the body musculature ( Thisse et al . , 2001 ) , was largely unaffected except for a small reduction at 11 dpf . Other myosin genes specific to muscle groups in the head and trunk ( Peng et al . , 2002; Elworthy et al . , 2008 ) , on the other hand , showed more complex differential expression patterns . Fluid accumulation in the form of edema is a hallmark indication of crude oil toxicity in fish embryos . Although patterns of edema formation vary across freshwater and marine fish species ( Incardona and Scholz , 2016 ) , it nearly always involves anatomical compartments adjacent to the heart and the yolk sac . However , marine embryos are hyposmolar to surrounding seawater , and they should therefore lose water along a diffusion gradient if osmoregulation is disrupted as a consequence of heart and circulatory failure . In fish embryos and yolk sac larvae , osmoregulation is controlled by MRCs in the epidermis and yolk sac membrane that actively secrete NaCl ( specifically Cl- ) to maintain an appropriate water and ion balance ( Hiroi et al . , 2005 ) . Genetic and pharmacologic studies have shown that circulation is required to maintain embryonic MRC cell function . For example , total body osmolality increased in seawater-adapted tilapia embryos with a ~50% reduction in total cardiac output ( Miyanishi et al . , 2013 ) . Therefore , edema formation in marine species with oil-induced circulatory defects is not a consequence of water moving into the embryo ( as it is for freshwater species ) but rather water moving along an internal osmotic gradient from peripheral tissues to the vicinity of the heart and yolk sac . Accordingly , dorsal anterior finfold defects in edematous embryos and larvae ( Figure 1H , I ) represent a visible phenotypic anchor for ionoregulatory disruption , a third distinct oil-induced adverse outcome pathway . Our analysis focused on key ionoregulatory proteins in MRCs and their associated genes , including Na+/K+ ATPase subunits ( at1 genes , e . g . at1a1-a4 , at1b1-b4 ) , a urea transporter ( ut1 ) , a Na+/K+/2Cl- co-transporter ( s12a2 ) , the sodium-hydrogen exchanger Nhe3 ( slc9a3 ) , and a chloride channel , the latter an ortholog to the human cystic fibrosis transmembrane conductance regulator ( cftr ) ( Hirose et al . , 2003 ) . The disruption of MRC function in oil-exposed haddock embryos corresponded to significantly lower levels of at1a1-3 , at1b2-3 , ut1 , s12a2 , sl9a3 , and cftr ( Figure 8 , Supplementary file 1K ) . This downregulation primarily spanned a developmental window between 6 dpf and hatching ( Figure 8 , Supplementary file 1K ) . We also found significant transcriptional modifications of genes encoding other pumps , channels , and transporters specific to the nervous system and other tissues , including the aquaporins ( aqp genes ) that rapidly transport water across cell membranes ( Supplementary file 1L ) . For example , there was a pronounced decrease in the expression of the primary neuronal water channel , aqp4 , from 6 dpf to hatching and a strong upregulation of aqp12 at hatch ( Figure 8 ) . Crude oil exposures therefore appear to cause osmotic stress in the developing embryonic nervous system . 10 . 7554/eLife . 20707 . 018Figure 8 . DEGs involved in osmoregulation . E1–E6: Embryonic exposure , L1–L5: Larval exposure . Black letters: during exposure , blue letters: after exposure . DOI: http://dx . doi . org/10 . 7554/eLife . 20707 . 018 During larval exposure , edema accumulated in different compartments from embryos , and there were corresponding differences in expression of genes related to ion and water balance . At late stages of larval exposure , edema accumulated in the peritoneal cavity , and the dorsal marginal finfold appeared increased rather than decreased as in embryos . Fewer ion transport genes were affected , with increased expression observed for only at1a3 at 14 dph , and one at1b2 paralogue and at1b3 at 14 and 18 dph . Similarly , aquaporin genes were affected differently . Whereas aqp4 was unaltered in larvae , expression of aqp7 and aqp9 was increased while aqp3 and aqp12 were decreased ( Figure 8 , Supplementary file 1K ) . Cholesterol is an essential structural component required for maintaining both the integrity and the fluidity of all metazoan cell membranes . Cholesterol is sourced from de novo cellular synthesis and from the uptake of external lipoprotein cholesterol from the circulation ( Bjorkhem and Meaney , 2004 ) . During fish development , cholesterol is mobilized from the yolk and distributed to cells during embryonic and larval yolk sac stages . Later , after the yolk is absorbed and larvae begin exogenous feeding , cholesterol is transported from the intestines . Crude-oil-exposed haddock embryos and larvae with the most severe morphological abnormalities were visibly unable to effectively mobilize yolk ( Figure 1J , K ) . Moreover , larvae from the highest exposure concentration had less visible food in their stomachs relative to controls . These observations together suggest that chemical components of crude oil may deprive developing tissues of externally available cholesterol . Of the 28 genes differentially regulated at all developmental time points , 5 are involved in cholesterol synthesis and feedback control ( Table 1 ) . These include 3-hydroxy-3-methylglutaryl-coenzyme A reductase , an enzyme encoded by hmdh that plays a primary feedback regulation role in the cholesterol biosynthetic pathway ( Brown and Goldstein , 2009 ) . Although reduced yolk absorption was not physically measureable in exposed embryos until after 3 dph , genes controlling cholesterol synthesis were upregulated much earlier , prior to visible cardiac circulation ( 6 dpf/E2 ) . We also detected complex regulation of apolipoproteins during and after exposure , with mainly down-regulation of apob paralogs before first heartbeat ( 6 dpf/E2 ) and up-regulation of apoa4 , apod apoeb and apoc2 after initiation of cardiac circulation ( Supporting dataset 1 , Sørhus et al . , 2017 ) . Scavenger receptor class B 1 ( encoded by scarb1 ) , a transcytotic receptor for cholesterol-containing high-density lipoprotein ( Acton et al . , 1996 ) , was also down-regulated in the exposed groups in haddock at 6 dpf ( Supporting dataset 1 , Sørhus et al . , 2017 ) . Pathway analysis was also consistent with a significant effect on cholesterol homeostasis ( Figure 3—figure supplement 1 ) . In the larval exposure , we also detected increased expression of hdmh , erg1 , cp51a , and npc2 ( encoding the proteins squalene epoxidase , and cytochrome P450 51A , Niemann-Pick disease , type C2 , respectively ) at the latest stages examined ( 14 and 18 dph ) . Conversely , pathways involved in digestion – that is , pancreatic secretion , protein digestion , and protein absorption – were suppressed . This includes the downregulation of genes encoding digestive enzymes such as trypsin and chymotrypsin ( Figure 4—figure supplement 1 ) . The stomachs of oil-exposed larvae at the final time point were relatively empty , and the associated loss of food-derived nutrients likely triggered the observed increase in endogenous cellular cholesterol synthesis . Whereas abnormal phenotypic traits corresponded to differential gene expression , genes associated with normal traits were unchanged . For example , the lateral line and liver appeared normal in the most severely affected embryos ( Figure 1—figure supplement 1 ) . Consistent with this , markers for the lateral line ( protein atonal homolog 1 , atoh1 ) ( Cai and Groves , 2015 ) , liver growth ( hepatocyte growth factor , met ) and differentiation ( genes encoding wnt2 and 2b protein ( wnt2 , wnt2b ) ( Wilkins and Pack , 2013 ) , hematopoietically-expressed homeobox protein ( hhex ) , and protein heg ( heg ) ) ( Supplementary file 1M ) were not significantly modified . While some markers for liver differentiation , including genes encoding transferrin ( tfr1 ) and fatty acid binding protein ( fa10a ) were differentially expressed , the changes were subtle and not consistent throughout development . Similarly , the related KEGG pathways that are inclusive of these genes were relatively unaffected by oil exposure at all time points . As noted above , IPA failed to identify significant enrichment of pathways related to phenotypically normal structures . Overall , we observed tight anchoring of temporal gene expression patterns to measurable phenotypes in crude oil-exposed haddock . First , the global changes in gene expression observed in the embryonic and larval exposures matched the general nature and severity of phenotypes . Embryonic exposure to crude oil or component cardiotoxic PAHs produces a coarse chemical phenocopy of the loss-of-function zebrafish mutants affecting heart function or development ( Incardona et al . , 2004 ) . Many aspects of the oil toxicity phenotype are secondary to a loss of circulation—that is , defects in non-cardiac tissues , such as the eye , that require circulation for normal organogenesis ( Incardona et al . , 2004 ) . In contrast , larval stages are primarily a period of rapid growth after major organogenesis is complete , and the injury phenotype in larvae is less severe . Consistent with this , embryos showed a larger number of DEGs than larvae , with a preponderance of down-regulation . Second , we identified specific changes in the expression of key genes involved in the function or morphogenesis of individual tissues and organs with visible abnormalities . Given unaltered gene expression and lack of statistically enriched pathways associated with apparently unaffected structures such as the liver , the DEGs in oil-exposed haddock indicate a disruption of specific developmental processes , as opposed to non-specific effects ( e . g . general developmental delay ) . This study demonstrates the ability to resolve changes in tissue-specific genes in a pool of total RNA from embryos and larvae , even for organs such as the heart that contribute a very small fraction to total tissue mass . A key finding is the general correlation of read count data with tissue specific patterns previously characterized in model species . Our findings have important implications for the utility of RNA-Seq and other quantitative measures of mRNA abundance in whole embryo or larval samples . For example , this demonstrates the feasibility of developing real-time monitoring tools based on quantification of gene expression in environmental samples collected following an oil spill . In addition , our extensive manual curation of the transcriptome groundtruths the utility of applications like IPA for use with non-model , non-mammalian organisms . Moreover , the changes in gene expression identified here represent significant information to be added to existing cardiotoxicity AOPs and novel AOPs associated with disruption of osmoregulation and lipid metabolism . Two major initiating events for crude-oil-associated cardiac defects during fish development are chemical blockade of IKr repolarizing potassium currents , ( encoded by kcnh2 ) and disruption of intracellular calcium handling , the latter culminating in sarcoplasmic reticulum ( SR ) calcium depletion through effects on either RyR or SERCA2 ( encoded by ryr2 and at2a2 , respectively ) ( Brette et al . , 2014 ) . In the fully formed heart , these pharmacologic effects impair cardiac function by inducing arrhythmia and reducing contractility ( Incardona et al . , 2009 , 2014 ) . However , rhythm and contractility defects during heart development lead to morphological defects ( Andrés-Delgado and Mercader , 2016 ) . In haddock embryos , these include poor chamber looping and outgrowth of the ventricle ( Sørhus et al . , 2016b ) . Our data demonstrate a transcriptional cascade that is tightly linked to these defects in cardiac function ( cardiomyocyte intracellular calcium cycling ) and form ( heart chamber growth ) through bmp10 . While chemical blockade of calcium cycling alone would be sufficient to induce the ventricular arrhythmias observed in oil-exposed embryolarval haddock , other elements of the E-C coupling physiological cascade were also selectively modified at the mRNA level . As shown previously using qPCR ( Sørhus et al . , 2016b ) , RNA-seq revealed a downregulation of genes encoding the Na/Ca exchanger ( nac1 ) and IKr ( kcnh2 ) in haddock embryos . Notably however , kcnh2 downregulation was only observed at later time points , in response to the highest oil exposures that caused the most severe phenotypes . There was no consistent decrease in the mRNA for a larger suite of proteins involved in cardiac E-C coupling . Assuming a broader transcriptional response in the heart was not masked by more abundant , normal expression of these genes in larger non-cardiac tissues , other non-specific mechanisms were unlikely to contribute to the formation of misshapen hearts . Moreover , the changing expression of key E-C coupling genes is a close match to the cardiac arrhythmia phenotype in both embryos and larvae . This includes a marked downregulation ( >5 fold ) of kcnj12 , which encodes a subunit of the repolarizing IK1 current and causes the same types of ventricular arrhythmias as a reduction of kcnh2 ( Domenighetti et al . , 2007 ) . At the same time , the up-regulation of E-C coupling genes following the chemical induction of arrhythmia in the larval exposures suggest that the more mature larval heart mounts a compensatory response . Intracellular calcium has multiple direct roles in regulating gene expression , including the process of excitation-transcription ( E-T ) coupling ( Wamhoff et al . , 2006 ) . Our findings suggest that E-T coupling may link reduced cardiomyocyte calcium cycling to structural defects in the haddock heart . Among vertebrates , bmp10 is expressed exclusively in the early tubular hearts of zebrafish , mouse , and chick embryos . The normal function of Bmp10 in the developing heart is primarily to drive ventricular cardiomyocyte proliferation during trabeculation ( Grego-Bessa et al . , 2007 ) , a relatively late process during embryogenesis ( around hatching stages in fish ) . Both loss of and excess Bmp10 leads to severe abnormalities in ventricular development in mouse ( Chen et al . , 2004 ) . In mice lacking the RyR-associated Fkbp12 protein , disruption of SR calcium handling leads to ventricular defects through elevated bmp10 transcription ( Shou et al . , 1998; Chen et al . , 2004 ) , probably through calcium-dependent activation of myocardin ( Wamhoff et al . , 2004 , 2006 ) , the transcriptional activator of bmp10 ( Huang et al . , 2012 ) . Moreover , Bmp10 is the most potent Bmp family member , showing greater resistance to Bmp antagonists ( e . g . Noggin ) than Bmp4 ( Lichtner et al . , 2013 ) , the primary cardiac Bmp family member at early stages . While bmp10 normally functions at late stages of cardiogenesis , bmp4 is normally expressed at the cardiac cone stage in zebrafish . At this stage , bmp4 levels shift from radially symmetric to elevated on the left side of the cone , to drive proper looping ( Chen et al . , 2004 ) . Loss of this asymmetry with ectopic bmp4 leads to un-looped hearts . Therefore , the premature up-regulation of a more potent family member , bmp10 , at the cone stage is very likely to underlie the looping defects observed in oil-exposed embryos . Further evidence for bmp10 overexpression initiating abnormal cardiac morphogenesis is the secondary up-regulation of the Bmp10 target gene nkx25 ( Chen et al . , 2004 ) . In zebrafish , nkx25 overexpression or loss of function ( Tu et al . , 2009 ) yields a reduced ventricle , and nkx25 must be down-regulated or antagonised in specific regions of the ventricle in order to form specialised conduction cells through the repressor action of Tbx3 ( Hoogaars et al . , 2004 ) . The higher levels of tbx3 that follow upregulation of nkx25 and subsequent downregulation of anf thus likely reflect an imbalance between myocardial and non-myocardial cell fates within the ventricle . Thus , normal heart development in zebrafish requires tight control over bmp10 , nkx25 , and tbx3 expression , and all three genes were dysregulated in oil-exposed haddock . The observed ventricular and looping defects may represent chemical phenocopies of the fkbp12 mutant , wherein reduced intracellular calcium transients are linked to altered bmp10 expression by E-T coupling , thereby changing cell fate and chamber formation in the developing heart . Calcium-mediated E-T coupling may also be a feedback mechanism for altering the expression of genes that encode repolarizing potassium channels . Although the haddock with craniofacial deformities superficially resemble certain zebrafish mutants , associated changes in gene expression suggest a more complex developmental perturbation than previously described . As is the case with Bmp10 , Edn1 is a strong morphogen that must be tightly regulated , as both too much and too little lead to craniofacial defects ( Sato et al . , 2008; Clouthier et al . , 2010 ) . Higher levels of edn1 observed here are thus highly likely to be related to the craniofacial defects , which is supported by the subsequent up-regulation of its target dlx genes . However , the phenotype does not appear to reflect changes in dorsal-ventral patterning , as expected for perturbation of edn1-dependent NCC identity . Most studies of craniofacial development in zebrafish and other vertebrates have focused on NCC-derived cartilaginous precursors . However , craniofacial skeletal elements develop in synchrony with their associated muscles ( Schilling and Kimmel , 1997 ) , and defects in muscle formation or function produce phenotypes that are in many ways indistinguishable from primary cartilage defects and appear more similar to the phenotypes observed here ( Hinits et al . , 2011; Shwartz et al . , 2012 ) . Mesodermal precursors of craniofacial muscle cells express edn1 , the downregulation of which is associated with terminal muscle differentiation ( Choudhry et al . , 2011 ) . A failure to downregulate edn1 is consistent with failure to up-regulate or maintain myh gene expression . It is unknown whether a failure of skeletal muscle to terminally differentiate would lead to reduced expression of genes associated with NCC development and cartilage growth and survival—that is foxi1 , Wnt pathway genes ( wnt9b , fzd7a , mycn ) , Fgf receptors and tgfb3 . On the other hand , the early expression of foxi1 has no clear linkage to craniofacial muscle development in the literature , but it is required for NCC survival indirectly by driving Fgf signaling ( Edlund et al . , 2014 ) . Sorting out the precise pathways of craniofacial malformation will thus require a concerted spatial localization of these DEGs . The regulation of genes controlling ion and water balance , combined with the collapse of the dorsal marginal finfold , provides a novel insight into the origins of edema in marine fish . Conversely , the genetic elements of this phenotype provide a starting point to study the molecular basis of buoyancy control in pelagic fish larvae . Our findings suggest that crude oil may disrupt MRC function , leading to decreased expression of MRC channel and transporter genes . These effects could be direct , indirect , or both . First , ion and water balance in embryos require cardiac circulation ( Miyanishi et al . , 2013 ) . Flow is essential for osmoregulatory counter-current exchange in the gills and kidney ( Somero , 1998; Grosell , 2011 ) , and similar principals should apply to the yolk sac epidermis . Second , epidermal MRCs are likely directly exposed to the highest PAH concentrations , because the highest Cyp1a upregulation in response to oil occurs in the skin of embryos and yolk sac larvae ( Sørhus et al . , 2016b ) . PAHs or their metabolites could block solute carriers in a similar manner as shown for cardiac calcium and potassium channels ( Brette et al . , 2014; Incardona et al . , 2014 ) . Lastly , MRC function could be impaired if the metabolic cost of PAH degradation competes with ion transport . Although the effects of oil exposure on salt and water balance in fish embryos have not been examined previously , exposing water-soluble oil fraction to adult Pacific herring ( Clupea pallasi ) showed an effect on ion homeostasis ( Kennedy and Farrell , 2005 ) . Genes involved in ion and water balance were differentially regulated in embryos and larvae , in close correspondence to the loss and expansion of the dorsal subdermal space , respectively . Edema flows along the path of least resistance and accumulates in expandable spaces . In haddock embryos , fluid moves from the dorsal subdermal space to the yolk sac . At the larval stage , the presence of fluid in the dorsal finfold suggests that a developed peritoneal cavity and body wall provide greater resistance than the thin , permeable yolk sac membrane . As a consequence , the central nervous system is likely deprived of water during embryonic development and turgidly stressed at the larval stages . Changes in cell volume can modify intra- and extracellular ionic concentrations , and thus the electrophysiological properties of neurons ( Pasantes-Morales et al . , 2000 ) . This may underlie the observed down-regulation of the main water channel in the brain , aquaporin 4 ( aqp4 ) , during and after embryonic exposure . Although not a focus of the current study , there are likely important mechanistic connections between a loss of embryonic MRC function , disrupted osmoregulation in the brain , and pathophysiological impacts on neuronal development . Our findings also provide new insights into fundamental transcriptional mechanisms of lipid mobilization from yolk . The delivery of yolk-derived cholesterol not been widely studied in early fish embryos , which are distinct from other vertebrates in that the yolk mass is contained separately from the vasculature by the yolk syncytial layer ( YSL , or periblast in older literature ) . The YSL is a multicellular membrane that forms during gastrulation and has been studied mostly for its role in early pattern formation ( Carvalho and Heisenberg , 2010 ) . At later embryonic and larval stages , the YSL transports yolk-derived nutrients into the circulation ( e . g , [Poupard et al . , 2000] ) . Although cholesterol is exported from the yolk prior to established circulation ( Fraher et al . , 2016 ) , the cellular basis for this is not known—for example , by trancytosis or direct membrane transport . Cellular sterol levels are tightly controlled by membrane-bound transcription factors , which are cleaved and activated when membrane cholesterol levels fall , leading to transcription of hmdh , the rate-limiting enzyme for cholesterol synthesis ( Brown and Goldstein , 2009 ) . Therefore , the brisk up-regulation of intrinsic cholesterol biosynthetic genes , especially hmdh , prior to a visible mobilization of yolk , indicates the importance of yolk-derived cholesterol for embryonic tissues . In zebrafish and other fish , apolipoprotein genes ( e . g . ApoB and ApoE ) as well as Scarb1 are first expressed in YSL ( Poupard et al . , 2000; Thisse et al . , 2001; Otis et al . , 2015 ) . The downregulation of multiple apob paralogs and scarb1 during early development suggest a specific defect in packaging and transporting of lipoprotein-cholesterol in the YSL , possibly involving Scarb1-dependent transcytosis . At stages subsequent to heartbeat initiation , the upregulation of intrinsic cholesterol biosynthetic genes likely reflects cholesterol deprivation as a consequence of the heart failing to deliver lipoproteins from the yolk ( embryos ) and intestine ( larvae ) . The identification of broader lipid metabolism pathways by IPA also suggests that oil exposure leads to more global derangements relating to either poor yolk absorption or dysfunction of the YSL . It is well known both that embryonic oil exposure leads to later growth impairment ( e . g . [Heintz , 2000; Incardona and Scholz , 2016] ) and that lipids provide critical fuel for marine fish larvae , particularly at the first-feeding stage ( Tocher et al . , 2003 ) . The consequences of disordered lipid metabolism for larval physiology and survival should be a focus for future studies . Importantly , the induction of cholesterol synthetic genes is a promising and novel indicator of crude oil toxicity to fish embryos . Finally , we consider the four individual genes that were consistently upregulated across all developmental stages: ccbe1 , rhag , foxq1 , and fgf7 . Elevated tissue pressure is a signal for lymphangiogenesis ( Schulte-Merker et al . , 2011 ) and the marked increase in ccbe1 expression is consistent with a compensatory formation of lymphatics to alleviate the physical effects of edema ( Planas-Paz et al . , 2012 ) . In zebrafish , the secreted Ccbe1 protein appears to function exclusively in lymphangiogenesis ( Hogan et al . , 2009; Le Guen et al . , 2014 ) by enhancing the activation of VEGF-C ( Le Guen et al . , 2014 ) . For future spills , quantitative measures of ccbe1 upregulation should serve as very sensitive indicators of edema formation in crude oil-exposed fish embryos . The Rh proteins are primarily structural components of erythrocyte membranes but were also recently identified as ammonia transporters ( Weiner and Verlander , 2014 ) . The rhag , rhbg , and rhcg genes are important for excretion of nitrogenous waste in fish ( Braun et al . , 2009 ) . The strong upregulation of rhag observed in haddock embryos and larvae might be a consequence of MRC dysfunction . The increase in rhag expression corresponded to a downregulation of urea transporters ut1 and ut2 in embryos but not larvae , suggesting metabolic defects relating to amino acid or protein catabolism and an increased demand for nitrogen excretion . Alternatively , rhag may play a novel role in embryolarval physiology . The last two highly responsive genes , foxq1 and fgf7 , may be linked to the craniofacial injury phenotype based on prior work in other species . In zebrafish and frog embryos , foxq1 is expressed in the pharyngeal region ( Choi et al . , 2006; Planchart and Mattingly , 2010 ) . In chick embryos , fgf7 is first expressed in the pharyngeal endoderm and head mesoderm ( Kumar and Chapman , 2012 ) . However , foxq1 has been shown to be a downstream target of AhR in other tissues ( Faust et al . , 2013 ) , and crude oil exposures strongly induced fgf7 expression in the livers of juvenile polar cod ( Andersen et al . , 2015 ) . Both genes are promising new biomarkers for future studies , but tissue localization during craniofacial development in embryonic fish is needed to confirm a role in this adverse outcome pathway . In zebrafish exposed to the dioxin TCDD , foxq1 was up-regulated in the branchial arches ( Planchart and Mattingly , 2010 ) , but TCDD-induced craniofacial defects have been shown to be entirely secondary to cardiotoxicity ( Lanham et al . , 2014 ) . Lastly , our findings show how transcriptomics can inform chemical genetics and environmental forensics in non-model organisms . First , we used known spatial mRNA distributions in model species ( primarily zebrafish ) to more accurately phenotypically anchor the transcriptome data for crude oil-exposed haddock . This accelerates the pace of discovery , particularly given difficulties in applying zebrafish methods for in situ hybridization to wild species . Second , our results reveal transcriptional aspects of chemical injury in fish , in response to a global pollution threat . Although endocrine disrupting compounds act on steroid hormone receptors or epigenetically modify DNA or histones ( Walker and Gore , 2011 ) , it has been much less clear whether crude oil , which acts on protein targets ( e . g . [Brette et al . , 2014] ) , also influences mRNA levels as part of a widely studied developmental injury phenotype . In conclusion , our findings greatly expand our understanding of crude oil impacts to fish early life stages at a molecular level . The scientific focus on highly vulnerable fish embryos and larvae began with the 1989 Exxon Valdez disaster in Prince William Sound , Alaska . Major crude oil spills have continued worldwide—for example , the 2002 Prestige spill in Spain , the 2007 Hebei Spirit spill in Korea , and the 2010 Deepwater Horizon blowout in the northern Gulf of Mexico . An enduring challenge spans these and other spills: namely , the accurate determination of fisheries losses as a consequence of oiled spawning habitats . Our results more clearly define the different categories of developmental injury . We have also identified responsive genes that hold promise as sensitive molecular indicators of physiological stress . These can be developed into novel assessment tools for diagnostic use in future spill zones . Lastly , differential patterns of gene expression in oil-exposed haddock provide insight into fundamental but still poorly understood developmental processes in marine fish , including calcium-mediated excitation-transcription coupling , fluid balance , lipid mobilization , and buoyancy . The samples transcriptome profiled here are the exact same samples described in Sørhus et al . ( 2016b ) , with the methodology for animal collection , maintenance , and crude oil exposure provided therein . Briefly , fertilized eggs were collected from tanks with wild captured Atlantic haddock ( collected spring 2013 ) and transferred to indoor egg incubators ( 7°C ) . At experiment start , eggs from the egg incubators were transferred to 50 L exposure tanks ( 7 . 8°C ) . Embryonic and larval stages of Atlantic haddock were exposed to two doses , low and high dose , in addition to a pulsed high dose ( see Figures 3 and 4 for details ) . Heidrun oil blend was artificially weathered by distillation and then pumped into the dispersion system using a HPLC pump ( Shimadzu , LC-20AD Liquid Chromatograph Pump ) with a flow of 5 μL/min together with a flow of seawater of 180 mL/min . The system described in Nordtug et . al 2011 ( Nordtug et al . , 2011 ) generates an oil dispersion with oil droplets in the low μm range with a nominal oil load of 26 mg/L ( stock solution ) . Water samples were collected from each exposure tank at the beginning of each experiment and subjected for detailed PAH analysis . At end of exposure , pooled samples of eggs and larvae were extracted by solid liquid extraction and purified by solid phase extraction prior to analysis by GC-MS/MS ( Sørensen et al . , 2015 ) to reveal the PAH content in animals . After end of exposure , the animals were transferred to new tanks with clean water . Images of the animals were collected from 12 and 8 stages during and after embryonic and larval exposure , respectively . Videos from 60 ( embryonic ) and 48 ( larval ) individual embryos/larvae per treatment per stage were collected from 9 dpf , 0 dph , and 3 dph ( embryonic exposure ) and 2 dph and 9 dph ( larval exposure ) . The experiments included four replicate tanks for each dose , and pooled samples from three replicate tanks for each dose were subjected for sequencing ( see details below ) . Total RNA was isolated from frozen pools of embryos and larvae using Trizol reagent ( Invitrogen ) per manufacturer instructions . This included a DNase treatment step using a TURBO DNA-free kit ( Life Technologies Corporation ) . RNA was quantified using a Nanodrop spectrophotometer ( NanoDrop Technologies ) , and confirmed using a 2100 Bioanalyzer ( Agilent Technologies ) . cDNA was subsequently generated using SuperScript VILO cDNA Synthesis Kit ( Life Technologies Corporation ) , according to the manufacturer’s instructions . The cDNA was normalized to obtain a concentration of 50 ng/µL . Six responsive genes from the transcriptome were validated by real-time quantitative PCR ( qPCR ) ( Figure 3—figure supplement 2 [Embryonic exposure] , Figure 4—figure supplement 2 [Larval exposure] ) . Specific primers and probes ( Supplementary file 1N ) for a reference gene ( ef1α ) and the six DEGs ( cp1a , wnt11 , kcnh2 , nac1 , cac1c , and at2a2 ) were designed with either Primer Express Software ( Applied Biosystems ) or Eurofins qPCR probe and primer design software ( Eurofins Scientific ) , according to the manufacturer’s guidelines . The two methods generally yielded the same quantitative trends . Primer and probe sequences are shown in Supplementary file 1J . TaqMan PCR assays were performed in duplicate , using 384-well optical plates on an ABI Prism Fast 7900HT Sequence Detection System ( Applied Biosystems ) with settings as follows: 50°C for 2 min , 95°C for 20 s , followed by a 40 cycles of 95°C for 1 s and 60°C for 20 s . For each 10 μL PCR reaction , a 2 μL cDNA 1:40 dilution was mixed with 200 nM fluorogenic probe , 900 nM sense primer , and 900 nM antisense primer in 1xTaqMan Fast Advanced Master Mix ( Applied Biosystems ) . Gene expression was calculated relative to the exposure time zero sample ( 2 dpf and 0 dph in embryonic and larval exposures , respectively ) using the ΔΔΔCt method , generating reference residuals ( Edmunds et al . , 2014 ) from ef1a and at2a2 . cDNA library preparation and sequencing was performed by the Norwegian Sequencing Centre ( NSC , Oslo , Norway ) using the Illumina TruSeq RNA Sample Preparation Kit . A total of 132 samples were sequenced and 126 were subjected for analysis: Three ( control , low and high dose ) or two ( pulse ) biological replicates for each dose from six stages during and after embryonic exposure and three biological replicates for each dose from five stages during larval exposure ( see Figures 3 and 4 ) . Paired-end libraries were sequenced on the Illumina HiSeq2500 platform . The raw data are available from the Sequence Read Archive ( SRA ) at NCBI ( Accession ID: PRJNA328092 ) . The high sequence similarity between the two species justified use of the cod template , and we chose a verified model over a reference-free de novo transcriptome approach to avoid fragmentation noise and false positives from un-collapsed genes . The average sequence similarity between mapped haddock reads and the previously verified cod genome was 98 . 4% . Moreover , of the 20 , 954 annotated cod genes , there were 18 , 990 ( 90 . 6% ) corresponding haddock genes with at least 10 reads in one sample . Thus , the RNA sequencing data were mapped to the coding sequences of the cod genes ( Star et al . , 2011 ) using the Bowtie aligner ( RRID:SCR_005476 ) ( Langmead et al . , 2009 ) and annotated as described in Sørhus et al . ( 2016a ) . Samtools idxstat ( RRID:SCR_002105 ) ( Li et al . , 2009 ) was used to extract the number of mapped reads which were then normalised to the total number of mapped sequences . NOISeqBIO ( RRID:SCR_003002 ) ( Tarazona et al . , 2011 , 2015 ) was used to identify differentially expressed genes ( DEGs , threshold of 0 . 95 ) . The total number of reads averaged 41 . 39 million per sample and the mapping efficiency averaged 32 . 69% , giving an overall average of 13 . 51 million mapped paired end reads ( 125 bp ) for each sample . Given the absence of untranslated regions ( UTRs ) and mitochondrial genes from the reference cod genome , reads with a UTR sequence and all reads for mitochondrial genes were excluded from the haddock analyses . Only genes with 10 reads or more in at least one of the samples were included for further analysis . Kyoto Encyclopedia of Genes and Genomes ( KEGG ) pathways analysis ( RRID:SCR_012773 ) ( Kanehisa et al . , 2012 ) was performed by mapping the KEGG annotated DEGs from NOISeqBIO to KEGG pathways as described in the KEGG Mapper tool . Heat maps were generated from fold change data in R ( R Core Team , 2013 ) and Venn diagrams were created using the web-tool , Venn ( http://bioinformatics . psb . ugent . be/webtools/Venn ) . Individual genes involved in cardiac and craniofacial development , osmoregulation and lipid metabolism , or not directly linked to KEGG pathways were curated manually in an intensive process that took a full year . In a previous effort , we characterized the transcriptome of normally developing haddock embryos and larvae ( Sørhus et al . , 2016a ) . Through extensive literature searches ( PubMed , Web of Science , and Google Scholar ) and reading , lists of roughly 150 key genes involved in cardiac and craniofacial development and cardiac function were assembled . After obtaining the oil exposure RNA-Seq dataset , these lists were expanded by further literature searches . Genes of interest were identified as ones that showed strong dose-dependency and had potential phenotypic association based on the literature . We relied heavily on the expression database at the Zebrafish Information Network ( www . zfin . org ) to determine whether individual genes of interest were expressed in the relevant tissues at the appropriate developmental stage to be associated with phenotypes , helping to narrow in on key linkages . This was performed both by searching individual gene names and by generating lists of all genes expressed in a specific tissue at a time relevant to the phenotypes ( e . g . all genes expressed in the ethmoid plate between segmentation and hatch ) . These lists were cross-referenced to the list of DEGs to identify candidates . If zfin . org lacked expression data , we searched for any papers describing tissue localization by in situ hybridization in other model or non-model species . Qiagen’s Ingenuity Pathway Analysis ( IPA version 01–04 ) ( RRID:SCR_008653 ) was used subsequent to our manual curation . The fold change and p values were extracted from the original NOISeqBIO output . Fold change values were multiplied by −1 to flip the direction ( opposite the convention used by IPA ) , and any fold change values between −1 and 1 was set to 1 , as non-significant values and those below 10 reads were originally set to 0 . 5 and 0 . 75 . The cod genes were then BLASTed ( version 2 . 3 . 0+ ) against the ENSEMBL zebrafish ( GRCz10 ) and human ( GCRh38 ) databases . The top match for each gene , filtered based on e-value ( cutoff 10−5 ) , was used to build a mapping table of the genes to the IPA database . For genes that lacked mapping at that point , SwissProt information was used to manually map . The mapping information was then combined with the fold change and p-value data and uploaded to IPA . This resulted in 17608 of 20954 genes ( 84% ) mapping to the IPA database . Data were then analyzed using a fold change cutoff of 2 . 0 for both up- and down-regulation , considering relationships only when confidence was equal to the experimentally observed level . For this analysis , IPA uses a built-in right-tailed Fisher Exact Test to calculate p-values for significant functions and pathways . After each time point was analyzed individually in this manner , data outputs were used in comparison analyses in which the control dataset was compared to each dose at each time point . The output of the comparison analyses included rankings of top pathways in a variety of categories and subcategories . These rankings were expanded in the IPA software interface and inspected manually to generate Table 3 and Supplementary file 1B and 1C . Supporting dataset 1: Expression of all genes in control , low dose , pulse dose , high dose during and after embryonic exposure . Highlighted FC: Prob ≥ 0 . 95; Not highlighted FC: Prob 0 . 8–0 . 95; FC = 0 . 5: Not significant; FC = 0 . 75: Less than 10 reads in both treatment and control . SP , swissprot; GB , genebank; E1 , 3 dpf; E2 , 6 dpf; E3 , 10 dpf; E4 , 11 dpf; E5 , 0 dph; E6 , 3 dph; C , control; L , low dose; P , pulse dose; H , high dose; FC; fold change; prob; probability BioNoiseq . ( Sørhus et al . , 2017 ) . Supporting dataset 2: Expression of all genes in control , low dose , pulse dose , high dose during larval exposure . Highlighted FC: Prob ≥ 0 . 95; Not highlighted FC: Prob 0 . 8–0 . 95; FC = 0 . 5: Not significant; FC = 0 . 75: Less than 10 reads in both treatment and control . SP , swissprot; GB , genebank; L1 , 1 dph; L2 , 3 dph; L3 , 9 dph; L4 , 14 dph; L5 , 18 dph; C , control; L , low dose; P , pulse dose; H , high dose; FC; fold change; prob; probability BioNoiseq . ( Sørhus et al . , 2017 ) .	Accidental oil spills are a worldwide threat to ocean life . Fish eggs and larvae are especially vulnerable; therefore oil spills in areas where fish spawn are of great concern . Fish embryos exposed to crude oil grow slower than normal as larvae and juveniles and often show defects in the heart , face and jaw . However , the underlying mechanisms behind these defects are largely unknown . Working with the Atlantic haddock ( Melanogrammus aeglefinus ) , Sørhus et al . have now examined fish embryos and larvae that had been exposed to crude oil , and identified those genes that were more active or less active than normal . The findings add further support to the idea that exposure to crude oil causes heart and face defects because it interferes with how the cells that develop into these structures use calcium ions . Signals sent via calcium ions are not only important for the contraction of muscle cells , but they are also essential for regulation of some genes . So , by interfering with the circulation of calcium ions , crude oil can have consequences for both how muscles work and how genes are regulated . Sørhus et al . also report two previously uncharacterized defects . Firstly , genes that help to regulate the ion and water content of the tissues were highly affected in young fish exposed to crude oil . Some of the genes were more active than normal , while others were less active . This finding in particular would explain why oil-exposed embryos often accumulate fluids , and suggests that the larvae may have altered buoyancy too . Secondly , the oil-exposed embryos showed signs of a shortage of cholesterol and other fatty molecules . This is most likely because they absorbed less material from their yolk , which could also explain why larvae exposed to crude oil grow more slowly than normal . Finally , in the future , these newly identified genes connected to crude oil toxicity could be used as diagnostic markers to confirm oil-induced injury in fish , and monitor the health of fish populations in the ocean .	[ "Abstract", "Introduction", "Results", "Discussion", "Materials", "and", "methods" ]	[ "ecology", "genetics", "and", "genomics" ]	2017	Novel adverse outcome pathways revealed by chemical genetics in a developing marine fish
Morphogen gradients induce sharply defined domains of gene expression in a concentration-dependent manner , yet how cells interpret these signals in the face of spatial and temporal noise remains unclear . Using fluorescence lifetime imaging microscopy ( FLIM ) and phasor analysis to measure endogenous retinoic acid ( RA ) directly in vivo , we have investigated the amplitude of noise in RA signaling , and how modulation of this noise affects patterning of hindbrain segments ( rhombomeres ) in the zebrafish embryo . We demonstrate that RA forms a noisy gradient during critical stages of hindbrain patterning and that cells use distinct intracellular binding proteins to attenuate noise in RA levels . Increasing noise disrupts sharpening of rhombomere boundaries and proper patterning of the hindbrain . These findings reveal novel cellular mechanisms of noise regulation , which are likely to play important roles in other aspects of physiology and disease . Cells responding to signals need to be able to distinguish these signals from random fluctuations ( i . e . , noise ) and presumably have evolved mechanisms to do so . Noise is inherent in biological systems , but until recently we have lacked the tools to study such complexity in vivo ( Gregor et al . , 2007; Holloway et al . , 2011 ) . Noise in signaling pathways arises from many sources , including stochastic variation in transcription , protein synthesis , and cellular environment ( Elowitz et al . , 2002 ) . Morphogens are long-range signals thought to induce different cell behaviors in a concentration-dependent manner , but how such graded signals can be established in the face of noise and how they specify sharp boundaries of target gene expression remain unclear . Retinoic acid ( RA ) is thought to act as a morphogen in the embryonic vertebrate hindbrain to pattern cells into a series of segments , called rhombomeres , which give rise to different domains of the adult brainstem ( White and Schilling , 2008; Niederreither and Dollé , 2008 ) . A small hydrophobic signaling molecule derived from dietary precursor Vitamin A , RA is produced in mesoderm near the head/trunk boundary and forms an anteriorly declining gradient across the hindbrain progenitor domain , in part through the activities of RA-degrading enzymes , Cyp26s ( Sirbu et al . , 2005; Hernandez et al . , 2007; White et al . , 2007 ) . In particular , self-enhanced degradation through induction of Cyp26a1 by RA itself was shown to be critical for gradient formation ( White et al . , 2007 ) . Excess RA during human embryogenesis can cause anterior-posterior ( A-P ) patterning defects , and RA has been implicated in the development and maintenance of numerous cell types as well as in cancers ( leukemia ) , stem cells ( pancreas ) and regenerating organs ( cardiomyocytes ) ( Duester , 2008; Tang and Gudas , 2011; Rhinn and Dolle , 2012 ) . Due to its hydrophobic nature , RA requires proteins to bind and transport it through extracellular and intracellular space , which enhances robustness ( Cai et al . , 2012 ) but also introduces various sources of noise ( Schilling et al . , 2012 ) . Our computational models suggest that noise in RA signaling can also play a positive role in hindbrain segmentation through noise-induced switches in gene expression at rhombomere boundaries ( Zhang et al . , 2012 ) , but until recently methods were lacking to test this hypothesis . Here we present a novel methodology utilizing fluorescence lifetime imaging microscopy ( FLIM ) and a phasor analysis ( phasor-FLIM ) to study the abundance of endogenous RA in vivo in zebrafish embryos . Using this new tool to visualize endogenous RA in living cells we quantify variability in RA levels and provide some of the first evidence that cells actively control the magnitude of noise in a signaling molecule in a multicellular system in vivo . We took advantage of the fact that RA is a fluorescent molecule to quantify its endogenous abundance in vivo in the developing zebrafish hindbrain . Due to the low abundance of RA in cells and its wide spectra of absorbance and emission , traditional fluorescence microscopic techniques fail to detect RA specifically . We opted instead to visualize RA by its unique fluorescence lifetime , rather than its fluorescence intensity . Focusing on the presumptive neural ectoderm of mid-gastrula stage embryos ( 8–8 . 5 hr post fertilization ) ( Kimmel et al . , 1995 ) we used FLIM to measure the relative abundances of RA as a function of cell position along the A-P axis . This revealed that intracellular free RA forms an anteriorly-declining gradient ( Figure 1A , B ) , similar to that previously reported with FRET reporters for RA ( Shimozono et al . , 2013 ) and suggested by the pattern of RARE-lacZ expression in late-gastrula mouse embryos ( Sirbu et al . , 2005 ) . Relative abundances were calculated using phasor-FLIM ( Digman et al . , 2008 ) , where within the phasor space , each individual fluorescent species is represented in a characteristic and invariable position . Mixtures of molecules generate a FLIM signature that lies along a line connecting the positions of the individual component species and the position in that line is weighted according to the relative abundances ( Figure 1—figure supplement 1A ) . Because we know the absolute position of pure RA ( in the lower right corner of phasor space ) ( Figure 1—figure supplement 1 ) ( Stringari et al . , 2011 ) , we can use the Cartesian distance within this space as a measure of the relative abundance of RA expressed as 1-dRA ( Figure 1—figure supplement 1B ) . We observed that 1-dRA increased progressively as measurements were taken further posteriorly within the hindbrain field , suggesting that our FLIM approach is sensitive enough to detect endogenous RA gradients . An ordinary least squares regression analysis of gradient shape could not distinguish between exponential , linear , and quadratic fits , but confirmed the presence of a gradient ( Figure 1—figure supplement 2 ) . Unfortunately FLIM is also sensitive to the fluorescence emitted by transgenic markers of rhombomeres or other landmarks in embryos , making it difficult to determine precise segmental locations within the hindbrain field . 10 . 7554/eLife . 14034 . 003Figure 1 . Measuring RA gradients in zebrafish embryos with Phasor-FLIM . ( A ) Example of a zebrafish embryo at mid-gastrula stage ( 8 . 5 hr post-fertilization ) with the imaging area ( black square – encompassing positions 230–330 in B-D ) in the neural ectoderm ( NE ) centered ~200 μm from the advancing blastoderm margin ( white line ) ( A: anterior , P: posterior , Y: yolk ) . Scale bar = 150 μm . ( B-D ) Plots of the relative abundance of RA ( as the difference 1-dRA ) versus position in μm along the A-P axis ( anterior to the left ) in WT ( B ) , DEAB-treated ( C ) , and DEAB-treated embryos co-treated with 0 . 7 nM exogenous RA ( D ) . Solid curves in ( B-D ) represent best fit . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 00310 . 7554/eLife . 14034 . 004Figure 1—figure supplement 1 . Phasor-FLIM detects relative levels of RA . ( A ) Schematic representation of how shifts in measurements displayed on the phasor plot reflect relative levels of each component in a complex mixture . Each plotted position is the weighted linear composition of the positions of all of the autofluorescent constituents ( rhodamine , fluorescein and RA ) of the mixture . ( B ) Graphic representation of the measured distance used to calculate the relative abundance of RA ( 1-dRA ) . The diameter of the phasor plot equals 1 and thus , the function 1-dRA can never reach zero ( even in total absence of RA ) . ( C ) Fluorescence intensity image of a zebrafish embryo with an implanted , RA-infused oil droplet to create an ectopic source ( S ) ; scale bar = 50 um . Red and purple circles represent the positions measured and graphed for the oil + RA embryo ( top row ) . The plots represent the relative abundance of RA ( as the difference 1-dRA ) with respect to distance from the source ( lines represent best fit curves ) . Analysis of the RA distribution from an ectopic RA source at the 3rd harmonic , corresponding to measurements at the 1st harmonic , reveals that the observed differences are due to changes in the relative abundance of RA . Embryos injected with oil alone ( bottom row ) show no detectable differences in the distribution or RA with distance . Plots presented are representative figures from 6–12 embryos from each condition , with three independent datasets . ( D ) mCherry fluorescence in r3 and r5 combined with DIC in a living Tg ( shhb:KalTA4 , UAS-E1b:mCherry ) transgenic zebrafish hindbrain ( anterior to the left ) used for FLIM measurements . White boxes represent areas assessed by FLIM in E and F . ( E ) Phasor plot showing typical FLIM signatures for these three areas in wildtype embryos ( blue ) , embryos injected with Aldh1a2-MO ( red ) or injected with Aldh1a2-MO and subsequently rescued by transplanting WT paraxial mesoderm ( green ) . The FLIM signatures in rhombomeres of MO-injected embryos lie further away from that of absolute RA , indicating reduced RA concentration . Transplanted WT mesoderm partially rescues the positions of FLIM-signatures for each rhombomere in the phasor plot . ( F ) Quantification of the distances of r2-r4 and r4-r6 in the phasor space . The separation of RA levels in these rhombomeres is significantly reduced in MO-injected embryos and rescued by transplanted WT mesoderm . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 00410 . 7554/eLife . 14034 . 005Figure 1—figure supplement 2 . Regression analysis of RA gradient shape . ( A ) Fit results for ordinary least squares regression analyses of the pooled dataset . Note that the red line for the exponential fit is hidden under the linear and quad2 fits . Column 2 shows the p-values for a t-test on each coefficient . In the quadratic case , the linear term was not significant , and thus it was re-fit as Quadratic 2 . Column 4 shows the correctness of the fit as the adjusted R2 . Each of the fits ( with all significant terms ) explains similar amounts of the variance , and thus it is hard to differentiate between curve types . However , in each case there is a highly-significant ( p<10–14 with associated F-test values similar ) increasing trend , indicating the presence of a gradient . ( B ) To compare anterior and posterior portions of the gradient , we split the dataset at x = 310 and ran the fits on the two portions . x < 310 . Results were almost identical to the full fit ( A ) . As before , the linear term in the two-term quadratic was not significant . x > 310 . Results were almost identical to the full fit ( A ) , including the exponential curve . Here the quadratic fit is clearly not statistically significant , indicating it is not a mix of behaviors . ( C ) The exponential curve is based on the linear fit of x with the logarithm of [RA] . Note that the magnitude of differences in [RA] is sufficiently small that the logarithmic change basically becomes a translation . Thus on the semi-translated data lower points in the middle cause the linear fit of the log to have a small slope , making the exponential of the coefficient small enough to make its resulting fit look linear . Thus , given measured differences in relative RA abundance , exponential , linear , and quadratic fits are indistinguishable , even at the tail end of the gradient . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 005 To confirm that with phasor-FLIM we could detect RA specifically and its relative levels , we treated embryos with 10 μM DEAB to prevent the enzymatic conversion of retinal to RA , which eliminated the gradient at mid-gastrula stage ( Figure 1C ) , and incubating these in 0 . 7 nM RA re-established the gradient as expected ( Figure 1D ) ( White et al . , 2007 ) . We generated artificial gradients of RA by injecting embryos with RA saturated mineral oil and found that the relative abundance of RA decreased as a function of the distance from the source in two orthogonal axes ( Figure 1—figure supplement 1C ) . Phasor analysis of the third harmonic of the laser pulse frequency ( 240 MHz instead of the standard 80 MHz ) revealed similar gradients . Analyzing a different harmonic de-couples dRA from any other component of the mixture – i . e . the locations of each component in the phasor plot will vary independently of changes in dRA - and thus provides an independent means of confirming the specificity of the RA phasor-FLIM signature ( Figure 1—figure supplement 1C ) . We next asked if we could detect RA gradients at later stages , when rhombomere boundaries are being established , by performing FLIM measurements in the transgenic line MÜ4127 ( Egr2b:mCherry ) , which labels rhombomeres 3 and 5 ( r3 , r5 ) , in regions devoid of transgene fluorescence to avoid interference ( Distel et al . , 2009 ) . We found a similar graded increase in dRA on the phasor plot in r4 and r2 relative to r6 at 24 hpf ( Figure 1—figure supplement 1D–F ) ( i . e . 1- increases posteriorly ) , suggesting a graded reduction in RA content anteriorly . Injection of embryos with morpholino oligonucleotides ( MOs ) targeting Aldh1a2 , to inhibit RA synthesis greatly reduced the separation between FLIM signatures in r2 , r4 and r6 , which was partially rescued by transplantation of wildtype ( WT ) paraxial mesoderm to restore the local RA source ( Figure 1—figure supplement 1D–F ) ( White et al . , 2007 ) . These results show that the RA gradient persists during gastrulation and establishment of rhombomeres . Because phasor-FLIM measures endogenous RA and is not biased by the Kd of a reporter , in contrast to the FRET method previously published ( Shimozono et al . , 2013 ) , it is more direct and more reliably reflects real-time RA dynamics . Thus we next applied this technique to measure stochastic fluctuations ( noise ) in RA levels across the embryonic hindbrain . Our models predict that the magnitude of such noise is large ( Lander , 2013 ) , as we have argued that these fluctuations help cells switch between stable states of gene expression and thereby sharpen gene expression boundaries , i . e . noise-induced switching ( Schilling et al . , 2012; Zhang et al . , 2012 ) , but direct evidence of such noise is lacking . To assess 'spatial noise' we analyzed five consecutive parallel rows of cells in which each cell within a row lies at the same A-P position within the hindbrain field . This revealed variability as high as 45% of the entire magnitude of the gradient among cells within a row ( Figure 2A ) , consistent with the levels of noise predicted by our stochastic mathematical models ( White et al . , 2007; Zhang et al . , 2012 ) ( Figure 2B; Supplementary file 1 ) . To assess 'temporal noise' we analyzed the same cells repeatedly at 12-second intervals , and also found that their RA levels were very noisy ( Figure 2C , D ) . 10 . 7554/eLife . 14034 . 006Figure 2 . RA gradients are noisy in space and time . ( A , B ) Spatial noise . Plots show relative abundance of RA in five parallel rows of cells ( each color corresponds to a different row ) along the A-P axis of the neural ectoderm within a single embryo . ( A ) Experimental – each point represents the integrated signal of 40 consecutive FLIM measurements ( 2 . 7 min ) ( solid line represents best fit ) . ( B ) Computational – line represents the mean of 500 model simulations . ( C , D ) Temporal noise . Graphs show variability in relative abundance of RA in five single cells ( each color corresponds to a different cell ) at equivalent A-P positions over time . ( C ) Experimental – FLIM measurements were taken every 12 s . ( D ) Computational – colors correspond to individual cells for each stochastic realization . See also Supplementary file 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 00610 . 7554/eLife . 14034 . 007Figure 2—figure supplement 1 . Instrument noise cannot account for noise in phasor-FLIM measurements of RA . Comparison of temporal noise in FLIM measurements of three different solutions ( fluorescein in aqueous KOH , pH 9 . 0 , rhodamine in water and RA in DMSO ) and single cells in 9 different wildtype ( WT ) embryos . The graph represents the deviation from the mean position on the phasor plot for the solutions and the deviation from the mean abundance of RA for the cells . Each point corresponds to values at a single time point and the lines represent the mean and standard deviation . The variability in the measurements from cells is larger than those for solutions and vastly exceeds the maximum theoretical uncertainty due to photon shot noise . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 007 In order to rule out the possibility that the noise in our measurements was introduced by systematic artifacts or the measurement itself , we compared the variance in FLIM measurements of pure solutions of fluorescein , rhodamine and RA with the noise measured in cells of 9 independent embryos and found that noise in embryos is two orders of magnitude greater ( Figure 2—figure supplement 1 ) . We also calculated the maximum theoretical uncertainty due to photon shot noise and verified that the noise we measured is significantly larger ( Colyer et al . , 2008 ) . Thus the fluctuations in RA levels that we observed in embryos are clearly biological in origin . Noise in RA levels could be largely irrelevant for downstream gene expression if its frequency is faster than cellular responses , and is therefore averaged out . To address this possibility we performed an autocorrelation analysis of our temporal noise measurements using a moving window on each cell to search for significant lags . This revealed significant correlations ( lags 13 and 14 ) corresponding to a predominant frequency on the order of 2 . 7 min . This is significantly slower than the half-life of the RA-Crabp2a complex , which is approximately 1 . 7 min ( Dong et al . , 1999 ) . Because Crabp2 helps deliver RA to its nuclear receptor , and considering the scale of noise in transcriptional activation , noise at this time scale in RA signaling could propagate downstream . Thus it seems likely that cells possess mechanisms to limit this noise propagation . If cells actively control noise in RA signaling , they likely do it through intracellular RA-binding proteins , Crabps , or RA-degrading enzymes , Cyp26s , that can rapidly alter freely available RA ( Kleywegt et al . , 1994 ) and both of which have been shown to play critical roles in RA signaling ( Sirbu et al . , 2005; Hernandez et al . , 2007; White et al . , 2007; Cai et al . , 2012 ) . To test these candidates we reduced the amount ( microinjected MOs ) or overexpressed ( microinjected mRNA ) Crabp2a and Cyp26a1 in zebrafish embryos and measured noise in RA at mid-gastrula stages . Strikingly , MO depletion of Crabp2a increased temporal noise in RA without altering the mean RA level at a given A-P position , while overexpression of Crabp2a decreased variability in RA , again without altering the mean levels of RA ( Figure 3 ) . In contrast depletion or overexpression of Cyp26a1 increased or decreased mean RA levels , respectively , without altering noise . These results agree with simulations using our stochastic mathematical model in which we altered the levels of Crabp2a or Cyp26a1 ( Figure 3—figure supplement 1 ) . These results reveal a novel , active role for Crabps in modulating noise in RA . 10 . 7554/eLife . 14034 . 008Figure 3 . Crabp2a but not Cyp26a1 attenuates level of noise in RA . Analysis of the temporal distribution of RA’s relative abundance in wildtype ( WT ) , Crabp2a morpholino ( MO ) -injected , Crabp2a mRNA-injected ( gain-of-function - GOF ) and Cyp26a1 MO and mRNA-injected zebrafish embryos . Each column shows the signal obtained for a single representative cell and each point corresponds to a single time point . Lines represent the mean and standard deviation . Embryos with reduced or increased levels of Crabp2a show increased and decreased variability in RA , respectively , while altering Cyp26a1 changes the mean concentrations but not the variance . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 00810 . 7554/eLife . 14034 . 009Figure 3—figure supplement 1 . Crabp2a actively modulates RA signal noise . Results of three realizations of our stochastic mathematical model analyzing the temporal distribution of RA’s relative abundance in wildtype ( WT ) , Crabp2a morpholino ( MO ) -injected , Crabp2a mRNA ( gain-of-function - GOF ) injected , Cyp26a1 MO-injected and Cyp26a1 mRNA-injected zebrafish embryos . Each column shows the signal obtained for a single realization for a single cell . Each point corresponds to a single time point . Lines represent the mean and standard deviation . Embryos with altered Crabp2a expression show changes in the variability in free intracellular RA , while embryos with altered Cyp26a1 expression show changes in the mean levels of RA as in the experimental case shown in Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 009 To determine how altering RA levels influences noise in gene expression within the hindbrain we disrupted Crabp2a or Cyp26a1 and assayed expression of krox20 in r3 and r5 . In situ Hybridization Chain Reaction ( HCR ) ( Choi et al . , 2014 ) allowed us to quantify krox20 expression levels by manually segmenting confocal images and measuring total fluorescence of each cell ( Figure 4A ) ( Video 1 ) . All HCR analyses were performed on raw 3D data , with Z-projections performed post-analysis . Either raising or lowering Crabp2a levels increased variance in krox20 expression from cell to cell ( Figure 4B , C ) when normalized for heterogeneity in gene expression from embryo to embryo , as confirmed by single embryo qPCR ( Figure 4—figure supplement 1A ) . In addition , it decreased the sharpness of boundaries of krox20 expression in r3 and r5 , using a sharpness index calculated as the ratio between the length of the theoretical sharp boundary and the actual measured length of the boundary ( Figure 4D , E ) ( Figure 4—figure supplement 1B ) ( Materials and methods ) . These results suggest that , in contrast to its effects on RA levels where Crabp2a appears to attenuate noise , an optimal range of Crabp2a is required to induce sharp boundaries of gene expression in rhombomeres and too much Crabp2a is also detrimental to the system ( White et al . , 2007; Zhang et al . , 2012; Cai et al . , 2012 ) . Similarly , either raising or lowering Cyp26a1 levels increased variance in krox20 expression ( Figure 4 ) . Thus , while Crabp2a may play a unique role in reducing noise in RA levels it appears to function together with Cyp26a1 and potentially other RA signaling components in allowing robust expression of downstream targets . 10 . 7554/eLife . 14034 . 010Figure 4 . Both Crabp2a and Cyp26a1 attenuate noise in krox20 expression and facilitate rhombomere boundary sharpening . ( A ) Representative Z projections of r3 and r5 ( dorsal views , anterior to the left ) analyzed by hybridization chain reaction ( HCR ) for krox20 ( r3 , rhombomere 3; r5 , rhombomere 5; A , anterior; P , posterior ) . We performed all HCR analyses on raw 3D data and later generated Z-projections and enhanced contrast to simplify presentation . Colors correspond to total krox20 RNA in each cell as measured by total fluorescence intensity bracketed for maximum and minimum for the 5 conditions and represented in a linear scale . ( B ) Mean-centered analysis of krox20 expression of a subset of cells for r3 and r5 from 3 randomly selected embryos for each condition . ( C ) Sharpness indices of the r3/r4 boundary ( blue ) and r4/r5 boundary ( red ) for embryos from each of the treatment conditions . Bars correspond to s . d . ( D ) Analysis of the variance in boundary sharpness from the quantification in ( C ) . All perturbations yielded significant differences from wild-type controls , as noted in the Statistical Analysis . Therefore no asterisks were included to indicate columns representing statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 01010 . 7554/eLife . 14034 . 011Figure 4—figure supplement 1 . Single embryos show highly variable mean levels of krox20 expression and boundary sharpness . ( A ) Embryos were injected with mRNA coding for the membrane-bound GFP-CAAX and morpholinos ( MO ) against Crabp2a or Cyp26a1 or mRNA coding for Crabp2a-Myc or Cyp26a1-Myc or water ( WT ) . Ef1α was used as the standard and a homogeneous collection of mRNA from 100 wildtype embryos was used as the reference for the △△Ct method . Experiments were run in triplicates and repeated 4 ( four ) times . ( B ) A sharpness index was calculated using the ratio between the measured length of the theoretical sharp boundary and the actual measured length of the boundary of krox20 expression , consistent with previous models ( Zhang et al . , 2012 ) . In this example , sharpness of the r3-r4 boundary is the ratio of a/b and the index for the r4-r5 boundary is the ratio c/d . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 01110 . 7554/eLife . 14034 . 012Video 1 . 3D rendering of HCR dataset . 3D rendering shows the specific HCR signal on rhombomeres 3 and 5 ( red ) with very low non-specific signal in surrounding tissue , which appears evenly distributed . DAPI signal ( blue ) demarcates nuclei of cells that are either Krox20 positive ( red ) or negative ( no signal ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 012 Most studies of morphogen gradients and transcriptional noise have focused on the bicoid-hunchback transcriptional network in the Drosophila embryo , prior to cellularization and onset of zygotic transcription ( He et al . , 2012 ) . The findings in that system indicate that due to the slow diffusion rate of the Bicoid protein , hunchback expression is mostly influenced by its own intrinsic noise and transcriptional noise in the bicoid gene does not propagate ( Gregor et al . , 2007; Okabe-Oho et al . , 2009; Holloway et al . , 2011 ) . In contrast , we show that noise in a secreted signal in the multicellular context of the vertebrate hindbrain influences noise in expression of its transcriptional targets . Our FLIM measurements demonstrate noisy concentration gradients of RA along the A-P axis and reveal a novel role for Crabp2a in noise-attenuation distinct from that of Cyp26a1 . Crabp2a could control noise in RA levels rapidly by binding RA and facilitating its entry into cells or buffering its availability within the cytoplasm ( Maden et al . , 1989; Boylan and Gudas , 1992 ) and our previous studies have demonstrated its critical roles in signal robustness ( Cai et al . , 2012 ) . In contrast , both Crabp2a and Cyp26a1 inhibit noise in downstream targets of RA . Previous studies have shown that transcriptional inhibitors act as noise filters within narrow levels of expression , since outside of this range , transcriptional noise in their target genes increases ( Dublanche et al . , 2006 ) . Such a biphasic response resembles our results with Crabp2a and Cyp26a1 . Retinoic acid receptors ( RARs ) often act as transcriptional repressors until they bind RA . Thus Crabp2a and Cyp26a1 may modulate noise in RA targets by altering this balance between activation and repression . As such , both must be present within a narrow optimal range ( Dublanche et al . , 2006; White and Schilling , 2008 ) . These mechanisms are likely to be similar in other signaling systems and critical for embryonic development and adult physiology , as well as defective in human diseases . Unless otherwise noted , all of the reagents were obtained from Sigma-Aldrich ( St . Louis , MO ) . All-Trans Retinoic Acid and 4-Diethylaminobenzaldehyde were dissolved at 10 mM and 100 mM , respectively , in anhydrous DMSO to create stocks and kept at -20°C in the dark until used . Morpholino Oligonucleotides ( MOs ) against aldh1a2 , Crabp2a and Cyp26a1 were obtained from Gene Tools ( Philomath , OR ) and used as previously described ( White et al . , 2007; Cai et al . , 2012 ) . HCR reagents were obtained from Molecular Tools ( Pasadena , CA ) . Restriction enzymes and SuperScriptII reverse transcriptase kit was obtained from NEB ( Ipswich , MA ) . LightCycler 480 SYBR Green I Master mix was obtained from Roche ( Indianapolis . IN ) . mMESSAGE mMACHINE kit , DAPI , Trizol reagent and fluorescein reference standard were obtained from Life Technologies ( Eugene , OR ) . All animal work was performed under the guidelines of UCI’s IACUC . Embryos were obtained by natural crosses , raised in embryo medium ( EM ) , and staged according to Kimmel et al . 1995 . The AB strain was used for WT experiments . MU4127 transgenics ( Tg[shhb:KalTA4 , UAS-E1b:mCherry] ) to visualize rhombomeres 3 and 5 were kindly provided by Dr . Köster ( HelmholtzZentrum , München ) . For synthesis of mRNA three constructs were generated as templates . pCS2+GFP-CAAX was generated by isothermal assembly of a pCS2+ backbone digested with EcoRI and an amplimer of GFP-CAAX generated by PCR with the primers ( forward ) 5’-ggatcccatcgattcgTGGACCATGGTGAGCAAG-3’ and ( reverse ) 5’-gctcgagaggccttgTCAGGAGAGCACACACTTG-3’ . Two C terminal Myc-tagged constructs were generated by traditional restriction-ligation procedure using a pCS2+MT as the backbone . Crabp2a was inserted between the BamHI and the ClaI sites of the proximal MCS and Cyp26a1 was inserted between the BamHI and ClaI sites of the proximal MCS . mRNA was synthesized by digestion of the constructs with NotI-HF and in vitro transcription with mMESSAGE mMACHINE SP6 transcription kit . Phasor-FLIM refers to a combination of fluorescence lifetime imaging microscopy ( FLIM ) and a methodology to analyze FLIM data . Rather than using the traditional intensity of fluorescence to analyze microscopic samples , FLIM measures the lifetime fluorescence decay of the fluorophore . Fluorophores possess a characteristic lifetime of fluorescence that represents the time that takes an excited electron to relax back to its basal state emitting a photon . This technique eliminates most sources of noise present in intensity-based fluorescence microscopy techniques . This is due to the fact that most sources of noise , like thermal flickering or dark current have no lifetimes ( Colyer et al . , 2008 ) . However photon shot noise remains a source of uncertainty , but this is inversely proportional to the square-root of the fluorescence signal intensity . FLIM also requires a high numerical aperture objective ( 40X/NA 1 . 2 ) , which intrinsically has a short working distance , making it impossible to perform the measurements at lower magnification . Representing FLIM data in a phasor plot instead of a time-delay histogram allows analysis of the entire image , rather than pixel by pixel . In addition , because each molecular species is represented in a defined area of the phasor space , it allows analysis of samples with multiple fluorescent species ( Digman et al . , 2008 ) . Individual fluorescent molecules have a constant lifetime , independent of concentration . Because the phasor space operates linearly , analysis of relative concentrations of fluorophores in samples with complex mixes can be performed . Samples with complex mixtures of fluorescent molecules generate a FLIM signature in the phasor plot that corresponds to the linear combination of the positions of the individual fluorescent species in a weighted manner . By calculating the Cartesian distance in the phasor space one can calculate the relative contributions of the different constituents ( Digman et al . , 2008; Stringari et al . , 2011 ) . An additional advantage of this method over the use of reporters is the direct measurement of the endogenous fluorophore of interest in vivo and without the potential artifacts introduced by genetic manipulations/transgenic reporters . Genetically encoded FRET reporters published previously for RA , bind RA proportionally to their association/dissociation constants ( ka/kd ) and either emit or stop emitting a signal . This binding biases the data ( both spatially and temporally ) according to the binding constant of the reporter . Embryos were dechorionated and mounted dorsally on #1 . 5 coverslips with 1% low-melt agarose in EM without methylene blue . Acquisition and analysis was performed as previously described ( Stringari et al . , 2011 ) . Briefly , the embryos were imaged for 2 . 7 min ( for spatial analysis ) or in single frames ( 4 sec –for temporal analysis ) on a Zeiss 710 confocal microscope with a Ti:Sapphire laser ( Spectra-Physics , Newport Beach , CA ) as a two photon excitation source and an ISS A320 FastFLIM box coupled to two H7422P-40 photo-multiplying tubes ( Hamamatsu , Japan ) . Data acquisition and analysis were performed using SimFCS software ( LFD , Irvine , CA ) . Images were acquired with a 40X 1 . 2 NA water immersion objective . The excitation frequency used was 760 nm and in order to enrich the signal for RA , the emission was filtered through a 495LP dichroic mirror . Solutions of Rhodamine in water and Fluorescein in 100 mM KOH ( pH 9 . 0 ) were used as references . MOs were injected at the one-cell stage and cell transplantations were performed as previously described ( White et al . , 2007; Cai et al . , 2012 ) . For the generation of ectopic retinoic acid ( RA ) sources , mineral oil was infused with all-trans RA to saturation . Embryos were dechorionated and temporarily mounted in 1% low-melt soft agar in EM over coverslips . Drops of the RA saturated oil or oil alone were then injected in 6–12 embryos using a mouth pipette and a capillary needle . The embryos were then released and left to heal for two hours when they were mounted for FLIM imaging . This experiment was repeated 3 ( three ) times . mRNA was injected into one-cell embryos with glass micropipettes and a Narishige IM 300 microinjector with 50 pg of GFP-CAAX , 50 pg of Crabp2a-Myc or 100 pg of Cyp26a1-Myc . Expression verification was performed by microscopic observation for GFP or by Western blot with anti-Myc antibody ( clone 9E10 ) for Crabp2a-Myc and Cyp26a1-Myc . One-cell stage embryos were injected with mRNA coding for GFP-CAAX to assist in later segmentation . The embryos were then divided into five experimental groups and injected with Crabp2a morpholinos ( MOs ) , Crabp2a-Myc mRNA , Cyp26a1 MOs , Cyp26a1-Myc mRNA or 500 pl of water ( WT ) . Embryos were incubated at 28C in EM until 11 hr postfertilization . Embryos were then treated as previously described ( Choi et al . , 2014 ) . Briefly , embryos were dechorionated and fixed with fresh 4% PFA at 4°C for 16 hr , washed in PBS and dehydrated with methanol for 1 hr followed by graded rehydration . Embryos were then pre-hybridized at 45°C for 30 min in hybridization buffer . Embryos were hybridized in hybridization buffer containing 1 pmol of each of 5 ( five ) different DNA probes designed against Krox20 containing the B1 double initiator arms and Hoxb1a containing the B2 double initiator arms ( Table 1 ) at 45°C for 16 hr . Probe specificity was verified by blast search and controlled by adding the hairpins but no initiator probe , which showed no non-specific signal , as well as single probe experiments . Excess probe was removed and embryos were gradually buffer exchanged to 5xSSCT and washed in 5xSSCT for 3 . 75 hr at 45°C . Samples were then pre-amplified in amplification buffer for 30 min at room temperature ( RT ) after which they were left to amplify in amplification buffer containing B1H1 and B1H2 snap-cooled hairpins conjugated to Alexa 594 and B2H1 and B2H2 snap-cooled hairpins conjugated to Alexa 647 at room temperature for 16 hr . Finally the embryos were washed in 5xSSCT and counterstained with DAPI before mounting in soft agar on number 1 . 5 thickness coverslips for confocal imaging . Samples were imaged with a Leica SP8 scanning confocal microscope acquiring z-stacks covering the entire hindbrain as 12 bit 512 X 512 images and analyzed using ImageJ software . Experiments were performed with 12 embryos per condition and repeated 4 ( four ) times . Microscope settings were kept constant throughout . Mean intensity values obtained from HCR experiments were of 190000 for the WT embryos , with an average cell size of 85 pixels , making the average signal 50% of the maximum . 10 . 7554/eLife . 14034 . 013Table 1 . Sequences of the probes used for HCR corresponding to the specific genes as indicated and flanked by the corresponding adaptor sequences ( B1 or B2 ) . P1 , P2 , etc . corresponds to the different probes used for each gene . DOI: http://dx . doi . org/10 . 7554/eLife . 14034 . 013ProbeSequenceKrox20_B1-P15'-GAGGAGGGCAGCAAACGGGAAGAGTCTTCCTTTACGATATTAGAAGTGGCTGGGGGAGACTGAGGATGCAGGTGACGAGGATGCTGAGGATATATAGCATTCTTTCTTGAGGAGGGCAGCAAACGGGAAGAG-3'Krox20_B1-P25'-GAGGAGGGCAGCAAACGGGAAGAGTCTTCCTTTACGATATTGTGGAAAGGAACGCAGACGGGTCTTGATAGACCTCTCCGCATCCAGAGTAATATAGCATTCTTTCTTGAGGAGGGCAGCAAACGGGAAGAG-3'Krox20_B1-P35'-GAGGAGGGCAGCAAACGGGAAGAGTCTTCCTTTACGATATTAGGTTGGAAAAAGCCGGCGTAGTCCGGGATTATAGGGAACAACCCAGAGTATATAGCATTCTTTCTTGAGGAGGGCAGCAAACGGGAAGAG-3'Krox20_B1-P45'-GAGGAGGGCAGCAAACGGGAAGAGTCTTCCTTTACGATATTGTTAGAGGAGGCGGTAATTTGAAAGAGTCCAGCGGGCAGGAGAACGGTTTATATAGCATTCTTTCTTGAGGAGGGCAGCAAACGGGAAGAG-3'Hoxb1a_B2-P15'-CCTCGTAAATCCTCATCAATCATCCAGTAAACCGCCAAAAAAGTGTGGAAAGGGCCCGGGAACGCCTGGTCCAAGTGGTGGTATCCAGCCTAAAAAAGCTCAGTCCATCCTCGTAAATCCTCATCAATCATC-3'Hoxb1a_B2-P25'-CCTCGTAAATCCTCATCAATCATCCAGTAAACCGCCAAAAACAGTTCCACCATAGGTAAGGCCCATGCCAGTTTGATTTTGGTGCTGGTGAAAAAAAGCTCAGTCCATCCTCGTAAATCCTCATCAATCATC-3'Hoxb1a_B2-P35'-CCTCGTAAATCCTCATCAATCATCCAGTAAACCGCCAAAAATGTTGAGCATAGTCCGAGTTGGCGCAGGCCTGTGTCCCATAACTTGTTGTAAAAAAGCTCAGTCCATCCTCGTAAATCCTCATCAATCATC-3'Hoxb1a_B2-P45'-CCTCGTAAATCCTCATCAATCATCCAGTAAACCGCCAAAAAAGTACGCACCGGCCATAGAGCCATAGTGTGGACTGGCATTTGATGTTGAAAAAAAAGCTCAGTCCATCCTCGTAAATCCTCATCAATCATC-3'Hoxb1a_B2-P55'-CCTCGTAAATCCTCATCAATCATCCAGTAAACCGCCAAAAAGAGTGATCAGATTGATCCTCGAGGTCTTTAGACGAAGTGGAGGAAGCAGGAAAAAAGCTCAGTCCATCCTCGTAAATCCTCATCAATCATC-3' We defined a sharpness index as the ratio between the length of a perfectly sharp boundary and the actual measured length of the boundary according to the following equation:S=AsharpAreal=∑n=1N ( dnsharp×z ) ∑n=1N ( dnxy×z ) = ( ∑n=1Ndnsharp ) ×z ( ∑n=1Ndnxy ) ×z=∑n=1Ndnsharp∑n=1Ndnxy Where S is the sharpness index , Asharp is the area of the theoretical sharp boundary , Areal is the real measured area of the boundary , n is each individual slice in the z-stack , N is the total number of slices of the z-stack , dsharp is the minimum distance between the rhombomere’s lateral edges ( Figure 4—figure supplement 1B ) in XY ( the theoretical sharp boundary ) , dxy is the measured distance in XY of the boundary and Z is the thickness of each slice . Eight embryos injected for HCR were separated after dechorionation and total RNA extracted with 150 μl of Trizol reagent . After chloroform addition and separation of the aqueous phase , the samples were concentrated using the DNA-Free RNA Kit ( Zymo Research , Irvine , CA ) . Poly A-RNA was transcribed using SuperScriptII . SYBR Green qPCR reactions were performed with primers 5’-ATCTATTCGGTGGACGAGC-3’ and 5’-TAATCAGGCCATCTCCTGC-3’ for Krox20 and 5’-CAAGGGATGGAAGATTGAGC-3’ and 5’-AACCATACCAGGCTTGAGGA-3’ for EF1α . Primer sets were tested and confirmed to have an amplification efficiency of 2 . In order to study the individual variability in gene expression on each embryo , a homogeneous standard was generated with RNA pooled from 100 embryos at the same stage ( 3 somites-11 hpf- ) of development and reactions of this 'standard' were run in parallel . The △△Ct method was used to analyze the samples . All samples were run in triplicates and the experiment was repeated 4 ( four ) times . Unless otherwise noted , statistical analysis was performed using Prism 5 ( GraphPad software , La Jolla , CA ) . In experiments where the graded distribution was analyzed ( Figure 1 , 2A; Figure 1—figure supplement 1C ) , a comparison of the best-fit lines was performed . In assays where the variance was analyzed ( Figure 2 , 3 , 4 and associated figure supplements ) a one-way ANOVA of the coefficients of variation was performed . In order to establish significance in the changes in variation , a Levene’s test was performed using MATLAB ( MathWorks , Natick , MA ) . To establish the significance of the changes in mean values , a Newman-Keuls test was used . For the rescue experiment ( Figure 1—figure supplement 1D–F ) a two-way ANOVA analysis was performed and significance was established after the Bonferroni post test correction . To establish the predominant frequency of the noise ( Figure 2C ) an analysis of the datasets for different cells was performed using a moving window and searching for lags for which a correlation function would provide significant p-values using MATLAB ( MathWorks , Natick , MA ) . Lags 13 and 14 gave p-values between 0 . 08 and 0 . 01 . Average of these lags corresponds to a frequency of about 2 . 7 min . Boundary sharpness was calculated as the ratio between the length of the theoretical sharp boundary and the actual measured length of the boundary ( Figure 4—figure supplement 1B ) . All perturbations yielded significant differences from wild-type controls . Thus no asterisks were included to indicate columns representing statistical significance .	Animal cells need to be able to communicate with each other so that they can work together in tissues and organs . To do so , cells release signaling molecules that can move around within a tissue and be detected by receptors on other cells . We tend to assume that the signaling molecules are evenly distributed across a tissue and affect all the receiving cells in the same way . However , random variations ( noise ) that affect how many of these molecules are produced , how they move through the space between cells and how they bind to receptors makes the reality much more complex . Cells responding to the signal somehow can ignore this noise and establish sharp boundaries between different cell types so that neighboring cells have distinct roles in the tissue . Few studies have attempted to measure such noise or address how cells manage to respond to noisy signals in a consistent manner . Retinoic acid is a signaling molecule that plays an important role in the development of the brain in animal embryos . It forms a gradient along the body of the embryo from the head end to the tail end , but it has proved difficult to measure this gradient directly . Sosnik et al . exploited the fact that this molecule is weakly fluorescent and used microscopy to directly detect it in zebrafish embryos . The experiments show that retinoic acid forms a gradient in the embryos , with high levels at the tail end and lower levels at the head end . Sosnik et al . also found that there is a large amount of noise in the retinoic acid gradient . Two cells in the same position can have very different retinoic acid levels , and the levels in a particular cell can vary from one minute to the next . The experiments also show that proteins that interact with retinoic acid help to reduce noise within a cell . This noise reduction is important for sharpening the boundaries between different brain regions in the embryo to allow the brain to develop normally . A future challenge will be to see if similar retinoic acid gradients and noise control occur in other tissues , and if the noise has any positive role to play in development .	[ "Abstract", "Introduction", "Results", "and", "discussion", "Materials", "and", "methods" ]	[ "developmental", "biology", "short", "report", "computational", "and", "systems", "biology" ]	2016	Noise modulation in retinoic acid signaling sharpens segmental boundaries of gene expression in the embryonic zebrafish hindbrain

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.